wip

add example
prepare for wan2.2 self-forcing
2025-10-31 20:54:49 +00:00 · 2025-10-31 20:54:00 +00:00 · 2025-10-31 20:52:31 +00:00
279 changed files with 3363 additions and 18487 deletions
@@ -222,14 +222,3 @@ steps:
                  - TEST_TYPE=unit_test
                agents:
                  queue: "default"
-            - path:
-                - "scripts/lora_extraction/**"
-                - "pyproject.toml"
-                - "docker/Dockerfile.python3.12"
-              config:
-                command: "timeout 90m .buildkite/scripts/pr_test.sh"
-                label: "LoRA Extraction Tests"
-                env:
-                  - TEST_TYPE=lora_extraction
-                agents:
-                  queue: "default"
@@ -31,9 +31,9 @@ log "Setting up Modal authentication from Buildkite secrets..."
 MODAL_TOKEN_ID=$(buildkite-agent secret get modal_token_id)
 MODAL_TOKEN_SECRET=$(buildkite-agent secret get modal_token_secret)

-# Retrieve other secrets
 WANDB_API_KEY=$(buildkite-agent secret get wandb_api_key)
-HF_API_KEY=$(buildkite-agent secret get hf_api_key)
+
+WANDB_API_KEY=$(buildkite-agent secret get wandb_api_key)

 if [ -n "$MODAL_TOKEN_ID" ] && [ -n "$MODAL_TOKEN_SECRET" ]; then
    log "Retrieved Modal credentials from Buildkite secrets"
@@ -63,19 +63,19 @@ MODAL_ENV="BUILDKITE_REPO=$BUILDKITE_REPO BUILDKITE_COMMIT=$BUILDKITE_COMMIT BUI
 case "$TEST_TYPE" in
    "encoder")
        log "Running encoder tests..."
-        MODAL_COMMAND="$MODAL_ENV HF_API_KEY=$HF_API_KEY python3 -m modal run $MODAL_TEST_FILE::run_encoder_tests"
+        MODAL_COMMAND="$MODAL_ENV python3 -m modal run $MODAL_TEST_FILE::run_encoder_tests"
        ;;
    "vae")
        log "Running VAE tests..."
-        MODAL_COMMAND="$MODAL_ENV HF_API_KEY=$HF_API_KEY python3 -m modal run $MODAL_TEST_FILE::run_vae_tests"
+        MODAL_COMMAND="$MODAL_ENV python3 -m modal run $MODAL_TEST_FILE::run_vae_tests"
        ;;
    "transformer")
        log "Running transformer tests..."
-        MODAL_COMMAND="$MODAL_ENV HF_API_KEY=$HF_API_KEY python3 -m modal run $MODAL_TEST_FILE::run_transformer_tests"
+        MODAL_COMMAND="$MODAL_ENV python3 -m modal run $MODAL_TEST_FILE::run_transformer_tests"
        ;;
    "ssim")
        log "Running SSIM tests..."
-        MODAL_COMMAND="$MODAL_ENV HF_API_KEY=$HF_API_KEY python3 -m modal run $MODAL_TEST_FILE::run_ssim_tests"
+        MODAL_COMMAND="$MODAL_ENV python3 -m modal run $MODAL_TEST_FILE::run_ssim_tests"
        ;;
    "training")
        log "Running training tests..."
@@ -126,10 +126,6 @@ case "$TEST_TYPE" in
        log "Running unit tests..."
        MODAL_COMMAND="$MODAL_ENV python3 -m modal run $MODAL_TEST_FILE::run_unit_test"
        ;;
-    "lora_extraction")
-        log "Running LoRA extraction tests..."
-        MODAL_COMMAND="$MODAL_ENV HF_API_KEY=$HF_API_KEY python3 -m modal run $MODAL_TEST_FILE::run_lora_extraction_tests"
-        ;;
    *)
        log "Error: Unknown test type: $TEST_TYPE"
        exit 1
@@ -1,65 +1,82 @@
-name: Deploy Documentation
+# Sample workflow for building and deploying a Hugo site to GitHub Pages
+name: Deploy FastVideo Docs to Pages

 on:
+  # Runs on pushes targeting the default branch
  push:
-    branches: [ main ]
+    branches:
+      - main
    paths:
-      - 'docs/**'
-      - 'mkdocs.yml'
-      - 'requirements-mkdocs.txt'
-      - '.github/workflows/docs.yml'
+      - "docs/**/*.md"
+      - "fastvideo/examples/**/*.py"
  pull_request:
-    branches: [ main ]
+    branches: 
+      - main
+    types: [opened, ready_for_review, synchronize, reopened]
    paths:
-      - 'docs/**'
-      - 'mkdocs.yml'
-      - 'requirements-mkdocs.txt'
-      - '.github/workflows/docs.yml'
+      - "docs/**/*.md"
+      - "fastvideo/examples/**/*.py"

+  # Allows you to run this workflow manually from the Actions tab
+  workflow_dispatch:
+
+# Sets permissions of the GITHUB_TOKEN to allow deployment to GitHub Pages
 permissions:
  contents: read
  pages: write
  id-token: write

+# Allow only one concurrent deployment, skipping runs queued between the run in-progress and latest queued.
+# However, do NOT cancel in-progress runs as we want to allow these production deployments to complete.
 concurrency:
  group: "pages"
  cancel-in-progress: false

+# Default to bash
+defaults:
+  run:
+    shell: bash
+
 jobs:
+  pre-commit:
+    uses: ./.github/workflows/pre-commit.yml
+
+  # Build job
  build:
    runs-on: ubuntu-latest
+    needs: pre-commit
    steps:
      - name: Checkout
        uses: actions/checkout@v4
-
-      - name: Setup Python
+      - name: Setup Pages
+        id: pages
+        uses: actions/configure-pages@v5
+      - name: Set up Python
        uses: actions/setup-python@v5
        with:
-          python-version: '3.12'
-
+          python-version: "3.10"
      - name: Install dependencies
        run: |
-          python -m pip install --upgrade pip
-          pip install -r requirements-mkdocs.txt
-
-      - name: Setup Pages
-        uses: actions/configure-pages@v4
-
-      - name: Build documentation
-        run: mkdocs build
-
+          cd docs
+          pip install -r requirements-docs.txt
+      - name: Build docs
+        run: |
+          cd docs
+          make clean
+          make html
      - name: Upload artifact
        uses: actions/upload-pages-artifact@v3
        with:
-          path: ./site
+          path: ./docs/build/html

+  # Deployment job
  deploy:
    environment:
      name: github-pages
      url: ${{ steps.deployment.outputs.page_url }}
+    if: ${{ github.event_name == 'push' }}
    runs-on: ubuntu-latest
    needs: build
-    if: github.ref == 'refs/heads/main'
    steps:
      - name: Deploy to GitHub Pages
        id: deployment
@@ -14,8 +14,6 @@ wandb/
 *.pt
 cache_dir/
 wandb/
-venv/
-.venv/
 runs/
 samples/
 *validation/
@@ -39,13 +37,12 @@ dist/
 eggs/
 .eggs/

-# MkDocs documentation
-site/
-docs/getting_started/examples/
-docs/inference/examples/
-docs/training/examples/
-docs/distillation/examples/
-!requirements-mkdocs.txt
+# Sphinx documentation
+docs/_build/
+docs/source/getting_started/examples/
+docs/source/inference/examples/
+docs/source/training/examples/
+docs/source/distillation/examples/

 # VSCode
 .vscode/
@@ -64,7 +61,7 @@ docs/distillation/examples/
 !fastvideo/tests/ssim/reference_videos/**/*.mp4

 # Static images
-!docs/assets/images/**/*.png
+!docs/source/_static/images/**/*.png
 !comfyui/assets/**/*.png
 !comfyui/assets/**/*.gif

@@ -10,7 +10,6 @@ exclude: |
    demo/.*|
    predict\.py|
    scripts/.*|
-    prompts/.*|
    fastvideo/data_preprocess/.*|
    fastvideo/dataset/.*|
    fastvideo/models/.*|
@@ -1,12 +1,13 @@
 <div align="center">
 <img src=assets/logos/logo.svg width="30%"/>
 </div>
+
 **FastVideo is a unified post-training and inference framework for accelerated video generation.**

 FastVideo features an end-to-end unified pipeline for accelerating diffusion models, starting from data preprocessing to model training, finetuning, distillation, and inference. FastVideo is designed to be modular and extensible, allowing users to easily add new optimizations and techniques. Whether it is training-free optimizations or post-training optimizations, FastVideo has you covered.

 <p align="center">
-    | 🕹️ <a href="https://fastwan.fastvideo.org/"<b>Online Demo</b></a> | <a href="https://hao-ai-lab.github.io/FastVideo"><b>Documentation</b></a> | <a href="https://hao-ai-lab.github.io/FastVideo/inference/inference_quick_start/"><b> Quick Start</b></a> | 🤗 <a href="https://huggingface.co/collections/FastVideo/fastwan-6886a305d9799c8cd1496408"  target="_blank"><b>FastWan</b></a>  | 🟣💬 <a href="https://join.slack.com/t/fastvideo/shared_invite/zt-3f4lao1uq-u~Ipx6Lt4J27AlD2y~IdLQ" target="_blank"> <b>Slack</b> </a> |  🟣💬 <a href="https://ibb.co/c7g1qdD" target="_blank"> <b> WeChat </b> </a> |
+    | 🕹️ <a href="https://fastwan.fastvideo.org/"<b>Online Demo</b></a> | <a href="https://hao-ai-lab.github.io/FastVideo"><b>Documentation</b></a> | <a href="https://hao-ai-lab.github.io/FastVideo/inference/inference_quick_start.html"><b> Quick Start</b></a> | 🤗 <a href="https://huggingface.co/collections/FastVideo/fastwan-6886a305d9799c8cd1496408"  target="_blank"><b>FastWan</b></a>  | 🟣💬 <a href="https://join.slack.com/t/fastvideo/shared_invite/zt-3csdw1isz-Euq8_Q8~baewG8hxjXs2gQ" target="_blank"> <b>Slack</b> </a> |  🟣💬 <a href="https://ibb.co/tMwknPLY" target="_blank"> <b> WeChat </b> </a> |
 </p>

 <div align="center">
@@ -14,8 +15,7 @@ FastVideo features an end-to-end unified pipeline for accelerating diffusion mod
 </div>

 ## NEWS
- ```2025/11/19```: Release [CausalWan2.2 I2V A14B Preview](https://huggingface.co/FastVideo/CausalWan2.2-I2V-A14B-Preview-Diffusers) models, [Blog](https://hao-ai-lab.github.io/blogs/fastvideo_causalwan_preview/) and [Inference Code!](https://github.com/hao-ai-lab/FastVideo/blob/main/examples/inference/basic/basic_self_forcing_causal_wan2_2_i2v.py)
- ```2025/08/04```: Release [FastWan](https://hao-ai-lab.github.io/FastVideo/distillation/dmd) models and [Sparse-Distillation](https://hao-ai-lab.github.io/blogs/fastvideo_post_training/).
+- ```2025/08/04```: Release [FastWan](https://hao-ai-lab.github.io/FastVideo/distillation/dmd.html) models and [Sparse-Distillation](https://hao-ai-lab.github.io/blogs/fastvideo_post_training/).
 - ```2025/06/14```: Release finetuning and inference code for [VSA](https://arxiv.org/pdf/2505.13389)
 - ```2025/04/24```: [FastVideo V1](https://hao-ai-lab.github.io/blogs/fastvideo/) is released!
 - ```2025/02/18```: Release the inference code for [Sliding Tile Attention](https://hao-ai-lab.github.io/blogs/sta/).
@@ -49,10 +49,10 @@ conda activate fastvideo
 pip install fastvideo
 ```

-Please see our [docs](https://hao-ai-lab.github.io/FastVideo/getting_started/installation/) for more detailed installation instructions.
+Please see our [docs](https://hao-ai-lab.github.io/FastVideo/getting_started/installation.html) for more detailed installation instructions.

 ## Sparse Distillation
-For our sparse distillation techniques, please see our [distillation docs](https://hao-ai-lab.github.io/FastVideo/distillation/dmd/) and check out our [blog](https://hao-ai-lab.github.io/blogs/fastvideo_post_training/).
+For our sparse distillation techniques, please see our [distillation docs](https://hao-ai-lab.github.io/FastVideo/distillation/dmd.html) and check out our [blog](https://hao-ai-lab.github.io/blogs/fastvideo_post_training/).

 See below for recipes and datasets:

@@ -64,7 +64,7 @@ See below for recipes and datasets:

 ## Inference
 ### Generating Your First Video
-Here's a minimal example to generate a video using the default settings. Make sure VSA kernels are [installed](https://hao-ai-lab.github.io/FastVideo/video_sparse_attention/installation/). Create a file called `example.py` with the following code:
+Here's a minimal example to generate a video using the default settings. Make sure VSA kernels are [installed](https://hao-ai-lab.github.io/FastVideo/video_sparse_attention/installation.html). Create a file called `example.py` with the following code:

 ```python
 import os
@@ -100,32 +100,35 @@ Run the script with:
 python example.py
 ```

-For a more detailed guide, please see our [inference quick start](https://hao-ai-lab.github.io/FastVideo/inference/inference_quick_start/).
+For a more detailed guide, please see our [inference quick start](https://hao-ai-lab.github.io/FastVideo/inference/inference_quick_start.html).

 ### Other docs:

- [Design Overview](https://hao-ai-lab.github.io/FastVideo/design/overview/)
- [Contribution Guide](https://hao-ai-lab.github.io/FastVideo/getting_started/installation/)
+- [Design Overview](https://hao-ai-lab.github.io/FastVideo/design/overview.html)
+- [Contribution Guide](https://hao-ai-lab.github.io/FastVideo/getting_started/installation.html)

 ## Distillation and Finetuning
- [Distillation Guide](https://hao-ai-lab.github.io/FastVideo/distillation/dmd/)
+- [Distillation Guide](https://hao-ai-lab.github.io/FastVideo/distillation/dmd.html)
 <!-- - [Finetuning Guide](https://hao-ai-lab.github.io/FastVideo/training/finetune.html) -->

-## Awesome work using FastVideo or our research projects
+## 📑 Development Plan
+<!-- - More distillation methods -->
+  <!-- - [ ] Add Distribution Matching Distillation -->
+More FastWan Models Coming Soon!
+- [ ] Add FastWan2.1-T2V-14B
+- [ ] Add FastWan2.2-T2V-14B
+- [ ] Add FastWan2.2-I2V-14B
+<!-- - Optimization features
+- Code updates -->
+  <!-- - [ ] fp8 support -->
+  <!-- - [ ] faster load model and save model support -->

- [SGLang](https://github.com/sgl-project/sglang/tree/main/python/sglang/multimodal_gen): SGLang's diffusion inference functionality is based  on a fork of FastVideo on Sept. 24, 2025. [![Star](https://img.shields.io/github/stars/sgl-project/sglang.svg?style=social&label=Star)](https://github.com/sgl-project/sglang)
-
- [DanceGRPO](https://github.com/XueZeyue/DanceGRPO): A  unified framework to adapt Group Relative Policy Optimization (GRPO) to visual generation paradigms. Code based on FastVideo. [![Star](https://img.shields.io/github/stars/XueZeyue/DanceGRPO.svg?style=social&label=Star)](https://github.com/XueZeyue/DanceGRPO)
- [SRPO](https://github.com/Tencent-Hunyuan/SRPO): A method to directly align the full diffusion trajectory with fine-grained human preference. Code based on FastVideo. [![Star](https://img.shields.io/github/stars/Tencent-Hunyuan/SRPO.svg?style=social&label=Star)](https://github.com/Tencent-Hunyuan/SRPO)
- [DCM](https://github.com/Vchitect/DCM): Dual-expert consistency model for efficient and high-quality video generation. Code based on FastVideo. [![Star](https://img.shields.io/github/stars/Vchitect/DCM.svg?style=social&label=Star)](https://github.com/Vchitect/DCM)
- [Hunyuan Video 1.5](https://github.com/Tencent-Hunyuan/HunyuanVideo-1.5): A leading lightweight video generation model, where they proposed SSTA based on Sliding Tile Attention. [![Star](https://img.shields.io/github/stars/Tencent-Hunyuan/HunyuanVideo-1.5.svg?style=social&label=Star)](https://github.com/Tencent-Hunyuan/HunyuanVideo-1.5)
- [Kandinsky-5.0](https://github.com/kandinskylab/kandinsky-5): A family of diffusion models for video & image generation, where their NABLA attention includes a Sliding Tile Attention branch. [![Star](https://img.shields.io/github/stars/kandinskylab/kandinsky-5.svg?style=social&label=Star)](https://github.com/kandinskylab/kandinsky-5)
- [LongCat Video](https://github.com/meituan-longcat/LongCat-Video): A foundational video generation model with 13.6B parameters with block-sparse attention similar to Video Sparse Attention. [![Star](https://img.shields.io/github/stars/meituan-longcat/LongCat-Video.svg?style=social&label=Star)](https://github.com/meituan-longcat/LongCat-Video)
+See details in [development roadmap](https://github.com/hao-ai-lab/FastVideo/issues/468).

 ## 🤝 Contributing

-We welcome all contributions. Please check out our guide [here](https://hao-ai-lab.github.io/FastVideo/contributing/overview/).
-See details in [development roadmap](https://github.com/hao-ai-lab/FastVideo/issues/899).
+We welcome all contributions. Please check out our guide [here](https://hao-ai-lab.github.io/FastVideo/contributing/overview.html)
+
 ## Acknowledgement
 We learned and reused code from the following projects:
 - [Wan-Video](https://github.com/Wan-Video)
@@ -1,103 +0,0 @@
-# FVD (Fréchet Video Distance) Benchmark
-
-Evaluate generated video quality using FVD with the I3D feature extractor.
-
-## Quick Start
-
-**Run the benchmark:**
-
-```bash
-bash benchmarks/scripts/run.sh
-```
-
-That's it! The script auto-installs dependencies and runs the benchmark.
-
-**To customize:** Edit `benchmarks/fvd/run_fvd.py` to change:
- Video paths (`real_dir`, `gen_dir`)
- Number of videos, frames, sampling strategy
- Device, batch size, caching, etc.
-
-## Advanced Usage (CLI)
-
-For more control without editing Python files, use the CLI.
-
-**First-time setup** (one-time per pod/environment):
-
-```bash
-bash benchmarks/scripts/setup_fvd.sh
-```
-
-Then run any configuration you want:
-
-```bash
-# Custom configuration
-python -m benchmarks.fvd.cli \
-    --real-path data/real/ \
-    --gen-path outputs/gen/ \
-    --num-videos 1024 \
-    --num-frames 32 \
-    --clip-strategy random \
-    --batch-size 32 \
-    --seed 42
-```
-
-**Standard protocols:**
-
-```bash
-# Use predefined protocols
-python -m benchmarks.fvd.cli \
-    --real-path data/real/ \
-    --gen-path outputs/gen/ \
-    --protocol fvd2048_16f  # or fvd2048_128f, quick_test, etc.
-```
-
-**Feature caching** (speed up repeated evaluations):
-
-```bash
-python -m benchmarks.fvd.cli \
-    --real-path data/real/ \
-    --gen-path outputs/gen/ \
-    --protocol fvd2048_16f \
-    --cache-real-features cache/real  # Directory path (will save/load cache/real/real_features.pkl)
-```
-
-Run `python -m benchmarks.fvd.cli --help` for all options.
-
-## Available Protocols
-
- `fvd2048_16f` - Standard (2048 videos, 16 frames)
- `fvd2048_128f` - Long videos (128 frames)
- `fvd2048_128f_subsample8` - Subsampled long videos
- `quick_test` - Fast testing (10 videos)
-
-## Configuration Options
-
-Key options in `FVDConfig`:
-
-```python
-num_videos=2048,              # Videos to evaluate
-num_frames_per_clip=16,       # Frames per clip
-clip_strategy='beginning',    # beginning|random|uniform|middle|sliding
-frame_stride=1,               # Frame subsampling
-batch_size=32,                # GPU batch size
-device='cuda',                # cuda|cpu
-cache_real_features=None,     # Cache path for speed
-seed=42,                      # Reproducibility
-```
-
-## Programmatic Usage
-
-```python
-from benchmarks.fvd import compute_fvd_with_config, FVDConfig
-
-config = FVDConfig.fvd2048_16f()  # or custom config
-results = compute_fvd_with_config('data/real/', 'outputs/gen/', config)
-print(f"FVD: {results['fvd']:.2f}")
-```
-
-## Notes
-
- I3D model auto-downloads from Hugging Face on first run
- Requires minimum 10 frames per clip
- Supports both video files (.mp4, .avi, etc.) and frame directories
- `--cache-real-features` expects a **directory path** (e.g., `cache/real`), it will automatically create/load `real_features.pkl` inside that directory
@@ -1,35 +0,0 @@
-"""
-FastVideo Frechet Video Distance (FVD) Benchmark Module.
-    >>> from fastvideo.benchmarks.fvd import compute_fvd_with_config, FVDConfig
-    >>> config = FVDConfig.fvd2048_16f()  # Standard protocol
-    >>> results = compute_fvd_with_config('data/real/', 'outputs/gen/', config)
-    >>> print(f"FVD: {results['fvd']:.2f}")
-"""
-
-from .fvd import (
-    compute_fvd,
-    compute_fvd_with_config,
-    compute_frechet_distance,
-    compute_statistics,
-    FVDConfig,
-)
-from .i3d_model import I3DFeatureExtractor
-from .video_utils import (
-    load_video_auto,
-    sample_clips_from_video,
-    load_video_clips_streaming,
-    ClipSamplingStrategy,
-)
-
-__all__ = [
-    'compute_fvd',
-    'compute_fvd_with_config',
-    'compute_frechet_distance',
-    'compute_statistics',
-    'FVDConfig',
-    'I3DFeatureExtractor',
-    'load_video_auto',
-    'sample_clips_from_video',
-    'load_video_clips_streaming',
-    'ClipSamplingStrategy',
-]
@@ -1,185 +0,0 @@
-import argparse
-import json
-import sys
-from pathlib import Path
-
-from .fvd import compute_fvd_with_config, FVDConfig
-
-
-def main() -> int:
-    parser = argparse.ArgumentParser(
-        description='Compute Fréchet Video Distance (FVD)',
-        formatter_class=argparse.RawDescriptionHelpFormatter,
-        epilog="""
-Examples:
-  # Standard FVD2048_16f protocol
-  python -m fastvideo.benchmarks.fvd.cli \\
-      --real-path data/real/ \\
-      --gen-path outputs/gen/ \\
-      --protocol fvd2048_16f
-
-  # Custom configuration
-  python -m fastvideo.benchmarks.fvd.cli \\
-      --real-path data/real/ \\
-      --gen-path outputs/gen/ \\
-      --num-videos 1024 \\
-      --num-frames 32 \\
-      --clip-strategy random \\
-      --frame-stride 2
-        """)
-
-    # Required arguments
-    parser.add_argument('--real-path',
-                        type=str,
-                        required=True,
-                        help='Path to real videos directory')
-    parser.add_argument('--gen-path',
-                        type=str,
-                        required=True,
-                        help='Path to generated videos directory')
-
-    # Reproducibility
-    parser.add_argument(
-        '--seed',
-        type=int,
-        default=None,
-        help='Random seed for reproducibility (np.random, random, torch)')
-
-    # Protocol presets
-    parser.add_argument('--protocol',
-                        type=str,
-                        default=None,
-                        choices=[
-                            'fvd2048_16f', 'fvd2048_128f',
-                            'fvd2048_128f_subsample8', 'quick_test'
-                        ],
-                        help='Use standard protocol (overrides other settings)')
-
-    # Video selection
-    parser.add_argument('--num-videos',
-                        type=int,
-                        default=2048,
-                        help='Number of videos to use (default: 2048)')
-
-    # Clip sampling
-    parser.add_argument('--num-frames',
-                        type=int,
-                        default=16,
-                        help='Number of frames per clip (default: 16)')
-    parser.add_argument('--num-clips',
-                        type=int,
-                        default=1,
-                        help='Number of clips per video (default: 1)')
-    parser.add_argument(
-        '--clip-strategy',
-        type=str,
-        default='beginning',
-        choices=['beginning', 'random', 'uniform', 'middle', 'sliding', 'all'],
-        help='Clip sampling strategy (default: beginning)')
-    parser.add_argument(
-        '--frame-stride',
-        type=int,
-        default=1,
-        help='Frame stride for FPS subsampling (default: 1, no subsampling)')
-    parser.add_argument('--temporal-stride',
-                        type=int,
-                        default=1,
-                        help='Temporal stride for sliding window (default: 1)')
-
-    # Data processing
-    parser.add_argument('--no-frame-dirs',
-                        action='store_true',
-                        help='Disable frame directory support')
-
-    # Computation
-    parser.add_argument('--batch-size',
-                        type=int,
-                        default=32,
-                        help='Batch size for feature extraction (default: 32)')
-    parser.add_argument('--device',
-                        type=str,
-                        default='cuda',
-                        choices=['cuda', 'cpu'],
-                        help='Device to use (default: cuda)')
-
-    # Caching
-    parser.add_argument('--cache-real-features',
-                        type=str,
-                        default=None,
-                        help='Path to cache real video features')
-    parser.add_argument('--i3d-model-path',
-                        type=str,
-                        default=None,
-                        help='Custom cache path for I3D model')
-
-    # Output
-    parser.add_argument('--output',
-                        type=str,
-                        default='fvd_results.json',
-                        help='Output JSON file (default: fvd_results.json)')
-    parser.add_argument('--quiet',
-                        action='store_true',
-                        help='Suppress progress output')
-
-    args = parser.parse_args()
-
-    # Create config
-    if args.protocol:
-        protocol_map = {
-            'fvd2048_16f': FVDConfig.fvd2048_16f,
-            'fvd2048_128f': FVDConfig.fvd2048_128f,
-            'fvd2048_128f_subsample8': FVDConfig.fvd2048_128f_subsample8,
-            'quick_test': FVDConfig.quick_test,
-        }
-        config = protocol_map[args.protocol]()
-
-        # Override device and caching from args
-        config.device = args.device
-        config.cache_real_features = args.cache_real_features
-        config.i3d_model_path = args.i3d_model_path
-        config.batch_size = args.batch_size
-        config.seed = args.seed
-    else:
-        # Custom config from args
-        config = FVDConfig(num_videos=args.num_videos,
-                           num_frames_per_clip=args.num_frames,
-                           num_clips_per_video=args.num_clips,
-                           clip_strategy=args.clip_strategy,
-                           frame_stride=args.frame_stride,
-                           temporal_stride=args.temporal_stride,
-                           support_frame_dirs=not args.no_frame_dirs,
-                           batch_size=args.batch_size,
-                           device=args.device,
-                           cache_real_features=args.cache_real_features,
-                           i3d_model_path=args.i3d_model_path,
-                           seed=args.seed)
-
-    # Compute FVD
-    try:
-        results = compute_fvd_with_config(real_videos=args.real_path,
-                                          gen_videos=args.gen_path,
-                                          config=config,
-                                          verbose=not args.quiet)
-
-        # Save results
-        output_path = Path(args.output)
-        output_path.parent.mkdir(parents=True, exist_ok=True)
-
-        with open(output_path, 'w') as f:
-            json.dump(results, f, indent=2)
-
-        print(f"\nResults saved to {output_path}")
-        print(f"FVD: {results['fvd']:.2f}")
-        print(f"Protocol: {results['protocol']}")
-
-        return 0
-
-    except Exception as e:
-        print(f"Error: {e}", file=sys.stderr)
-        import traceback
-        traceback.print_exc()
-        return 1
-
-
-if __name__ == '__main__':
-    sys.exit(main())
@@ -1,447 +0,0 @@
-import numpy as np
-import scipy.linalg
-import torch
-from pathlib import Path
-from collections.abc import Iterator
-import pickle
-from dataclasses import dataclass, field
-
-from .i3d_model import I3DFeatureExtractor
-from .video_utils import ClipSamplingStrategy, load_video_clips_streaming
-
-
-def compute_statistics(features: np.ndarray) -> tuple[np.ndarray, np.ndarray]:
-    """Compute mean and covariance."""
-    mu = np.mean(features, axis=0)
-    sigma = np.cov(features, rowvar=False)
-    return mu, sigma
-
-
-def compute_frechet_distance(mu1: np.ndarray,
-                             sigma1: np.ndarray,
-                             mu2: np.ndarray,
-                             sigma2: np.ndarray,
-                             eps: float = 1e-6) -> float:
-    """
-    Compute Fréchet distance between two Gaussians.
-    """
-    sigma1 = sigma1 + eps * np.eye(sigma1.shape[0])
-    sigma2 = sigma2 + eps * np.eye(sigma2.shape[0])
-
-    diff = mu1 - mu2
-    mean_distance = np.sum(diff**2)
-
-    trace_sum = np.trace(sigma1 + sigma2)
-
-    covmean = scipy.linalg.sqrtm(sigma1 @ sigma2)
-
-    if np.iscomplexobj(covmean):
-        if not np.allclose(np.diagonal(covmean).imag, 0, atol=1e-3):
-            print(
-                f"Warning: Imaginary component: {np.max(np.abs(covmean.imag))}")
-        covmean = covmean.real
-
-    trace_product = np.trace(covmean)
-
-    fvd = mean_distance + trace_sum - 2 * trace_product
-
-    return float(fvd)
-
-
-@dataclass
-class FVDConfig:
-    # default configuration for FVD computation:
-
-    # Video selection
-    num_videos: int = 2048
-
-    # Clip sampling
-    num_frames_per_clip: int = 16
-    num_clips_per_video: int = 1
-    clip_strategy: str | ClipSamplingStrategy = 'beginning'
-
-    # Temporal subsampling
-    frame_stride: int = 1  # 1=no subsampling, 2=every 2nd, 8=every 8th
-    temporal_stride: int = 1  # For sliding window clips
-
-    # Data processing
-    video_extensions: list[str] = field(
-        default_factory=lambda: ['.mp4', '.avi', '.mov', '.mkv'])
-    support_frame_dirs: bool = True
-
-    # Computation
-    batch_size: int = 32
-    device: str = 'cuda'
-
-    use_streaming: bool = True
-    resize_before_extraction: bool = True
-
-    # Caching
-    cache_real_features: str | None = None
-    i3d_model_path: str | None = None
-
-    # Reproducibility
-    seed: int | None = None
-
-    @classmethod
-    def fvd2048_16f(cls) -> 'FVDConfig':
-        """
-        Standard FVD protocol: 2048 videos, 16 frames, beginning clip.
-        
-        most common FVD configuration used in papers
-        """
-        return cls(num_videos=2048,
-                   num_frames_per_clip=16,
-                   clip_strategy='beginning',
-                   use_streaming=True)
-
-    @classmethod
-    def fvd2048_128f(cls) -> 'FVDConfig':
-        """Long video protocol: 2048 videos, 128 frames."""
-        return cls(num_videos=2048,
-                   num_frames_per_clip=128,
-                   clip_strategy='beginning',
-                   use_streaming=True)
-
-    @classmethod
-    def fvd2048_128f_subsample8(cls) -> 'FVDConfig':
-        """
-        Long video with FPS subsampling: 2048 videos, 128 frames (every 8th).
-        Used for very long videos - samples every 8th frame
-        """
-        return cls(num_videos=2048,
-                   num_frames_per_clip=16,
-                   frame_stride=8,
-                   clip_strategy='beginning',
-                   use_streaming=True)
-
-    @classmethod
-    def quick_test(cls) -> 'FVDConfig':
-        """Quick test config: 100 videos, 16 frames."""
-        return cls(num_videos=100,
-                   num_frames_per_clip=16,
-                   clip_strategy='beginning')
-
-    def to_dict(self) -> dict:
-        """Export config to dict for logging"""
-        return {
-            'num_videos': self.num_videos,
-            'num_frames_per_clip': self.num_frames_per_clip,
-            'num_clips_per_video': self.num_clips_per_video,
-            'clip_strategy': str(self.clip_strategy),
-            'frame_stride': self.frame_stride,
-            'temporal_stride': self.temporal_stride,
-            'batch_size': self.batch_size,
-            'device': self.device,
-            'seed': self.seed,
-            'use_streaming': self.use_streaming,
-        }
-
-    def __str__(self) -> str:
-        """Human-readable protocol name"""
-        desc = f"FVD{self.num_videos}_{self.num_frames_per_clip}f"
-        if self.frame_stride > 1:
-            desc += f"_subsample{self.frame_stride}"
-        if self.num_clips_per_video > 1:
-            desc += f"_{self.num_clips_per_video}clips"
-        if self.clip_strategy != 'beginning':
-            desc += f"_{self.clip_strategy}"
-        return desc
-
-
-def extract_features_streaming(video_generator: Iterator[torch.Tensor],
-                               extractor: I3DFeatureExtractor,
-                               batch_size: int = 32,
-                               max_clips: int | None = None,
-                               verbose: bool = True) -> np.ndarray:
-    """
-    Extract features from a video clip generator using streaming.
-
-    Args:
-        video_generator: Iterator yielding clips [T, C, H, W]
-        extractor: I3D feature extractor
-        batch_size: Batch size for processing
-        max_clips: Maximum clips to process (for validation)
-        verbose: Show progress
-    
-    Returns:
-        features: [N, 400] numpy array
-    """
-    all_features = []
-    batch = []
-    clip_count = 0
-
-    if verbose:
-        print(f"Extracting features with batch_size={batch_size}...")
-
-    for clip_count, clip in enumerate(video_generator):
-        batch.append(clip)
-
-        # Process batch when full
-        if len(batch) == batch_size:
-            batch_tensor = torch.stack(batch).to(extractor.device)
-            features = extractor.extract_features(batch_tensor,
-                                                  batch_size=batch_size,
-                                                  verbose=False)
-            all_features.append(features.cpu().numpy())
-
-            batch = []  # Clear batch
-
-            if verbose and clip_count % (batch_size * 10) == 0:
-                print(f"Processed {clip_count} clips...")
-
-        # Stop if we've reached max_clips
-        if max_clips is not None and clip_count >= max_clips:
-            break
-
-    # Process remaining clips
-    if len(batch) > 0:
-        batch_tensor = torch.stack(batch).to(extractor.device)
-        features = extractor.extract_features(batch_tensor,
-                                              batch_size=len(batch),
-                                              verbose=False)
-        all_features.append(features.cpu().numpy())
-
-    if len(all_features) == 0:
-        raise RuntimeError("No features extracted - check video loading")
-
-    features = np.concatenate(all_features, axis=0)
-
-    if verbose:
-        print(f"Extracted {len(features)} feature vectors")
-
-    return features
-
-
-def load_or_compute_features(videos: str | Path | torch.Tensor,
-                             extractor: I3DFeatureExtractor,
-                             config: FVDConfig,
-                             cache_path: str | None = None,
-                             cache_name: str = "real_features") -> np.ndarray:
-    """Load features from cache or compute (with streaming support)"""
-
-    if cache_path is not None:
-        cache_file = Path(cache_path) / f"{cache_name}.pkl"
-        if cache_file.exists():
-            print(f"Loading cached features from {cache_file}")
-            with open(cache_file, 'rb') as f:
-                features = pickle.load(f)
-
-            # Validate and limit based on config
-            max_features = config.num_videos * config.num_clips_per_video
-            if len(features) < max_features:
-                print(
-                    f"WARNING: Cache has {len(features)} features but need {max_features}"
-                )
-                print("Recomputing features...")
-            elif len(features) > max_features:
-                features = features[:max_features]
-                return features
-            else:
-                return features
-
-    # Compute features
-    if isinstance(videos, str | Path):
-        target_size = (224, 224) if config.resize_before_extraction else None
-
-        video_generator = load_video_clips_streaming(
-            videos,
-            num_frames=config.num_frames_per_clip,
-            max_videos=config.num_videos,
-            clip_strategy=config.clip_strategy,
-            frame_stride=config.frame_stride,
-            num_clips_per_video=config.num_clips_per_video,
-            video_extensions=config.video_extensions,
-            support_frame_dirs=config.support_frame_dirs,
-            target_size=target_size,
-            verbose=True)
-
-        max_clips = config.num_videos * config.num_clips_per_video
-        features = extract_features_streaming(video_generator,
-                                              extractor,
-                                              batch_size=config.batch_size,
-                                              max_clips=max_clips,
-                                              verbose=True)
-
-    else:
-        # Already a tensor
-        print(f"Extracting features from {len(videos)} video tensors...")
-        features = extractor.extract_features(videos,
-                                              batch_size=config.batch_size,
-                                              verbose=True)
-        features = features.numpy()
-
-    # Validate feature count
-    expected_count = config.num_videos * config.num_clips_per_video
-    if len(features) < expected_count:
-        raise ValueError(
-            f"ERROR: Only extracted {len(features)} features, but need {expected_count}!\n"
-            f"Found fewer videos than expected. Check your video directory.")
-    elif len(features) > expected_count:
-        print(f"Truncating {len(features)} features to {expected_count}")
-        features = features[:expected_count]
-
-    # Cache features if requested
-    if cache_path is not None:
-        cache_dir = Path(cache_path)
-        cache_dir.mkdir(parents=True, exist_ok=True)
-        cache_file = cache_dir / f"{cache_name}.pkl"
-        print(f"Caching features to {cache_file}")
-        with open(cache_file, 'wb') as f:
-            pickle.dump(features, f)
-
-    return features
-
-
-def compute_fvd(real_videos: str | Path | torch.Tensor,
-                gen_videos: str | Path | torch.Tensor,
-                num_frames: int = 16,
-                batch_size: int = 32,
-                device: str = 'cuda',
-                num_videos: int | None = 2048,
-                cache_real_features: str | None = None,
-                i3d_model_path: str | None = None,
-                seed: int | None = None,
-                verbose: bool = True) -> float:
-    """
-    Compute Fréchet Video Distance (FVD)
-    
-    For advanced control, use compute_fvd_with_config() instead.
-    
-    Args:
-        real_videos: Path to real videos or tensor [N, T, C, H, W]
-        gen_videos: Path to generated videos or tensor [N, T, C, H, W]
-        num_frames: Frames per video (default: 16)
-        batch_size: Batch size (default: 32)
-        device: 'cuda' or 'cpu' (default: 'cuda')
-        num_videos: Max videos (default: 2048)
-        cache_real_features: Cache path for real features
-        i3d_model_path: Custom I3D model cache path
-        seed: Random seed for reproducibility
-        verbose: Print progress
-    
-    Returns:
-        FVD score (float). Lower is better.
-    """
-    num_videos = num_videos if num_videos is not None else 2048
-
-    config = FVDConfig(
-        num_videos=num_videos,
-        num_frames_per_clip=num_frames,
-        batch_size=batch_size,
-        device=device,
-        cache_real_features=cache_real_features,
-        i3d_model_path=i3d_model_path,
-        seed=seed,
-    )
-
-    result = compute_fvd_with_config(real_videos, gen_videos, config, verbose)
-    return result['fvd']
-
-
-def compute_fvd_with_config(real_videos: str | Path | torch.Tensor,
-                            gen_videos: str | Path | torch.Tensor,
-                            config: FVDConfig,
-                            verbose: bool = True) -> dict:
-    """
-    Compute FVD using a standardized configuration.
-    
-    This is the recommended way to compute FVD for reproducibility.
-    
-    Args:
-        real_videos: Path or tensors
-        gen_videos: Path or tensors
-        config: FVDConfig specifying protocol
-        verbose: Print progress
-    
-    Returns:
-        results: Dictionary with:
-            - 'fvd': FVD score (float)
-            - 'protocol': Protocol name (str)
-            - 'config': Configuration dict
-    
-    Example:
-        >>> config = FVDConfig.fvd2048_16f()
-        >>> results = compute_fvd_with_config('data/real/', 'outputs/gen/', config)
-        >>> print(f"FVD: {results['fvd']:.2f}")
-        >>> print(f"Protocol: {results['protocol']}")  # "FVD2048_16f"
-    """
-    # Seed for reproducibility
-    if config.seed is not None:
-        import random as _rnd
-        _rnd.seed(config.seed)
-        np.random.seed(config.seed)
-        torch.manual_seed(config.seed)
-        if torch.cuda.is_available():
-            torch.cuda.manual_seed_all(config.seed)
-
-    if verbose:
-        print("=" * 70)
-        print(f"Computing FVD with protocol: {config}")
-        print("=" * 70)
-        print("\nConfiguration:")
-        for key, value in config.to_dict().items():
-            print(f"  {key}: {value}")
-        print()
-
-    # Initialize I3D
-    if verbose:
-        print(f"\nInitializing I3D model on {config.device}...")
-
-    extractor = I3DFeatureExtractor(device=config.device,
-                                    cache_dir=config.i3d_model_path)
-
-    # Extract features
-    if verbose:
-        print(f"\n{'='*70}")
-        print("Extracting REAL video features...")
-        print(f"{'='*70}")
-
-    real_features = load_or_compute_features(
-        videos=real_videos,
-        extractor=extractor,
-        config=config,
-        cache_path=config.cache_real_features,
-        cache_name="real_features")
-
-    if verbose:
-        print(f"\n{'='*70}")
-        print("Extracting GENERATED video features...")
-        print(f"{'='*70}")
-
-    gen_features = load_or_compute_features(videos=gen_videos,
-                                            extractor=extractor,
-                                            config=config,
-                                            cache_path=None,
-                                            cache_name="gen_features")
-
-    if verbose:
-        print(f"\nReal videos/clips: {len(real_features)}")
-        print(f"Generated videos/clips: {len(gen_features)}")
-        print(f"\n{'='*70}")
-        print("Computing statistics...")
-        print(f"{'='*70}")
-
-    mu_real, sigma_real = compute_statistics(real_features)
-    mu_gen, sigma_gen = compute_statistics(gen_features)
-
-    if verbose:
-        print(f"\n{'='*70}")
-        print("Computing Fréchet distance...")
-        print(f"{'='*70}")
-
-    fvd = compute_frechet_distance(mu_real, sigma_real, mu_gen, sigma_gen)
-
-    if verbose:
-        print(f"\n{'='*70}")
-        print(f"FVD Score: {fvd:.4f}")
-        print(f"Protocol: {config}")
-        print(f"{'='*70}\n")
-
-    results = {
-        'fvd': fvd,
-        'protocol': str(config),
-        'config': config.to_dict(),
-    }
-
-    return results
@@ -1,142 +0,0 @@
-"""I3D Feature Extractor for FVD Computation"""
-
-import torch
-import torch.nn as nn
-import torch.nn.functional as F
-from pathlib import Path
-from huggingface_hub import hf_hub_download
-from tqdm import tqdm
-from contextlib import suppress
-
-
-class I3DFeatureExtractor(nn.Module):
-    """
-    I3D feature extractor for FVD computation.
-    Extracts 400-dimensional features from videos using I3D model
-    trained on Kinetics-400.
-    """
-
-    REPO_ID = 'flateon/FVD-I3D-torchscript'
-    MODEL_FILENAME = 'i3d_torchscript.pt'
-
-    def __init__(self,
-                 device: str = 'cuda',
-                 cache_dir: str | Path | None = None):
-        super().__init__()
-
-        self.device_str = device
-        if device == 'cuda' and not torch.cuda.is_available():
-            print(
-                "Warning: CUDA requested but not available – falling back to CPU"
-            )
-            self.device = torch.device('cpu')
-        else:
-            self.device = torch.device(device)
-
-        self.cache_dir: str | None
-        if cache_dir is not None:
-            self.cache_dir = str(Path(cache_dir).resolve())
-        else:
-            self.cache_dir = None  # Use HF default cache
-
-        self.model = self._load_model()
-        self.model.eval()
-
-        with suppress(Exception):
-            self.model.to(self.device)
-
-    def _load_model(self) -> torch.nn.Module:
-        """Download and load I3D TorchScript model from Hugging Face Hub."""
-        print(f"Loading I3D model from Hugging Face Hub ({self.REPO_ID})...")
-
-        try:
-            # Download model from Hugging Face Hub
-            model_path = hf_hub_download(repo_id=self.REPO_ID,
-                                         filename=self.MODEL_FILENAME,
-                                         cache_dir=self.cache_dir)
-
-            # Load directly to chosen device
-            model = torch.jit.load(model_path, map_location=self.device)
-            print("I3D model loaded successfully")
-            return model
-
-        except Exception as e:
-            raise RuntimeError(
-                f"Failed to load I3D model from Hugging Face Hub. Error: {e}\n"
-                f"Ensure you have internet connection and huggingface_hub installed:\n"
-                f"pip install huggingface_hub") from e
-
-    def preprocess(self, videos: torch.Tensor) -> torch.Tensor:
-        """
-        Preprocess videos for I3D.
-        
-        Args:
-            videos: [B, T, C, H, W], values in [0, 255]
-        
-        Returns:
-            Preprocessed videos [B, C, T, 224, 224] (normalized and resized)
-        """
-        B, T, C, H, W = videos.shape
-
-        if T < 10:
-            raise ValueError(f"I3D requires at least 10 frames, got {T}")
-
-        # Normalize to [0, 1] if needed
-        if videos.max() > 1.0:
-            videos = videos / 255.0
-
-        # Resize to 224x224 if needed
-        if H != 224 or W != 224:
-            videos = videos.reshape(B * T, C, H, W)
-            videos = F.interpolate(videos,
-                                   size=(224, 224),
-                                   mode='bilinear',
-                                   align_corners=False)
-            videos = videos.reshape(B, T, C, 224, 224)
-
-        # Convert to [B, C, T, H, W] format
-        videos = videos.permute(0, 2, 1, 3, 4).contiguous()
-
-        return videos
-
-    @torch.no_grad()
-    def extract_features(self,
-                         videos: torch.Tensor,
-                         batch_size: int = 32,
-                         verbose: bool = True) -> torch.Tensor:
-        """
-        Extract I3D features
-        
-        Args:
-            videos: [N, T, C, H, W], values in [0, 255]
-            batch_size: Batch size for processing
-            verbose: Show progress bar
-        
-        Returns:
-            Features [N, 400]
-        """
-        N = len(videos)
-        all_features = []
-
-        iterator = range(0, N, batch_size)
-        if verbose:
-            iterator = tqdm(iterator, desc="Extracting I3D features")
-
-        for i in iterator:
-            batch = videos[i:i + batch_size].to(self.device)
-            batch = self.preprocess(batch)  # Now returns [B, C, T, H, W]
-
-            # Use the HF model without rescale/resize (we handle it in preprocess)
-            features = self.model(batch,
-                                  rescale=False,
-                                  resize=False,
-                                  return_features=True)
-
-            all_features.append(features.cpu())
-
-        return torch.cat(all_features, dim=0)
-
-    def __call__(self,
-                 videos: torch.Tensor,
-                 batch_size: int = 32) -> torch.Tensor:
-        return self.extract_features(videos, batch_size=batch_size)
@@ -1,34 +0,0 @@
-import sys
-from pathlib import Path
-from benchmarks.fvd.fvd import FVDConfig, compute_fvd_with_config
-
-root_dir = Path(__file__).parent.parent.parent
-sys.path.insert(0, str(root_dir))
-
-
-def main() -> None:
-    # Get script directory
-    script_dir = Path(__file__).parent.resolve()
-
-    clip_strategy = 'beginning'  # Options: 'uniform', 'random', 'beginning', 'end', 'all'
-    cfg = FVDConfig(
-        num_videos=650,
-        num_frames_per_clip=16,
-        num_clips_per_video=1,
-        clip_strategy=clip_strategy,
-        frame_stride=1,
-        batch_size=32,
-        device='cuda',
-        seed=42,
-        cache_real_features=str(script_dir / f'fvd-cache/{clip_strategy}'),
-    )
-
-    real_dir = "benchmarks/data/real_videos"
-    gen_dir = "benchmarks/data/generated_videos"
-
-    results = compute_fvd_with_config(real_dir, gen_dir, cfg, verbose=True)
-    print(f"FVD = {results['fvd']:.2f}")
-
-
-if __name__ == '__main__':
-    main()
@@ -1,97 +0,0 @@
-#!/usr/bin/env python3
-import sys
-from pathlib import Path
-import shutil
-import random
-from fvd import compute_fvd_with_config, FVDConfig
-
-script_path = Path(__file__).resolve()
-fastvideo_root = script_path.parent.parent.parent
-sys.path.insert(0, str(fastvideo_root))
-
-
-def split_videos(video_dir: Path, n_per_subset: int = 128, seed: int = 42):
-    subset_a = video_dir.parent / 'bair_full_subset_A'
-    subset_b = video_dir.parent / 'bair_full_subset_B'
-
-    if subset_a.exists():
-        shutil.rmtree(subset_a)
-    if subset_b.exists():
-        shutil.rmtree(subset_b)
-
-    subset_a.mkdir(parents=True)
-    subset_b.mkdir(parents=True)
-
-    videos = sorted(video_dir.glob('*.mp4'))
-
-    random.seed(seed)
-    shuffled = list(videos)
-    random.shuffle(shuffled)
-
-    needed = n_per_subset * 2
-    if len(shuffled) > needed:
-        shuffled = shuffled[:needed]
-
-    mid = len(shuffled) // 2
-
-    print(f"\nSplitting {len(shuffled)} BAIR FULL videos:")
-    print(f"  Subset A: {mid} videos")
-    print(f"  Subset B: {len(shuffled) - mid} videos")
-
-    for v in shuffled[:mid]:
-        shutil.copy2(v, subset_a / v.name)
-
-    for v in shuffled[mid:]:
-        shutil.copy2(v, subset_b / v.name)
-
-    return subset_a, subset_b, mid
-
-
-def validate_fvd(subset_a: Path, subset_b: Path, num_videos: int):
-    config = FVDConfig(num_videos=num_videos,
-                       num_frames_per_clip=16,
-                       clip_strategy='beginning',
-                       batch_size=8,
-                       device='cuda',
-                       seed=42)
-
-    print("\n" + "=" * 70)
-    print("TEST 1: Identity Test")
-    print("=" * 70)
-
-    result1 = compute_fvd_with_config(real_videos=str(subset_a),
-                                      gen_videos=str(subset_a),
-                                      config=config,
-                                      verbose=False)
-    fvd_identity = result1['fvd']
-    print(f"\nIdentity FVD: {fvd_identity:.2f}")
-
-    print("\n" + "=" * 70)
-    print("TEST 2: Real vs Real")
-    print("=" * 70)
-
-    result2 = compute_fvd_with_config(real_videos=str(subset_a),
-                                      gen_videos=str(subset_b),
-                                      config=config,
-                                      verbose=False)
-    fvd_real = result2['fvd']
-    print(f"\nReal vs Real FVD: {fvd_real:.2f}")
-
-    print("\n" + "=" * 70)
-    print("RESULTS")
-    print("=" * 70)
-    print(f"Identity:        {fvd_identity:.2f}")
-    print(f"Real vs Real:    {fvd_real:.2f}")
-
-
-def main() -> None:
-    bair_dir = Path('benchmarks/data/bair_full_videos')
-
-    subset_a, subset_b, count = split_videos(bair_dir,
-                                             n_per_subset=128,
-                                             seed=42)
-    validate_fvd(subset_a, subset_b, count)
-
-
-if __name__ == '__main__':
-    main()
@@ -1,490 +0,0 @@
-import torch
-import cv2
-import numpy as np
-from pathlib import Path
-from collections.abc import Iterator
-from tqdm import tqdm
-from enum import Enum
-
-
-class ClipSamplingStrategy(Enum):
-    """Clip sampling strategies for FVD evaluation."""
-    BEGINNING = 'beginning'  # Take first N frames (most common)
-    RANDOM = 'random'  # Random N consecutive frames
-    UNIFORM = 'uniform'  # Uniformly spaced frames across video
-    MIDDLE = 'middle'  # Middle N frames
-    SLIDING = 'sliding'  # Multiple sliding windows
-    ALL = 'all'  # All possible clips
-
-
-def _load_video_cv2(video_path: str | Path,
-                    num_frames: int | None = 16,
-                    sample_strategy: str = 'uniform') -> torch.Tensor:
-    """
-    Load video from video file using OpenCV.
-    
-    Args:
-        video_path: Path to video file (MP4, AVI, MOV, MKV)
-        num_frames: Number of frames to extract
-        sample_strategy: 'uniform' or 'random'
-    
-    Returns:
-        video: [T, C, H, W]
-    """
-    video_path = str(video_path)
-    cap = cv2.VideoCapture(video_path)
-
-    if not cap.isOpened():
-        raise RuntimeError(f"Cannot open video: {video_path}")
-
-    frames = []
-    total_frames = int(cap.get(cv2.CAP_PROP_FRAME_COUNT))
-
-    if num_frames is None:
-        # Read all available frames
-        while True:
-            ret, frame = cap.read()
-            if not ret:
-                break
-            frame = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
-            frames.append(frame)
-
-        cap.release()
-        if len(frames) == 0:
-            raise RuntimeError(f"Video has 0 frames: {video_path}")
-
-        frames = np.stack(frames)  # [T, H, W, C]
-        frames = torch.from_numpy(frames).permute(0, 3, 1,
-                                                  2).float()  # [T, C, H, W]
-        return frames
-
-    if total_frames == 0:
-        raise RuntimeError(f"Video has 0 frames: {video_path}")
-
-    # Determine frame indices for sampling
-    if total_frames < num_frames:
-        frame_indices = list(range(
-            total_frames)) + [total_frames - 1] * (num_frames - total_frames)
-    elif sample_strategy == 'uniform':
-        frame_indices = np.linspace(0, total_frames - 1, num_frames,
-                                    dtype=int).tolist()
-    elif sample_strategy == 'random':
-        frame_indices = sorted(
-            np.random.choice(total_frames, num_frames, replace=False))
-    else:
-        raise ValueError(f"Unknown sample_strategy: {sample_strategy}")
-
-    # Extract frames
-    for idx in frame_indices:
-        cap.set(cv2.CAP_PROP_POS_FRAMES, idx)
-        ret, frame = cap.read()
-
-        if not ret:
-            if len(frames) > 0:
-                frames.append(frames[-1].copy())
-            else:
-                h = int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT))
-                w = int(cap.get(cv2.CAP_PROP_FRAME_WIDTH))
-                frames.append(np.zeros((h, w, 3), dtype=np.uint8))
-            continue
-
-        frame = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
-        frames.append(frame)
-
-    cap.release()
-
-    frames = np.stack(frames)  # [T, H, W, C]
-    frames = torch.from_numpy(frames).permute(0, 3, 1,
-                                              2).float()  # [T, C, H, W]
-
-    return frames
-
-
-def _load_video_from_frames(
-        frame_dir: str | Path,
-        num_frames: int | None = 16,
-        sample_strategy: str = 'uniform',
-        frame_extensions: list[str] | None = None) -> torch.Tensor:
-    """
-    Load video from directory of frame images.
-    
-    Args:
-        frame_dir: Directory containing frames
-        num_frames: Number of frames to sample
-        sample_strategy: 'uniform' or 'random'
-        frame_extensions: Image file extensions to look for
-    
-    Returns:
-        video: [T, C, H, W]
-    """
-    if frame_extensions is None:
-        frame_extensions = ['.jpg', '.png', '.jpeg', '.bmp']
-
-    frame_dir = Path(frame_dir)
-
-    if not frame_dir.exists():
-        raise FileNotFoundError(f"Frame directory not found: {frame_dir}")
-
-    # Find all frames
-    frame_files: list[Path] = []
-    for ext in frame_extensions:
-        frame_files.extend(frame_dir.glob(f"*{ext}"))
-
-    if len(frame_files) == 0:
-        raise ValueError(
-            f"No frames found in {frame_dir} with extensions {frame_extensions}"
-        )
-
-    frame_files = sorted(frame_files, key=lambda x: x.name)
-    total_frames = len(frame_files)
-
-    # Determine frame indices
-    if num_frames is None:
-        frame_indices = list(range(total_frames))
-    else:
-        if total_frames < num_frames:
-            frame_indices = list(range(total_frames)) + [total_frames - 1] * (
-                num_frames - total_frames)
-        elif sample_strategy == 'uniform':
-            frame_indices = np.linspace(0,
-                                        total_frames - 1,
-                                        num_frames,
-                                        dtype=int).tolist()
-        elif sample_strategy == 'random':
-            frame_indices = sorted(
-                np.random.choice(total_frames, num_frames, replace=False))
-        else:
-            raise ValueError(f"Unknown sample_strategy: {sample_strategy}")
-
-    # Load frames
-    frames = []
-    for idx in frame_indices:
-        frame_path = frame_files[idx]
-        frame = cv2.imread(str(frame_path))
-
-        if frame is None:
-            raise RuntimeError(f"Failed to load frame: {frame_path}")
-
-        frame = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
-        frames.append(frame)
-
-    # Stack and convert to tensor
-    frames = np.stack(frames)  # [T, H, W, C]
-    frames = torch.from_numpy(frames).permute(0, 3, 1,
-                                              2).float()  # [T, C, H, W]
-
-    return frames
-
-
-def _detect_video_format(path: str | Path) -> str:
-    """
-    Detect if path is a video file or frame directory.
-    
-    Returns:
-        'video_file', 'frame_directory', or 'unknown'
-    """
-    path = Path(path)
-
-    if path.is_file():
-        return 'video_file'
-    elif path.is_dir():
-        # Check if contains image files
-        image_extensions = ['.jpg', '.jpeg', '.png', '.bmp']
-        for ext in image_extensions:
-            if list(path.glob(f"*{ext}")):
-                return 'frame_directory'
-        return 'unknown'
-    else:
-        raise ValueError(f"Path does not exist: {path}")
-
-
-def load_video_auto(video_path: str | Path,
-                    num_frames: int | None = 16,
-                    sample_strategy: str = 'uniform') -> torch.Tensor:
-    """
-    Automatically detect format and load video.
-    
-    Supports:
-    - Video files (MP4, AVI, MOV, MKV)
-    - Frame directories (JPG, PNG)
-    
-    Args:
-        video_path: Path to video file or frame directory
-        num_frames: Number of frames to extract
-        sample_strategy: 'uniform' or 'random'
-    
-    Returns:
-        video: [T, C, H, W]
-    """
-    format_type = _detect_video_format(video_path)
-
-    if format_type == 'video_file':
-        return _load_video_cv2(video_path, num_frames, sample_strategy)
-    elif format_type == 'frame_directory':
-        return _load_video_from_frames(video_path, num_frames, sample_strategy)
-    else:
-        raise ValueError(f"Unknown video format at {video_path}")
-
-
-def sample_clips_from_video(
-        video: torch.Tensor,
-        num_frames_per_clip: int = 16,
-        num_clips: int = 1,
-        strategy: str | ClipSamplingStrategy = ClipSamplingStrategy.BEGINNING,
-        frame_stride: int = 1,
-        temporal_stride: int = 1) -> list[torch.Tensor]:
-    """
-    Sample clips from a video with various strategies.
-        
-    Args:
-        video: [T, C, H, W] full video
-        num_frames_per_clip: Frames per clip
-        num_clips: Number of clips to extract
-        strategy: ClipSamplingStrategy or string ('beginning', 'random', etc.)
-        frame_stride: Skip frames (FPS control: 1=all, 2=every 2nd, 8=every 8th)
-        temporal_stride: Stride between clips for sliding window
-    
-    Returns:
-        List of clips, each [num_frames_per_clip, C, H, W]
-    
-    Examples:
-        >>> # Beginning clip (most common for FVD)
-        >>> clips = sample_clips_from_video(video, 16, strategy='beginning')
-        
-        >>> # Multiple random clips
-        >>> clips = sample_clips_from_video(video, 16, num_clips=4, strategy='random')
-        
-        >>> # Subsample FPS by 2x (every 2nd frame)
-        >>> clips = sample_clips_from_video(video, 16, frame_stride=2)
-        
-        >>> # Sliding window with overlap
-        >>> clips = sample_clips_from_video(video, 16, strategy='sliding', temporal_stride=8)
-    """
-    # Convert string to enum if needed
-    if isinstance(strategy, str):
-        strategy = ClipSamplingStrategy(strategy)
-
-    T, C, H, W = video.shape
-
-    # Apply frame stride (FPS subsampling)
-    if frame_stride > 1:
-        video = video[::frame_stride]
-        T = len(video)
-
-    effective_clip_length = num_frames_per_clip
-
-    # Handle videos shorter than clip length
-    if effective_clip_length > T:
-        pad_length = effective_clip_length - T
-        last_frame = video[-1:].repeat(pad_length, 1, 1, 1)
-        video = torch.cat([video, last_frame], dim=0)
-        T = len(video)
-
-    clips = []
-
-    if strategy == ClipSamplingStrategy.BEGINNING:
-        # Take first clip (most common for FVD evaluation)
-        clip = video[:effective_clip_length]
-        clips.append(clip)
-
-    elif strategy == ClipSamplingStrategy.MIDDLE:
-        # Take middle clip
-        start = (T - effective_clip_length) // 2
-        clip = video[start:start + effective_clip_length]
-        clips.append(clip)
-
-    elif strategy == ClipSamplingStrategy.RANDOM:
-        # Sample N random clips
-        for _ in range(num_clips):
-            if effective_clip_length == T:
-                start = 0
-            else:
-                start = np.random.randint(0, T - effective_clip_length + 1)
-            clip = video[start:start + effective_clip_length]
-            clips.append(clip)
-
-    elif strategy == ClipSamplingStrategy.UNIFORM:
-        # Uniformly spaced clips
-        if num_clips == 1:
-            # Single clip from middle
-            start = (T - effective_clip_length) // 2
-            clip = video[start:start + effective_clip_length]
-            clips.append(clip)
-        else:
-            # Multiple uniformly spaced clips
-            step = (T - effective_clip_length) / (num_clips -
-                                                  1) if num_clips > 1 else 0
-            for i in range(num_clips):
-                start = int(i * step)
-                start = min(start, T - effective_clip_length)
-                clip = video[start:start + effective_clip_length]
-                clips.append(clip)
-
-    elif strategy == ClipSamplingStrategy.SLIDING:
-        # Sliding window with stride
-        for start in range(0, T - effective_clip_length + 1, temporal_stride):
-            clip = video[start:start + effective_clip_length]
-            clips.append(clip)
-            if len(clips) >= num_clips:
-                break
-
-    elif strategy == ClipSamplingStrategy.ALL:
-        # All possible clips (overlapping)
-        for start in range(T - effective_clip_length + 1):
-            clip = video[start:start + effective_clip_length]
-            clips.append(clip)
-
-    else:
-        raise ValueError(f"Unknown strategy: {strategy}")
-
-    return clips
-
-
-def load_video_clips_streaming(directory: str | Path,
-                               num_frames: int = 16,
-                               max_videos: int | None = None,
-                               clip_strategy: str
-                               | ClipSamplingStrategy = 'beginning',
-                               frame_stride: int = 1,
-                               num_clips_per_video: int = 1,
-                               video_extensions: list[str] | None = None,
-                               support_frame_dirs: bool = True,
-                               target_size: tuple[int, int] | None = (224, 224),
-                               verbose: bool = True) -> Iterator[torch.Tensor]:
-    """
-    This generator yields clips one-by-one instead of loading all videos into RAM.
-    Perfect for large datasets where memory is limited.
-    
-    Args:
-        directory: Path to directory with videos
-        num_frames: Frames per clip
-        max_videos: Max videos to load
-        clip_strategy: 'beginning', 'random', 'uniform', etc.
-        frame_stride: Frame skip (1=all, 2=every 2nd, 8=every 8th)
-        num_clips_per_video: Number of clips per video
-        video_extensions: Video file extensions
-        support_frame_dirs: Also load frame directories
-        target_size: Resize clips to (H, W). If None, keep original size.
-        verbose: Show progress
-    
-    Yields:
-        clip: [T, C, H, W] individual clips
-    
-    Example:
-        >>> for clip in load_video_clips_streaming('data/videos/', num_frames=16):
-        >>>     features = model.extract_features(clip.unsqueeze(0))
-        >>>     # Process one clip at a time - low memory usage!
-    """
-    if video_extensions is None:
-        video_extensions = ['.mp4', '.avi', '.mov', '.mkv']
-
-    directory = Path(directory)
-
-    if not directory.exists():
-        raise FileNotFoundError(f"Directory not found: {directory}")
-
-    # Find video paths
-    video_paths: list[Path] = []
-
-    # Find video files
-    for ext in video_extensions:
-        video_paths.extend(directory.glob(f"**/*{ext}"))
-
-    # Find frame directories if enabled
-    if support_frame_dirs:
-        for subdir in directory.iterdir():
-            if subdir.is_dir():
-                # Check if it contains frames
-                image_extensions = ['.jpg', '.jpeg', '.png', '.bmp']
-                for ext in image_extensions:
-                    if list(subdir.glob(f"*{ext}")):
-                        video_paths.append(subdir)
-                        break
-
-    if len(video_paths) == 0:
-        raise ValueError(f"No videos found in {directory}")
-
-    video_paths = sorted(video_paths)
-
-    if max_videos is not None:
-        video_paths = video_paths[:max_videos]
-
-    if verbose:
-        print(f"Found {len(video_paths)} videos in {directory}")
-        if num_clips_per_video > 1:
-            print(f"Extracting {num_clips_per_video} clips per video...")
-        if frame_stride > 1:
-            print(f"Subsampling frames with stride {frame_stride}...")
-        if target_size:
-            print(f"Resizing clips to {target_size}...")
-
-    # Track statistics
-    failed_count = 0
-    total_clips = 0
-
-    iterator = tqdm(video_paths,
-                    desc="Loading videos") if verbose else video_paths
-
-    for video_path in iterator:
-        try:
-            # Load full video
-            video = load_video_auto(video_path,
-                                    num_frames=None,
-                                    sample_strategy='uniform')
-
-            # Sample clips from video
-            clips = sample_clips_from_video(video,
-                                            num_frames_per_clip=num_frames,
-                                            num_clips=num_clips_per_video,
-                                            strategy=clip_strategy,
-                                            frame_stride=frame_stride)
-
-            if target_size is not None:
-                resized_clips = []
-                for clip in clips:
-                    T, C, H, W = clip.shape
-                    if target_size != (H, W):
-                        # Resize to target size
-                        clip = clip.contiguous(
-                        )  # Fix non-contiguous tensors first
-                        clip_flat = clip.view(T * C, H,
-                                              W).unsqueeze(0)  # [1, T*C, H, W]
-                        clip_resized = torch.nn.functional.interpolate(
-                            clip_flat,
-                            size=target_size,
-                            mode='bilinear',
-                            align_corners=False)
-                        clip = clip_resized.squeeze(0).view(
-                            T, C, target_size[0],
-                            target_size[1])  # Back to [T, C, H, W]
-                    resized_clips.append(clip)
-                clips = resized_clips
-
-            # Yield clips one by one
-            for clip in clips:
-                yield clip
-                total_clips += 1
-
-            # Free memory
-            del video, clips
-
-        except Exception as e:
-            failed_count += 1
-            if verbose:
-                print(f"\nWarning: Failed to load {video_path}: {e}")
-            continue
-
-    # Validate
-    if total_clips == 0:
-        raise RuntimeError(f"Failed to load any videos from {directory}")
-
-    failure_rate = failed_count / len(video_paths)
-    if failure_rate > 0.1:  # More than 10% failed
-        print(
-            f"\nWARNING: {failure_rate:.1%} of videos failed to load ({failed_count}/{len(video_paths)})"
-        )
-
-    if verbose:
-        print(
-            f"\nSuccessfully loaded {total_clips} clips from {len(video_paths) - failed_count} videos"
-        )
@@ -1,7 +0,0 @@
-#!/bin/bash
-
-# 1. Install missing dependency
-pip install -q opencv-python-headless
-
-# 2. Run FVD script
-python benchmarks/fvd/run_fvd.py
@@ -1,4 +0,0 @@
-#!/bin/bash
-
-# 1. Install missing dependency
-pip install -q opencv-python-headless
@@ -77,7 +77,7 @@ https://github.com/user-attachments/assets/f3b6dd79-7b43-4b60-a0fa-3d6495ec5747
 Here is a diagram of how the window is configured and passed through the FastVideo pipeline:

 <div align="center">
-<img src="../../../docs/assets/images/STA_configuration.png" width="80%"/>
+<img src="../../../docs/source/_static/images/STA_configuration.png" width="80%"/>
 </div>


@@ -34,16 +34,16 @@ def pytorch_test(Q, K, V, block_sparse_mask, dO):
    )


-def block_sparse_kernel_test(Q, K, V, block_sparse_mask, variable_block_sizes, q_non_pad_index, kv_non_pad_index, q_num_blocks, kv_num_blocks, dO):
+def block_sparse_kernel_test(Q, K, V, block_sparse_mask, variable_block_sizes, non_pad_index, dO):
    Q = Q.detach().requires_grad_()
    K = K.detach().requires_grad_()
    V = V.detach().requires_grad_()
    
-    q_padded = vsa_pad(Q, q_non_pad_index, q_num_blocks, BLOCK_M)
-    k_padded = vsa_pad(K, kv_non_pad_index, kv_num_blocks, BLOCK_M)
-    v_padded = vsa_pad(V, kv_non_pad_index, kv_num_blocks, BLOCK_M)
+    q_padded = vsa_pad(Q, non_pad_index, variable_block_sizes.shape[0], BLOCK_M)
+    k_padded = vsa_pad(K, non_pad_index, variable_block_sizes.shape[0], BLOCK_M)
+    v_padded = vsa_pad(V, non_pad_index, variable_block_sizes.shape[0], BLOCK_M)
    output, _= block_sparse_attn(q_padded, k_padded, v_padded, block_sparse_mask, variable_block_sizes)
-    output = output[:, :, q_non_pad_index, :]
+    output = output[:, :, non_pad_index, :]
    output.backward(dO)
    return output, Q.grad, K.grad, V.grad

@@ -64,7 +64,7 @@ def generate_tensor(shape, dtype, device):
    tensor = torch.randn(shape, dtype=dtype, device=device)
    return tensor

-def generate_variable_block_sizes(num_blocks, min_size=16, max_size=64, device="cuda"):
+def generate_variable_block_sizes(num_blocks, min_size=32, max_size=64, device="cuda"):
    return torch.randint(min_size, max_size + 1, (num_blocks,), device=device, dtype=torch.int32)


@@ -86,21 +86,19 @@ def check_correctness(h, d, num_blocks, k,  num_iterations=20, error_mode='all')
    S = int(variable_block_sizes.sum().item())
    padded_S = num_blocks * BLOCK_M
    non_pad_index = get_non_pad_index(variable_block_sizes, num_blocks, BLOCK_M)
-    block_mask = generate_block_sparse_mask_for_function(h, num_blocks, num_blocks, k, device)
-    full_mask = create_full_mask_from_block_mask(block_mask, variable_block_sizes, variable_block_sizes, device)
+    block_mask = generate_block_sparse_mask_for_function(h, num_blocks, k, device)
+    full_mask = create_full_mask_from_block_mask(block_mask, variable_block_sizes, device)
    for _ in range(num_iterations):
        Q = generate_tensor((1, h, S, d),  torch.bfloat16, device)
        K = generate_tensor((1, h, S, d), torch.bfloat16, device)
        V = generate_tensor((1, h, S, d),  torch.bfloat16, device)
        dO = generate_tensor((1, h, S, d), torch.bfloat16, device)
-        
-        # print(Q.shape, K.shape, V.shape, dO.shape)

        # dO_padded = torch.zeros_like(dO_padded)
        # dO_padded[:, :, non_pad_index, :] = dO

        pt_o, pt_qg, pt_kg, pt_vg = pytorch_test(Q, K, V, full_mask, dO)
-        bs_o, bs_qg, bs_kg, bs_vg = block_sparse_kernel_test(Q, K, V, block_mask.unsqueeze(0), variable_block_sizes, non_pad_index, non_pad_index, num_blocks, num_blocks, dO)
+        bs_o, bs_qg, bs_kg, bs_vg = block_sparse_kernel_test(Q, K, V, block_mask.unsqueeze(0), variable_block_sizes,non_pad_index, dO)
        for name, (pt, bs) in zip(['gQ', 'gK', 'gV', 'gO'], [(pt_qg, bs_qg), (pt_kg, bs_kg), (pt_vg, bs_vg), (pt_o, bs_o)]):
            if bs is not None:
                diff = pt - bs
@@ -120,60 +118,6 @@ def check_correctness(h, d, num_blocks, k,  num_iterations=20, error_mode='all')

    return results

-def check_correctness_qkdiff(h, d, num_q_blocks, num_kv_blocks, k, num_iterations=20, error_mode='all'):
-    results = {
-        'gO': {'sum_diff': 0.0, 'sum_abs': 0.0, 'max_diff': 0.0},
-        'gQ': {'sum_diff': 0.0, 'sum_abs': 0.0, 'max_diff': 0.0},
-        'gK': {'sum_diff': 0.0, 'sum_abs': 0.0, 'max_diff': 0.0},
-        'gV': {'sum_diff': 0.0, 'sum_abs': 0.0, 'max_diff': 0.0},
-    }
-
-    device = "cuda" if torch.cuda.is_available() else "cpu"
-    q_variable_block_sizes = generate_variable_block_sizes(num_q_blocks, device=device)
-    kv_variable_block_sizes = generate_variable_block_sizes(num_kv_blocks, device=device)
-    
-    S_q = int(q_variable_block_sizes.sum().item())
-    S_kv = int(kv_variable_block_sizes.sum().item())
-    
-    q_non_pad_index = get_non_pad_index(q_variable_block_sizes, num_q_blocks, BLOCK_M)
-    kv_non_pad_index = get_non_pad_index(kv_variable_block_sizes, num_kv_blocks, BLOCK_M)
-    
-    block_mask = generate_block_sparse_mask_for_function(h, num_q_blocks, num_kv_blocks, k, device)
-    full_mask = create_full_mask_from_block_mask(block_mask, q_variable_block_sizes, kv_variable_block_sizes, device)
-    
-    for _ in range(num_iterations):
-        Q = generate_tensor((1, h, S_q, d),  torch.bfloat16, device)
-        K = generate_tensor((1, h, S_kv, d), torch.bfloat16, device)
-        V = generate_tensor((1, h, S_kv, d),  torch.bfloat16, device)
-        dO = generate_tensor((1, h, S_q, d), torch.bfloat16, device)
-        
-        # print(Q.shape, K.shape, V.shape, dO.shape)
-
-        pt_o, pt_qg, pt_kg, pt_vg = pytorch_test(Q, K, V, full_mask, dO)
-        bs_o, bs_qg, bs_kg, bs_vg = block_sparse_kernel_test(Q, K, V, block_mask.unsqueeze(0), kv_variable_block_sizes, q_non_pad_index, kv_non_pad_index, num_q_blocks, num_kv_blocks, dO)
-        
-        for name, (pt, bs) in zip(['gQ', 'gK', 'gV', 'gO'], [(pt_qg, bs_qg), (pt_kg, bs_kg), (pt_vg, bs_vg), (pt_o, bs_o)]):
-            if bs is not None:
-                diff = pt - bs
-                abs_diff = torch.abs(diff)
-                results[name]['sum_diff'] += torch.sum(abs_diff).item()
-                results[name]['sum_abs'] += torch.sum(torch.abs(pt)).item()
-                rel_max_diff = torch.max(abs_diff) / torch.mean(torch.abs(pt))
-                results[name]['max_diff'] = max(results[name]['max_diff'], rel_max_diff.item())
-        if torch.cuda.is_available():
-            torch.cuda.empty_cache()
-
-    total_elements_q = h * S_q * d * num_iterations
-    total_elements_kv = h * S_kv * d * num_iterations
-    
-    for name, data in results.items():
-        total_elements = total_elements_q if name in ['gQ', 'gO'] else total_elements_kv
-        avg_diff = data['sum_diff'] / total_elements
-        max_diff = data['max_diff']
-        results[name] = {'avg_diff': avg_diff, 'max_diff': max_diff}
-
-    return results
-
 def generate_error_graphs(h, d, error_mode='all'):
    test_configs = [
        {"num_blocks": 16, "k": 2, "description": "Small sequence"},
@@ -203,43 +147,10 @@ def generate_error_graphs(h, d, error_mode='all'):

    print("-" * 150)

-def generate_error_graphs_qkdiff(h, d, error_mode='all'):
-    test_configs = [
-        {"num_q_blocks": 16, "num_kv_blocks": 32, "k": 2, "description": "Small Q, Med KV"},
-        {"num_q_blocks": 32, "num_kv_blocks": 16, "k": 4, "description": "Med Q, Small KV"},
-        {"num_q_blocks": 53, "num_kv_blocks": 32, "k": 6, "description": "Large Q, Med KV"},
-        {"num_q_blocks": 16, "num_kv_blocks": 48, "k": 2, "description": "Small Q, Large KV"},
-        {"num_q_blocks": 48, "num_kv_blocks": 16, "k": 2, "description": "Large Q, Small KV"},
-    ]
-    
-    print(f"\nError Analysis (QK Diff) for h={h}, d={d}, mode={error_mode}")
-    print("=" * 150)
-    print(f"{'Config':<20} {'Q Blks':<8} {'KV Blks':<8} {'K':<4} "
-          f"{'gQ Avg':<12} {'Rel gQ Max':<12} "
-          f"{'gK Avg':<12} {'Rel gK Max':<12} "
-          f"{'gV Avg':<12} {'Rel gV Max':<12} "
-          f"{'gO Avg':<12} {'Rel gO Max':<12}")
-    print("-" * 150)
-
-    for config in test_configs:
-        num_q_blocks = config["num_q_blocks"]
-        num_kv_blocks = config["num_kv_blocks"]
-        k = config["k"]
-        description = config["description"]
-        results = check_correctness_qkdiff(h, d, num_q_blocks, num_kv_blocks, k, error_mode=error_mode)
-        print(f"{description:<20} {num_q_blocks:<8} {num_kv_blocks:<8} {k:<4} "
-              f"{results['gQ']['avg_diff']:<12.6e} {results['gQ']['max_diff']:<12.6e} "
-              f"{results['gK']['avg_diff']:<12.6e} {results['gK']['max_diff']:<12.6e} "
-              f"{results['gV']['avg_diff']:<12.6e} {results['gV']['max_diff']:<12.6e} "
-              f"{results['gO']['avg_diff']:<12.6e} {results['gO']['max_diff']:<12.6e}")
-
-    print("-" * 150)
-
 if __name__ == "__main__":
    h, d = 16, 128
    print("Block Sparse Attention with Variable Block Sizes Analysis")
    print("=" * 60)
    for mode in ['backward']:
        generate_error_graphs(h, d, error_mode=mode)
-        generate_error_graphs_qkdiff(h, d, error_mode=mode)
-    print("\nAnalysis completed for all modes.")
+    print("\nAnalysis completed for all modes.")
@@ -1,236 +0,0 @@
-import os
-import sys
-from typing import Tuple
-
-import torch
-
-# Make sure we can import from the project root (`vsa`, `tests.utils`, etc.)
-CURRENT_DIR = os.path.dirname(os.path.abspath(__file__))
-PROJECT_ROOT = os.path.dirname(CURRENT_DIR)
-if PROJECT_ROOT not in sys.path:
-    sys.path.append(PROJECT_ROOT)
-if CURRENT_DIR not in sys.path:
-    sys.path.append(CURRENT_DIR)
-
-from tests.utils import (
-    generate_block_sparse_mask_for_function,
-    create_full_mask_from_block_mask,
-)
-from vsa import block_sparse_attn, BLOCK_M
-import test_vsa as ref  # reuse helper functions from backward test
-
-
-def pytorch_forward(
-    Q: torch.Tensor,
-    K: torch.Tensor,
-    V: torch.Tensor,
-    block_sparse_mask: torch.Tensor,
-) -> torch.Tensor:
-    """
-    Dense PyTorch reference forward:
-    - Q: [1, h, S_q, d]
-    - K,V: [1, h, S_kv, d]
-    - block_sparse_mask: [h, S_q, S_kv] bool
-    """
-    q = Q.clone().float()
-    k = K.clone().float()
-    v = V.clone().float()
-
-    attn = torch.matmul(q, k.transpose(-2, -1))  # [1, h, S_q, S_kv]
-    attn = attn / (q.size(-1) ** 0.5)
-    attn = attn.masked_fill(~block_sparse_mask.unsqueeze(0), float("-inf"))
-    attn = torch.nn.functional.softmax(attn, dim=-1)
-    out = torch.matmul(attn, v)  # [1, h, S_q, d]
-    return out.to(torch.bfloat16)
-
-
-def block_sparse_forward_test(
-    Q: torch.Tensor,
-    K: torch.Tensor,
-    V: torch.Tensor,
-    block_sparse_mask: torch.Tensor,
-    variable_block_sizes: torch.Tensor,
-    q_non_pad_index: torch.Tensor,
-    kv_non_pad_index: torch.Tensor,
-    q_num_blocks: int,
-    kv_num_blocks: int,
-) -> torch.Tensor:
-    """
-    Forward-only wrapper around `block_sparse_attn`, mirroring `block_sparse_kernel_test`
-    but without any backward / grad logic.
-    """
-    Q = Q.detach()
-    K = K.detach()
-    V = V.detach()
-
-    q_padded = ref.vsa_pad(Q, q_non_pad_index, q_num_blocks, BLOCK_M)
-    k_padded = ref.vsa_pad(K, kv_non_pad_index, kv_num_blocks, BLOCK_M)
-    v_padded = ref.vsa_pad(V, kv_non_pad_index, kv_num_blocks, BLOCK_M)
-
-    out_padded, _ = block_sparse_attn(
-        q_padded, k_padded, v_padded, block_sparse_mask, variable_block_sizes
-    )
-    # Remove padding on the query side
-    out = out_padded[:, :, q_non_pad_index, :]
-    return out
-
-
-def run_forward_equal_qk(
-    h: int = 16,
-    d: int = 128,
-    num_blocks: int = 16,
-    k: int = 2,
-    num_iterations: int = 5,
-) -> Tuple[float, float]:
-    """
-    Forward-only correctness test for the case S_q == S_kv.
-    Mirrors `check_correctness` but only compares forward outputs.
-    """
-    assert torch.cuda.is_available(), "VSA kernels require CUDA"
-    device = "cuda"
-
-    variable_block_sizes = ref.generate_variable_block_sizes(
-        num_blocks, device=device
-    )
-    S = int(variable_block_sizes.sum().item())
-    non_pad_index = ref.get_non_pad_index(
-        variable_block_sizes, num_blocks, BLOCK_M
-    )
-
-    block_mask = generate_block_sparse_mask_for_function(
-        h, num_blocks, num_blocks, k, device
-    )
-    full_mask = create_full_mask_from_block_mask(
-        block_mask, variable_block_sizes, variable_block_sizes, device
-    )
-    print(f"[qkequal] h: {h}, d: {d}, num_blocks: {num_blocks}, k: {k}")
-    print(f"[qkequal] variable_block_sizes: {variable_block_sizes}, non_pad_index: {non_pad_index.shape}, block_mask: {block_mask.shape}, full_mask: {full_mask.shape}")
-    sum_diff = 0.0
-    sum_abs = 0.0
-    max_rel_diff = 0.0
-
-    for i in range(num_iterations):
-        Q = ref.generate_tensor((1, h, S, d), torch.bfloat16, device)
-        K = ref.generate_tensor((1, h, S, d), torch.bfloat16, device)
-        V = ref.generate_tensor((1, h, S, d), torch.bfloat16, device)
-
-        if i == 0: print(f"[qkequal] Q: {Q.shape}, K: {K.shape}, V: {V.shape}, full_mask: {full_mask.shape}")
-        if i == 0: print(f"[qkequal] block_mask: {block_mask.shape}")
-
-        pt_o = pytorch_forward(Q, K, V, full_mask)
-        bs_o = block_sparse_forward_test(
-            Q,
-            K,
-            V,
-            block_mask.unsqueeze(0),
-            variable_block_sizes,
-            non_pad_index,
-            non_pad_index,
-            num_blocks,
-            num_blocks,
-        )
-
-        diff = (pt_o - bs_o).abs()
-        sum_diff += diff.sum().item()
-        sum_abs += pt_o.abs().sum().item()
-        rel_max = diff.max() / (pt_o.abs().mean() + 1e-6)
-        max_rel_diff = max(max_rel_diff, rel_max.item())
-
-    total_elems = h * S * d * num_iterations
-    avg_abs_err = sum_diff / total_elems
-    return avg_abs_err, max_rel_diff
-
-
-def run_forward_qk_diff(
-    h: int = 16,
-    d: int = 128,
-    num_q_blocks: int = 16,
-    num_kv_blocks: int = 32,
-    k: int = 2,
-    num_iterations: int = 5,
-) -> Tuple[float, float]:
-    """
-    Forward-only correctness test for the case S_q != S_kv.
-
-    NOTE:
-      - The Triton backend supports different Q/KV logical lengths via padding.
-      - The SM90 (H100) CUDA backend currently assumes the same number of blocks
-        for Q and KV, so we skip this test there.
-    """
-    assert torch.cuda.is_available(), "VSA kernels require CUDA"
-
-    device = "cuda"
-
-    q_variable_block_sizes = ref.generate_variable_block_sizes(
-        num_q_blocks, device=device
-    )
-    kv_variable_block_sizes = ref.generate_variable_block_sizes(
-        num_kv_blocks, device=device
-    )
-
-    S_q = int(q_variable_block_sizes.sum().item())
-    S_kv = int(kv_variable_block_sizes.sum().item())
-
-    q_non_pad_index = ref.get_non_pad_index(
-        q_variable_block_sizes, num_q_blocks, BLOCK_M
-    )
-    kv_non_pad_index = ref.get_non_pad_index(
-        kv_variable_block_sizes, num_kv_blocks, BLOCK_M
-    )
-
-    block_mask = generate_block_sparse_mask_for_function(
-        h, num_q_blocks, num_kv_blocks, k, device
-    )
-    full_mask = create_full_mask_from_block_mask(
-        block_mask, q_variable_block_sizes, kv_variable_block_sizes, device
-    )
-
-    sum_diff = 0.0
-    sum_abs = 0.0
-    max_rel_diff = 0.0
-
-    for _ in range(num_iterations):
-        Q = ref.generate_tensor((1, h, S_q, d), torch.bfloat16, device)
-        K = ref.generate_tensor((1, h, S_kv, d), torch.bfloat16, device)
-        V = ref.generate_tensor((1, h, S_kv, d), torch.bfloat16, device)
-
-        pt_o = pytorch_forward(Q, K, V, full_mask)
-        bs_o = block_sparse_forward_test(
-            Q,
-            K,
-            V,
-            block_mask.unsqueeze(0),
-            kv_variable_block_sizes,
-            q_non_pad_index,
-            kv_non_pad_index,
-            num_q_blocks,
-            num_kv_blocks,
-        )
-
-        diff = (pt_o - bs_o).abs()
-        sum_diff += diff.sum().item()
-        sum_abs += pt_o.abs().sum().item()
-        rel_max = diff.max() / (pt_o.abs().mean() + 1e-6)
-        max_rel_diff = max(max_rel_diff, rel_max.item())
-
-    total_elems = h * S_q * d * num_iterations
-    avg_abs_err = sum_diff / total_elems
-    return avg_abs_err, max_rel_diff
-
-
-if __name__ == "__main__":
-    h, d = 16, 128
-    print("Forward Block Sparse Attention Check (QK Equal)")
-    print("=" * 80)
-    avg_err_eq, max_rel_eq = run_forward_equal_qk(h, d, num_blocks=32, k=2)
-    print(f"QK equal: avg |ΔO| = {avg_err_eq:.6e}, max rel ΔO = {max_rel_eq:.6e}")
-
-    print("\nForward Block Sparse Attention Check (QK Different)")
-    print("=" * 80)
-    avg_err_diff, max_rel_diff = run_forward_qk_diff(
-        h, d, num_q_blocks=32, num_kv_blocks=48, k=2
-    )
-    print(
-        f"QK diff:  avg |ΔO| = {avg_err_diff:.6e}, max rel ΔO = {max_rel_diff:.6e}"
-    )
-
@@ -1,60 +1,54 @@
 import torch

-def generate_block_sparse_mask_for_function(h, num_q_blocks, num_kv_blocks, k, device="cuda"):
+def generate_block_sparse_mask_for_function(h, num_blocks, k, device="cuda"):
    """
-    Generate block sparse mask of shape [h, num_q_blocks, num_kv_blocks].
+    Generate block sparse mask of shape [h, num_blocks, num_blocks].
    
    Args:
        h: number of heads
-        num_q_blocks: number of query blocks
-        num_kv_blocks: number of key/value blocks
+        num_blocks: number of blocks
        k: number of kv blocks each q block attends to
        device: device to create tensors on

    Returns:
-        block_sparse_mask: [h, num_q_blocks, num_kv_blocks] bool tensor
+        block_sparse_mask: [h, num_blocks, num_blocks] bool tensor
    """
-    k = min(k, num_kv_blocks)
-    scores = torch.rand(h, num_q_blocks, num_kv_blocks, device=device)
+    k = min(k, num_blocks)
+    scores = torch.rand(h, num_blocks, num_blocks, device=device)
    _, indices = torch.topk(scores, k, dim=-1)
-    block_sparse_mask = torch.zeros(h, num_q_blocks, num_kv_blocks, dtype=torch.bool, device=device)
+    block_sparse_mask = torch.zeros(h, num_blocks, num_blocks, dtype=torch.bool, device=device)

    block_sparse_mask = block_sparse_mask.scatter_(2, indices, 1).bool()
    return block_sparse_mask


-def create_full_mask_from_block_mask(block_sparse_mask, q_variable_block_sizes, 
-                                     kv_variable_block_sizes, device="cuda"):
+def create_full_mask_from_block_mask(block_sparse_mask, variable_block_sizes, device="cuda"):
    """
    Convert block-level sparse mask to full attention mask.
    
    Args:
-        block_sparse_mask: [h, num_q_blocks, num_kv_blocks] bool tensor
-        q_variable_block_sizes: [num_q_blocks] tensor
-        kv_variable_block_sizes: [num_kv_blocks] tensor
+        block_sparse_mask: [h, num_blocks, num_blocks] bool tensor
+        variable_block_sizes: [num_blocks] tensor
        device: device to create tensors on

    Returns:
-        full_mask: [h, S_q, S_kv] bool tensor where S = total sequence length
+        full_mask: [h, S, S] bool tensor where S = total sequence length
    """
-    h, num_q_blocks, num_kv_blocks = block_sparse_mask.shape
-    total_q_seq_len = q_variable_block_sizes.sum().item()
-    total_kv_seq_len = kv_variable_block_sizes.sum().item()
-    
-    q_cumsum = torch.cat([torch.tensor([0], device=device), q_variable_block_sizes.cumsum(dim=0)[:-1]])
-    kv_cumsum = torch.cat([torch.tensor([0], device=device), kv_variable_block_sizes.cumsum(dim=0)[:-1]])
+    h, num_blocks, _ = block_sparse_mask.shape
+    total_seq_len = variable_block_sizes.sum().item()
+    cumsum = torch.cat([torch.tensor([0], device=device), variable_block_sizes.cumsum(dim=0)[:-1]])

-    full_mask = torch.zeros(h, total_q_seq_len, total_kv_seq_len, dtype=torch.bool, device=device)
+    full_mask = torch.zeros(h, total_seq_len, total_seq_len, dtype=torch.bool, device=device)

    for head in range(h):
-        for q_block in range(num_q_blocks):
-            q_start = q_cumsum[q_block]
-            q_end = q_start + q_variable_block_sizes[q_block]
+        for q_block in range(num_blocks):
+            q_start = cumsum[q_block]
+            q_end = q_start + variable_block_sizes[q_block]

-            for kv_block in range(num_kv_blocks):
+            for kv_block in range(num_blocks):
                if block_sparse_mask[head, q_block, kv_block]:
-                    kv_start = kv_cumsum[kv_block]
-                    kv_end = kv_start + kv_variable_block_sizes[kv_block]
+                    kv_start = cumsum[kv_block]
+                    kv_end = kv_start + variable_block_sizes[kv_block]
                    full_mask[head, q_start:q_end, kv_start:kv_end] = True

    return full_mask
@@ -250,9 +250,8 @@ def _attn_bwd_dq(dq, q, K, V,  #
    
    
    for blk_idx in range(kv_blocks*2):
-        kv_idx = tl.load(kv_ptr + blk_idx//2).to(tl.int32)
-        block_size = tl.load(variable_block_sizes + kv_idx) -  (blk_idx % 2) * step_n
-        block_sparse_offset = (kv_idx*2 + blk_idx%2) * step_n * stride_tok
+        block_sparse_offset = (tl.load(kv_ptr + blk_idx//2).to(tl.int32)*2 + blk_idx%2) *step_n * stride_tok
+        block_size = tl.load(variable_block_sizes + blk_idx//2) - (blk_idx%2) * step_n
        kT = tl.load(kT_ptrs + block_sparse_offset)
        vT = tl.load(vT_ptrs + block_sparse_offset)
        qk = tl.dot(q, kT)
@@ -672,32 +672,23 @@ block_sparse_attention_forward(
    torch::Tensor v, 
    torch::Tensor q2k_block_sparse_index, 
    torch::Tensor q2k_block_sparse_num, 
-    torch::Tensor kv_block_size
+    torch::Tensor block_size
 )
 {
    CHECK_INPUT(q);
    CHECK_INPUT(k);
    CHECK_INPUT(v);

-    // q shape: (batch, qo_heads, q_seq_len, head_dim)
-    // k shape: (batch, kv_heads, kv_seq_len, head_dim)
-    // v shape: (batch, kv_heads, kv_seq_len, head_dim)
-    // q2k_block_sparse_index shape: (batch, qo_heads, num_q_blocks, max_kv_blocks_per_q)
-    // q2k_block_sparse_num shape: (batch, qo_heads, num_q_blocks)
-    // kv_block_size shape: (num_kv_blocks) This does not need other dimensions because across all batch/heads the padding is the same.
-    
    auto batch    = q.size(0);
-    auto q_seq_len  = q.size(2);
-    auto kv_seq_len = k.size(2);
+    auto seq_len  = q.size(2); 
    auto head_dim = q.size(3); 
    auto qo_heads = q.size(1);
    auto kv_heads = k.size(1);
    auto max_kv_blocks_per_q = q2k_block_sparse_index.size(3);
-    auto num_q_blocks = q2k_block_sparse_index.size(2);
-    auto num_kv_blocks = kv_block_size.size(0);
+    auto num_q_blocks = block_size.size(0);
    TORCH_CHECK(batch==1, "Batch size dim will be removed in the future, please set batch to 1");
-    TORCH_CHECK(num_q_blocks * BLOCK_M == q_seq_len, "This kernel supports variable q block size, but it assumes the input sequence is properly padded.");
-    TORCH_CHECK(num_kv_blocks * BLOCK_M == kv_seq_len, "This kernel supports variable kv block size, but it assumes the input sequence is properly padded.");
+    TORCH_CHECK(num_q_blocks * 64 == seq_len, "This kernel supports variable block size, but it assumes the input sequence is properly padded.");
+    TORCH_CHECK(num_q_blocks == q2k_block_sparse_index.size(2), "Number of Q blocks does not match between q2k_block_sparse_index and block_size");
    // check to see that these dimensions match for all inputs
    TORCH_CHECK(q.size(0) == batch, "Q batch dimension - idx 0 - must match for all inputs");
    TORCH_CHECK(k.size(0) == batch, "K batch dimension - idx 0 - must match for all inputs");
@@ -705,8 +696,11 @@ block_sparse_attention_forward(
    TORCH_CHECK(q2k_block_sparse_index.size(0) == batch, "q2k_block_sparse_index batch dimension - idx 0 - must match for all inputs");
    TORCH_CHECK(q2k_block_sparse_num.size(0) == batch, "q2k_block_sparse_num batch dimension - idx 0 - must match for all inputs");

-    TORCH_CHECK(v.size(2) == kv_seq_len, "V sequence length dimension - idx 2 - must match K inputs");
-    TORCH_CHECK(q2k_block_sparse_num.size(2) == num_q_blocks, "q2k_block_sparse_num idx 2 - must match num_q_blocks");
+    TORCH_CHECK(q.size(2) == seq_len, "Q sequence length dimension - idx 2 - must match for all inputs");
+    TORCH_CHECK(k.size(2) == seq_len, "K sequence length dimension - idx 2 - must match for all inputs");
+    TORCH_CHECK(v.size(2) == seq_len, "V sequence length dimension - idx 2 - must match for all inputs");
+    TORCH_CHECK(q2k_block_sparse_index.size(2) == seq_len / BLOCK_M, "q2k_block_sparse_index idx 2 - must match seq_len / BLOCK_M");
+    TORCH_CHECK(q2k_block_sparse_num.size(2) == seq_len / BLOCK_M, "q2k_block_sparse_num idx 2 - must match seq_len / BLOCK_M");


    TORCH_CHECK(q.size(3) == head_dim, "Q head dimension - idx 3 - must match for all non-vector inputs");
@@ -733,12 +727,12 @@ block_sparse_attention_forward(
    // for the returned outputs
    torch::Tensor o     = torch::empty({static_cast<const uint>(batch), 
                                        static_cast<const uint>(qo_heads), 
-                                        static_cast<const uint>(q_seq_len), 
+                                        static_cast<const uint>(seq_len), 
                                        static_cast<const uint>(head_dim)}, v.options());
    
    torch::Tensor l_vec = torch::empty({static_cast<const uint>(batch), 
                                        static_cast<const uint>(qo_heads), 
-                                        static_cast<const uint>(q_seq_len), 
+                                        static_cast<const uint>(seq_len), 
                                        static_cast<const uint>(1)}, 
                                        torch::TensorOptions().dtype(torch::kFloat).device(q.device()).memory_format(at::MemoryFormat::Contiguous));
        
@@ -768,11 +762,11 @@ block_sparse_attention_forward(

        using globals      = fwd_globals<64>;

-        q_global qg_arg{d_q, static_cast<unsigned int>(batch), static_cast<unsigned int>(qo_heads), static_cast<unsigned int>(q_seq_len), 64U};
-        k_global kg_arg{d_k, static_cast<unsigned int>(batch), static_cast<unsigned int>(kv_heads), static_cast<unsigned int>(kv_seq_len), 64U};
-        v_global vg_arg{d_v, static_cast<unsigned int>(batch), static_cast<unsigned int>(kv_heads), static_cast<unsigned int>(kv_seq_len), 64U};
-        l_global lg_arg{d_l, static_cast<unsigned int>(batch), static_cast<unsigned int>(qo_heads), 1U,   static_cast<unsigned int>(q_seq_len)};
-        o_global og_arg{d_o, static_cast<unsigned int>(batch), static_cast<unsigned int>(qo_heads), static_cast<unsigned int>(q_seq_len), 64U};
+        q_global qg_arg{d_q, static_cast<unsigned int>(batch), static_cast<unsigned int>(qo_heads), static_cast<unsigned int>(seq_len), 64U};
+        k_global kg_arg{d_k, static_cast<unsigned int>(batch), static_cast<unsigned int>(kv_heads), static_cast<unsigned int>(seq_len), 64U};
+        v_global vg_arg{d_v, static_cast<unsigned int>(batch), static_cast<unsigned int>(kv_heads), static_cast<unsigned int>(seq_len), 64U};
+        l_global lg_arg{d_l, static_cast<unsigned int>(batch), static_cast<unsigned int>(qo_heads), 1U,   static_cast<unsigned int>(seq_len)};
+        o_global og_arg{d_o, static_cast<unsigned int>(batch), static_cast<unsigned int>(qo_heads), static_cast<unsigned int>(seq_len), 64U};

        globals g{
            qg_arg, 
@@ -780,17 +774,17 @@ block_sparse_attention_forward(
            vg_arg, 
            lg_arg, 
            og_arg, 
-            static_cast<int>(q_seq_len), 
+            static_cast<int>(seq_len), 
            static_cast<int>(hr), 
            static_cast<int>(max_kv_blocks_per_q), 
            reinterpret_cast<int32_t*>(q2k_block_sparse_index.data_ptr()), 
            reinterpret_cast<int32_t*>(q2k_block_sparse_num.data_ptr()),
-            reinterpret_cast<int32_t*>(kv_block_size.data_ptr())
+            reinterpret_cast<int32_t*>(block_size.data_ptr())
        };

        constexpr int mem_size = 54000;

-        dim3 grid(q_seq_len/(BLOCK_M), qo_heads, batch);
+        dim3 grid(seq_len/(64), qo_heads, batch);

        cudaFuncSetAttribute(
            fwd_attend_ker<64>,
@@ -819,11 +813,11 @@ block_sparse_attention_forward(

        using globals      = fwd_globals<128>;

-        q_global qg_arg{d_q, static_cast<unsigned int>(batch), static_cast<unsigned int>(qo_heads), static_cast<unsigned int>(q_seq_len), 128U};
-        k_global kg_arg{d_k, static_cast<unsigned int>(batch), static_cast<unsigned int>(kv_heads), static_cast<unsigned int>(kv_seq_len), 128U};
-        v_global vg_arg{d_v, static_cast<unsigned int>(batch), static_cast<unsigned int>(kv_heads), static_cast<unsigned int>(kv_seq_len), 128U};
-        l_global lg_arg{d_l, static_cast<unsigned int>(batch), static_cast<unsigned int>(qo_heads), 1U,   static_cast<unsigned int>(q_seq_len)};
-        o_global og_arg{d_o, static_cast<unsigned int>(batch), static_cast<unsigned int>(qo_heads), static_cast<unsigned int>(q_seq_len), 128U};
+        q_global qg_arg{d_q, static_cast<unsigned int>(batch), static_cast<unsigned int>(qo_heads), static_cast<unsigned int>(seq_len), 128U};
+        k_global kg_arg{d_k, static_cast<unsigned int>(batch), static_cast<unsigned int>(kv_heads), static_cast<unsigned int>(seq_len), 128U};
+        v_global vg_arg{d_v, static_cast<unsigned int>(batch), static_cast<unsigned int>(kv_heads), static_cast<unsigned int>(seq_len), 128U};
+        l_global lg_arg{d_l, static_cast<unsigned int>(batch), static_cast<unsigned int>(qo_heads), 1U,   static_cast<unsigned int>(seq_len)};
+        o_global og_arg{d_o, static_cast<unsigned int>(batch), static_cast<unsigned int>(qo_heads), static_cast<unsigned int>(seq_len), 128U};

        globals g{
            qg_arg, 
@@ -831,17 +825,17 @@ block_sparse_attention_forward(
            vg_arg, 
            lg_arg, 
            og_arg, 
-            static_cast<int>(q_seq_len), 
+            static_cast<int>(seq_len), 
            static_cast<int>(hr), 
            static_cast<int>(max_kv_blocks_per_q), 
            reinterpret_cast<int32_t*>(q2k_block_sparse_index.data_ptr()), 
            reinterpret_cast<int32_t*>(q2k_block_sparse_num.data_ptr()),
-            reinterpret_cast<int32_t*>(kv_block_size.data_ptr())
+            reinterpret_cast<int32_t*>(block_size.data_ptr())
        };

        constexpr int mem_size = 54000;

-        dim3 grid(q_seq_len/(BLOCK_M), qo_heads, batch);
+        dim3 grid(seq_len/(64), qo_heads, batch);

        cudaFuncSetAttribute(
            fwd_attend_ker<128>,
@@ -868,7 +862,7 @@ block_sparse_attention_backward(torch::Tensor q,
                   torch::Tensor og,
                   torch::Tensor k2q_block_sparse_index,
                   torch::Tensor k2q_block_sparse_num,
-                   torch::Tensor kv_block_size)
+                   torch::Tensor block_size)
 {
    CHECK_INPUT(q);
    CHECK_INPUT(k);
@@ -877,23 +871,11 @@ block_sparse_attention_backward(torch::Tensor q,
    CHECK_INPUT(o);
    CHECK_INPUT(og);

-    // q: [batch, qo_heads, q_seq_len, head_dim]
-    // k: [batch, kv_heads, kv_seq_len, head_dim]
-    // v: [batch, kv_heads, kv_seq_len, head_dim]
-    // o: [batch, qo_heads, q_seq_len, head_dim]
-    // l_vec: [batch, qo_heads, q_seq_len, 1]
-    // og: [batch, qo_heads, q_seq_len, head_dim]
-    // k2q_block_sparse_index: [batch, kv_heads, num_kv_blocks, max_num_q_blocks]
-    // k2q_block_sparse_num: [batch, kv_heads, num_kv_blocks]
-    // kv_block_size: [num_kv_blocks]
-
    auto batch    = q.size(0);
-    auto q_seq_len  = q.size(2);
-    auto kv_seq_len = k.size(2);
+    auto seq_len  = q.size(2);
    auto head_dim = q.size(3);
    auto max_q_blocks_per_kv = k2q_block_sparse_index.size(3);
-    auto num_kv_blocks = kv_block_size.size(0);
-    TORCH_CHECK(k2q_block_sparse_index.size(2) == num_kv_blocks, "k2q_block_sparse_index.size(2) must match num_kv_blocks (kv_block_size.size(0))");
+    TORCH_CHECK(k2q_block_sparse_index.size(2) == block_size.size(0), "k2q_block_sparse_index.size(2) must match block_size.size(0)");
    // check to see that these dimensions match for all inputs
    TORCH_CHECK(q.size(0)     == batch, "Q  batch dimension - idx 0 - must match for all inputs");
    TORCH_CHECK(k.size(0)     == batch, "K  batch dimension - idx 0 - must match for all inputs");
@@ -904,18 +886,23 @@ block_sparse_attention_backward(torch::Tensor q,
    TORCH_CHECK(k2q_block_sparse_index.size(0) == batch, "k2q_block_sparse_index batch dimension - idx 0 - must match for all inputs");
    TORCH_CHECK(k2q_block_sparse_num.size(0) == batch, "k2q_block_sparse_num batch dimension - idx 0 - must match for all inputs");

-    TORCH_CHECK(v.size(2)     == kv_seq_len, "V  sequence length dimension - idx 2 - must match K sequence length");
-    TORCH_CHECK(l_vec.size(2) == q_seq_len, "L  sequence length dimension - idx 2 - must match Q sequence length");
-    TORCH_CHECK(o.size(2)     == q_seq_len, "O  sequence length dimension - idx 2 - must match Q sequence length");
-    TORCH_CHECK(og.size(2)    == q_seq_len, "OG sequence length dimension - idx 2 - must match Q sequence length");
-    TORCH_CHECK(k2q_block_sparse_index.size(2) == num_kv_blocks, "k2q_block_sparse_index idx 2 - must match num_kv_blocks (kv_block_size.size(0))");
-    TORCH_CHECK(k2q_block_sparse_num.size(2) == num_kv_blocks, "k2q_block_sparse_num idx 2 - must match num_kv_blocks (kv_block_size.size(0))");
+
+    TORCH_CHECK(q.size(2)     == seq_len, "Q  sequence length dimension - idx 2 - must match for all inputs");
+    TORCH_CHECK(k.size(2)     == seq_len, "K  sequence length dimension - idx 2 - must match for all inputs");
+    TORCH_CHECK(v.size(2)     == seq_len, "V  sequence length dimension - idx 2 - must match for all inputs");
+    TORCH_CHECK(l_vec.size(2) == seq_len, "L  sequence length dimension - idx 2 - must match for all inputs");
+    TORCH_CHECK(o.size(2)     == seq_len, "O  sequence length dimension - idx 2 - must match for all inputs");
+    TORCH_CHECK(og.size(2)    == seq_len, "OG sequence length dimension - idx 2 - must match for all inputs");
+    TORCH_CHECK(k2q_block_sparse_index.size(2) == seq_len / BLOCK_N, "k2q_block_sparse_index idx 2 - must match seq_len / BLOCK_N");
+    TORCH_CHECK(k2q_block_sparse_num.size(2) == seq_len / BLOCK_N, "k2q_block_sparse_num idx 2 - must match seq_len / BLOCK_N");

    TORCH_CHECK(q.size(3)  == head_dim, "Q  head dimension - idx 3 - must match for all non-vector inputs");
    TORCH_CHECK(k.size(3)  == head_dim, "K  head dimension - idx 3 - must match for all non-vector inputs");
    TORCH_CHECK(v.size(3)  == head_dim, "V  head dimension - idx 3 - must match for all non-vector inputs");
    TORCH_CHECK(o.size(3)  == head_dim, "O  head dimension - idx 3 - must match for all non-vector inputs");
    TORCH_CHECK(og.size(3) == head_dim, "OG head dimension - idx 3 - must match for all non-vector inputs");
+    
+

    auto qo_heads = q.size(1);
    auto kv_heads = k.size(1);
@@ -942,20 +929,20 @@ block_sparse_attention_backward(torch::Tensor q,

    torch::Tensor qg = torch::zeros({static_cast<const uint>(batch), 
                                     static_cast<const uint>(qo_heads), 
-                                     static_cast<const uint>(q_seq_len), 
+                                     static_cast<const uint>(seq_len), 
                                     static_cast<const uint>(head_dim)},   l_vec.options());
    torch::Tensor kg = torch::zeros({static_cast<const uint>(batch), 
                                     static_cast<const uint>(kv_heads), 
-                                     static_cast<const uint>(kv_seq_len), 
+                                     static_cast<const uint>(seq_len), 
                                     static_cast<const uint>(head_dim)},   l_vec.options());
    torch::Tensor vg = torch::zeros({static_cast<const uint>(batch), 
                                     static_cast<const uint>(kv_heads), 
-                                     static_cast<const uint>(kv_seq_len), 
+                                     static_cast<const uint>(seq_len), 
                                     static_cast<const uint>(head_dim)},   l_vec.options());
    
    torch::Tensor d_vec = torch::empty({static_cast<const uint>(batch), 
                                        static_cast<const uint>(qo_heads), 
-                                        static_cast<const uint>(q_seq_len), 
+                                        static_cast<const uint>(seq_len), 
                                        static_cast<const uint>(1)},       l_vec.options());

    float*         qg_ptr = qg.data_ptr<float>();
@@ -984,7 +971,7 @@ block_sparse_attention_backward(torch::Tensor q,
    //  cudaStreamSynchronize(stream);

    // TORCH_CHECK(seq_len % (4*kittens::TILE_DIM*4) == 0, "sequence length must be divisible by 256");
-    dim3 grid_bwd(q_seq_len/(PREP_NUM_WARPS*kittens::TILE_ROW_DIM<bf16>*4), qo_heads, batch);
+    dim3 grid_bwd(seq_len/(PREP_NUM_WARPS*kittens::TILE_ROW_DIM<bf16>*4), qo_heads, batch);

    if (head_dim == 64)  {
        using og_tile = st_bf<4*16, 64>;
@@ -997,9 +984,9 @@ block_sparse_attention_backward(torch::Tensor q,

        using bwd_prep_globals = bwd_prep_globals<64>;

-        og_global prep_og_arg{d_og, static_cast<unsigned int>(batch), static_cast<unsigned int>(qo_heads), static_cast<unsigned int>(q_seq_len), 64U};
-        o_global  prep_o_arg {d_o,  static_cast<unsigned int>(batch), static_cast<unsigned int>(qo_heads), static_cast<unsigned int>(q_seq_len), 64U};
-        d_global  prep_d_arg {d_d,  static_cast<unsigned int>(batch), static_cast<unsigned int>(qo_heads), 1U,  static_cast<unsigned int>(q_seq_len)};
+        og_global prep_og_arg{d_og, static_cast<unsigned int>(batch), static_cast<unsigned int>(qo_heads), static_cast<unsigned int>(seq_len), 64U};
+        o_global  prep_o_arg {d_o,  static_cast<unsigned int>(batch), static_cast<unsigned int>(qo_heads), static_cast<unsigned int>(seq_len), 64U};
+        d_global  prep_d_arg {d_d,  static_cast<unsigned int>(batch), static_cast<unsigned int>(qo_heads), 1U,  static_cast<unsigned int>(seq_len)};

        bwd_prep_globals bwd_g{prep_og_arg, prep_o_arg, prep_d_arg};

@@ -1036,15 +1023,15 @@ block_sparse_attention_backward(torch::Tensor q,

        using bwd_global_args = bwd_globals<64>;

-        bwd_q_global  bwd_q_arg {d_q,  static_cast<unsigned int>(batch), static_cast<unsigned int>(qo_heads), static_cast<unsigned int>(q_seq_len), 64U};
-        bwd_k_global  bwd_k_arg {d_k,  static_cast<unsigned int>(batch), static_cast<unsigned int>(kv_heads), static_cast<unsigned int>(kv_seq_len), 64U};
-        bwd_v_global  bwd_v_arg {d_v,  static_cast<unsigned int>(batch), static_cast<unsigned int>(kv_heads), static_cast<unsigned int>(kv_seq_len), 64U};
-        bwd_og_global bwd_og_arg{d_og, static_cast<unsigned int>(batch), static_cast<unsigned int>(qo_heads), static_cast<unsigned int>(q_seq_len), 64U};
-        bwd_qg_global bwd_qg_arg{d_qg, static_cast<unsigned int>(batch), static_cast<unsigned int>(qo_heads), static_cast<unsigned int>(q_seq_len), 64U};
-        bwd_kg_global bwd_kg_arg{d_kg, static_cast<unsigned int>(batch), static_cast<unsigned int>(kv_heads), static_cast<unsigned int>(kv_seq_len), 64U};
-        bwd_vg_global bwd_vg_arg{d_vg, static_cast<unsigned int>(batch), static_cast<unsigned int>(kv_heads), static_cast<unsigned int>(kv_seq_len), 64U};
-        bwd_l_global  bwd_l_arg {d_l,  static_cast<unsigned int>(batch), static_cast<unsigned int>(qo_heads), 1U,  static_cast<unsigned int>(q_seq_len)};
-        bwd_d_global  bwd_d_arg {d_d,  static_cast<unsigned int>(batch), static_cast<unsigned int>(qo_heads), 1U,  static_cast<unsigned int>(q_seq_len)};
+        bwd_q_global  bwd_q_arg {d_q,  static_cast<unsigned int>(batch), static_cast<unsigned int>(qo_heads), static_cast<unsigned int>(seq_len), 64U};
+        bwd_k_global  bwd_k_arg {d_k,  static_cast<unsigned int>(batch), static_cast<unsigned int>(kv_heads), static_cast<unsigned int>(seq_len), 64U};
+        bwd_v_global  bwd_v_arg {d_v,  static_cast<unsigned int>(batch), static_cast<unsigned int>(kv_heads), static_cast<unsigned int>(seq_len), 64U};
+        bwd_og_global bwd_og_arg{d_og, static_cast<unsigned int>(batch), static_cast<unsigned int>(qo_heads), static_cast<unsigned int>(seq_len), 64U};
+        bwd_qg_global bwd_qg_arg{d_qg, static_cast<unsigned int>(batch), static_cast<unsigned int>(qo_heads), static_cast<unsigned int>(seq_len), 64U};
+        bwd_kg_global bwd_kg_arg{d_kg, static_cast<unsigned int>(batch), static_cast<unsigned int>(kv_heads), static_cast<unsigned int>(seq_len), 64U};
+        bwd_vg_global bwd_vg_arg{d_vg, static_cast<unsigned int>(batch), static_cast<unsigned int>(kv_heads), static_cast<unsigned int>(seq_len), 64U};
+        bwd_l_global  bwd_l_arg {d_l,  static_cast<unsigned int>(batch), static_cast<unsigned int>(qo_heads), 1U,  static_cast<unsigned int>(seq_len)};
+        bwd_d_global  bwd_d_arg {d_d,  static_cast<unsigned int>(batch), static_cast<unsigned int>(qo_heads), 1U,  static_cast<unsigned int>(seq_len)};

        bwd_global_args bwd_global{bwd_q_arg, 
                        bwd_k_arg, 
@@ -1055,14 +1042,14 @@ block_sparse_attention_backward(torch::Tensor q,
                        bwd_vg_arg, 
                        bwd_l_arg, 
                        bwd_d_arg, 
-                        static_cast<int>(kv_seq_len), // N is not used in the kernel
+                        static_cast<int>(seq_len), 
                        static_cast<int>(hr),
                        static_cast<int>(max_q_blocks_per_kv),
                        reinterpret_cast<int32_t*>(k2q_block_sparse_index.data_ptr()),
                        reinterpret_cast<int32_t*>(k2q_block_sparse_num.data_ptr()),
-                        reinterpret_cast<int32_t*>(kv_block_size.data_ptr())};
+                        reinterpret_cast<int32_t*>(block_size.data_ptr())};

-        dim3 grid_bwd_2(kv_seq_len/BLOCK_N, qo_heads, batch);
+        dim3 grid_bwd_2(seq_len/64, qo_heads, batch);
        threads = 128;

        //cudadevicesynchronize();
@@ -1101,9 +1088,9 @@ block_sparse_attention_backward(torch::Tensor q,

        using bwd_prep_globals = bwd_prep_globals<128>;

-        og_global prep_og_arg{d_og, static_cast<unsigned int>(batch), static_cast<unsigned int>(qo_heads), static_cast<unsigned int>(q_seq_len), 128U};
-        o_global  prep_o_arg {d_o,  static_cast<unsigned int>(batch), static_cast<unsigned int>(qo_heads), static_cast<unsigned int>(q_seq_len), 128U};
-        d_global  prep_d_arg {d_d,  static_cast<unsigned int>(batch), static_cast<unsigned int>(qo_heads), 1U,  static_cast<unsigned int>(q_seq_len)};
+        og_global prep_og_arg{d_og, static_cast<unsigned int>(batch), static_cast<unsigned int>(qo_heads), static_cast<unsigned int>(seq_len), 128U};
+        o_global  prep_o_arg {d_o,  static_cast<unsigned int>(batch), static_cast<unsigned int>(qo_heads), static_cast<unsigned int>(seq_len), 128U};
+        d_global  prep_d_arg {d_d,  static_cast<unsigned int>(batch), static_cast<unsigned int>(qo_heads), 1U,  static_cast<unsigned int>(seq_len)};

        bwd_prep_globals bwd_g{prep_og_arg, prep_o_arg, prep_d_arg};

@@ -1140,15 +1127,15 @@ block_sparse_attention_backward(torch::Tensor q,

        using bwd_global_args = bwd_globals<128>;

-        bwd_q_global  bwd_q_arg {d_q,  static_cast<unsigned int>(batch), static_cast<unsigned int>(qo_heads), static_cast<unsigned int>(q_seq_len), 128U};
-        bwd_k_global  bwd_k_arg {d_k,  static_cast<unsigned int>(batch), static_cast<unsigned int>(kv_heads), static_cast<unsigned int>(kv_seq_len), 128U};
-        bwd_v_global  bwd_v_arg {d_v,  static_cast<unsigned int>(batch), static_cast<unsigned int>(kv_heads), static_cast<unsigned int>(kv_seq_len), 128U};
-        bwd_og_global bwd_og_arg{d_og, static_cast<unsigned int>(batch), static_cast<unsigned int>(qo_heads), static_cast<unsigned int>(q_seq_len), 128U};
-        bwd_qg_global bwd_qg_arg{d_qg, static_cast<unsigned int>(batch), static_cast<unsigned int>(qo_heads), static_cast<unsigned int>(q_seq_len), 128U};
-        bwd_kg_global bwd_kg_arg{d_kg, static_cast<unsigned int>(batch), static_cast<unsigned int>(kv_heads), static_cast<unsigned int>(kv_seq_len), 128U};
-        bwd_vg_global bwd_vg_arg{d_vg, static_cast<unsigned int>(batch), static_cast<unsigned int>(kv_heads), static_cast<unsigned int>(kv_seq_len), 128U};
-        bwd_l_global  bwd_l_arg {d_l,  static_cast<unsigned int>(batch), static_cast<unsigned int>(qo_heads), 1U,  static_cast<unsigned int>(q_seq_len)};
-        bwd_d_global  bwd_d_arg {d_d,  static_cast<unsigned int>(batch), static_cast<unsigned int>(qo_heads), 1U,  static_cast<unsigned int>(q_seq_len)};
+        bwd_q_global  bwd_q_arg {d_q,  static_cast<unsigned int>(batch), static_cast<unsigned int>(qo_heads), static_cast<unsigned int>(seq_len), 128U};
+        bwd_k_global  bwd_k_arg {d_k,  static_cast<unsigned int>(batch), static_cast<unsigned int>(kv_heads), static_cast<unsigned int>(seq_len), 128U};
+        bwd_v_global  bwd_v_arg {d_v,  static_cast<unsigned int>(batch), static_cast<unsigned int>(kv_heads), static_cast<unsigned int>(seq_len), 128U};
+        bwd_og_global bwd_og_arg{d_og, static_cast<unsigned int>(batch), static_cast<unsigned int>(qo_heads), static_cast<unsigned int>(seq_len), 128U};
+        bwd_qg_global bwd_qg_arg{d_qg, static_cast<unsigned int>(batch), static_cast<unsigned int>(qo_heads), static_cast<unsigned int>(seq_len), 128U};
+        bwd_kg_global bwd_kg_arg{d_kg, static_cast<unsigned int>(batch), static_cast<unsigned int>(kv_heads), static_cast<unsigned int>(seq_len), 128U};
+        bwd_vg_global bwd_vg_arg{d_vg, static_cast<unsigned int>(batch), static_cast<unsigned int>(kv_heads), static_cast<unsigned int>(seq_len), 128U};
+        bwd_l_global  bwd_l_arg {d_l,  static_cast<unsigned int>(batch), static_cast<unsigned int>(qo_heads), 1U,  static_cast<unsigned int>(seq_len)};
+        bwd_d_global  bwd_d_arg {d_d,  static_cast<unsigned int>(batch), static_cast<unsigned int>(qo_heads), 1U,  static_cast<unsigned int>(seq_len)};

        bwd_global_args bwd_global{bwd_q_arg, 
                        bwd_k_arg, 
@@ -1159,14 +1146,14 @@ block_sparse_attention_backward(torch::Tensor q,
                        bwd_vg_arg, 
                        bwd_l_arg, 
                        bwd_d_arg, 
-                        static_cast<int>(kv_seq_len), // N is not used in the kernel
+                        static_cast<int>(seq_len), 
                        static_cast<int>(hr),
                        static_cast<int>(max_q_blocks_per_kv),
                        reinterpret_cast<int32_t*>(k2q_block_sparse_index.data_ptr()),
                        reinterpret_cast<int32_t*>(k2q_block_sparse_num.data_ptr()),
-                        reinterpret_cast<int32_t*>(kv_block_size.data_ptr())};
+                        reinterpret_cast<int32_t*>(block_size.data_ptr())};

-        dim3 grid_bwd_2(kv_seq_len/BLOCK_N, qo_heads, batch);
+        dim3 grid_bwd_2(seq_len/64, qo_heads, batch);
        threads = 128;

        //cudadevicesynchronize();
@@ -1,4 +1,4 @@
-FROM nvidia/cuda:12.8.0-cudnn-devel-ubuntu22.04
+FROM nvidia/cuda:12.8.0-devel-ubuntu22.04

 ENV DEBIAN_FRONTEND=noninteractive

@@ -43,7 +43,7 @@ RUN source $HOME/.local/bin/env && \
    source /opt/venv/bin/activate && \
    uv pip install --no-cache-dir --upgrade pip && \
    uv pip install --no-cache-dir .[dev] && \
-    uv pip install --no-cache-dir https://github.com/mjun0812/flash-attention-prebuild-wheels/releases/download/v0.5.4/flash_attn-2.8.3%2Bcu128torch2.9-cp310-cp310-linux_x86_64.whl
+    uv pip install --no-cache-dir flash-attn==2.8.3 --no-build-isolation

 COPY . .

@@ -1,4 +1,4 @@
-FROM nvidia/cuda:12.8.0-cudnn-devel-ubuntu22.04
+FROM nvidia/cuda:12.8.0-devel-ubuntu22.04

 ENV DEBIAN_FRONTEND=noninteractive

@@ -43,7 +43,7 @@ RUN source $HOME/.local/bin/env && \
    source /opt/venv/bin/activate && \
    uv pip install --no-cache-dir --upgrade pip && \
    uv pip install --no-cache-dir .[dev] && \
-    uv pip install --no-cache-dir https://github.com/mjun0812/flash-attention-prebuild-wheels/releases/download/v0.5.4/flash_attn-2.8.3%2Bcu128torch2.9-cp311-cp311-linux_x86_64.whl
+    uv pip install --no-cache-dir flash-attn==2.8.3 --no-build-isolation

 COPY . .

@@ -43,7 +43,7 @@ RUN source $HOME/.local/bin/env && \
    source /opt/venv/bin/activate && \
    uv pip install --no-cache-dir --upgrade pip && \
    uv pip install --no-cache-dir .[dev] && \
-    uv pip install --no-cache-dir https://github.com/mjun0812/flash-attention-prebuild-wheels/releases/download/v0.5.4/flash_attn-2.8.3%2Bcu128torch2.9-cp312-cp312-linux_x86_64.whl
+    uv pip install --no-cache-dir flash-attn==2.8.3 --no-build-isolation

 COPY . .

@@ -0,0 +1,26 @@
+# Minimal makefile for Sphinx documentation
+#
+
+# You can set these variables from the command line, and also
+# from the environment for the first two.
+SPHINXOPTS    ?=
+SPHINXBUILD   ?= sphinx-build
+SOURCEDIR     = source
+BUILDDIR      = build
+
+# Put it first so that "make" without argument is like "make help".
+help:
+	@$(SPHINXBUILD) -M help "$(SOURCEDIR)" "$(BUILDDIR)" $(SPHINXOPTS) $(O)
+
+.PHONY: help Makefile
+
+# Catch-all target: route all unknown targets to Sphinx using the new
+# "make mode" option.  $(O) is meant as a shortcut for $(SPHINXOPTS).
+%: Makefile
+	@$(SPHINXBUILD) -M $@ "$(SOURCEDIR)" "$(BUILDDIR)" $(SPHINXOPTS) $(O)
+
+clean:
+	@$(SPHINXBUILD) -M clean "$(SOURCEDIR)" "$(BUILDDIR)" $(SPHINXOPTS) $(O)
+	rm -rf "$(SOURCEDIR)/getting_started/examples"
+	rm -rf "$(SOURCEDIR)/inference/examples"
+	rm -rf "$(SOURCEDIR)/training/examples"
@@ -1,39 +1,20 @@
-# FastVideo Documentation
+# FastVideo documents

-This directory contains the FastVideo documentation built with MkDocs.
-
-## Build the docs locally
+## Build the docs

 ```bash
-# Install dependencies
-pip install -r docs/requirements-mkdocs.txt
+# Install dependencies.
+pip install -r requirements-docs.txt

-# Serve docs with live reload (recommended for development)
-mkdocs serve
-
-# Or build static site
-mkdocs build
+# Build the docs.
+make clean
+make html
 ```

-## View the docs
-
-### Development server (with live reload)
+## Open the docs with your browser

 ```bash
-mkdocs serve
+python -m http.server -d build/html/
 ```

-Then open your browser to: http://127.0.0.1:8000
-
-### Static build
-
-```bash
-mkdocs build
-python -m http.server -d site/
-```
-
-Then open your browser to: http://localhost:8000
-
-## Automatic Deployment
-
-Documentation is automatically built and deployed to GitHub Pages when changes are pushed to the `main` branch via the `.github/workflows/docs.yml` workflow.
+Launch your browser and open localhost:8000.
@@ -1,248 +0,0 @@
-# FastVideo API Reference
-
-This page contains the complete API reference for the FastVideo library.
-
-## fastvideo
-
-### Modules
-
-| Name | Description |
-|------|-------------|
-| [attention](#fastvideoattention) | Attention mechanisms and backends for video generation |
-| [configs](#fastvideoconfigs) | Configuration classes for pipelines, models, and sampling |
-| [distributed](#fastvideodistributed) | Distributed execution and communication utilities |
-| [entrypoints](#fastvideoentrypoints) | Main API entry points for video generation |
-| [models](#fastvideomodels) | Model implementations (transformers, VAEs, schedulers) |
-| [pipelines](#fastvideopipelines) | Core pipeline classes for video diffusion |
-| [training](#fastvideotraining) | Training utilities and helpers |
-| [workflow](#fastvideoworkflow) | Workflow management and orchestration |
-| [dataset](#fastvideodataset) | Dataset handling and preprocessing |
-| [layers](#fastvideolayers) | Custom neural network layers |
-| [platforms](#fastvideoplatforms) | Platform-specific implementations |
-| [utils](#fastvideoutils) | Utility functions and helpers |
-| [worker](#fastvideoworker) | Execution workers for video generation |
-
-## fastvideo.attention
-
-::: fastvideo.attention
-    options:
-      show_source: true
-      show_root_heading: true
-      show_root_toc_entry: true
-      show_submodules: true
-      heading_level: 3
-
-## fastvideo.configs
-
-::: fastvideo.configs
-    options:
-      show_source: true
-      show_root_heading: true
-      show_root_toc_entry: true
-      show_submodules: true
-      heading_level: 3
-
-### Submodules
-
-#### fastvideo.configs.pipelines
-
-::: fastvideo.configs.pipelines
-    options:
-      show_source: true
-      show_root_heading: true
-      show_root_toc_entry: true
-      heading_level: 4
-
-#### fastvideo.configs.models
-
-::: fastvideo.configs.models
-    options:
-      show_source: true
-      show_root_heading: true
-      show_root_toc_entry: true
-      heading_level: 4
-
-#### fastvideo.configs.sample
-
-::: fastvideo.configs.sample
-    options:
-      show_source: true
-      show_root_heading: true
-      show_root_toc_entry: true
-      heading_level: 4
-
-## fastvideo.distributed
-
-::: fastvideo.distributed
-    options:
-      show_source: true
-      show_root_heading: true
-      show_root_toc_entry: true
-      show_submodules: true
-      heading_level: 3
-
-## fastvideo.entrypoints
-
-::: fastvideo.entrypoints
-    options:
-      show_source: true
-      show_root_heading: true
-      show_root_toc_entry: true
-      show_submodules: true
-      heading_level: 3
-
-## fastvideo.models
-
-::: fastvideo.models
-    options:
-      show_source: true
-      show_root_heading: true
-      show_root_toc_entry: true
-      show_submodules: true
-      heading_level: 3
-
-### Submodules
-
-#### fastvideo.models.registry
-
-::: fastvideo.models.registry
-    options:
-      show_source: true
-      show_root_heading: true
-      show_root_toc_entry: true
-      heading_level: 4
-
-#### fastvideo.models.loader
-
-::: fastvideo.models.loader
-    options:
-      show_source: true
-      show_root_heading: true
-      show_root_toc_entry: true
-      heading_level: 4
-
-## fastvideo.pipelines
-
-::: fastvideo.pipelines
-    options:
-      show_source: true
-      show_root_heading: true
-      show_root_toc_entry: true
-      show_submodules: true
-      heading_level: 3
-
-### Submodules
-
-#### fastvideo.pipelines.composed_pipeline_base
-
-::: fastvideo.pipelines.composed_pipeline_base
-    options:
-      show_source: true
-      show_root_heading: true
-      show_root_toc_entry: true
-      heading_level: 4
-
-#### fastvideo.pipelines.lora_pipeline
-
-::: fastvideo.pipelines.lora_pipeline
-    options:
-      show_source: true
-      show_root_heading: true
-      show_root_toc_entry: true
-      heading_level: 4
-
-#### fastvideo.pipelines.pipeline_batch_info
-
-::: fastvideo.pipelines.pipeline_batch_info
-    options:
-      show_source: true
-      show_root_heading: true
-      show_root_toc_entry: true
-      heading_level: 4
-
-#### fastvideo.pipelines.pipeline_registry
-
-::: fastvideo.pipelines.pipeline_registry
-    options:
-      show_source: true
-      show_root_heading: true
-      show_root_toc_entry: true
-      heading_level: 4
-
-#### fastvideo.pipelines.stages
-
-::: fastvideo.pipelines.stages
-    options:
-      show_source: true
-      show_root_heading: true
-      show_root_toc_entry: true
-      heading_level: 4
-
-## fastvideo.training
-
-::: fastvideo.training
-    options:
-      show_source: true
-      show_root_heading: true
-      show_root_toc_entry: true
-      show_submodules: true
-      heading_level: 3
-
-## fastvideo.workflow
-
-::: fastvideo.workflow
-    options:
-      show_source: true
-      show_root_heading: true
-      show_root_toc_entry: true
-      show_submodules: true
-      heading_level: 3
-
-## fastvideo.dataset
-
-::: fastvideo.dataset
-    options:
-      show_source: true
-      show_root_heading: true
-      show_root_toc_entry: true
-      show_submodules: true
-      heading_level: 3
-
-## fastvideo.layers
-
-::: fastvideo.layers
-    options:
-      show_source: true
-      show_root_heading: true
-      show_root_toc_entry: true
-      show_submodules: true
-      heading_level: 3
-
-## fastvideo.platforms
-
-::: fastvideo.platforms
-    options:
-      show_source: true
-      show_root_heading: true
-      show_root_toc_entry: true
-      show_submodules: true
-      heading_level: 3
-
-## fastvideo.utils
-
-::: fastvideo.utils
-    options:
-      show_source: true
-      show_root_heading: true
-      show_root_toc_entry: true
-      heading_level: 3
-
-## fastvideo.worker
-
-::: fastvideo.worker
-    options:
-      show_source: true
-      show_root_heading: true
-      show_root_toc_entry: true
-      show_submodules: true
-      heading_level: 3
@@ -1,27 +0,0 @@
-# API Summary
-
-This page provides a quick overview of the main FastVideo API components.
-
-## Video Generator
-
-::: fastvideo.VideoGenerator
-    options:
-      show_root_heading: false
-      show_source: false
-      heading_level: 3
-
-## Initialization Configuration
-
-::: fastvideo.PipelineConfig
-    options:
-      show_root_heading: false
-      show_source: false
-      heading_level: 3
-
-## Sampling Configuration
-
-::: fastvideo.SamplingParam
-    options:
-      show_root_heading: false
-      show_source: false
-      heading_level: 3
@@ -1,41 +0,0 @@
-.vertical-table-header th.head:not(.stub) {
-    writing-mode: sideways-lr;
-    white-space: nowrap;
-    max-width: 0;
-    p {
-       margin: 0;
-    }
-}
-
-/* Image sizing classes */
-.image-small {
-    max-width: 200px;
-    height: auto;
-}
-
-.image-medium {
-    max-width: 400px;
-    height: auto;
-}
-
-.image-large {
-    max-width: 600px;
-    height: auto;
-}
-
-.image-full {
-    max-width: 100%;
-    height: auto;
-}
-
-/* Responsive images */
-img {
-    max-width: 100%;
-    height: auto;
-}
-
-/* Center images */
-.image-center {
-    display: block;
-    margin: 0 auto;
-}
@@ -1,6 +0,0 @@
-<svg width="160" height="93" viewBox="0 0 160 93" fill="none" xmlns="http://www.w3.org/2000/svg">
-<path d="M28.8511 91.66L57.6319 1.86368H64.5394L35.7585 91.66H28.8511Z" fill="#356CFF" stroke="#356CFF" stroke-width="2.30244"/>
-<path d="M15.0376 91.66L43.8185 1.86368H46.1209L17.3401 91.66H15.0376Z" fill="#356CFF" stroke="#356CFF" stroke-width="2.30244"/>
-<path d="M1.22217 91.66L30.003 1.86366H31.1543L2.3734 91.66H1.22217Z" fill="#356CFF" stroke="#356CFF" stroke-width="1.15122"/>
-<path d="M71.4465 1.86483L42.666 91.6599H69.144L78.3538 58.2746H123.251L129.007 39.855H84.1099L89.866 22.5868H152.032L157.788 1.86483H71.4465Z" fill="#356CFF" stroke="#356CFF" stroke-width="2.30244"/>
-</svg>
@@ -1,18 +0,0 @@
-<svg width="252" height="105" viewBox="0 0 252 105" fill="none" xmlns="http://www.w3.org/2000/svg">
-<path d="M89.4843 55.5457H101.361L87.7028 101H74.638L89.4843 55.5457Z" fill="#356CFF"/>
-<path fill-rule="evenodd" clip-rule="evenodd" d="M96.0167 1.00057H112.645L118.583 48.273H104.924L103.737 39.7882H79.9827L67.5117 55.5457H85.3273L43.1638 101H28.3174L22.3789 55.5457H33.6621L38.4129 91.3031L58.604 68.2729H44.3515L96.0167 1.00057ZM100.768 13.1217L87.7028 29.4852H103.143L100.768 13.1217Z" fill="#356CFF"/>
-<path d="M37.2252 1.00057L22.3789 48.273H36.0375L40.7884 30.6974L62.6727 30.6974L69.6727 21.0005L43.7576 21.0004L46.7269 11.9096L77.6727 11.9096L86 1.00057L37.2252 1.00057Z" fill="#356CFF"/>
-<path fill-rule="evenodd" clip-rule="evenodd" d="M108.488 55.5457L94.2351 101C94.2351 101 105.518 101 120.959 101C136.399 101 144.078 93.0133 148.276 79.788C152.432 68.0157 153.027 55.5457 136.399 55.5457C119.771 55.5457 108.488 55.5457 108.488 55.5457ZM109.081 90.697L116.802 65.8487C116.802 65.8487 120.959 65.8487 132.242 65.8487C143.525 65.8487 137.586 78.5759 135.211 84.0304C133.307 88.4021 127.491 90.697 122.74 90.697C117.989 90.697 109.081 90.697 109.081 90.697Z" fill="#356CFF"/>
-<path d="M173.188 1.00056L168.625 11.9096C168.625 11.9096 149.386 11.9092 142.525 11.9095C135.664 11.9098 136.586 20.3944 141.337 20.3944H159.747C168.654 20.3944 166.961 33.6899 163.904 38.5761C160.467 44.0675 157.371 48.273 148.463 48.273L125.188 48.273L124 37.97L147.87 37.97C153.808 37.97 156.184 29.4852 151.433 29.4852H131.836C120.142 29.4852 125.897 1.00043 141.337 1.00043L173.188 1.00056Z" fill="#356CFF"/>
-<path d="M179.938 1.00056L175.688 11.9096L191.221 11.9096L179.938 48.273H192.409L203.692 11.9096L219.132 11.9095L223.289 1.00043L179.938 1.00056Z" fill="#356CFF"/>
-<path d="M161.341 55.5457H202.845L198.5 65.8487H169.654L167.279 73.7268H188.5L184.749 82.8177H164.31L161.934 90.697H190.251L186.624 101H146.494L161.341 55.5457Z" fill="#356CFF"/>
-<path fill-rule="evenodd" clip-rule="evenodd" d="M230.821 54.9391C255.169 54.9391 251.776 67.0602 249.231 77.9692C246.686 88.8783 240.917 101 217.757 101C194.596 101 195.606 88.8783 199.347 77.9692C203.089 67.0602 206.473 54.9391 230.821 54.9391ZM237.948 77.9692C239.984 70.6965 240.917 65.242 228.446 65.242C215.975 65.242 211.818 71.9087 210.037 77.9692C208.255 84.0298 208.255 91.3025 219.538 91.3025C230.821 91.3025 235.911 85.2419 237.948 77.9692Z" fill="#356CFF"/>
-<path d="M173.188 1.00056L168.625 11.9096C168.625 11.9096 149.386 11.9092 142.525 11.9095C135.664 11.9098 136.586 20.3944 141.337 20.3944M173.188 1.00056C173.188 1.00056 156.777 1.00043 141.337 1.00043M173.188 1.00056L141.337 1.00043M141.337 20.3944C146.088 20.3944 150.839 20.3944 159.747 20.3944M141.337 20.3944H159.747M159.747 20.3944C168.654 20.3944 166.961 33.6899 163.904 38.5761C160.467 44.0675 157.371 48.273 148.463 48.273M148.463 48.273C139.556 48.273 125.188 48.273 125.188 48.273M148.463 48.273L125.188 48.273M125.188 48.273L124 37.97M124 37.97C124 37.97 141.931 37.97 147.87 37.97M124 37.97L147.87 37.97M147.87 37.97C153.808 37.97 156.184 29.4852 151.433 29.4852M151.433 29.4852C146.682 29.4852 138.962 29.4852 131.836 29.4852M151.433 29.4852H131.836M131.836 29.4852C120.142 29.4852 125.897 1.00043 141.337 1.00043M37.2252 1.00057L22.3789 48.273H36.0375L40.7884 30.6974L62.6727 30.6974L69.6727 21.0005L43.7576 21.0004L46.7269 11.9096L77.6727 11.9096L86 1.00057L37.2252 1.00057ZM96.0167 1.00057H112.645L118.583 48.273H104.924L103.737 39.7882H79.9827L67.5117 55.5457H85.3273L43.1638 101H28.3174L22.3789 55.5457H33.6621L38.4129 91.3031L58.604 68.2729H44.3515L96.0167 1.00057ZM87.7028 29.4852L100.768 13.1217L103.143 29.4852H87.7028ZM89.4843 55.5457H101.361L87.7028 101H74.638L89.4843 55.5457ZM108.488 55.5457L94.2351 101C94.2351 101 105.518 101 120.959 101C136.399 101 144.078 93.0133 148.276 79.788C152.432 68.0157 153.027 55.5457 136.399 55.5457C119.771 55.5457 108.488 55.5457 108.488 55.5457ZM116.802 65.8487L109.081 90.697C109.081 90.697 117.989 90.697 122.74 90.697C127.491 90.697 133.307 88.4021 135.211 84.0304C137.586 78.5759 143.525 65.8487 132.242 65.8487C120.959 65.8487 116.802 65.8487 116.802 65.8487ZM179.938 1.00056L175.688 11.9096L191.221 11.9096L179.938 48.273H192.409L203.692 11.9096L219.132 11.9095L223.289 1.00043L179.938 1.00056ZM161.341 55.5457H202.845L198.5 65.8487H169.654L167.279 73.7268H188.5L184.749 82.8177H164.31L161.934 90.697H190.251L186.624 101H146.494L161.341 55.5457ZM230.821 54.9391C255.169 54.9391 251.776 67.0602 249.231 77.9692C246.686 88.8783 240.917 101 217.757 101C194.596 101 195.606 88.8783 199.347 77.9692C203.089 67.0602 206.473 54.9391 230.821 54.9391ZM228.446 65.242C240.917 65.242 239.984 70.6965 237.948 77.9692C235.911 85.2419 230.821 91.3025 219.538 91.3025C208.255 91.3025 208.255 84.0298 210.037 77.9692C211.818 71.9087 215.975 65.242 228.446 65.242Z" stroke="#356CFF" stroke-width="1.18771"/>
-<path d="M15.2524 55.5451L21.191 100.999L24.7541 100.999L18.8156 55.5451L15.2524 55.5451Z" fill="#356CFF" stroke="#356CFF" stroke-width="1.18771"/>
-<path d="M8.12646 55.5451L14.065 100.999L15.2527 100.999L9.31417 55.5451L8.12646 55.5451Z" fill="#356CFF" stroke="#356CFF" stroke-width="1.18771"/>
-<path d="M1 55.5451L6.93853 100.999L7.53239 100.999L1.59385 55.5451L1 55.5451Z" fill="#356CFF" stroke="#356CFF" stroke-width="0.593853"/>
-<path d="M15.2524 48.2724L30.0988 1H33.6619L18.8156 48.2724H15.2524Z" fill="#356CFF" stroke="#356CFF" stroke-width="1.18771"/>
-<path d="M8.12646 48.2724L22.9728 1H24.1605L9.31417 48.2724H8.12646Z" fill="#356CFF" stroke="#356CFF" stroke-width="1.18771"/>
-<path d="M1 48.2724L15.8463 1H16.4402L1.59385 48.2724H1Z" fill="#356CFF" stroke="#356CFF" stroke-width="0.593853"/>
-<path d="M85.3271 55.5457H67.5116L87 12.7363L44.3513 68.2729H58.6038L43.1636 101L85.3271 55.5457Z" fill="#FDC717" stroke="#FDC717" stroke-width="1.18771" stroke-miterlimit="16"/>
-</svg>
@@ -1,129 +0,0 @@
-# Testing in FastVideo
-
-This guide explains how to add and run tests in FastVideo. The testing suite is divided into several categories to ensure correctness across components, training workflows, and inference quality.
-
-## Test Types
-
-* **Unit Tests**: Located in `fastvideo/tests/dataset`, `fastvideo/tests/entrypoints`, and `fastvideo/tests/workflow`. These test individual functions and classes.
-* **Component Tests**: Located in `fastvideo/tests/encoders`, `fastvideo/tests/transformers`, and `fastvideo/tests/vaes`. These verify the loading and basic functionality of model components.
-* **SSIM Tests**: Located in `fastvideo/tests/ssim`. These are regression tests that compare generated videos against reference videos using the Structural Similarity Index Measure (SSIM) to detect quality degradation.
-* **Training Tests**: Located in `fastvideo/tests/training`. These validate training loops, loss calculations, and specific training techniques like LoRA, Distillation, and VSA.
-* **Inference Tests**: Located in `fastvideo/tests/inference`. These test specialized inference pipelines and optimizations (e.g., STA, V-MoBA).
-
-For now, we will focus on **SSIM Tests**.
-
-## SSIM Tests
-
-SSIM tests are located in `fastvideo/tests/ssim`. These tests generate videos using specific models and parameters, and compare them against reference videos to ensure that changes in the codebase do not degrade generation quality or alter the output unexpectedly.
-
-!!! note
-    If you are adding an SSIM test, this serves as a safeguard. Any future code changes that break or cause errors with the specific arguments and configurations you defined will trigger a failure. Therefore, it is important to include multiple settings and arguments that cover the core features of your new pipeline to ensure robust regression testing.
-
-### Directory Structure
-
-```
-fastvideo/tests/ssim/
-├── <GPU>_reference_videos/   # Reference videos organized by GPU type (e.g., L40S_reference_videos)
-│   ├── <Model_Name>/
-│   │   ├── <Backend>/        # e.g., FLASH_ATTN, TORCH_SDPA
-│   │   │   └── <Video_File>
-├── test_causal_similarity.py
-├── test_inference_similarity.py
-├── update_reference_videos.sh
-└── ...
-```
-
-### Adding a New SSIM Test
-
-To add a new SSIM test, follow these steps:
-
-1. **Create or Update a Test File**: You can add a new test function to an existing file (like `test_inference_similarity.py`) or create a new one if testing a distinct category of models.
-
-2. **Define Model Parameters**: Define the configuration for the model you want to test. This includes model path, dimensions, inference steps, and other generation parameters. **Note:** Consider using lower `num_inference_steps` or reduced resolution (e.g., 480p instead of 720p) to keep test execution time reasonable, provided it doesn't compromise the test's ability to detect regression.
-
-   ```python
-   MY_MODEL_PARAMS = {
-       "num_gpus": 1,
-       "model_path": "organization/model-name",
-       "height": 480,
-       "width": 832,
-       "num_frames": 45,
-       "num_inference_steps": 20,
-       # ... other parameters
-   }
-   ```
-
-3. **Implement the Test Function**:
-   * Use `pytest.mark.parametrize` to run the test with different prompts, backends, and models.
-   * Set the attention backend environment variable.
-   * Initialize the `VideoGenerator`.
-   * Generate the video.
-   * Compare the generated video with the reference video using `compute_video_ssim_torchvision`.
-
-   Example structure:
-
-   ```python
-   @pytest.mark.parametrize("prompt", TEST_PROMPTS)
-   @pytest.mark.parametrize("ATTENTION_BACKEND", ["FLASH_ATTN"])
-   def test_my_model_similarity(prompt, ATTENTION_BACKEND):
-       # Setup output directories
-       # ...
-
-       # Initialize Generator
-       generator = VideoGenerator.from_pretrained(...)
-       generator.generate_video(prompt, ...)
-
-       # Compare with Reference
-       ssim_values = compute_video_ssim_torchvision(reference_path, generated_path, use_ms_ssim=True)
-        assert ssim_values[0] >= 0.98  # Threshold
-    ```
-
-4. **Reference Videos**:
-   * When running the test for the first time (or when updating the reference), the test will fail because the reference video is missing. The generated video will be saved in `fastvideo/tests/ssim/generated_videos`.
-   * Inspect the generated video to ensure it meets quality expectations.
-   * Move the generated video to the appropriate reference folder: `fastvideo/tests/ssim/<GPU>_reference_videos/<Model>/<Backend>/`.
-   * You can use the helper script `update_reference_videos.sh` to automate copying videos from `generated_videos` to `L40S_reference_videos`. Note: Check the script to ensure paths match your environment (it defaults to `L40S_reference_videos`).
-
-### Running Tests Locally
-
-To run the SSIM tests locally:
-
-```bash
-pytest fastvideo/tests/ssim/ -vs
-```
-
-Ensure you have the necessary GPUs available as defined in your test parameters.
-
-## Modal Workflow
-
-FastVideo uses [Modal](https://modal.com/) for running tests in a CI environment. The workflow scripts are located in `fastvideo/tests/modal/`.
-
-### `pr_test.py`
-
-The main entry point for CI tests is `fastvideo/tests/modal/pr_test.py`. This script defines Modal functions that execute the pytest suites on specific hardware (e.g., L40S, H100).
-
-### Updating Modal Configuration
-
-If you add a new test that requires:
-* **Different GPU Hardware**: You may need to change the `@app.function(gpu=...)` decorator.
-* **Longer Execution Time**: Increase the `timeout` parameter.
-* **New Environment Variables/Secrets**: Add them to `secrets=[...]` or the image environment. For example, if your model is gated on Hugging Face, ensure `HF_API_KEY` is passed.
-
-For SSIM tests, the `run_ssim_tests` function in `pr_test.py` currently runs:
-
-```python
-@app.function(gpu="L40S:2", image=image, timeout=2700, secrets=[modal.Secret.from_dict({"HF_API_KEY": os.environ.get("HF_API_KEY", "")})])
-def run_ssim_tests():
-    run_test("hf auth login --token $HF_API_KEY && pytest ./fastvideo/tests/ssim -vs")
-```
-
-If your new test file is inside `fastvideo/tests/ssim`, it will automatically be picked up by this command. However, ensure that the `gpu="L40S:2"` configuration is sufficient for your model. If your model requires more GPUs (e.g., 4 or 8), you might need to create a separate Modal function or update the existing one.
-
-### Workflow Scripts
-
-The shell script that triggers these tests in the CI pipeline is located at `.buildkite/scripts/pr_test.sh`. If you add a new test category (e.g., a new folder outside of `ssim`), you will need to:
-1. Add a new function in `fastvideo/tests/modal/pr_test.py`.
-2. Add a new case in `.buildkite/scripts/pr_test.sh` to handle the new test type.
-
-!!! note
-    If you are a maintainer, you'll need to finally manually update the workflow script in Buildkite. Otherwise, a maintainer will help you update.
@@ -1,12 +0,0 @@
-# 💡 Examples
-
-A collection of examples demonstrating usage of FastVideo.
-
-All documented examples are autogenerated using [generate_examples.py](https://github.com/hao-ai-lab/FastVideo/blob/main/docs/generate_examples.py) from examples found in the [examples](https://github.com/hao-ai-lab/FastVideo/tree/main/examples) directory.
-
-## Examples
-
- [Examples Distillation Index](distillation/examples/examples_distillation_index.md)
- [Examples Training Index](training/examples/examples_training_index.md)
- [Examples Inference Index](inference/examples/examples_inference_index.md)
-
@@ -1,41 +0,0 @@
-
-# 🔧 Installation
-
-FastVideo supports the following hardware platforms:
-
- [NVIDIA CUDA](installation/gpu.md)
- [Apple silicon](installation/mps.md)
-
-## Quick Installation
-
-### Using pip
-
-```bash
-pip install fastvideo
-```
-
-### Using conda
-
-```bash
-conda install -c conda-forge fastvideo
-```
-
-### From source
-
-```bash
-git clone https://github.com/hao-ai-lab/FastVideo.git
-cd FastVideo
-pip install -e .
-```
-
-## Hardware Requirements
-
- **NVIDIA GPUs**: CUDA 11.8+ with compute capability 7.0+
- **Apple Silicon**: macOS 12.0+ with M1/M2/M3 chips
- **CPU**: x86_64 architecture (for CPU-only inference)
-
-## Next Steps
-
- [Quick Start Guide](quick_start.md) - Get started with your first video generation
- [Configuration](../inference/configuration.md) - Learn about configuration options
- [Examples](../inference/examples/) - Explore example scripts and notebooks
@@ -1,55 +0,0 @@
-# 🚀 Quick Start
-
-Get up and running with FastVideo in minutes!
-
-## Installation
-
-First, install FastVideo:
-
-```bash
-pip install fastvideo
-```
-
-## Basic Usage
-
-### Text-to-Video Generation
-
-```python
-from fastvideo import FastVideoPipeline
-
-# Initialize the pipeline
-pipe = FastVideoPipeline.from_pretrained("wan2.1-t2v-1.3B")
-
-# Generate a video
-prompt = "A cat playing with a ball of yarn"
-video = pipe(prompt, num_frames=16, height=512, width=512)
-
-# Save the video
-video.save("output.mp4")
-```
-
-### Image-to-Video Generation
-
-```python
-from fastvideo import FastVideoPipeline
-from PIL import Image
-
-# Load an image
-image = Image.open("input.jpg")
-
-# Initialize the pipeline
-pipe = FastVideoPipeline.from_pretrained("wan2.1-i2v-14B-480p")
-
-# Generate a video from the image
-video = pipe(image, num_frames=16, height=480, width=480)
-
-# Save the video
-video.save("output.mp4")
-```
-
-## Next Steps
-
- [Installation Guide](installation.md) - Detailed installation instructions
- [Configuration](../inference/configuration.md) - Learn about configuration options
- [Examples](../inference/examples/) - Explore more examples
- [Optimizations](../inference/optimizations.md) - Performance optimization tips
@@ -1,42 +0,0 @@
-# V1 API
-
-FastVideo's V1 API provides a streamlined interface for video generation tasks with powerful customization options. This page documents the primary components of the API.
-
-## Video Generator
-
-This class will be the primary Python API for generating videos and images.
-
-::: fastvideo.entrypoints.video_generator.VideoGenerator
-    options:
-      show_root_heading: true
-      show_source: false
-      members:
-        - from_pretrained
-      heading_level: 3
-
-`VideoGenerator.from_pretrained()` should be the primary way of creating a new video generator.
-
-## Configuring FastVideo
-
-The following two classes `PipelineConfig` and `SamplingParam` are used to configure initialization and sampling parameters, respectively.
-
-### PipelineConfig
-
-::: fastvideo.configs.pipelines.base.PipelineConfig
-    options:
-      show_root_heading: true
-      show_source: false
-      members:
-        - from_pretrained
-        - dump_to_json
-      heading_level: 4
-
-### SamplingParam
-
-::: fastvideo.configs.sample.base.SamplingParam
-    options:
-      show_root_heading: true
-      show_source: false
-      members:
-        - from_pretrained
-      heading_level: 4
@@ -1,66 +0,0 @@
-# Compatibility Matrix
-
-The table below shows every supported model and optimizations supported for them.
-
-The symbols used have the following meanings:
-
- ✅ = Full compatibility
- ❌ = No compatibility
- ⭕ = Does not apply to this model
-
-## Models x Optimization
-
-The `HuggingFace Model ID` can be directly pass to `from_pretrained()` methods and FastVideo will use the optimal default parameters when initializing and generating videos.
-
-<style>
-  /* Target tables in this section */
-  #models-x-optimization + p + table {
-    display: block;
-    overflow-x: auto;
-    width: 100%;
-    font-size: 0.85rem;
-  }
-  
-  #models-x-optimization + p + table td,
-  #models-x-optimization + p + table th {
-    text-align: center;
-    white-space: nowrap;
-    padding: 0.5em;
-  }
-  
-  /* First two columns can wrap */
-  #models-x-optimization + p + table td:nth-child(1),
-  #models-x-optimization + p + table td:nth-child(2) {
-    white-space: normal;
-    min-width: 120px;
-  }
-  
-  #models-x-optimization + p + table td:nth-child(2) code {
-    font-size: 0.75rem;
-  }
-</style>
-
-| Model Name | HuggingFace Model ID | Resolutions | TeaCache | Sliding Tile Attn | Sage Attn | VSA |
-|------------|---------------------|-------------|----------|-------------------|-----------|-----|
-| FastWan2.1 T2V 1.3B | `FastVideo/FastWan2.1-T2V-1.3B-Diffusers` | 480P | ⭕ | ⭕ | ⭕ | ✅ |
-| FastWan2.2 TI2V 5B Full Attn* | `FastVideo/FastWan2.2-TI2V-5B-FullAttn-Diffusers` | 720P | ⭕ | ⭕ | ⭕ | ✅ |
-| Wan2.2 TI2V 5B | `Wan-AI/Wan2.2-TI2V-5B-Diffusers` | 720P | ⭕ | ⭕ | ✅ | ⭕ |
-| Wan2.2 T2V A14B | `Wan-AI/Wan2.2-T2V-A14B-Diffusers` | 480P<br>720P | ❌ | ❌ | ✅ | ⭕ |
-| Wan2.2 I2V A14B | `Wan-AI/Wan2.2-I2V-A14B-Diffusers` | 480P<br>720P | ❌ | ❌ | ✅ | ⭕ |
-| HunyuanVideo | `hunyuanvideo-community/HunyuanVideo` | 720px1280p<br>544px960p | ❌ | ✅ | ✅ | ⭕ |
-| FastHunyuan | `FastVideo/FastHunyuan-diffusers` | 720px1280p<br>544px960p | ❌ | ✅ | ✅ | ⭕ |
-| Wan2.1 T2V 1.3B | `Wan-AI/Wan2.1-T2V-1.3B-Diffusers` | 480P | ✅ | ✅* | ✅ | ⭕ |
-| Wan2.1 T2V 14B | `Wan-AI/Wan2.1-T2V-14B-Diffusers` | 480P, 720P | ✅ | ✅* | ✅ | ⭕ |
-| Wan2.1 I2V 480P | `Wan-AI/Wan2.1-I2V-14B-480P-Diffusers` | 480P | ✅ | ✅* | ✅ | ⭕ |
-| Wan2.1 I2V 720P | `Wan-AI/Wan2.1-I2V-14B-720P-Diffusers` | 720P | ✅ | ✅ | ✅ | ⭕ |
-| StepVideo T2V | `FastVideo/stepvideo-t2v-diffusers` | 768px768px204f<br>544px992px204f<br>544px992px136f | ❌ | ❌ | ✅ | ⭕ |
-
-**Note**: Wan2.2 TI2V 5B has some quality issues when performing I2V generation. We are working on fixing this issue.
-
-## Special requirements
-
-### StepVideo T2V
- The self-attention in text-encoder (step_llm) only supports CUDA capabilities sm_80 sm_86 and sm_90
-
-### Sliding Tile Attention
- Currently only Hopper GPUs (H100s) are supported.
@@ -0,0 +1,35 @@
+@ECHO OFF
+
+pushd %~dp0
+
+REM Command file for Sphinx documentation
+
+if "%SPHINXBUILD%" == "" (
+	set SPHINXBUILD=sphinx-build
+)
+set SOURCEDIR=source
+set BUILDDIR=build
+
+%SPHINXBUILD% >NUL 2>NUL
+if errorlevel 9009 (
+	echo.
+	echo.The 'sphinx-build' command was not found. Make sure you have Sphinx
+	echo.installed, then set the SPHINXBUILD environment variable to point
+	echo.to the full path of the 'sphinx-build' executable. Alternatively you
+	echo.may add the Sphinx directory to PATH.
+	echo.
+	echo.If you don't have Sphinx installed, grab it from
+	echo.https://www.sphinx-doc.org/
+	exit /b 1
+)
+
+if "%1" == "" goto help
+
+%SPHINXBUILD% -M %1 %SOURCEDIR% %BUILDDIR% %SPHINXOPTS% %O%
+goto end
+
+:help
+%SPHINXBUILD% -M help %SOURCEDIR% %BUILDDIR% %SPHINXOPTS% %O%
+
+:end
+popd
@@ -0,0 +1,15 @@
+sphinx==7.4.7
+sphinx-argparse==0.5.2
+sphinx-autodoc2==0.5.0
+sphinx-book-theme==1.1.4
+sphinx-copybutton==0.5.2
+sphinx-design==0.6.1
+sphinx-togglebutton==0.3.2
+myst-parser==3.0.1
+msgspec
+commonmark # Required by sphinx-argparse when using :markdownhelp:
+
+# packages to install to build the documentation
+cachetools
+# -f https://download.pytorch.org/whl/cpu
+torch
@@ -0,0 +1,51 @@
+# Seed Parameter Behavior in vLLM
+
+## Overview
+
+The `seed` parameter in vLLM is used to control the random states for various random number generators. This parameter can affect the behavior of random operations in user code, especially when working with models in vLLM.
+
+## Default Behavior
+
+By default, the `seed` parameter is set to `None`. When the `seed` parameter is `None`, the global random states for `random`, `np.random`, and `torch.manual_seed` are not set. This means that the random operations will behave as expected, without any fixed random states.
+
+## Specifying a Seed
+
+If a specific seed value is provided, the global random states for `random`, `np.random`, and `torch.manual_seed` will be set accordingly. This can be useful for reproducibility, as it ensures that the random operations produce the same results across multiple runs.
+
+## Example Usage
+
+### Without Specifying a Seed
+
+```python
+import random
+from vllm import LLM
+
+# Initialize a vLLM model without specifying a seed
+model = LLM(model="Qwen/Qwen2.5-0.5B-Instruct")
+
+# Try generating random numbers
+print(random.randint(0, 100))  # Outputs different numbers across runs
+```
+
+### Specifying a Seed
+
+```python
+import random
+from vllm import LLM
+
+# Initialize a vLLM model with a specific seed
+model = LLM(model="Qwen/Qwen2.5-0.5B-Instruct", seed=42)
+
+# Try generating random numbers
+print(random.randint(0, 100))  # Outputs the same number across runs
+```
+
+## Important Notes
+
+- If the `seed` parameter is not specified, the behavior of global random states remains unaffected.
+- If a specific seed value is provided, the global random states for `random`, `np.random`, and `torch.manual_seed` will be set to that value.
+- This behavior can be useful for reproducibility but may lead to non-intuitive behavior if the user is not explicitly aware of it.
+
+## Conclusion
+
+Understanding the behavior of the `seed` parameter in vLLM is crucial for ensuring the expected behavior of random operations in your code. By default, the `seed` parameter is set to `None`, which means that the global random states are not affected. However, specifying a seed value can help achieve reproducibility in your experiments.
@@ -0,0 +1,8 @@
+.vertical-table-header th.head:not(.stub) {
+    writing-mode: sideways-lr;
+    white-space: nowrap;
+    max-width: 0;
+    p {
+       margin: 0;
+    }
+}
@@ -0,0 +1,39 @@
+<style>
+  .notification-bar {
+    width: 100vw;
+    display: flex;
+    justify-content: center;
+    align-items: center;
+    font-size: 16px;
+    padding: 0 6px 0 6px;
+  }
+  .notification-bar p {
+    margin: 0;
+  }
+  .notification-bar a {
+    font-weight: bold;
+    text-decoration: none;
+  }
+
+  /* Light mode styles (default) */
+  .notification-bar {
+    background-color: #fff3cd;
+    color: #856404;
+  }
+  .notification-bar a {
+    color: #d97706;
+  }
+
+  /* Dark mode styles */
+  html[data-theme=dark] .notification-bar {
+    background-color: #333;
+    color: #ddd;
+  }
+  html[data-theme=dark] .notification-bar a {
+    color: #ffa500; /* Brighter color for visibility */
+  }
+</style>
+
+<!-- <div class="notification-bar">
+  <p>You are viewing the latest developer preview docs. <a href="https://docs.vllm.ai/en/stable/">Click here</a> to view docs for the latest stable release.</p>
+</div> -->
@@ -0,0 +1,19 @@
+# Summary
+
+## Video Generator
+
+```{autodoc2-summary}
+    fastvideo.VideoGenerator
+```
+
+## Initialization Configuration
+
+```{autodoc2-summary}
+    fastvideo.configs.pipelines.PipelineConfig
+```
+
+## Sampling Configuration
+
+```{autodoc2-summary}
+    fastvideo.configs.sample.SamplingParam
+```
@@ -0,0 +1,22 @@
+# type: ignore
+# SPDX-License-Identifier: Apache-2.0
+from docutils import nodes
+from myst_parser.parsers.sphinx_ import MystParser
+from sphinx.ext.napoleon import docstring
+
+
+class NapoleonParser(MystParser):
+
+    def parse(self, input_string: str, document: nodes.document) -> None:
+        # Get the Sphinx configuration
+        config = document.settings.env.config
+
+        parsed_content = str(
+            docstring.GoogleDocstring(
+                str(docstring.NumpyDocstring(input_string, config)),
+                config,
+            ))
+        return super().parse(parsed_content, document)
+
+
+Parser = NapoleonParser
@@ -0,0 +1,275 @@
+# SPDX-License-Identifier: Apache-2.0
+
+# Configuration file for the Sphinx documentation builder.
+#
+# This file only contains a selection of the most common options. For a full
+# list see the documentation:
+# https://www.sphinx-doc.org/en/master/usage/configuration.html
+
+# -- Path setup --------------------------------------------------------------
+
+# If extensions (or modules to document with autodoc) are in another directory,
+# add these directories to sys.path here. If the directory is relative to the
+# documentation root, use os.path.abspath to make it absolute, like shown here.
+
+import datetime
+import logging
+import os
+import re
+import sys
+from pathlib import Path
+
+import requests
+
+logger = logging.getLogger(__name__)
+REPO_ROOT = Path(__file__).resolve().parent.parent.parent
+print(os.path.abspath(REPO_ROOT))
+sys.path.append(os.path.abspath(REPO_ROOT))
+
+# -- Project information -----------------------------------------------------
+
+project = 'FastVideo'
+copyright = f'{datetime.datetime.now().year}, FastVideo Team'
+author = 'the FastVideo Team'
+
+# -- General configuration ---------------------------------------------------
+
+# Add any Sphinx extension module names here, as strings. They can be
+# extensions coming with Sphinx (named 'sphinx.ext.*') or your custom
+# ones.
+extensions = [
+    "sphinx.ext.napoleon",
+    "sphinx.ext.linkcode",
+    "sphinx.ext.intersphinx",
+    "sphinx_copybutton",
+    "autodoc2",
+    "myst_parser",
+    "sphinxarg.ext",
+    "sphinx_design",
+    "sphinx_togglebutton",
+]
+myst_enable_extensions = [
+    "colon_fence",
+    "fieldlist",
+]
+autodoc2_packages = [
+    {
+        "path": "../../fastvideo",
+        "exclude_dirs": ["__pycache__", "third_party"],
+    },
+]
+autodoc2_output_dir = "api"
+autodoc2_render_plugin = "myst"
+autodoc2_hidden_objects = ["dunder", "private", "inherited"]
+autodoc2_docstring_parser_regexes = [
+    (".*", "docs.source.autodoc2_docstring_parser"),
+]
+autodoc2_sort_names = True
+autodoc2_index_template = None
+autodoc2_skip_module_regexes = [
+    "fastvideo.dataset",
+    "fastvideo.distill",
+    "fastvideo.data_preprocess",
+    "fastvideo.models",
+    "fastvideo.sample",
+    "fastvideo.utils",
+    "fastvideo.distill_adv",
+    "fastvideo.train",
+]
+
+# Add any paths that contain templates here, relative to this directory.
+templates_path = ['_templates']
+
+# List of patterns, relative to source directory, that match files and
+# directories to ignore when looking for source files.
+# This pattern also affects html_static_path and html_extra_path.
+exclude_patterns: list[str] = ["**/*.template.md", "**/*.inc.md"]
+
+# Exclude the prompt "$" when copying code
+copybutton_prompt_text = r"\$ "
+copybutton_prompt_is_regexp = True
+
+# -- Options for HTML output -------------------------------------------------
+
+# The theme to use for HTML and HTML Help pages.  See the documentation for
+# a list of builtin themes.
+#
+html_title = project
+html_theme = 'sphinx_book_theme'
+html_logo = '../../assets/logos/icon_simple.svg'
+html_theme_options = {
+    'path_to_docs': 'docs/source',
+    'repository_url': 'https://github.com/hao-ai-lab/FastVideo/',
+    'use_repository_button': True,
+    'use_edit_page_button': True,
+    # Prevents the full API being added to the left sidebar of every page.
+    # Reduces build time by 2.5x and reduces build size from ~225MB to ~95MB.
+    'collapse_navbar': True,
+    # Makes API visible in the right sidebar on API reference pages.
+    'show_toc_level': 3,
+}
+# Add any paths that contain custom static files (such as style sheets) here,
+# relative to this directory. They are copied after the builtin static files,
+# so a file named "default.css" will overwrite the builtin "default.css".
+html_static_path = ["_static"]
+html_js_files = ["custom.js"]
+html_css_files = ["custom.css"]
+
+myst_url_schemes = {
+    'http': None,
+    'https': None,
+    'mailto': None,
+    'ftp': None,
+    "gh-issue": {
+        "url":
+        "https://github.com/hao-ai-lab/FastVideo/issues/{{path}}#{{fragment}}",
+        "title": "Issue #{{path}}",
+        "classes": ["github"],
+    },
+    "gh-pr": {
+        "url":
+        "https://github.com/hao-ai-lab/FastVideo/pull/{{path}}#{{fragment}}",
+        "title": "Pull Request #{{path}}",
+        "classes": ["github"],
+    },
+    "gh-dir": {
+        "url": "https://github.com/hao-ai-lab/FastVideo/tree/main/{{path}}",
+        "title": "{{path}}",
+        "classes": ["github"],
+    },
+    "gh-file": {
+        "url": "https://github.com/hao-ai-lab/FastVideo/blob/main/{{path}}",
+        "title": "{{path}}",
+        "classes": ["github"],
+    },
+}
+
+# see https://docs.readthedocs.io/en/stable/reference/environment-variables.html # noqa
+READTHEDOCS_VERSION_TYPE = os.environ.get('READTHEDOCS_VERSION_TYPE')
+if READTHEDOCS_VERSION_TYPE == "tag":
+    # remove the warning banner if the version is a tagged release
+    header_file = os.path.join(os.path.dirname(__file__),
+                               "_templates/sections/header.html")
+    # The file might be removed already if the build is triggered multiple times
+    # (readthedocs build both HTML and PDF versions separately)
+    if os.path.exists(header_file):
+        os.remove(header_file)
+
+
+# Generate additional rst documentation here.
+def setup(app):
+    from docs.source.generate_examples import generate_examples
+    generate_examples()
+
+
+_cached_base: str = ""
+_cached_branch: str = ""
+
+
+def get_repo_base_and_branch(pr_number: str) -> tuple[str | None, str | None]:
+    global _cached_base, _cached_branch
+    if _cached_base and _cached_branch:
+        return _cached_base, _cached_branch
+
+    url = f"https://api.github.com/repos/hao-ai-lab/FastVideo/pulls/{pr_number}"
+    response = requests.get(url)
+    if response.status_code == 200:
+        data = response.json()
+        _cached_base = data['head']['repo']['full_name']
+        _cached_branch = data['head']['ref']
+        return _cached_base, _cached_branch
+    else:
+        logger.error("Failed to fetch PR details: %s", response)
+        return None, None
+
+
+def linkcode_resolve(domain, info):
+    if domain != 'py':
+        return None
+    if not info['module']:
+        return None
+
+    # Get path from module name
+    file = Path(f"{info['module'].replace('.', '/')}.py")
+    path = REPO_ROOT / file
+    if not path.exists():
+        path = REPO_ROOT / file.with_suffix("") / "__init__.py"
+    if not path.exists():
+        return None
+
+    # Get the line number of the object
+    with open(path) as f:
+        lines = f.readlines()
+    name = info['fullname'].split(".")[-1]
+    pattern = fr"^( {{4}})*((def|class) )?{name}\b.*"
+    for lineno, line in enumerate(lines, 1):
+        if not line or line.startswith("#"):
+            continue
+        if re.match(pattern, line):
+            break
+
+    # If the line number is not found, return None
+    if lineno == len(lines):
+        return None
+
+    # If the line number is found, create the URL
+    filename = path.relative_to(REPO_ROOT)
+    if "checkouts" in path.parts:
+        # a PR build on readthedocs
+        pr_number = REPO_ROOT.name
+        base, branch = get_repo_base_and_branch(pr_number)
+        if base and branch:
+            return f"https://github.com/{base}/blob/{branch}/{filename}#L{lineno}"
+    # Otherwise, link to the source file on the main branch
+    return f"https://github.com/hao-ai-lab/FastVideo/blob/main/{filename}#L{lineno}"
+
+
+# Mock out external dependencies here, otherwise the autodoc pages may be blank.
+autodoc_mock_imports = [
+    "blake3",
+    "compressed_tensors",
+    "cpuinfo",
+    "cv2",
+    "torch",
+    "huggingface_hub",
+    "torchvision",
+    "transformers",
+    "psutil",
+    "prometheus_client",
+    "sentencepiece",
+    "vllm._C",
+    "PIL",
+    "numpy",
+    'triton',
+    "tqdm",
+    "tensorizer",
+    "pynvml",
+    "outlines",
+    "xgrammar",
+    "librosa",
+    "soundfile",
+    "gguf",
+    "lark",
+    "decord",
+]
+
+for mock_target in autodoc_mock_imports:
+    if mock_target in sys.modules:
+        logger.info(
+            "Potentially problematic mock target (%s) found; "
+            "autodoc_mock_imports cannot mock modules that have already "
+            "been loaded into sys.modules when the sphinx build starts.",
+            mock_target)
+
+intersphinx_mapping = {
+    "python": ("https://docs.python.org/3", None),
+    "typing_extensions":
+    ("https://typing-extensions.readthedocs.io/en/latest", None),
+    "aiohttp": ("https://docs.aiohttp.org/en/stable", None),
+    "pillow": ("https://pillow.readthedocs.io/en/stable", None),
+    "numpy": ("https://numpy.org/doc/stable", None),
+    "torch": ("https://pytorch.org/docs/stable", None),
+    "psutil": ("https://psutil.readthedocs.io/en/stable", None),
+}
+
+navigation_with_keys = False
@@ -1,4 +1,4 @@
-
+(docker)=
 # 🐳 Using the FastVideo Docker Image

 If you prefer a containerized development environment or want to avoid managing dependencies manually, you can use our prebuilt Docker image:
@@ -3,3 +3,11 @@
 # 🧰 Developer Environment

 Accelerate your FastVideo development workflow by leveraging Docker images and cloud GPUs for efficient experimentation and reproducible environments.
+
+:::{toctree}
+:caption: Contents
+:maxdepth: 1
+
+docker
+runpod
+:::
@@ -1,3 +1,4 @@
+(runpod)=

 # 📦 Developing FastVideo on RunPod

@@ -9,7 +10,7 @@ Choose a GPU that supports CUDA 12.8

 Pick 1 or 2 L40S GPU(s)

-![RunPod CUDA selection](../../assets/images/runpod_cuda.png)
+![RunPod CUDA selection](../../_static/images/runpod_cuda.png)

 When creating your pod template, use this image:

@@ -23,11 +24,11 @@ Paste Container Start Command to support SSH ([RunPod Docs](https://docs.runpod.
 bash -c "apt update;DEBIAN_FRONTEND=noninteractive apt-get install openssh-server -y;mkdir -p ~/.ssh;cd $_;chmod 700 ~/.ssh;echo \"$PUBLIC_KEY\" >> authorized_keys;chmod 700 authorized_keys;service ssh start;sleep infinity"
 ```

-![RunPod template configuration](../../assets/images/runpod_template.png)
+![RunPod template configuration](../../_static/images/runpod_template.png)

 After deploying, the pod will take a few minutes to pull the image and start the SSH service.

-![RunPod ssh](../../assets/images/runpod_ssh.png)
+![RunPod ssh](../../_static/images/runpod_ssh.png)

 ## Working with the pod

@@ -1,3 +1,4 @@
+(developer-overview)=

 # 🛠️ Contributing to FastVideo

@@ -70,7 +71,3 @@ uv pip install ninja

 python setup.py install
 ```
-
-## Testing
-
-Please refer to the [Testing Guide](testing.md) for more information on how to add and run tests in FastVideo.
@@ -29,6 +29,7 @@ FastVideo separates model components from execution logic with these principles:
 - **Custom Attention Backends**: Components can support and use different Attention implementations
 - **Pipeline Abstraction**: Consistent interface across diffusion models

+(design-fastvideo-args)=
 ## FastVideoArgs

 The `FastVideoArgs` class in `fastvideo/fastvideo_args.py` serves as the central configuration system for FastVideo. It contains all parameters needed to control model loading, inference configuration, performance optimization settings, and more.
@@ -60,6 +61,7 @@ with set_current_fastvideo_args(fastvideo_args):
    result = generate_video()
 ```

+(design-pipeline-system)=
 ## Pipeline System

 ### `ComposedPipelineBase`
@@ -106,8 +108,7 @@ def forward(self, batch: ForwardBatch, fastvideo_args: FastVideoArgs) -> Forward
    return batch
 ```

-![Pipeline execution and data flow](../assets/images/pipeline.png)
-
+(design-forwardbatch)=
 ### ForwardBatch

 Defined in `fastvideo/pipelines/pipeline_batch_info.py`, `ForwardBatch` encapsulates the data payload passed between pipeline stages. It typically holds:
@@ -119,10 +120,12 @@ Defined in `fastvideo/pipelines/pipeline_batch_info.py`, `ForwardBatch` encapsul

 This structure facilitates clear state transitions between stages.

+(design-model-components)=
 ## Model Components

 The `fastvideo/models/` directory contains implementations of the core neural network models used in video diffusion:

+(design-transformer-models)=
 ### Transformer Models

 Transformer networks perform the actual denoising during diffusion:
@@ -149,6 +152,7 @@ def forward(
    return noise_pred  # Predicted noise residual
 ```

+(design-vae-variational-auto-encoder)=
 ### VAE (Variational Auto-Encoder)

 VAEs handle conversion between pixel space and latent space:
@@ -166,6 +170,7 @@ FastVideo's VAE implementations include:
 - Optional tiling for large frames
 - Distributed weight support

+(design-text-and-image-encoders)=
 ### Text and Image Encoders

 Encoders process conditioning inputs into embeddings:
@@ -183,6 +188,7 @@ FastVideo implements optimizations such as:
 - Caching for common prompts
 - Precision-tuned computation

+(design-schedulers)=
 ### Schedulers

 Schedulers manage the diffusion sampling process:
@@ -210,10 +216,7 @@ def step(
    return prev_sample
 ```

-This diagram shows how models are discovered, validated, and loaded across entrypoints, executors, pipelines, and model loaders.
-
-![Model loading flow](../assets/images/load_models.png)
-
+(design-optimized-attention)=
 ## Optimized Attention

 The `fastvideo/attention/` directory contains optimized attention implementations crucial for efficient video diffusion:
@@ -237,17 +240,17 @@ self.attn = LocalAttention(
 # export FASTVIDEO_ATTENTION_BACKEND=FLASH_ATTN
 ```

-![Attention backend selector design](../assets/images/attention_backend.png)
-
 ### Attention Patterns
 Supports various patterns with memory optimization techniques:
 - **Cross/Self/Temporal/Global-Local Attention**
 - Chunking, progressive computation, optimized masking

+(design-distributed-processing)=
 ## Distributed Processing

 The `fastvideo/distributed/` directory contains implementations for distributed model execution:

+(design-tensor-parallelism)=
 ### Tensor Parallelism

 Tensor parallelism splits model weights across devices:
@@ -304,6 +307,7 @@ Efficient communication primitives minimize distributed overhead:
 - **Tensor-Parallel AllReduce**: Combines partial results
 - **Distributed Synchronization**: Coordinates execution

+(design-forwardcontext)=
 ## Forward Context Management

 ### ForwardContext
@@ -326,6 +330,7 @@ with set_forward_context(current_timestep, attn_metadata, fastvideo_args):
    output = model(inputs)
 ```

+(design-executor-and-worker-abstractions)=
 ## Executor and Worker System

 The `fastvideo/worker/` directory contains the distributed execution framework:
@@ -352,6 +357,7 @@ Each GPU worker:

 This design allows FastVideo to efficiently utilize multiple GPUs while providing a simple, unified interface for model execution.

+(design-platforms)=
 ## Platforms

 The `fastvideo/platforms/` directory provides hardware platform abstractions that enable FastVideo to run efficiently on different hardware configurations:
@@ -382,6 +388,7 @@ else:

 The platform system is designed to be extensible for future hardware targets.

+(design-logger)=
 ## Logger
 See [PR](https://github.com/hao-ai-lab/FastVideo/pull/356)

@@ -1,3 +1,4 @@
+(v0-data-preprocess)=

 # 🧱 Data Preprocess for Distillation

@@ -13,7 +13,7 @@ We provide two distilled models:
 Both models are trained on **61×448×832** resolution but support generating videos with **any resolution** (1.3B  model mainly support 480P, 14B model support 480P and 720P, quality may degrade for different resolutions).

 ## ⚙️ Inference
-First install [VSA](https://hao-ai-lab.github.io/FastVideo/video_sparse_attention/installation). Set `MODEL_BASE` to your own model path and run:
+First install [VSA](https://hao-ai-lab.github.io/FastVideo/video_sparse_attention/installation.html). Set `MODEL_BASE` to your own model path and run:

 ```bash
 bash scripts/inference/v1_inference_wan_dmd.sh
@@ -6,11 +6,10 @@ import re
 from dataclasses import dataclass, field
 from pathlib import Path

-ROOT_DIR = Path(__file__).parent.parent.resolve()
-ROOT_DIR_RELATIVE = '../..'
+ROOT_DIR = Path(__file__).parent.parent.parent.resolve()
+ROOT_DIR_RELATIVE = '../../../..'
 EXAMPLE_DIR = ROOT_DIR / "examples"
-EXAMPLE_DOC_DIR = ROOT_DIR / "docs/getting_started/examples"
-GITHUB_REPO = "hao-ai-lab/FastVideo"  # Update this to your repo
+EXAMPLE_DOC_DIR = ROOT_DIR / "docs/source/getting_started/examples"


 def fix_case(text: str) -> str:
@@ -72,16 +71,9 @@ class Index:

    def generate(self) -> str:
        content = f"# {self.title}\n\n{self.description}\n\n"
-        if self.caption:
-            content += f"## {self.caption}\n\n"
-        # Generate a simple list of links for MkDocs
-        for doc in self.documents:
-            # Convert document path to proper link
-            doc_link = doc.replace("\\", "/")
-            # Get just the filename for the link text
-            doc_title = fix_case(Path(doc).stem.replace("_", " ").title())
-            content += f"- [{doc_title}]({doc_link}.md)\n"
-        content += "\n"
+        content += ":::{toctree}\n"
+        content += f":caption: {self.caption}\n:maxdepth: {self.maxdepth}\n"
+        content += "\n".join(self.documents) + "\n:::\n"
        return content


@@ -150,66 +142,30 @@ class Example:
        return fix_case(self.path.stem.replace("_", " ").title())

    def generate(self) -> str:
-        # Create GitHub link to source
-        github_path = str(self.path.relative_to(ROOT_DIR)).replace("\\", "/")
-        github_url = f"https://github.com/{GITHUB_REPO}/blob/main/{github_path}"
-        content = f"**Source:** [{github_path}]({github_url})\n\n"
+        # Convert the path to a relative path from __file__
+        make_relative = lambda path: ROOT_DIR_RELATIVE / path.relative_to(
+            ROOT_DIR)

-        # Add title for code files
-        if self.main_file.suffix != ".md":
+        content = f"Source <gh-file:{self.path.relative_to(ROOT_DIR)}>.\n\n"
+        include = "include" if self.main_file.suffix == ".md" else \
+            "literalinclude"
+        if include == "literalinclude":
            content += f"# {self.title}\n\n"
-
-        # Include main file content
-        if self.main_file.suffix == ".md":
-            # For markdown files, include the content directly
-            with open(self.main_file, encoding='utf-8') as f:
-                content += f.read() + "\n\n"
-        else:
-            # For code files, use code blocks
-            language = self.main_file.suffix[1:] if self.main_file.suffix else ""
-            with open(self.main_file, encoding='utf-8') as f:
-                file_content = f.read()
-            content += f"```{language}\n{file_content}\n```\n\n"
+        content += f":::{{{include}}} {make_relative(self.main_file)}\n"  # type: ignore[no-untyped-call]
+        if include == "literalinclude":
+            content += f":language: {self.main_file.suffix[1:]}\n"
+        content += ":::\n\n"

        if not self.other_files:
            return content

-        content += "## Additional Files\n\n"
-        # Define binary/non-text file extensions to skip
-        binary_extensions = {
-            '.mp4', '.avi', '.mov', '.mkv', '.gif', '.jpg', '.jpeg', '.png',
-            '.webp', '.bmp', '.pdf', '.zip', '.tar', '.gz', '.mp3', '.wav'
-        }
-
+        content += "## Example materials\n\n"
        for file in sorted(self.other_files):
-            # Skip binary files
-            if file.suffix.lower() in binary_extensions:
-                continue
-
-            file_rel_path = file.relative_to(self.path)
-            # Use collapsible admonition syntax for MkDocs
-            content += f"??? note \"{file_rel_path}\"\n\n"
-
-            try:
-                if file.suffix == ".md":
-                    # Include markdown content with indentation
-                    with open(file, encoding='utf-8') as f:
-                        for line in f:
-                            content += f"    {line}"
-                else:
-                    # Include code with proper formatting
-                    language = file.suffix[1:] if file.suffix else ""
-                    with open(file, encoding='utf-8') as f:
-                        file_content = f.read()
-                    # Indent the code block for the admonition
-                    content += f"    ```{language}\n"
-                    for line in file_content.split('\n'):
-                        content += f"    {line}\n"
-                    content += "    ```\n"
-                content += "\n"
-            except UnicodeDecodeError:
-                # Skip files that can't be decoded as UTF-8
-                continue
+            include = "include" if file.suffix == ".md" else "literalinclude"
+            content += f":::{{admonition}} {file.relative_to(self.path)}\n"
+            content += ":class: dropdown\n\n"
+            content += f":::{{{include}}} {make_relative(file)}\n:::\n"  # type: ignore[no-untyped-call]
+            content += ":::\n\n"

        return content

@@ -239,7 +195,7 @@ class NestedStructure:

 def create_category_indices() -> dict[str, Index]:
    """Create category indices with their respective configurations."""
-    main_index_dir = ROOT_DIR / "docs/examples"
+    main_index_dir = ROOT_DIR / "docs/source/examples"
    if not main_index_dir.exists():
        main_index_dir.mkdir(parents=True)

@@ -247,16 +203,17 @@ def create_category_indices() -> dict[str, Index]:
        "inference":
        Index(
            path=ROOT_DIR /
-            "docs/inference/examples/examples_inference_index.md",
+            "docs/source/inference/examples/examples_inference_index.md",
            title="🚀 Examples",
            description=
-            "Inference examples demonstrate how to use FastVideo inference. We recommend starting with [basic.md](basic.md).",
+            "Inference examples demonstrate how to use FastVideo inference. We recommend starting with <project:basic.md>.",
            caption="Examples",
            maxdepth=1,
        ),
        "training":
        Index(
-            path=ROOT_DIR / "docs/training/examples/examples_training_index.md",
+            path=ROOT_DIR /
+            "docs/source/training/examples/examples_training_index.md",
            title="🚀 Examples",
            description=
            "Training examples demonstrate how to use FastVideo training.",
@@ -266,7 +223,7 @@ def create_category_indices() -> dict[str, Index]:
        "distillation":
        Index(
            path=ROOT_DIR /
-            "docs/distillation/examples/examples_distillation_index.md",
+            "docs/source/distillation/examples/examples_distillation_index.md",
            title="🚀 Examples",
            description=
            "Distillation examples demonstrate how to use FastVideo distillation.",
@@ -289,21 +246,9 @@ def find_examples(category_indices: dict[str, Index],
    examples = []
    glob_patterns = ["*.py", "*.md", "*.sh"]

-    # Map category names to actual directory names
-    category_dir_mapping = {
-        "distillation": "distill",  # examples/distill/ -> distillation category
-    }
-
    # Find categorised examples
    for category in category_indices:
-        # Use mapped directory name if available, otherwise use category name
-        dir_name = category_dir_mapping.get(category, category)
-        category_dir = EXAMPLE_DIR / dir_name
-
-        # Skip if directory doesn't exist
-        if not category_dir.exists():
-            continue
-
+        category_dir = EXAMPLE_DIR / category
        globs = [category_dir.glob(pattern) for pattern in glob_patterns]
        for path in itertools.chain(*globs):
            examples.append(Example(path, category))
@@ -334,18 +279,11 @@ def create_nested_structures(
                                                dict[str,
                                                     NestedStructure]]]] = {}

-    # Map category names to actual directory names
-    category_dir_mapping = {
-        "distillation": "distill",
-    }
-
    for example in examples:
        if example.category not in ["training", "distillation"]:
            continue

-        # Use mapped directory name if available
-        dir_name = category_dir_mapping.get(example.category, example.category)
-        category_dir = EXAMPLE_DIR / dir_name
+        category_dir = EXAMPLE_DIR / example.category
        relative_path = example.path.relative_to(category_dir)
        path_parts = relative_path.parts

@@ -477,7 +415,7 @@ def generate_nested_examples(nested_structures: dict[str, dict[str, dict[
            category_index.documents.append(method)


-def generate_examples(generate_main_index: bool = False) -> None:
+def generate_examples(generate_main_index=False):
    """
    Generate example documentation.
    
@@ -491,14 +429,12 @@ def generate_examples(generate_main_index: bool = False) -> None:
    # Create the main examples index only if requested
    examples_index = None
    if generate_main_index:
-        main_index_dir = ROOT_DIR / "docs/examples"
+        main_index_dir = ROOT_DIR / "docs/source/examples"
        examples_index = Index(
            path=main_index_dir / "examples_index.md",
            title="💡 Examples",
            description=
-            "A collection of examples demonstrating usage of FastVideo.\n\n"
-            f"All documented examples are autogenerated using [generate_examples.py](https://github.com/{GITHUB_REPO}/blob/main/docs/generate_examples.py) "
-            f"from examples found in the [examples](https://github.com/{GITHUB_REPO}/tree/main/examples) directory.",
+            "A collection of examples demonstrating usage of FastVideo.\nAll documented examples are autogenerated using <gh-file:docs/source/generate_examples.py> from examples found in <gh-file:examples>.",
            caption="Examples",
            maxdepth=2)

@@ -535,19 +471,3 @@ def generate_examples(generate_main_index: bool = False) -> None:
    if generate_main_index and examples_index:
        with open(examples_index.path, "w+") as f:
            f.write(examples_index.generate())
-
-
-def on_pre_build_hook(config, **kwargs):
-    """
-    MkDocs hook to generate examples before building the documentation.
-    This function is called automatically by the mkdocs-simple-hooks plugin.
-    """
-    print("Generating example documentation...")
-    generate_examples(generate_main_index=True)
-    print("Example documentation generated successfully!")
-
-
-if __name__ == "__main__":
-    print("Generating example documentation...")
-    generate_examples(generate_main_index=True)
-    print("Example documentation generated successfully!")
@@ -0,0 +1,18 @@
+(installation-index)=
+
+# 🔧 Installation
+
+FastVideo supports the following hardware platforms:
+
+:::{toctree}
+:maxdepth: 1
+:hidden:
+
+installation/gpu
+installation/mps
+:::
+
+- <project:installation/gpu.md>
+  - NVIDIA CUDA
+- <project:installation/mps.md>
+  - Apple silicon
@@ -30,8 +30,17 @@ conda create -n fastvideo python=3.12 -y
 conda activate fastvideo
 ```

+:::{note}
+[PyTorch has deprecated the conda release channel](https://github.com/pytorch/pytorch/issues/138506). If you use `conda`, please only use it to create Python environment rather than installing packages.
+:::
+
 #### uv

+:::{tip}
+We highly recommend using `uv` to install FastVideo. In our experience, `uv` speeds up installation by at least 3x.
+Note that you can also use `uv` to install FastVideo in a Conda environment.
+:::
+
 Or you can create a new Python environment using [uv](https://docs.astral.sh/uv/), a very fast Python environment manager. Please follow the [documentation](https://docs.astral.sh/uv/#getting-started) to install `uv`. After installing `uv`, you can create a new Python environment using the following command:

 ```console
@@ -31,8 +31,17 @@ conda create -n fastvideo python=3.12.4 -y
 conda activate fastvideo
 ```

+:::{note}
+[PyTorch has deprecated the conda release channel](https://github.com/pytorch/pytorch/issues/138506). If you use `conda`, please only use it to create Python environment rather than installing packages.
+:::
+
 #### uv

+:::{tip}
+We highly recommend using `uv` to install FastVideo. In our experience, `uv` speeds up installation by at least 3x.
+Note that you can also use `uv` to install FastVideo in a Conda environment.
+:::
+
 Or you can create a new Python environment using [uv](https://docs.astral.sh/uv/), a very fast Python environment manager. Please follow the [documentation](https://docs.astral.sh/uv/#getting-started) to install `uv`. After installing `uv`, you can create a new Python environment using the following command:

 ```console
@@ -0,0 +1,83 @@
+# V1 API
+
+FastVideo's V1 API provides a streamlined interface for video generation tasks with powerful customization options. This page documents the primary components of the API.
+
+## Video Generator
+
+This class will be the primary Python API for generating videos and images.
+
+```{autodoc2-summary}
+    fastvideo.VideoGenerator
+```
+
+`````{py:class} VideoGenerator(fastvideo_args: fastvideo.fastvideo_args.FastVideoArgs, executor_class: type[fastvideo.worker.executor.Executor], log_stats: bool)
+:canonical: fastvideo.entrypoints.video_generator.VideoGenerator
+
+```{autodoc2-docstring} fastvideo.entrypoints.video_generator.VideoGenerator
+:parser: docs.source.autodoc2_docstring_parser
+```
+
+`VideoGenerator.from_pretrained()` should be the primary way of creating a new video generator.
+
+````{py:method} from_pretrained(model_path: str, device: typing.Optional[str] = None, torch_dtype: typing.Optional[torch.dtype] = None, pipeline_config: typing.Optional[typing.Union[str | fastvideo.configs.pipelines.PipelineConfig]] = None, **kwargs) -> fastvideo.entrypoints.video_generator.VideoGenerator
+:canonical: fastvideo.entrypoints.video_generator.VideoGenerator.from_pretrained
+:classmethod:
+
+```{autodoc2-docstring} fastvideo.entrypoints.video_generator.VideoGenerator.from_pretrained
+:parser: docs.source.autodoc2_docstring_parser
+```
+
+
+## Configuring FastVideo
+
+The follow two classes `PipelineConfig` and `SamplingParam` are used to configure initialization and sampling parameters, respectively.
+
+### PipelineConfig
+```{autodoc2-summary}
+    fastvideo.PipelineConfig
+```
+
+`````{py:class} PipelineConfig
+:canonical: fastvideo.configs.pipelines.base.PipelineConfig
+
+```{autodoc2-docstring} fastvideo.configs.pipelines.base.PipelineConfig
+:parser: docs.source.autodoc2_docstring_parser
+```
+
+````{py:method} from_pretrained(model_path: str) -> fastvideo.configs.pipelines.base.PipelineConfig
+:canonical: fastvideo.configs.pipelines.base.PipelineConfig.from_pretrained
+:classmethod:
+
+```{autodoc2-docstring} fastvideo.configs.pipelines.base.PipelineConfig.from_pretrained
+:parser: docs.source.autodoc2_docstring_parser
+```
+
+
+````{py:method} dump_to_json(file_path: str)
+:canonical: fastvideo.configs.pipelines.base.PipelineConfig.dump_to_json
+
+```{autodoc2-docstring} fastvideo.configs.pipelines.base.PipelineConfig.dump_to_json
+:parser: docs.source.autodoc2_docstring_parser
+```
+
+
+### SamplingParam
+
+```{autodoc2-summary}
+    fastvideo.SamplingParam
+```
+
+`````{py:class} SamplingParam
+:canonical: fastvideo.configs.sample.base.SamplingParam
+
+```{autodoc2-docstring} fastvideo.configs.sample.base.SamplingParam
+:parser: docs.source.autodoc2_docstring_parser
+```
+
+````{py:method} from_pretrained(model_path: str) -> fastvideo.configs.sample.base.SamplingParam
+:canonical: fastvideo.configs.sample.base.SamplingParam.from_pretrained
+:classmethod:
+
+```{autodoc2-docstring} fastvideo.configs.sample.base.SamplingParam.from_pretrained
+:parser: docs.source.autodoc2_docstring_parser
+```
@@ -1,52 +1,132 @@
 # Welcome to FastVideo

-<div style="text-align: center;">
-  <img src="assets/logos/logo.svg" alt="FastVideo" style="width: 60%;" />
-</div>
+:::{figure} ../../assets/logos/logo.svg
+:align: center
+:alt: FastVideo
+:class: no-scaled-link
+:width: 60%
+:::

-<div style="text-align: center;">
-<strong>FastVideo is a unified inference and post-training framework for accelerated video generation.</strong>
-</div>
+:::{raw} html
+<p style="text-align:center">
+<strong>FastVideo is a unified inference and post-training framework for accelerated video generation.
+</strong>
+</p>

-<div style="text-align: center;">
+<p style="text-align:center">
 <script async defer src="https://buttons.github.io/buttons.js"></script>
 <a class="github-button" href="https://github.com/hao-ai-lab/FastVideo/" data-show-count="true" data-size="large" aria-label="Star">Star</a>
 <a class="github-button" href="https://github.com/hao-ai-lab/FastVideo/subscription" data-icon="octicon-eye" data-size="large" aria-label="Watch">Watch</a>
 <a class="github-button" href="https://github.com/hao-ai-lab/FastVideo/fork" data-icon="octicon-repo-forked" data-size="large" aria-label="Fork">Fork</a>
-</div>
+</p>
+:::

 FastVideo is an inference and post-training framework for diffusion models. It features an end-to-end unified pipeline for accelerating diffusion models, starting from data preprocessing to model training, finetuning, distillation, and inference. FastVideo is designed to be modular and extensible, allowing users to easily add new optimizations and techniques. Whether it is training-free optimizations or post-training optimizations, FastVideo has you covered.

 <div style="text-align: center;">
-  <img src="assets/images/fastwan.png" style="width: 100%;"/>
+  <img src=_static/images/fastwan.png width="100%"/>
 </div>

 ## Key Features

 FastVideo has the following features:
-
 - State-of-the-art performance optimizations for inference
  - [Sliding Tile Attention](https://arxiv.org/pdf/2502.04507)
  - [TeaCache](https://arxiv.org/pdf/2411.19108)
  - [Sage Attention](https://arxiv.org/abs/2410.02367)
 - E2E post-training support
-  - Data preprocessing pipeline for video data
+  - Data preprocessing pipeline for video data.
  - [Sparse distillation](https://hao-ai-lab.github.io/blogs/fastvideo_post_training/) for Wan2.1 and Wan2.2 using [Video Sparse Attention](https://arxiv.org/pdf/2505.13389) and [Distribution Matching Distillation](https://tianweiy.github.io/dmd2/)
  - Support full finetuning and LoRA finetuning for state-of-the-art open video DiTs.
  - Scalable training with FSDP2, sequence parallelism, and selective activation checkpointing, with near linear scaling to 64 GPUs.

 ## Documentation

-Welcome to FastVideo! This documentation will help you get started with our unified inference and post-training framework for accelerated video generation.
+% How to start using FastVideo?

-Use the navigation menu on the left to explore different sections:
+:::{toctree}
+:caption: Getting Started
+:maxdepth: 1

- **Getting Started**: Installation and quick start guides
- **Inference**: Learn how to use FastVideo for video generation
- **Training**: Data preprocessing and fine-tuning workflows
- **Distillation**: Post-training optimization techniques
- **Sliding Tile Attention**: Advanced attention mechanisms
- **Video Sparse Attention**: Efficient attention for video models
- **Design**: Framework architecture and design principles
- **Developer Guide**: Contributing and development setup
- **API Reference**: Complete API documentation
+getting_started/installation
+<!-- getting_started/v1_api -->
+:::
+
+:::{toctree}
+:caption: Inference
+:maxdepth: 1
+
+inference/inference_quick_start
+inference/examples/examples_inference_index
+inference/configuration
+inference/optimizations
+inference/comfyui
+inference/support_matrix
+inference/cli
+inference/add_pipeline
+:::
+
+:::{toctree}
+:caption: Training
+:maxdepth: 1
+
+training/examples/examples_training_index
+training/data_preprocess
+<!-- training/finetune -->
+:::
+
+:::{toctree}
+:caption: Distillation
+:maxdepth: 1
+
+distillation/examples/examples_distillation_index
+distillation/data_preprocess
+distillation/dmd
+:::
+
+% What is STA Kernel?
+
+:::{toctree}
+:caption: Sliding Tile Attention
+:maxdepth: 1
+
+sliding_tile_attention/installation
+sliding_tile_attention/demo
+:::
+
+% What is VSA Kernel?
+
+:::{toctree}
+:caption: Video Sparse Attention
+:maxdepth: 1
+
+video_sparse_attention/installation
+:::
+
+:::{toctree}
+:caption: Design
+:maxdepth: 1
+design/overview
+:::
+
+:::{toctree}
+:caption: Developer Guide
+:maxdepth: 2
+
+contributing/overview
+contributing/developer_env/index
+contributing/profiling
+:::
+
+:::{toctree}
+:caption: API Reference
+:maxdepth: 2
+
+<!-- api/summary -->
+api/fastvideo/fastvideo
+:::
+
+## Indices and tables
+
+- {ref}`genindex`
+- {ref}`modindex`
@@ -1,3 +1,4 @@
+(add-pipeline)=

 # 🏗️ Adding a New Pipeline

@@ -1,4 +1,4 @@
-
+(inference-configuration)=
 # Configuration

 ## Multi-GPU Setup
@@ -1,3 +1,4 @@
+(inference-optimizations)=

 # Optimizations

@@ -15,6 +16,8 @@ This page describes the various options for speeding up generation times in Fast
 - Caching Techniques
  - [TeaCache](#optimizations-teacache)

+(optimizations-backends)=
+
 ## Attention Backends

 ### Available Backends
@@ -46,6 +49,8 @@ You can also set the environment variable on the command line:
 FASTVIDEO_ATTENTION_BACKEND=SAGE_ATTN python example.py
 ```

+(optimizations-flash)=
+
 ### Flash Attention

 **`FLASH_ATTN`**
@@ -66,6 +71,12 @@ pip install ninja
 python setup.py install
 ```

+:::{note}
+FastVideo will automatically detect and use `FA3` if it is installed when using `FLASH_ATTN` backend.
+:::
+
+(optimizations-sta)=
+
 ### Sliding Tile Attention

 **`SLIDING_TILE_ATTN`**
@@ -76,6 +87,8 @@ pip install st_attn==0.0.4

 Please see [this page](#sta-installation) for more installation instructions.

+(optimizations-vsa)=
+
 ### Video Sparse Attention

 **`VIDEO_SPARSE_ATTN`**
@@ -87,6 +100,8 @@ python setup_vsa.py install

 Please see [this page](#vsa-installation) for more installation instructions.

+(optimizations-sage)=
+
 ### Sage Attention

 **`SAGE_ATTN`**
@@ -99,6 +114,8 @@ cd sageattention
 python setup.py install  # or pip install -e .
 ```

+(optimizations-sage3)=
+
 ### Sage Attention 3

 **`SAGE_ATTN_THREE`**
@@ -119,6 +136,8 @@ To use Sage Attention 3 in FastVideo, first get access to the SageAttention3 cod
 python setup.py install
 ```

+(optimizations-teacache)=
+
 ## Teacache

 TeaCache is an optimization technique supported in FastVideo that can significantly speed up video generation by skipping redundant calculations across diffusion steps. This guide explains how to enable and configure TeaCache for optimal performance in FastVideo.
@@ -0,0 +1,136 @@
+(support-matrix)=
+# Compatibility Matrix
+The table below shows every supported model and optimizations supported for them.
+
+The symbols used have the following meanings:
+
+- ✅ = Full compatibility
+- ❌ = No compatibility
+- ⭕ = Does not apply to this model
+
+## Models x Optimization
+The `HuggingFace Model ID` can be directly pass to `from_pretrained()` methods and FastVideo will use the optimal default parameters when initializing and generating videos.
+
+:::{raw} html
+<style>
+  /* Make smaller to try to improve readability  */
+  td {
+    font-size: 0.9rem;
+    text-align: center;
+  }
+
+  th {
+    text-align: center;
+    font-size: 0.9rem;
+  }
+</style>
+:::
+
+:::{list-table}
+:header-rows: 1
+:stub-columns: 3
+:widths: auto
+:class: vertical-table-header
+
+- * Model Name
+  * HuggingFace Model ID
+  * Resolutions
+  * TeaCache
+  * Sliding Tile Attn
+  * Sage Attn
+  * Video Sparse Attention (VSA)
+- * FastWan2.1 T2V 1.3B
+  * `FastVideo/FastWan2.1-T2V-1.3B-Diffusers`
+  * 480P
+  * ⭕
+  * ⭕
+  * ⭕
+  * ✅
+- * FastWan2.2 TI2V 5B Full Attn*
+  * `FastVideo/FastWan2.2-TI2V-5B-FullAttn-Diffusers`
+  * 720P
+  * ⭕
+  * ⭕
+  * ⭕
+  * ✅
+- * Wan2.2 TI2V 5B 
+  * `Wan-AI/Wan2.2-TI2V-5B-Diffusers`
+  * 720P
+  * ⭕
+  * ⭕
+  * ✅
+  * ⭕
+- * Wan2.2 T2V A14B
+  * `Wan-AI/Wan2.2-T2V-A14B-Diffusers`
+  * 480P<br>720P
+  * ❌
+  * ❌
+  * ✅
+  * ⭕
+- * Wan2.2 I2V A14B
+  * `Wan-AI/Wan2.2-I2V-A14B-Diffusers`
+  * 480P<br>720P
+  * ❌
+  * ❌
+  * ✅
+  * ⭕
+- * HunyuanVideo
+  * `hunyuanvideo-community/HunyuanVideo`
+  * 720px1280p<br>544px960p
+  * ❌
+  * ✅
+  * ✅
+  * ⭕
+- * FastHunyuan
+  * `FastVideo/FastHunyuan-diffusers`
+  * 720px1280p<br>544px960p
+  * ❌
+  * ✅
+  * ✅
+  * ⭕
+- * Wan2.1 T2V 1.3B
+  * `Wan-AI/Wan2.1-T2V-1.3B-Diffusers`
+  * 480P
+  * ✅
+  * ✅*
+  * ✅
+  * ⭕
+- * Wan2.1 T2V 14B
+  * `Wan-AI/Wan2.1-T2V-14B-Diffusers`
+  * 480P, 720P
+  * ✅
+  * ✅*
+  * ✅
+  * ⭕
+- * Wan2.1 I2V 480P
+  * `Wan-AI/Wan2.1-I2V-14B-480P-Diffusers`
+  * 480P
+  * ✅
+  * ✅*
+  * ✅
+  * ⭕
+- * Wan2.1 I2V 720P
+  * `Wan-AI/Wan2.1-I2V-14B-720P-Diffusers`
+  * 720P
+  * ✅
+  * ✅
+  * ✅
+  * ⭕
+- * StepVideo T2V
+  * `FastVideo/stepvideo-t2v-diffusers`
+  * 768px768px204f<br>544px992px204f<br>544px992px136f
+  * ❌
+  * ❌
+  * ✅
+  * ⭕
+:::
+
+**Note**: Wan2.2 TI2V 5B has some quality issues when performing I2V generation. We are working on fixing this issue.
+
+## Special requirements
+
+### StepVideo T2V
+- The self-attention in text-encoder (step_llm) only supports CUDA capabilities sm_80 sm_86 and sm_90
+
+### Sliding Tile Attention
+- Currently only Hopper GPUs (H100s) are supported.
@@ -1,3 +1,4 @@
+(sta-demo)=

 # 🔍 Demo
 This is is a demo for 2D STA with window size (6,6) operating on a (10, 10) image.
@@ -1,3 +1,4 @@
+(sta-installation)=

 # 🔧 Installation
 You can install the Sliding Tile Attention package using
@@ -1,3 +1,4 @@
+(v0-data-preprocess)=

 # 🧱 Data Preprocess

@@ -1,4 +1,4 @@
-
+(v0-finetune)=
 # 🧠 Finetune
 ## ⚡ Full Finetune
 Ensure your data is prepared and preprocessed in the format specified in [data_preprocess.md](#v0-data-preprocess). For convenience, we also provide a mochi preprocessed Black Myth Wukong data that can be downloaded directly:
@@ -1,3 +1,4 @@
+(vsa-installation)=

 # 🔧 Installation
 You can install the Video Sparse Attention package using
@@ -6,7 +6,7 @@ The `VideoGenerator` class provides the primary Python interface for doing offli
 - Python 3.10-3.12

 ## Installation
-If you have not installed FastVideo, please following these [instructions](https://hao-ai-lab.github.io/FastVideo/getting_started/installation) first.
+If you have not installed FastVideo, please following these [instructions](https://hao-ai-lab.github.io/FastVideo/getting_started/installation.html) first.

 ## Usage
 The first script in this example shows the most basic usage of FastVideo. If you are new to Python and FastVideo, you should start here.
@@ -1,42 +0,0 @@
-from fastvideo import VideoGenerator
-import json
-# from fastvideo.configs.sample import SamplingParam
-
-OUTPUT_PATH = "video_samples_hy15"
-def main():
-    # FastVideo will automatically use the optimal default arguments for the
-    # model.
-    # If a local path is provided, FastVideo will make a best effort
-    # attempt to identify the optimal arguments.
-    generator = VideoGenerator.from_pretrained(
-        "hunyuanvideo-community/HunyuanVideo-1.5-Diffusers-480p_t2v",
-        # FastVideo will automatically handle distributed setup
-        num_gpus=1,
-        use_fsdp_inference=True,
-        dit_cpu_offload=True,
-        vae_cpu_offload=True,
-        text_encoder_cpu_offload=True,
-        pin_cpu_memory=True, # set to false if low CPU RAM or hit obscure "CUDA error: Invalid argument"
-        # image_encoder_cpu_offload=False,
-    )
-
-    prompt = (
-        "A curious raccoon peers through a vibrant field of yellow sunflowers, its eyes "
-        "wide with interest. The playful yet serene atmosphere is complemented by soft "
-        "natural light filtering through the petals. Mid-shot, warm and cheerful tones."
-    )
-
-    video = generator.generate_video(prompt, output_path=OUTPUT_PATH, save_video=True, negative_prompt="", num_frames=81, fps=16)
-
-    prompt2 = (
-        "A majestic lion strides across the golden savanna, its powerful frame "
-        "glistening under the warm afternoon sun. The tall grass ripples gently in "
-        "the breeze, enhancing the lion's commanding presence. The tone is vibrant, "
-        "embodying the raw energy of the wild. Low angle, steady tracking shot, "
-        "cinematic.")
-
-    video2 = generator.generate_video(prompt2, output_path=OUTPUT_PATH, save_video=True, negative_prompt="", num_frames=81, fps=16)
-
-
-if __name__ == "__main__":
-    main()
@@ -1,75 +0,0 @@
-from fastvideo import VideoGenerator
-from fastvideo.configs.pipelines.wan import MatrixGameI2V480PConfig
-from fastvideo.models.dits.matrix_game.utils import create_action_presets
-
-import torch
-
-# Available variants: "base_distilled_model", "gta_distilled_model", "templerun_distilled_model"
-# Each variant has different keyboard_dim:
-#   - base_distilled_model: keyboard_dim=4
-#   - gta_distilled_model: keyboard_dim=2
-#   - templerun_distilled_model: keyboard_dim=7 (keyboard only, no mouse)
-MODEL_VARIANT = "base_distilled_model"
-
-# Variant-specific settings
-VARIANT_CONFIG = {
-    "base_distilled_model": {
-        "model_path": "FastVideo/Matrix-Game-2.0-Base-Diffusers",
-        "keyboard_dim": 4,
-        "image_url": "https://raw.githubusercontent.com/SkyworkAI/Matrix-Game/main/Matrix-Game-2/demo_images/universal/0000.png",
-    },
-    "gta_distilled_model": {
-        "model_path": "FastVideo/Matrix-Game-2.0-GTA-Diffusers",
-        "keyboard_dim": 2,
-        "image_url": "https://raw.githubusercontent.com/SkyworkAI/Matrix-Game/main/Matrix-Game-2/demo_images/gta_drive/0000.png",
-    },
-    "templerun_distilled_model": {
-        "model_path": "FastVideo/Matrix-Game-2.0-TempleRun-Diffusers",
-        "keyboard_dim": 7,
-        "image_url": "https://raw.githubusercontent.com/SkyworkAI/Matrix-Game/main/Matrix-Game-2/demo_images/temple_run/0000.png",
-    },
-}
-
-
-OUTPUT_PATH = "video_samples_matrixgame2"
-def main():
-    # FastVideo will automatically use the optimal default arguments for the
-    # model.
-    # If a local path is provided, FastVideo will make a best effort
-    # attempt to identify the optimal arguments.
-    config = VARIANT_CONFIG[MODEL_VARIANT]
-
-    generator = VideoGenerator.from_pretrained(
-        config["model_path"],
-        # FastVideo will automatically handle distributed setup
-        num_gpus=1,
-        use_fsdp_inference=True,
-        dit_cpu_offload=True, # DiT need to be offloaded for MoE
-        vae_cpu_offload=False,
-        text_encoder_cpu_offload=True,
-        # Set pin_cpu_memory to false if CPU RAM is limited and there're no frequent CPU-GPU transfer
-        pin_cpu_memory=True,
-        # image_encoder_cpu_offload=False,
-    )
-
-    num_frames = 597
-    actions = create_action_presets(num_frames, keyboard_dim=config["keyboard_dim"])
-    grid_sizes = torch.tensor([150, 44, 80])
-
-    generator.generate_video(
-        prompt="",
-        image_path=config["image_url"],
-        mouse_cond=actions["mouse"].unsqueeze(0),
-        keyboard_cond=actions["keyboard"].unsqueeze(0),
-        grid_sizes=grid_sizes,
-        num_frames=num_frames,
-        height=352,
-        width=640,
-        num_inference_steps=50,
-        output_path=OUTPUT_PATH,
-        save_video=True,
-    )
-
-
-if __name__ == "__main__":
-    main()
@@ -1,44 +0,0 @@
-# NOTE: This is still a work in progress, and the checkpoints are not released yet.
-
-from fastvideo import VideoGenerator, SamplingParam
-import json
-# from fastvideo.configs.sample import SamplingParam
-
-OUTPUT_PATH = "video_samples_self_forcing_causal_wan2_2_14B_i2v"
-def main():
-    # FastVideo will automatically use the optimal default arguments for the
-    # model.
-    # If a local path is provided, FastVideo will make a best effort
-    # attempt to identify the optimal arguments.
-    generator = VideoGenerator.from_pretrained(
-        "FastVideo/SFWan2.2-I2V-A14B-Preview-Diffusers",
-        # FastVideo will automatically handle distributed setup
-        num_gpus=1,
-        use_fsdp_inference=True,
-        dit_cpu_offload=True, # DiT need to be offloaded for MoE
-        dit_precision="fp32",
-        vae_cpu_offload=False,
-        text_encoder_cpu_offload=True,
-        dmd_denoising_steps=[1000, 850, 700, 550, 350, 275, 200, 125],
-        # Set pin_cpu_memory to false if CPU RAM is limited and there're no frequent CPU-GPU transfer
-        pin_cpu_memory=True,
-        # image_encoder_cpu_offload=False,
-    )
-
-    sampling_param = SamplingParam.from_pretrained("FastVideo/SFWan2.2-I2V-A14B-Preview-Diffusers")
-    sampling_param.num_frames = 81
-    sampling_param.width = 832
-    sampling_param.height = 480
-    sampling_param.seed = 1000
-
-    with open("prompts/mixkit_i2v.jsonl", "r") as f:
-        prompt_image_pairs = json.load(f)
-
-    for prompt_image_pair in prompt_image_pairs:
-        prompt = prompt_image_pair["prompt"]
-        image_path = prompt_image_pair["image_path"]
-        _ = generator.generate_video(prompt, image_path=image_path, output_path=OUTPUT_PATH, save_video=True, sampling_param=sampling_param)
-
-
-if __name__ == "__main__":
-    main()
@@ -3,9 +3,6 @@ import os
 import requests
 import base64
 import time
-import json
-from pathlib import Path
-import tempfile

 import gradio as gr

@@ -15,7 +12,6 @@ from fastvideo.configs.sample.base import SamplingParam
 MODEL_PATH_MAPPING = {
    "FastWan2.1-T2V-1.3B": "FastVideo/FastWan2.1-T2V-1.3B-Diffusers",
    "FastWan2.2-TI2V-5B-FullAttn": "FastVideo/FastWan2.2-TI2V-5B-FullAttn-Diffusers",
-    "CausalWan2.2-I2V-A14B-Preview": "FastVideo/SFWan2.2-I2V-A14B-Preview-Diffusers",
 }


@@ -41,7 +37,7 @@ class RayServeClient:
                f"{self.backend_url}/generate_video",
                json=request_data,
                headers=headers,
-                timeout=900  # 15 minutes timeout for longer video generation
+                timeout=300
            )
            
            round_trip_time = time.time() - start_time
@@ -85,78 +81,49 @@ def save_video_from_base64(video_data: str, output_dir: str, prompt: str) -> str
        return None


-def encode_image_to_base64(image_path: str) -> str:
-    """Encode an image file to base64 string."""
-    if not image_path or not os.path.exists(image_path):
-        return None
+def create_timing_display(inference_time, encoding_time, network_time, total_time, stage_execution_times, num_frames):
+    dit_denoising_time = f"{stage_execution_times[5]:.2f}s" if len(stage_execution_times) > 5 else "N/A"
    
-    try:
-        with open(image_path, 'rb') as f:
-            image_bytes = f.read()
-        
-        image_base64 = base64.b64encode(image_bytes).decode('utf-8')
-        
-        # Determine image type from extension
-        ext = os.path.splitext(image_path)[1].lower()
-        mime_types = {
-            '.jpg': 'image/jpeg',
-            '.jpeg': 'image/jpeg',
-            '.png': 'image/png',
-            '.gif': 'image/gif',
-            '.webp': 'image/webp',
-        }
-        mime_type = mime_types.get(ext, 'image/jpeg')
-        
-        return f"data:{mime_type};base64,{image_base64}"
-        
-    except Exception as e:
-        print(f"Failed to encode image: {e}")
-        return None
-
-
-# def create_timing_display(inference_time, encoding_time, network_time, total_time, stage_execution_times, num_frames):
-#     dit_denoising_time = f"{stage_execution_times[5]:.2f}s" if len(stage_execution_times) > 5 else "N/A"
-#     
-#     timing_html = f"""
-#     <div style="margin: 10px 0;">
-#         <h3 style="text-align: center; margin-bottom: 10px;">⏱️ Timing Breakdown</h3>
-#         <div style="display: grid; grid-template-columns: repeat(5, 1fr); gap: 10px; margin-bottom: 10px;">
-#             <div class="timing-card timing-card-highlight">
-#                 <div style="font-size: 20px;">🚀</div>
-#                 <div style="font-weight: bold; margin: 3px 0; font-size: 14px;">DiT Denoising</div>
-#                 <div style="font-size: 16px; color: #ffa200; font-weight: bold;">{dit_denoising_time}</div>
-#             </div>
-#             <div class="timing-card">
-#                 <div style="font-size: 20px;">🧠</div>
-#                 <div style="font-weight: bold; margin: 3px 0; font-size: 14px;">E2E (w. vae/text encoder)</div>
-#                 <div style="font-size: 16px; color: #2563eb;">{inference_time:.2f}s</div>
-#             </div>
-#             <div class="timing-card">
-#                 <div style="font-size: 20px;">🎬</div>
-#                 <div style="font-weight: bold; margin: 3px 0; font-size: 14px;">Video Encoding</div>
-#                 <div style="font-size: 16px; color: #dc2626;">{encoding_time:.2f}s</div>
-#             </div>
-#             <div class="timing-card">
-#                 <div style="font-size: 20px;">🌐</div>
-#                 <div style="font-weight: bold; margin: 3px 0; font-size: 14px;">Network Transfer</div>
-#                 <div style="font-size: 16px; color: #059669;">{network_time:.2f}s</div>
-#             </div>
-#             <div class="timing-card">
-#                 <div style="font-size: 20px;">📊</div>
-#                 <div style="font-weight: bold; margin: 3px 0; font-size: 14px;">Total Processing</div>
-#                 <div style="font-size: 18px; color: #0277bd;">{total_time:.2f}s</div>
-#             </div>
-#         </div>"""
-#     
-#     if inference_time > 0:
-#         fps = num_frames / inference_time
-#         timing_html += f"""
-#         <div class="performance-card" style="margin-top: 15px;">
-#             <span style="font-weight: bold;">Generation Speed: </span>
-#             <span style="font-size: 18px; color: #6366f1; font-weight: bold;">{fps:.1f} frames/second</span>
-#         </div>"""
-#     
-#     return timing_html + "</div>"
+    timing_html = f"""
+    <div style="margin: 10px 0;">
+        <h3 style="text-align: center; margin-bottom: 10px;">⏱️ Timing Breakdown</h3>
+        <div style="display: grid; grid-template-columns: repeat(5, 1fr); gap: 10px; margin-bottom: 10px;">
+            <div class="timing-card timing-card-highlight">
+                <div style="font-size: 20px;">🚀</div>
+                <div style="font-weight: bold; margin: 3px 0; font-size: 14px;">DiT Denoising</div>
+                <div style="font-size: 16px; color: #ffa200; font-weight: bold;">{dit_denoising_time}</div>
+            </div>
+            <div class="timing-card">
+                <div style="font-size: 20px;">🧠</div>
+                <div style="font-weight: bold; margin: 3px 0; font-size: 14px;">E2E (w. vae/text encoder)</div>
+                <div style="font-size: 16px; color: #2563eb;">{inference_time:.2f}s</div>
+            </div>
+            <div class="timing-card">
+                <div style="font-size: 20px;">🎬</div>
+                <div style="font-weight: bold; margin: 3px 0; font-size: 14px;">Video Encoding</div>
+                <div style="font-size: 16px; color: #dc2626;">{encoding_time:.2f}s</div>
+            </div>
+            <div class="timing-card">
+                <div style="font-size: 20px;">🌐</div>
+                <div style="font-weight: bold; margin: 3px 0; font-size: 14px;">Network Transfer</div>
+                <div style="font-size: 16px; color: #059669;">{network_time:.2f}s</div>
+            </div>
+            <div class="timing-card">
+                <div style="font-size: 20px;">📊</div>
+                <div style="font-weight: bold; margin: 3px 0; font-size: 14px;">Total Processing</div>
+                <div style="font-size: 18px; color: #0277bd;">{total_time:.2f}s</div>
+            </div>
+        </div>"""
+    
+    if inference_time > 0:
+        fps = num_frames / inference_time
+        timing_html += f"""
+        <div class="performance-card" style="margin-top: 15px;">
+            <span style="font-weight: bold;">Generation Speed: </span>
+            <span style="font-size: 18px; color: #6366f1; font-weight: bold;">{fps:.1f} frames/second</span>
+        </div>"""
+    
+    return timing_html + "</div>"


 def load_example_prompts():
@@ -177,83 +144,26 @@ def load_example_prompts():
            print(f"Warning: Could not read {filepath}: {e}")
        return prompts, labels
    
-    # Load prompts from prompts.txt
-    examples, example_labels = load_from_file("examples/inference/gradio/serving/prompts.txt")
+    examples, example_labels = load_from_file("prompts/prompts_final.txt")
    
    if not examples:
        examples = ["A crowded rooftop bar buzzes with energy, the city skyline twinkling like a field of stars in the background."]
        example_labels = ["Crowded rooftop bar at night"]
    
-    # Load image mappings from JSON file
-    prompt_to_image = {}
-    # Try to find the JSON file relative to project root
-    possible_json_paths = [
-        Path("prompts/mixkit_i2v.jsonl"),
-        Path(__file__).parent.parent.parent.parent / "prompts" / "mixkit_i2v.jsonl",
-    ]
-    json_path = None
-    for path in possible_json_paths:
-        if path.exists():
-            json_path = path
-            break
-    
-    if json_path and json_path.exists():
-        try:
-            with open(json_path, "r", encoding='utf-8') as f:
-                data = json.load(f)
-                # Get the project root directory (parent of prompts directory)
-                project_root = json_path.parent.parent
-                for item in data:
-                    prompt_text = item.get("prompt", "").strip()
-                    image_path = item.get("image_path", "")
-                    if prompt_text and image_path:
-                        # Resolve image path relative to project root
-                        full_image_path = project_root / image_path
-                        if full_image_path.exists():
-                            prompt_to_image[prompt_text] = str(full_image_path.absolute())
-        except Exception as e:
-            print(f"Warning: Could not load image mappings from {json_path}: {e}")
-    
-    # Create image paths list matching the prompts
-    example_images = []
-    for prompt in examples:
-        # Try exact match first
-        image_path = prompt_to_image.get(prompt)
-        if not image_path:
-            # Try fuzzy match (case-insensitive, whitespace normalized)
-            normalized_prompt = " ".join(prompt.split())
-            for json_prompt, img_path in prompt_to_image.items():
-                normalized_json = " ".join(json_prompt.split())
-                if normalized_prompt.lower() == normalized_json.lower():
-                    image_path = img_path
-                    break
-        example_images.append(image_path if image_path and os.path.exists(image_path) else None)
-    
-    return examples, example_labels, example_images
+    return examples, example_labels


 def create_gradio_interface(backend_url: str, default_params: dict[str, SamplingParam]):
    
    client = RayServeClient(backend_url)
    
-    def is_i2v_model(model_name: str) -> bool:
-        """Check if the model is an I2V model."""
-        return "I2V" in model_name
-    
    def generate_video(
-        prompt, negative_prompt, use_negative_prompt, guidance_scale,
-        num_frames, height, width, model_selection, input_image, progress
+        prompt, negative_prompt, use_negative_prompt, seed, guidance_scale,
+        num_frames, height, width, randomize_seed, model_selection, progress
    ):
-        # Use default seed value (randomize_seed disabled)
-        seed = 1000
-        randomize_seed = False
        if not client.check_health():
            return None, f"Backend is not available. Please check if Ray Serve is running at {backend_url}", ""
        
-        # Check if I2V model requires an image
-        if is_i2v_model(model_selection) and not input_image:
-            return None, "I2V models require an input image. Please upload an image.", ""
-        
        # Validate dimensions
        max_pixels = 720 * 1280
        if height * width > max_pixels:
@@ -262,15 +172,6 @@ def create_gradio_interface(backend_url: str, default_params: dict[str, Sampling
        if progress:
            progress(0.1, desc="Checking backend health...")
        
-        # Encode image if provided
-        image_data = None
-        if input_image:
-            if progress:
-                progress(0.2, desc="Encoding input image...")
-            image_data = encode_image_to_base64(input_image)
-            if not image_data:
-                return None, "Failed to encode input image", ""
-        
        request_data = {
            "prompt": prompt,
            "negative_prompt": negative_prompt,
@@ -282,7 +183,7 @@ def create_gradio_interface(backend_url: str, default_params: dict[str, Sampling
            "width": width,
            "randomize_seed": randomize_seed,
            "return_frames": False,
-            "image_data": image_data,
+            "image_path": None,
            "model_path": MODEL_PATH_MAPPING.get(model_selection, "FastVideo/FastWan2.1-T2V-1.3B-Diffusers")
        }
        
@@ -297,16 +198,16 @@ def create_gradio_interface(backend_url: str, default_params: dict[str, Sampling
        if response.get("success", False):
            video_data = response.get("video_data", "")
            used_seed = response.get("seed", seed)
-            # inference_time = response.get("inference_time", 0.0)
-            # encoding_time = response.get("encoding_time", 0.0)
-            # total_time = response.get("total_time", 0.0)
-            # network_time = response.get("network_time", 0.0)
-            # stage_execution_times = response.get("stage_execution_times", [])
+            inference_time = response.get("inference_time", 0.0)
+            encoding_time = response.get("encoding_time", 0.0)
+            total_time = response.get("total_time", 0.0)
+            network_time = response.get("network_time", 0.0)
+            stage_execution_times = response.get("stage_execution_times", [])
            
-            # timing_details = create_timing_display(
-            #     inference_time, encoding_time, network_time, total_time, 
-            #     stage_execution_times, num_frames
-            # )
+            timing_details = create_timing_display(
+                inference_time, encoding_time, network_time, total_time, 
+                stage_execution_times, num_frames
+            )
            
            if video_data:
                if progress:
@@ -318,7 +219,7 @@ def create_gradio_interface(backend_url: str, default_params: dict[str, Sampling
                    progress(1.0, desc="Generation complete!")
                
                if video_path and os.path.exists(video_path):
-                    return video_path, used_seed, ""
+                    return video_path, used_seed, timing_details
                else:
                    return None, "Failed to save video", ""
            else:
@@ -327,7 +228,7 @@ def create_gradio_interface(backend_url: str, default_params: dict[str, Sampling
            error_msg = response.get("error_message", "Unknown error occurred")
            return None, f"Generation failed: {error_msg}", ""
    
-    examples, example_labels, example_images = load_example_prompts()
+    examples, example_labels = load_example_prompts()
    
    theme = gr.themes.Base().set(
        button_primary_background_fill="#2563eb",
@@ -338,39 +239,33 @@ def create_gradio_interface(backend_url: str, default_params: dict[str, Sampling
    )
    
    def get_default_values(model_name):
-        # model_path = MODEL_PATH_MAPPING.get(model_name)
-        # if model_path and model_path in default_params:
-        #     params = default_params[model_path]
-        #     return {
-        #         'height': params.height,
-        #         'width': params.width,
-        #         'num_frames': params.num_frames,
-        #         'guidance_scale': params.guidance_scale,
-        #     }
+        model_path = MODEL_PATH_MAPPING.get(model_name)
+        if model_path and model_path in default_params:
+            params = default_params[model_path]
+            return {
+                'height': params.height,
+                'width': params.width,
+                'num_frames': params.num_frames,
+                'guidance_scale': params.guidance_scale,
+                'seed': params.seed,
+            }
        
        return {
-            'height': 480,
+            'height': 448,
            'width': 832,
-            'num_frames': 73,
+            'num_frames': 61,
+            'guidance_scale': 3.0,
+            'seed': 1024,
        }
    
-    # Get available models based on what's loaded
-    available_models = []
-    for model_name, model_path in MODEL_PATH_MAPPING.items():
-        if model_path in default_params:
-            available_models.append(model_name)
+    initial_values = get_default_values("FastWan2.1-T2V-1.3B")
    
-    # Select first available model as default
-    default_model = available_models[0] if available_models else "FastWan2.1-T2V-1.3B"
-    initial_values = get_default_values(default_model)
-    initial_show_image = is_i2v_model(default_model)
-    
-    with gr.Blocks(title="CausalWan", theme=theme) as demo:
+    with gr.Blocks(title="FastWan", theme=theme) as demo:
        gr.Image("assets/logos/logo.svg", show_label=False, container=False, height=80)
        gr.HTML("""
        <div style="text-align: center; margin-bottom: 10px;">
            <p style="font-size: 18px;"> Make Video Generation Go Blurrrrrrr </p>
-            <p style="font-size: 18px;"> <a href="https://github.com/hao-ai-lab/FastVideo/tree/main" target="_blank">Code</a> | <a href="https://hao-ai-lab.github.io/blogs/fastvideo_causalwan_preview/" target="_blank">Blog</a> | <a href="https://hao-ai-lab.github.io/FastVideo/" target="_blank">Docs</a>  </p>
+            <p style="font-size: 18px;"> <a href="https://github.com/hao-ai-lab/FastVideo/tree/main" target="_blank">Code</a> | <a href="https://hao-ai-lab.github.io/blogs/fastvideo_post_training/" target="_blank">Blog</a> | <a href="https://hao-ai-lab.github.io/FastVideo/" target="_blank">Docs</a>  </p>
        </div>
        """)
        
@@ -385,8 +280,8 @@ def create_gradio_interface(backend_url: str, default_params: dict[str, Sampling
        
        with gr.Row():
            model_selection = gr.Dropdown(
-                choices=available_models,
-                value=default_model,
+                choices=list(MODEL_PATH_MAPPING.keys()),
+                value="FastWan2.1-T2V-1.3B",
                label="Select Model",
                interactive=True
            )
@@ -417,70 +312,69 @@ def create_gradio_interface(backend_url: str, default_params: dict[str, Sampling
        with gr.Row():
            with gr.Column():
                error_output = gr.Text(label="Error", visible=False)
-                # timing_display = gr.Markdown(label="Timing Breakdown", visible=False)
+                timing_display = gr.Markdown(label="Timing Breakdown", visible=False)

-        with gr.Row(equal_height=False):
-            with gr.Column(scale=1):
-                with gr.Tabs():
-                    with gr.Tab("Input Image", visible=initial_show_image) as image_tab:
-                        gr.Markdown("**Please make sure you upload a 480x832 image**")
-                        input_image = gr.Image(
-                            label="",
-                            type="filepath",
-                            height=400,
+        with gr.Row(equal_height=True, elem_classes="main-content-row"):
+            with gr.Column(scale=1, elem_classes="advanced-options-column"):
+                with gr.Group():
+                    gr.HTML("<div style='margin: 0 0 15px 0; text-align: center; font-size: 16px;'>Advanced Options</div>")
+                    with gr.Row():
+                        height = gr.Number(
+                            label="Height",
+                            value=initial_values['height'],
+                            interactive=False,
+                            container=True
+                        )
+                        width = gr.Number(
+                            label="Width",
+                            value=initial_values['width'],
+                            interactive=False,
+                            container=True
                        )
                    
-                    with gr.Tab("Advanced Options"):
-                        with gr.Group():
-                            with gr.Row():
-                                height = gr.Number(
-                                    label="Height",
-                                    value=initial_values['height'],
-                                    interactive=False,
-                                    container=True
-                                )
-                                width = gr.Number(
-                                    label="Width",
-                                    value=initial_values['width'],
-                                    interactive=False,
-                                    container=True
-                                )
-                            
-                            with gr.Row():
-                                num_frames = gr.Number(
-                                    label="Number of Frames",
-                                    value=initial_values['num_frames'],
-                                    interactive=False,
-                                    container=True
-                                )
-                                guidance_scale = gr.Number(
-                                    label="Guidance Scale",
-                                    value=1.0,
-                                    interactive=False,
-                                    container=True
-                                )
-                            
-                            with gr.Row():
-                                use_negative_prompt = gr.Checkbox(
-                                    label="Use negative prompt", value=False)
-                                negative_prompt = gr.Text(
-                                    label="Negative prompt",
-                                    max_lines=3,
-                                    lines=3,
-                                    placeholder="Enter a negative prompt",
-                                    visible=False,
-                                )
+                    with gr.Row():
+                        num_frames = gr.Number(
+                            label="Number of Frames",
+                            value=initial_values['num_frames'],
+                            interactive=False,
+                            container=True
+                        )
+                        guidance_scale = gr.Slider(
+                            label="Guidance Scale",
+                            minimum=1,
+                            maximum=12,
+                            value=initial_values['guidance_scale'],
+                        )
+                    
+                    with gr.Row():
+                        use_negative_prompt = gr.Checkbox(
+                            label="Use negative prompt", value=False)
+                        negative_prompt = gr.Text(
+                            label="Negative prompt",
+                            max_lines=3,
+                            lines=3,
+                            placeholder="Enter a negative prompt",
+                            visible=False,
+                        )

-                            # randomize_seed = gr.Checkbox(label="Randomize seed", value=False)
-                            seed_output = gr.Number(label="Used Seed", value=1000)
+                    seed = gr.Slider(
+                        label="Seed",
+                        minimum=0,
+                        maximum=1000000,
+                        step=1,
+                        value=initial_values['seed'],
+                    )
+                    randomize_seed = gr.Checkbox(label="Randomize seed", value=False)
+                    seed_output = gr.Number(label="Used Seed")
        
-            with gr.Column(scale=1):
+            with gr.Column(scale=1, elem_classes="video-column"):
                result = gr.Video(
                    label="Generated Video", 
                    show_label=True,
-                    height=500,
+                    height=466,
+                    width=600,
                    container=True,
-                    autoplay=True,
+                    elem_classes="video-component"
                )
        
        gr.HTML("""
@@ -493,10 +387,116 @@ def create_gradio_interface(backend_url: str, default_params: dict[str, Sampling
        }
        
        .gradio-container {
-            max-width: 1400px !important;
+            max-width: 1200px !important;
            margin: 0 auto !important;
        }
        
+        .main {
+            max-width: 1200px !important;
+            margin: 0 auto !important;
+        }
+        
+        .gr-form, .gr-box, .gr-group {
+            max-width: 1200px !important;
+        }
+        
+        .gr-video {
+            max-width: 500px !important;
+            margin: 0 auto !important;
+        }
+        
+        .main-content-row {
+            display: flex !important;
+            align-items: flex-start !important;
+            min-height: 500px !important;
+            gap: 20px !important;
+        }
+        
+        .advanced-options-column,
+        .video-column {
+            display: flex !important;
+            flex-direction: column !important;
+            flex: 1 !important;
+            min-height: 400px !important;
+            align-items: stretch !important;
+        }
+        
+        .video-column > * {
+            margin-top: 0 !important;
+        }
+        
+        .video-column .gr-video,
+        .video-component {
+            margin-top: 0 !important;
+            padding-top: 0 !important;
+        }
+        
+        .video-column .gr-video .gr-form {
+            margin-top: 0 !important;
+        }
+        
+        .advanced-options-column .gr-group,
+        .video-column .gr-video {
+            margin-top: 0 !important;
+            vertical-align: top !important;
+        }
+        
+        .advanced-options-column > *:last-child,
+        .video-column > *:last-child {
+            flex-grow: 0 !important;
+        }
+        
+        @media (max-width: 1400px) {
+            .main-content-row {
+                min-height: 600px !important;
+            }
+            
+            .advanced-options-column,
+            .video-column {
+                min-height: 600px !important;
+            }
+        }
+        
+        @media (max-width: 1200px) {
+            .main-content-row {
+                flex-direction: column !important;
+                align-items: stretch !important;
+            }
+            
+            .advanced-options-column,
+            .video-column {
+                min-height: auto !important;
+                width: 100% !important;
+            }
+        }
+        
+        .timing-card {
+            background: var(--background-fill-secondary) !important;
+            border: 1px solid var(--border-color-primary) !important;
+            color: var(--body-text-color) !important;
+            padding: 10px;
+            border-radius: 8px;
+            text-align: center;
+            min-height: 80px;
+            display: flex;
+            flex-direction: column;
+            justify-content: center;
+        }
+        
+        .timing-card-highlight {
+            background: var(--background-fill-primary) !important;
+            border: 2px solid var(--color-accent) !important;
+        }
+        
+        .performance-card {
+            background: var(--background-fill-secondary) !important;
+            border: 1px solid var(--border-color-primary) !important;
+            color: var(--body-text-color) !important;
+            padding: 10px;
+            border-radius: 6px;
+            text-align: center;
+        }
+        
        .gr-number input[readonly] {
            background-color: var(--background-fill-secondary) !important;
            border: 1px solid var(--border-color-primary) !important;
@@ -511,20 +511,18 @@ def create_gradio_interface(backend_url: str, default_params: dict[str, Sampling
        def on_example_select(example_label):
            if example_label and example_label in example_labels:
                index = example_labels.index(example_label)
-                selected_prompt = examples[index]
-                selected_image = example_images[index] if index < len(example_images) else None
-                return selected_prompt, selected_image
-            return "", None
+                return examples[index]
+            return ""
        
        example_dropdown.change(
            fn=on_example_select,
            inputs=example_dropdown,
-            outputs=[prompt, input_image],
+            outputs=prompt,
        )
        
        gr.HTML("""
        <div style="text-align: center; margin-top: 10px; margin-bottom: 15px;">
-            <p style="font-size: 16px; margin: 0;">The compute for this demo is generously provided by <a href="https://www.gmicloud.ai/" target="_blank">GMI Cloud</a>.  Note that this demo is meant as a preview of our distilled I2V model. Outside of few-step distillation, we have not yet fully optimized it for speed. Stay tuned for updates!</p>
+            <p style="font-size: 16px; margin: 0;">The compute for this demo is generously provided by <a href="https://www.gmicloud.ai/" target="_blank">GMI Cloud</a>. Note that this demo is meant to showcase FastWan's quality and that under a large number of requests, generation speed may be affected. We are also rate-limiting users to 3 requests per minute.</p>
        </div>
        """)
        
@@ -539,7 +537,6 @@ def create_gradio_interface(backend_url: str, default_params: dict[str, Sampling
                selected_model = "FastWan2.1-T2V-1.3B"
            
            model_path = MODEL_PATH_MAPPING.get(selected_model)
-            show_image_input = is_i2v_model(selected_model)
            
            if model_path and model_path in default_params:
                params = default_params[model_path]
@@ -548,29 +545,29 @@ def create_gradio_interface(backend_url: str, default_params: dict[str, Sampling
                    gr.update(value=params.width),
                    gr.update(value=params.num_frames),
                    gr.update(value=params.guidance_scale),
-                    gr.update(visible=show_image_input),
+                    gr.update(value=params.seed),
                )
            
            return (
                gr.update(value=448),
                gr.update(value=832),
-                gr.update(value=20),
+                gr.update(value=61),
                gr.update(value=3.0),
-                gr.update(visible=show_image_input),
+                gr.update(value=1024),
            )
        
        model_selection.change(
            fn=on_model_selection_change,
            inputs=model_selection,
-            outputs=[height, width, num_frames, guidance_scale, image_tab],
+            outputs=[height, width, num_frames, guidance_scale, seed],
        )
        
        def handle_generation(*args, progress=None, request: gr.Request = None):
-            model_selection, prompt, negative_prompt, use_negative_prompt, guidance_scale, num_frames, height, width, input_image = args
+            model_selection, prompt, negative_prompt, use_negative_prompt, seed, guidance_scale, num_frames, height, width, randomize_seed = args
            
-            result_path, seed_or_error, _ = generate_video(
-                prompt, negative_prompt, use_negative_prompt, guidance_scale, 
-                num_frames, height, width, model_selection, input_image, progress
+            result_path, seed_or_error, timing_details = generate_video(
+                prompt, negative_prompt, use_negative_prompt, seed, guidance_scale, 
+                num_frames, height, width, randomize_seed, model_selection, progress
            )
            
            if result_path and os.path.exists(result_path):
@@ -578,12 +575,14 @@ def create_gradio_interface(backend_url: str, default_params: dict[str, Sampling
                    result_path, 
                    seed_or_error, 
                    gr.update(visible=False),
+                    gr.update(visible=True, value=timing_details),
                )
            else:
                return (
                    None, 
                    seed_or_error, 
                    gr.update(visible=True, value=seed_or_error),
+                    gr.update(visible=False),
                )
        
        run_button.click(
@@ -593,14 +592,14 @@ def create_gradio_interface(backend_url: str, default_params: dict[str, Sampling
                prompt,
                negative_prompt,
                use_negative_prompt,
+                seed,
                guidance_scale,
                num_frames,
                height,
                width,
-                # randomize_seed,
-                input_image,
+                randomize_seed,
            ],
-            outputs=[result, seed_output, error_output],  # timing_display removed
+            outputs=[result, seed_output, error_output, timing_display],
            concurrency_limit=20,
        )
    
@@ -612,11 +611,8 @@ def main():
    parser.add_argument("--backend_url", type=str, default="http://localhost:8000",
                        help="URL of the Ray Serve backend")
    parser.add_argument("--t2v_model_paths", type=str,
-                        default="",
+                        default="FastVideo/FastWan2.1-T2V-1.3B-Diffusers,FastVideo/FastWan2.1-T2V-14B-Diffusers",
                        help="Comma separated list of paths to the T2V model(s)")
-    parser.add_argument("--i2v_model_paths", type=str,
-                        default="FastVideo/SFWan2.2-I2V-A14B-Preview-Diffusers",
-                        help="Comma separated list of paths to the I2V model(s)")
    parser.add_argument("--host", type=str, default="0.0.0.0",
                        help="Host to bind to")
    parser.add_argument("--port", type=int, default=7860,
@@ -625,15 +621,8 @@ def main():
    args = parser.parse_args()
    
    default_params = {}
-    
-    # Load T2V model params
-    t2v_paths = [p.strip() for p in args.t2v_model_paths.split(",") if p.strip()]
-    for model_path in t2v_paths:
-        default_params[model_path] = SamplingParam.from_pretrained(model_path)
-    
-    # Load I2V model params
-    i2v_paths = [p.strip() for p in args.i2v_model_paths.split(",") if p.strip()]
-    for model_path in i2v_paths:
+    model_paths = args.t2v_model_paths.split(",")
+    for model_path in model_paths:
        default_params[model_path] = SamplingParam.from_pretrained(model_path)
    
    demo = create_gradio_interface(args.backend_url, default_params)
@@ -641,8 +630,6 @@ def main():
    print(f"Starting Gradio frontend at http://{args.host}:{args.port}")
    print(f"Backend URL: {args.backend_url}")
    print(f"T2V Models: {args.t2v_model_paths}")
-    if args.i2v_model_paths:
-        print(f"I2V Models: {args.i2v_model_paths}")
    
    from fastapi import FastAPI, Request, HTTPException
    from fastapi.responses import HTMLResponse, FileResponse
@@ -687,23 +674,23 @@ def main():
            <meta charset="UTF-8" />
            <meta name="viewport" content="width=device-width, initial-scale=1.0" />
            
-            <title>CausalWan</title>
-            <meta name="title" content="CausalWan">
+            <title>FastWan</title>
+            <meta name="title" content="FastWan">
            <meta name="description" content="Make video generation go blurrrrrrr">
-            <meta name="keywords" content="FastVideo, video generation, AI, machine learning, CausalWan">
+            <meta name="keywords" content="FastVideo, video generation, AI, machine learning, FastWan">
            
            <meta property="og:type" content="website">
            <meta property="og:url" content="{base_url}/">
-            <meta property="og:title" content="CausalWan">
+            <meta property="og:title" content="FastWan">
            <meta property="og:description" content="Make video generation go blurrrrrrr">
            <meta property="og:image" content="{base_url}/logo.svg">
            <meta property="og:image:width" content="1200">
            <meta property="og:image:height" content="630">
-            <meta property="og:site_name" content="CausalWan">
+            <meta property="og:site_name" content="FastWan">
            
            <meta property="twitter:card" content="summary_large_image">
            <meta property="twitter:url" content="{base_url}/">
-            <meta property="twitter:title" content="CausalWan">
+            <meta property="twitter:title" content="FastWan">
            <meta property="twitter:description" content="Make video generation go blurrrrrrr">
            <meta property="twitter:image" content="{base_url}/logo.svg">
            <link rel="icon" type="image/png" sizes="32x32" href="/favicon.ico">
@@ -733,14 +720,7 @@ def main():
        app, 
        demo, 
        path="/gradio",
-        allowed_paths=[
-            os.path.abspath("outputs"), 
-            os.path.abspath("fastvideo-logos"),
-            os.path.abspath("prompts"),
-            os.path.abspath("images"),
-            os.path.abspath(tempfile.gettempdir()),
-            os.path.abspath(os.path.join(tempfile.gettempdir(), "gradio")),
-        ]
+        allowed_paths=[os.path.abspath("outputs"), os.path.abspath("fastvideo-logos")]
    )
    
    uvicorn.run(app, host=args.host, port=args.port)
@@ -1,15 +0,0 @@
-Some friends dancing and having fun together in circles, at a party surrounded by colored lights at a party, in a fancy old place, in a view from below them.
-A man wearing grey shorts jumps rope in a gym, weights and gym equipment in the background.
-Flying over a peninsula covered in bushy trees, while discovering the sea around it, painted a beautiful turquoise blue, on a sunny day.
-Skillful cyclist doing a wheelie on a bike while riding through a forest, on a dirt road, surrounded by many trees, in the morning.
-In the video, a lone rider guides a majestic horse across an expansive, open field as the sun sets in the background. The rider, dressed in a classic blue shirt and wide-brimmed hat, sits confidently in the saddle, silhouetted against the warm glow of the evening sky. The horse moves gracefully, its mane and tail flowing with each step, creating a sense of harmony between horse and rider. Surrounding the pair, towering trees form a natural border, their leaves gently rustling in the breeze. The shadows lengthen on the ground, accentuating the serene and timeless feel of the scene. The distant hills and wooden fences frame the horizon, adding depth to the tranquil landscape. A few horses graze peacefully in the background, blending into the pastoral setting. The overall ambiance evokes a sense of calmness and quietude, capturing a perfect moment in the golden light of dusk.
-A silver SUV drives along a winding, snow-covered mountain road, with dense pine trees blanketed in snow lining both sides. The scene is serene, with the vehicle moving smoothly, possibly on a winter journey or vacation. As the SUV disappears around the bend, another, darker SUV follows, creating a sense of motion and perspective on the snow-dusted asphalt. The towering, snow-laden rock formation to the right contrasts with the dark green of the pines, highlighting the peacefulness of the wintry landscape.
-A man and a woman playing in a field with grass, during a bright afternoon, while cars pass by in the distance.
-A saxophonist wearing a blazer dances while playing a song in a park.
-Romantic couple embracing and looking at each other in the middle of a forest, during a break on a road trip through nature.
-Young woman cleaning her house decorated with plants and decorations, while dancing happily to music in her headphones.
-Man dressed in 80's style dances very happily in his kitchen while listening to music on his radio and drinking wine.
-Pair of jazz musicians performing a song with their saxophone and trombone on an abandoned train.
-A young woman with short hair wearing pink sunglasses chews gum and makes a bubble gum with the city in the background.
-Natural aerial landscape with a relief covered with abundant trees and vegetation and a thick layer of mist.
-Loving couple sitting on a log on the shore of a lake outside, sharing an affectionate hug.
@@ -26,7 +26,6 @@ SEED_RANGE_MAX = 1_000_000
 SUPPORTED_MODELS = [
    "FastVideo/FastWan2.1-T2V-1.3B-Diffusers", 
    "FastVideo/FastWan2.2-TI2V-5B-FullAttn-Diffusers",
-    "FastVideo/SFWan2.2-I2V-A14B-Preview-Diffusers",
 ]

 MODEL_CONFIGS = {
@@ -43,13 +42,6 @@ MODEL_CONFIGS = {
        "dit_cpu_offload": True,
        "vae_cpu_offload": False,
        "VSA_sparsity": 0.9,
-    },
-    "I2V-A14B": {
-        "num_cpus": 15,
-        "text_encoder_cpu_offload": True,
-        "dit_cpu_offload": True,
-        "vae_cpu_offload": False,
-        "VSA_sparsity": 0.0,
    }
 }

@@ -66,7 +58,6 @@ class VideoGenerationRequest(BaseModel):
    randomize_seed: bool = False
    return_frames: bool = False
    model_path: Optional[str] = None
-    image_data: Optional[str] = None  # Base64 encoded image for I2V


 class VideoGenerationResponse(BaseModel):
@@ -100,38 +91,11 @@ def encode_video_to_base64(frames: List[np.ndarray], fps: int = DEFAULT_FPS) ->
        return ""


-def save_image_from_base64(image_data: str, output_dir: str) -> Optional[str]:
-    """Save base64 image data to a temporary file and return the path."""
-    if not image_data:
-        return None
-    
-    try:
-        # Remove data URL prefix if present
-        if image_data.startswith('data:image/'):
-            image_data = image_data.split(',')[1]
-        
-        image_bytes = base64.b64decode(image_data)
-        
-        # Save to temporary file
-        os.makedirs(output_dir, exist_ok=True)
-        temp_image_path = os.path.join(output_dir, f"temp_input_{int(time.time() * 1000)}.png")
-        
-        with open(temp_image_path, 'wb') as f:
-            f.write(image_bytes)
-        
-        return temp_image_path
-        
-    except Exception as e:
-        print(f"Warning: Failed to save image: {e}")
-        return None
-
-
 def setup_model_environment(model_path: str) -> None:
-    # if "fullattn" in model_path.lower():
-    #     os.environ["FASTVIDEO_ATTENTION_BACKEND"] = "FLASH_ATTN"
-    # else:
-    #     os.environ["FASTVIDEO_ATTENTION_BACKEND"] = "VIDEO_SPARSE_ATTN"
-    os.environ["FASTVIDEO_ATTENTION_BACKEND"] = "FLASH_ATTN"
+    if "fullattn" in model_path.lower():
+        os.environ["FASTVIDEO_ATTENTION_BACKEND"] = "FLASH_ATTN"
+    else:
+        os.environ["FASTVIDEO_ATTENTION_BACKEND"] = "VIDEO_SPARSE_ATTN"
    os.environ["FASTVIDEO_STAGE_LOGGING"] = "1"


@@ -193,41 +157,22 @@ class BaseModelDeployment:
            num_gpus=1,
            use_fsdp_inference=True,
            text_encoder_cpu_offload=config["text_encoder_cpu_offload"],
-            dmd_denoising_steps=[1000, 850, 700, 550, 350, 275, 200, 125], # TODO: hardocde for I2V
-            dit_precision="fp32",  # TODO: hardocde for I2V
            dit_cpu_offload=config["dit_cpu_offload"],
            vae_cpu_offload=config["vae_cpu_offload"],
            VSA_sparsity=config["VSA_sparsity"],
            enable_stage_verification=False,
        )
        self.default_params = SamplingParam.from_pretrained(self.model_path)
-        self.default_params.seed = 1000
-        self.default_params.num_frames = 73
-        self.default_params.width = 832
-        self.default_params.height = 480

    def generate_video(self, video_request: VideoGenerationRequest) -> VideoGenerationResponse:
        total_start_time = time.time()
        
        params = prepare_sampling_params(video_request, self.default_params)

-        # Save image if provided (for I2V)
-        image_path = None
-        if video_request.image_data:
-            image_path = save_image_from_base64(video_request.image_data, self.output_path)
-            if image_path is None:
-                return VideoGenerationResponse(
-                    video_data=None,
-                    seed=params.seed,
-                    success=False,
-                    error_message="Failed to save input image",
-                )
-
        inference_start_time = time.time()
        result = self.generator.generate_video(
            prompt=video_request.prompt,
            sampling_param=params,
-            image_path=image_path,
            save_video=False,
            return_frames=False,
        )
@@ -240,13 +185,6 @@ class BaseModelDeployment:
        encoding_time = time.time() - encoding_start_time
        
        total_time = time.time() - total_start_time
-        
-        # Clean up temporary image file
-        if image_path and os.path.exists(image_path):
-            try:
-                os.remove(image_path)
-            except Exception as e:
-                print(f"Warning: Failed to remove temporary image file {image_path}: {e}")

        return VideoGenerationResponse(
            video_data=video_data,
@@ -262,7 +200,7 @@ class BaseModelDeployment:


@serve.deployment(
-    ray_actor_options={"num_cpus": 2, "num_gpus": 1, "runtime_env": {"conda": "demo-fv"}},
+    ray_actor_options={"num_cpus": 2, "num_gpus": 1, "runtime_env": {"conda": "fv"}},
 )
 class T2VModelDeployment(BaseModelDeployment):
    def __init__(self, t2v_model_path: str, output_path: str = "outputs"):
@@ -272,7 +210,7 @@ class T2VModelDeployment(BaseModelDeployment):


@serve.deployment(
-    ray_actor_options={"num_cpus": 16, "num_gpus": 1, "runtime_env": {"conda": "demo-fv"}},
+    ray_actor_options={"num_cpus": 16, "num_gpus": 1, "runtime_env": {"conda": "fv"}},
 )
 class T2V14BModelDeployment(BaseModelDeployment):
    def __init__(self, t2v_14b_model_path: str, output_path: str = "outputs"):
@@ -283,32 +221,18 @@ class T2V14BModelDeployment(BaseModelDeployment):
        print("✅ T2V 14B model initialized successfully")


-@serve.deployment(
-    ray_actor_options={"num_cpus": 15, "num_gpus": 1, "runtime_env": {"conda": "demo-fv"}},
-)
-class I2VModelDeployment(BaseModelDeployment):
-    def __init__(self, i2v_model_path: str, output_path: str = "outputs"):
-        super().__init__(i2v_model_path, output_path)
-        # Override environment for I2V model
-        os.environ["FASTVIDEO_ATTENTION_BACKEND"] = "FLASH_ATTN"
-        self._initialize_generator(MODEL_CONFIGS["I2V-A14B"])
-        print("✅ I2V model initialized successfully")
-
-
 app = FastAPI()
 limiter = Limiter(key_func=get_remote_address)
 app.state.limiter = limiter
 app.add_exception_handler(RateLimitExceeded, _rate_limit_exceeded_handler)


-@serve.deployment(num_replicas=1, ray_actor_options={"num_cpus": 1})
+@serve.deployment(num_replicas=50, ray_actor_options={"num_cpus": 2})
@serve.ingress(app)
 class FastVideoAPI:

-    def __init__(self, t2v_deployments: Dict[str, DeploymentHandle], i2v_deployments: Dict[str, DeploymentHandle] = None):
+    def __init__(self, t2v_deployments: Dict[str, DeploymentHandle]):
        self.t2v_deployments = t2v_deployments
-        self.i2v_deployments = i2v_deployments or {}
-        self.all_deployments = {**self.t2v_deployments, **self.i2v_deployments}
        
        # Initialize Prometheus metrics
        self.request_count = Counter('fastvideo_requests_total', 'Total FastVideo requests', ['model_type', 'status'])
@@ -333,10 +257,10 @@ class FastVideoAPI:
        model_name = self._get_model_name(video_request.model_path)
        
        try:
-            if video_request.model_path not in self.all_deployments:
+            if video_request.model_path not in self.t2v_deployments:
                raise ValueError(f"Model {video_request.model_path} not found")
            
-            response_ref = self.all_deployments[video_request.model_path].generate_video.remote(video_request)
+            response_ref = self.t2v_deployments[video_request.model_path].generate_video.remote(video_request)
            response = await response_ref
            
            self._record_metrics(model_name, "success", time.time() - start_time, response)
@@ -367,21 +291,18 @@ class FastVideoAPI:


 def validate_configuration(model_paths: List[str], replicas: List[int]) -> None:
-    assert len(model_paths) > 0, "At least one model must be specified"
    assert len(model_paths) == len(replicas), "Number of models and replicas must match"
    assert sum(replicas) <= NUM_GPUS, f"Total replicas ({sum(replicas)}) must be <= {NUM_GPUS}"
    
    for model, replica_count in zip(model_paths, replicas):
-        assert model in SUPPORTED_MODELS, f"Model {model} not supported. Supported models: {SUPPORTED_MODELS}"
+        assert model in SUPPORTED_MODELS, f"Model {model} not supported"
        assert replica_count > 0, f"Replicas must be greater than 0"


 def start_ray_serve(
    *,
-    t2v_model_paths: str = "",
-    t2v_model_replicas: str = "",
-    i2v_model_paths: str = "",
-    i2v_model_replicas: str = "",
+    t2v_model_paths: str,
+    t2v_model_replicas: str,
    output_path: str = "outputs",
    host: str = "0.0.0.0",
    port: int = 8000,
@@ -389,39 +310,21 @@ def start_ray_serve(
    if not ray.is_initialized():
        ray.init()

-    # Parse T2V models
-    t2v_paths = [p.strip() for p in t2v_model_paths.split(",") if p.strip()]
-    t2v_reps = [int(r.strip()) for r in t2v_model_replicas.split(",") if r.strip()] if t2v_model_replicas else []
-    
-    # Parse I2V models
-    i2v_paths = [p.strip() for p in i2v_model_paths.split(",") if p.strip()]
-    i2v_reps = [int(r.strip()) for r in i2v_model_replicas.split(",") if r.strip()] if i2v_model_replicas else []
-    
-    # Validate configurations
-    all_paths = t2v_paths + i2v_paths
-    all_replicas = t2v_reps + i2v_reps
-    validate_configuration(all_paths, all_replicas)
+    model_paths = t2v_model_paths.split(",")
+    replicas = [int(r) for r in t2v_model_replicas.split(",")]
+    validate_configuration(model_paths, replicas)

-    # Create T2V deployments
    t2v_deps = {}
-    for model_path, replica_count in zip(t2v_paths, t2v_reps):
+    for model_path, replica_count in zip(model_paths, replicas):
        t2v_dep = T2VModelDeployment.options(num_replicas=replica_count).bind(model_path, output_path)
        t2v_deps[model_path] = t2v_dep

-    # Create I2V deployments
-    i2v_deps = {}
-    for model_path, replica_count in zip(i2v_paths, i2v_reps):
-        i2v_dep = I2VModelDeployment.options(num_replicas=replica_count).bind(model_path, output_path)
-        i2v_deps[model_path] = i2v_dep
-
-    api = FastVideoAPI.bind(t2v_deps, i2v_deps)
+    api = FastVideoAPI.bind(t2v_deps)
    serve.run(api, route_prefix="/", name="fast_video")

    print(f"Ray Serve backend started at http://{host}:{port}")
-    for model_path, replica_count in zip(t2v_paths, t2v_reps):
+    for model_path, replica_count in zip(model_paths, replicas):
        print(f"T2V Model: {model_path} | Replicas: {replica_count}")
-    for model_path, replica_count in zip(i2v_paths, i2v_reps):
-        print(f"I2V Model: {model_path} | Replicas: {replica_count}")
    print(f"Health check: http://{host}:{port}/health")
    print(f"Video generation endpoint: http://{host}:{port}/generate_video")

@@ -437,20 +340,12 @@ if __name__ == "__main__":
    parser = argparse.ArgumentParser(description="FastVideo Ray Serve Backend")
    parser.add_argument("--t2v_model_paths",
                        type=str,
-                        default="",
+                        default="FastVideo/FastWan2.1-T2V-1.3B-Diffusers,FastVideo/FastWan2.2-TI2V-5B-FullAttn-Diffusers",
                        help="Comma separated list of paths to the T2V model(s)")
    parser.add_argument("--t2v_model_replicas",
                        type=str,
-                        default="",
+                        default="4,4",
                        help="Comma separated list of number of replicas for the T2V model(s)")
-    parser.add_argument("--i2v_model_paths",
-                        type=str,
-                        default="FastVideo/SFWan2.2-I2V-A14B-Preview-Diffusers",
-                        help="Comma separated list of paths to the I2V model(s)")
-    parser.add_argument("--i2v_model_replicas",
-                        type=str,
-                        default="1",
-                        help="Comma separated list of number of replicas for the I2V model(s)")
    parser.add_argument("--output_path",
                        type=str,
                        default="outputs",
@@ -466,21 +361,13 @@ if __name__ == "__main__":
    
    args = parser.parse_args()

-    # Parse and validate all models
-    t2v_paths = [p.strip() for p in args.t2v_model_paths.split(",") if p.strip()]
-    t2v_reps = [int(r.strip()) for r in args.t2v_model_replicas.split(",") if r.strip()] if args.t2v_model_replicas else []
-    i2v_paths = [p.strip() for p in args.i2v_model_paths.split(",") if p.strip()]
-    i2v_reps = [int(r.strip()) for r in args.i2v_model_replicas.split(",") if r.strip()] if args.i2v_model_replicas else []
-    
-    all_paths = t2v_paths + i2v_paths
-    all_replicas = t2v_reps + i2v_reps
-    validate_configuration(all_paths, all_replicas)
+    model_paths = args.t2v_model_paths.split(",")
+    replicas = [int(r) for r in args.t2v_model_replicas.split(",")]
+    validate_configuration(model_paths, replicas)
    
    start_ray_serve(
        t2v_model_paths=args.t2v_model_paths,
        t2v_model_replicas=args.t2v_model_replicas,
-        i2v_model_paths=args.i2v_model_paths,
-        i2v_model_replicas=args.i2v_model_replicas,
        output_path=args.output_path,
        host=args.host,
        port=args.port,
@@ -489,4 +376,4 @@ if __name__ == "__main__":
    setup_signal_handlers()
    print("✅ FastVideo backend is running. Press Ctrl-C to stop.")
    while True:
-        time.sleep(3600) 
+        time.sleep(3600) 
@@ -1,5 +1,3 @@
-python examples/inference/gradio/serving/start_ray_serve_app.py \
--t2v_model_paths "" \
--t2v_model_replicas "" \
--i2v_model_paths "FastVideo/SFWan2.2-I2V-A14B-Preview-Diffusers" \
--i2v_model_replicas "1"
+python examples/inference/gradio/start_ray_serve_app.py \
+--t2v_model_paths "FastVideo/FastWan2.1-T2V-1.3B-Diffusers,FastVideo/FastWan2.2-TI2V-5B-FullAttn-Diffusers" \
+--t2v_model_replicas "4,4"
@@ -20,10 +20,8 @@ DEFAULT_BACKEND_PORT = 8000
 DEFAULT_FRONTEND_HOST = "0.0.0.0"
 DEFAULT_FRONTEND_PORT = 7860
 DEFAULT_OUTPUT_PATH = "outputs"
-DEFAULT_T2V_MODELS = ""
-DEFAULT_T2V_REPLICAS = ""
-DEFAULT_I2V_MODELS = "FastVideo/SFWan2.2-I2V-A14B-Preview-Diffusers"
-DEFAULT_I2V_REPLICAS = "1"
+DEFAULT_T2V_MODELS = "FastVideo/FastWan2.1-T2V-1.3B-Diffusers,FastVideo/FastWan2.2-TI2V-5B-FullAttn-Diffusers"
+DEFAULT_T2V_REPLICAS = "4,4"

 HEALTH_CHECK_TIMEOUT = 5
 HEALTH_CHECK_MAX_RETRIES = 100
@@ -102,12 +100,6 @@ class ServiceManager:
            "port": self.args.backend_port
        }
        
-        # Add I2V parameters if provided
-        if self.args.i2v_model_paths:
-            backend_args["i2v_model_paths"] = self.args.i2v_model_paths
-        if self.args.i2v_model_replicas:
-            backend_args["i2v_model_replicas"] = self.args.i2v_model_replicas
-        
        self.backend_process = self._start_service("ray_serve_backend.py", backend_args, "backend")
        return self.backend_process
    
@@ -119,10 +111,6 @@ class ServiceManager:
            "port": self.args.frontend_port
        }
        
-        # Add I2V parameters if provided
-        if self.args.i2v_model_paths:
-            frontend_args["i2v_model_paths"] = self.args.i2v_model_paths
-        
        self.frontend_process = self._start_service("gradio_frontend.py", frontend_args, "frontend")
        return self.frontend_process
    
@@ -185,9 +173,6 @@ def print_startup_info(args: argparse.Namespace) -> None:
    print("=" * 50)
    print(f"T2V Models: {args.t2v_model_paths}")
    print(f"T2V Model Replicas: {args.t2v_model_replicas}")
-    if args.i2v_model_paths:
-        print(f"I2V Models: {args.i2v_model_paths}")
-        print(f"I2V Model Replicas: {args.i2v_model_replicas}")
    print(f"Output: {args.output_path}")
    print(f"Backend: http://{args.backend_host}:{args.backend_port}")
    print(f"Frontend: http://{args.frontend_host}:{args.frontend_port}")
@@ -205,14 +190,6 @@ def parse_arguments() -> argparse.Namespace:
                        type=str, 
                        default=DEFAULT_T2V_REPLICAS,
                        help="Comma separated list of number of replicas for the T2V model(s)")
-    parser.add_argument("--i2v_model_paths", 
-                        type=str,
-                        default=DEFAULT_I2V_MODELS,
-                        help="Comma separated list of paths to the I2V model(s)")
-    parser.add_argument("--i2v_model_replicas", 
-                        type=str, 
-                        default=DEFAULT_I2V_REPLICAS,
-                        help="Comma separated list of number of replicas for the I2V model(s)")
    parser.add_argument("--output_path", 
                        type=str, 
                        default=DEFAULT_OUTPUT_PATH,
@@ -5,12 +5,12 @@ These are e2e example scripts for finetuning Wan2.1 T2V 1.3B on the crush-smol d

 ### Download crush-smol dataset:

-`bash examples/training/finetune/wan_t2v_1.3B/crush_smol/download_dataset.sh`
+`bash examples/training/finetune/wan_t2v_1_3b/crush_smol/download_dataset.sh`

 ### Preprocess the videos and captions into latents:

-`bash examples/training/finetune/wan_t2v_1.3B/crush_smol/preprocess_wan_data_t2v.sh`
+`bash examples/training/finetune/wan_t2v_1_3b/crush_smol/preprocess_wan_data_t2v.sh`

 ### Edit the following file and run finetuning:

-`bash examples/training/finetune/wan_t2v_1.3B/crush_smol/finetune_t2v.sh`
+`bash examples/training/finetune/wan_t2v_1_3b/crush_smol/finetune_t2v.sh`
@@ -54,7 +54,7 @@ validation_args=(
  --validation_dataset_file $VALIDATION_DATASET_FILE
  --validation_steps 200
  --validation_sampling_steps "50" 
-  --validation_guidance_scale "3.0"
+  --validation_guidance_scale "6.0"
 )

 # Optimizer arguments
@@ -1,107 +0,0 @@
-#!/bin/bash
-
-export WANDB_BASE_URL="https://api.wandb.ai"
-export WANDB_MODE=online
-export TOKENIZERS_PARALLELISM=false
-# export FASTVIDEO_ATTENTION_BACKEND=TORCH_SDPA
-
-MODEL_PATH="Wan-AI/Wan2.1-T2V-1.3B-Diffusers"
-DATA_DIR="data/crush-smol_processed_t2v/combined_parquet_dataset/"
-VALIDATION_DATASET_FILE="$(dirname "$0")/validation.json"
-NUM_GPUS=4
-# export CUDA_VISIBLE_DEVICES=4,5
-
-mkdir -p ../profiler_traces/wan_t2v_finetune/
-
-# Torch Profiler Configuration
-export FASTVIDEO_TORCH_PROFILE_REGIONS="profiler_region_training_train_one_step"
-export FASTVIDEO_TORCH_PROFILER_RECORD_SHAPES=1
-export FASTVIDEO_TORCH_PROFILER_WITH_STACK=1  
-export FASTVIDEO_TORCH_PROFILER_WITH_FLOPS=1
-export FASTVIDEO_TORCH_PROFILER_WITH_PROFILE_MEMORY=1
-export FASTVIDEO_TORCH_PROFILER_WAIT_STEPS=2
-export FASTVIDEO_TORCH_PROFILER_WARMUP_STEPS=1
-export FASTVIDEO_TORCH_PROFILER_ACTIVE_STEPS=1
-export FASTVIDEO_TORCH_PROFILER_DIR="../profiler_traces/wan_t2v_finetune/"
-
-# Training arguments
-training_args=(
-  --tracker_project_name "wan_t2v_finetune"
-  --output_dir "checkpoints/wan_t2v_finetune"
-  --max_train_steps 5000
-  --train_batch_size 1
-  --train_sp_batch_size 1
-  --gradient_accumulation_steps 8
-  --num_latent_t 20
-  --num_height 480
-  --num_width 832
-  --num_frames 77
-  --enable_gradient_checkpointing_type "full"
-)
-
-# Parallel arguments
-parallel_args=(
-  --num_gpus $NUM_GPUS 
-  --sp_size $NUM_GPUS 
-  --tp_size 1
-  --hsdp_replicate_dim 1
-  --hsdp_shard_dim $NUM_GPUS
-)
-
-# Model arguments
-model_args=(
-  --model_path $MODEL_PATH
-  --pretrained_model_name_or_path $MODEL_PATH
-)
-
-# Dataset arguments
-dataset_args=(
-  --data_path $DATA_DIR
-  --dataloader_num_workers 1
-)
-
-# Validation arguments
-validation_args=(
-  --log_validation 
-  --validation_dataset_file $VALIDATION_DATASET_FILE
-  --validation_steps 200
-  --validation_sampling_steps "50" 
-  --validation_guidance_scale "6.0"
-)
-
-# Optimizer arguments
-optimizer_args=(
-  --learning_rate 5e-5
-  --mixed_precision "bf16"
-  --weight_only_checkpointing_steps 1000
-  --training_state_checkpointing_steps 1000
-  --weight_decay 1e-4
-  --max_grad_norm 1.0
-)
-
-# Miscellaneous arguments
-miscellaneous_args=(
-  --inference_mode False
-  --checkpoints_total_limit 3
-  --training_cfg_rate 0.1
-  --multi_phased_distill_schedule "4000-1"
-  --not_apply_cfg_solver
-  --dit_precision "fp32"
-  --num_euler_timesteps 50
-  --ema_start_step 0
-  --enable_gradient_checkpointing_type "full"
-  # --resume_from_checkpoint "checkpoints/wan_t2v_finetune/checkpoint-2500"
-)
-
-torchrun \
-  --nnodes 1 \
-  --nproc_per_node $NUM_GPUS \
-  --master_port 29502 \
-    fastvideo/training/wan_training_pipeline.py \
-    "${parallel_args[@]}" \
-    "${model_args[@]}" \
-    "${dataset_args[@]}" \
-    "${training_args[@]}" \
-    "${optimizer_args[@]}" \
-    "${validation_args[@]}" \
-    "${miscellaneous_args[@]}"
@@ -3,4 +3,4 @@ from fastvideo.configs.sample import SamplingParam
 from fastvideo.entrypoints.video_generator import VideoGenerator
 from fastvideo.version import __version__

-__all__ = ["VideoGenerator", "PipelineConfig", "SamplingParam", "__version__"]
+__all__ = ["VideoGenerator", "PipelineConfig", "SamplingParam", "__version__"]
@@ -1,9 +1,8 @@
 # SPDX-License-Identifier: Apache-2.0

 import torch
-import torch.nn.functional as F
 from flash_attn import flash_attn_func as flash_attn_2_func
-from dataclasses import dataclass
+
 try:
    from flash_attn_interface import flash_attn_func as flash_attn_3_func

@@ -47,29 +46,6 @@ class FlashAttentionBackend(AttentionBackend):
        raise NotImplementedError


-@dataclass
-class FlashAttnMetadata(AttentionMetadata):
-    current_timestep: int
-    attn_mask: torch.Tensor | None = None
-
-
-class FlashAttnMetadataBuilder(AttentionMetadataBuilder):
-
-    def __init__(self):
-        pass
-
-    def prepare(self):
-        pass
-
-    def build(  # type: ignore
-        self,
-        current_timestep: int,
-        attn_mask: torch.Tensor,
-    ) -> FlashAttnMetadata:
-        return FlashAttnMetadata(current_timestep=current_timestep,
-                                 attn_mask=attn_mask)
-
-
 class FlashAttentionImpl(AttentionImpl):

    def __init__(
@@ -90,27 +66,12 @@ class FlashAttentionImpl(AttentionImpl):
        query: torch.Tensor,
        key: torch.Tensor,
        value: torch.Tensor,
-        attn_metadata: FlashAttnMetadata,
+        attn_metadata: AttentionMetadata,
    ):
-        if attn_metadata is not None and hasattr(
-                attn_metadata,
-                "attn_mask") and attn_metadata.attn_mask is not None:
-            from fastvideo.attention.utils.flash_attn_no_pad import flash_attn_no_pad
-            attn_mask = attn_metadata.attn_mask
-            qkv = torch.stack([query, key, value], dim=2)
-
-            attn_mask = F.pad(attn_mask, (qkv.shape[1] - attn_mask.shape[1], 0),
-                              value=True)
-            output = flash_attn_no_pad(qkv,
-                                       attn_mask,
-                                       causal=False,
-                                       dropout_p=0,
-                                       softmax_scale=None)
-        else:
-            output = flash_attn_func(
-                query,  # type: ignore[no-untyped-call]
-                key,
-                value,
-                softmax_scale=self.softmax_scale,
-                causal=self.causal)
+        output = flash_attn_func(
+            query,  # type: ignore[no-untyped-call]
+            key,
+            value,
+            softmax_scale=self.softmax_scale,
+            causal=self.causal)
        return output
@@ -1,10 +1,9 @@
 # SPDX-License-Identifier: Apache-2.0

 import torch
-from dataclasses import dataclass
+
 from fastvideo.attention.backends.abstract import (  # FlashAttentionMetadata,
-    AttentionBackend, AttentionImpl, AttentionMetadata,
-    AttentionMetadataBuilder)
+    AttentionBackend, AttentionImpl, AttentionMetadata)
 from fastvideo.logger import init_logger

 logger = init_logger(__name__)
@@ -31,29 +30,6 @@ class SDPABackend(AttentionBackend):
    #     return FlashAttentionMetadata


-@dataclass
-class SDPAMetadata(AttentionMetadata):
-    current_timestep: int
-    attn_mask: torch.Tensor | None = None
-
-
-class SDPAMetadataBuilder(AttentionMetadataBuilder):
-
-    def __init__(self):
-        pass
-
-    def prepare(self):
-        pass
-
-    def build(  # type: ignore
-        self,
-        current_timestep: int,
-        attn_mask: torch.Tensor,
-    ) -> SDPAMetadata:
-        return SDPAMetadata(current_timestep=current_timestep,
-                            attn_mask=attn_mask)
-
-
 class SDPAImpl(AttentionImpl):

    def __init__(
@@ -75,15 +51,14 @@ class SDPAImpl(AttentionImpl):
        query: torch.Tensor,
        key: torch.Tensor,
        value: torch.Tensor,
-        attn_metadata: SDPAMetadata,
+        attn_metadata: AttentionMetadata,
    ) -> torch.Tensor:
        # transpose to bs, heads, seq_len, head_dim
        query = query.transpose(1, 2)
        key = key.transpose(1, 2)
        value = value.transpose(1, 2)
-        attn_mask = attn_metadata.attn_mask if attn_metadata is not None else None
        attn_kwargs = {
-            "attn_mask": attn_mask,
+            "attn_mask": None,
            "dropout_p": self.dropout,
            "is_causal": self.causal,
            "scale": self.softmax_scale
@@ -11,7 +11,6 @@ from fastvideo.distributed.parallel_state import (get_sp_parallel_rank,
 from fastvideo.forward_context import ForwardContext, get_forward_context
 from fastvideo.platforms import AttentionBackendEnum
 from fastvideo.utils import get_compute_dtype
-from fastvideo.layers.rotary_embedding import _apply_rotary_emb


 class DistributedAttention(nn.Module):
@@ -65,8 +64,6 @@ class DistributedAttention(nn.Module):
        replicated_q: torch.Tensor | None = None,
        replicated_k: torch.Tensor | None = None,
        replicated_v: torch.Tensor | None = None,
-        freqs_cis: tuple[torch.Tensor, torch.Tensor] | None = None,
-        attention_mask: torch.Tensor | None = None,
    ) -> tuple[torch.Tensor, torch.Tensor | None]:
        """Forward pass for distributed attention.
        
@@ -77,7 +74,6 @@ class DistributedAttention(nn.Module):
            replicated_q (Optional[torch.Tensor]): Replicated query tensor, typically for text tokens
            replicated_k (Optional[torch.Tensor]): Replicated key tensor
            replicated_v (Optional[torch.Tensor]): Replicated value tensor
-            attention_mask (Optional[torch.Tensor]): Attention mask [batch_size, seq_len]
            
        Returns:
            Tuple[torch.Tensor, Optional[torch.Tensor]]: A tuple containing:
@@ -95,30 +91,12 @@ class DistributedAttention(nn.Module):
        ctx_attn_metadata = forward_context.attn_metadata

        # Stack QKV
-        qkv = torch.cat([q, k, v],
-                        dim=0)  # [3*batch, seq_len, num_heads, head_dim]
+        qkv = torch.cat([q, k, v], dim=0)  # [3, seq_len, num_heads, head_dim]

        # Redistribute heads across sequence dimension
        qkv = sequence_model_parallel_all_to_all_4D(qkv,
                                                    scatter_dim=2,
                                                    gather_dim=1)
-
-        # After all-to-all, each rank has the full sequence but only a subset of heads
-        # The attention mask should now apply to the full sequence length
-        # Since mask is [batch, full_seq_len], it's already in the correct format
-
-        # LOAY TODO, instead of slicing repeatedly maintain an original qkv and rewrite into that
-        valid_seq_len = None
-        if attention_mask is not None:
-            valid_seq_len = (attention_mask[0] == 1).sum().item()
-            qkv = qkv[:, :valid_seq_len, :, :]
-
-        if freqs_cis is not None:
-            cos, sin = freqs_cis
-            qkv[:batch_size * 2] = _apply_rotary_emb(qkv[:batch_size * 2],
-                                                     cos,
-                                                     sin,
-                                                     is_neox_style=False)
        # Apply backend-specific preprocess_qkv
        qkv = self.attn_impl.preprocess_qkv(qkv, ctx_attn_metadata)

@@ -141,23 +119,17 @@ class DistributedAttention(nn.Module):
        # Redistribute back if using sequence parallelism
        replicated_output = None
        if replicated_q is not None:
-            split_idx = seq_len * world_size if valid_seq_len is None else valid_seq_len
-            replicated_output = output[:, split_idx:]
-            output = output[:, :split_idx]
+            replicated_output = output[:, seq_len * world_size:]
+            output = output[:, :seq_len * world_size]
            # TODO: make this asynchronous
            replicated_output = sequence_model_parallel_all_gather(
                replicated_output.contiguous(), dim=2)
        # Apply backend-specific postprocess_output
        output = self.attn_impl.postprocess_output(output, ctx_attn_metadata)

-        if attention_mask is not None:
-            pad_len = (attention_mask[0] == 0).sum().item()
-            output = torch.nn.functional.pad(output, (0, 0, 0, 0, 0, pad_len))
-
        output = sequence_model_parallel_all_to_all_4D(output,
                                                       scatter_dim=1,
                                                       gather_dim=2)
-
        return output, replicated_output


@@ -175,8 +147,6 @@ class DistributedAttention_VSA(DistributedAttention):
        replicated_k: torch.Tensor | None = None,
        replicated_v: torch.Tensor | None = None,
        gate_compress: torch.Tensor | None = None,
-        freqs_cis: tuple[torch.Tensor, torch.Tensor] | None = None,
-        attention_mask: torch.Tensor | None = None,
    ) -> tuple[torch.Tensor, torch.Tensor | None]:
        """Forward pass for distributed attention.
        
@@ -188,7 +158,6 @@ class DistributedAttention_VSA(DistributedAttention):
            replicated_q (Optional[torch.Tensor]): Replicated query tensor, typically for text tokens
            replicated_k (Optional[torch.Tensor]): Replicated key tensor
            replicated_v (Optional[torch.Tensor]): Replicated value tensor
-            attention_mask (Optional[torch.Tensor]): Attention mask [batch_size, seq_len]
            
        Returns:
            Tuple[torch.Tensor, Optional[torch.Tensor]]: A tuple containing:
@@ -204,32 +173,15 @@ class DistributedAttention_VSA(DistributedAttention):
        forward_context: ForwardContext = get_forward_context()
        ctx_attn_metadata = forward_context.attn_metadata

-        batch_size, seq_len, num_heads, head_dim = q.shape
        # Stack QKV
        qkvg = torch.cat([q, k, v, gate_compress],
-                         dim=0)  # [4*batch, seq_len, num_heads, head_dim]
+                         dim=0)  # [3, seq_len, num_heads, head_dim]

        # Redistribute heads across sequence dimension
-        # Before: [4*batch, shard_seq_len, num_heads, head_dim]
-        # After:  [4*batch, full_seq_len, shard_num_heads, head_dim]
        qkvg = sequence_model_parallel_all_to_all_4D(qkvg,
                                                     scatter_dim=2,
                                                     gather_dim=1)

-        # After all-to-all, each rank has the full sequence but only a subset of heads
-        # The attention mask should now apply to the full sequence length
-
-        if attention_mask is not None:
-            valid_seq_len = (attention_mask[0] == 1).sum().item()
-            qkvg = qkvg[:, :valid_seq_len, :, :]
-
-        if freqs_cis is not None:
-            cos, sin = freqs_cis
-            qkvg[:batch_size * 2] = _apply_rotary_emb(qkvg[:batch_size * 2],
-                                                      cos,
-                                                      sin,
-                                                      is_neox_style=False)
-
        qkvg = self.attn_impl.preprocess_qkv(qkvg, ctx_attn_metadata)

        q, k, v, gate_compress = qkvg.chunk(4, dim=0)
@@ -242,10 +194,6 @@ class DistributedAttention_VSA(DistributedAttention):
        # Apply backend-specific postprocess_output
        output = self.attn_impl.postprocess_output(output, ctx_attn_metadata)

-        if attention_mask is not None:
-            pad_len = (attention_mask[0] == 0).sum().item()
-            output = torch.nn.functional.pad(output, (0, 0, 0, 0, 0, pad_len))
-
        output = sequence_model_parallel_all_to_all_4D(output,
                                                       scatter_dim=1,
                                                       gather_dim=2)
@@ -296,7 +244,6 @@ class LocalAttention(nn.Module):
        q: torch.Tensor,
        k: torch.Tensor,
        v: torch.Tensor,
-        freqs_cis: tuple[torch.Tensor, torch.Tensor] | None = None,
    ) -> torch.Tensor:
        """
        Apply local attention between query, key and value tensors.
@@ -316,10 +263,5 @@ class LocalAttention(nn.Module):
        forward_context: ForwardContext = get_forward_context()
        ctx_attn_metadata = forward_context.attn_metadata

-        if freqs_cis is not None:
-            cos, sin = freqs_cis
-            q = _apply_rotary_emb(q, cos, sin, is_neox_style=False)
-            k = _apply_rotary_emb(k, cos, sin, is_neox_style=False)
-
        output = self.attn_impl.forward(q, k, v, ctx_attn_metadata)
        return output
@@ -1,99 +0,0 @@
-# Licensed under the TENCENT HUNYUAN COMMUNITY LICENSE AGREEMENT (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     https://github.com/Tencent-Hunyuan/HunyuanVideo-1.5/blob/main/LICENSE
-#
-# Unless and only to the extent required by applicable law, the Tencent Hunyuan works and any
-# output and results there from are provided "AS IS" without any express or implied warranties of
-# any kind including any warranties of title, merchantability, noninfringement, course of dealing,
-# usage of trade, or fitness for a particular purpose. You are solely responsible for determining the
-# appropriateness of using, reproducing, modifying, performing, displaying or distributing any of
-# the Tencent Hunyuan works or outputs and assume any and all risks associated with your or a
-# third party's use or distribution of any of the Tencent Hunyuan works or outputs and your exercise
-# of rights and permissions under this agreement.
-# See the License for the specific language governing permissions and limitations under the License.
-
-from einops import rearrange
-
-
-def flash_attn_no_pad(qkv,
-                      key_padding_mask,
-                      causal=False,
-                      dropout_p=0.0,
-                      softmax_scale=None,
-                      deterministic=False):
-    from flash_attn import flash_attn_varlen_qkvpacked_func
-    from flash_attn.bert_padding import pad_input, unpad_input
-    batch_size = qkv.shape[0]
-    seqlen = qkv.shape[1]
-    nheads = qkv.shape[-2]
-    x = rearrange(qkv, "b s three h d -> b s (three h d)")
-    x_unpad, indices, cu_seqlens, max_s, used_seqlens_in_batch = unpad_input(
-        x, key_padding_mask)
-
-    x_unpad = rearrange(x_unpad,
-                        "nnz (three h d) -> nnz three h d",
-                        three=3,
-                        h=nheads)
-    output_unpad = flash_attn_varlen_qkvpacked_func(
-        x_unpad,
-        cu_seqlens,
-        max_s,
-        dropout_p,
-        softmax_scale=softmax_scale,
-        causal=causal,
-        deterministic=deterministic,
-    )
-    output = rearrange(
-        pad_input(rearrange(output_unpad, "nnz h d -> nnz (h d)"), indices,
-                  batch_size, seqlen),
-        "b s (h d) -> b s h d",
-        h=nheads,
-    )
-    return output
-
-
-def flash_attn_no_pad_v3(qkv,
-                         key_padding_mask,
-                         causal=False,
-                         dropout_p=0.0,
-                         softmax_scale=None,
-                         deterministic=False):
-    from flash_attn.bert_padding import pad_input, unpad_input
-    from flash_attn_interface import flash_attn_varlen_func as flash_attn_varlen_func_v3
-
-    if flash_attn_varlen_func_v3 is None:
-        raise ImportError("FlashAttention V3 backend not available")
-
-    batch_size, seqlen, _, nheads, head_dim = qkv.shape
-    query, key, value = qkv.unbind(dim=2)
-
-    query_unpad, indices, cu_seqlens_q, max_seqlen_q, _ = unpad_input(
-        rearrange(query, "b s h d -> b s (h d)"), key_padding_mask)
-    key_unpad, _, cu_seqlens_k, _, _ = unpad_input(
-        rearrange(key, "b s h d -> b s (h d)"), key_padding_mask)
-    value_unpad, _, _, _, _ = unpad_input(
-        rearrange(value, "b s h d -> b s (h d)"), key_padding_mask)
-
-    query_unpad = rearrange(query_unpad, "nnz (h d) -> nnz h d", h=nheads)
-    key_unpad = rearrange(key_unpad, "nnz (h d) -> nnz h d", h=nheads)
-    value_unpad = rearrange(value_unpad, "nnz (h d) -> nnz h d", h=nheads)
-
-    output_unpad = flash_attn_varlen_func_v3(query_unpad,
-                                             key_unpad,
-                                             value_unpad,
-                                             cu_seqlens_q,
-                                             cu_seqlens_k,
-                                             max_seqlen_q,
-                                             max_seqlen_q,
-                                             softmax_scale=softmax_scale,
-                                             causal=causal,
-                                             deterministic=deterministic)
-
-    output = rearrange(pad_input(
-        rearrange(output_unpad, "nnz h d -> nnz (h d)"), indices, batch_size,
-        seqlen),
-                       "b s (h d) -> b s h d",
-                       h=nheads)
-    return output
@@ -1,11 +1,5 @@
-from fastvideo.configs.models.dits.cosmos import CosmosVideoConfig
-from fastvideo.configs.models.dits.cosmos2_5 import Cosmos25VideoConfig
 from fastvideo.configs.models.dits.hunyuanvideo import HunyuanVideoConfig
-from fastvideo.configs.models.dits.hunyuanvideo15 import HunyuanVideo15Config
 from fastvideo.configs.models.dits.stepvideo import StepVideoConfig
 from fastvideo.configs.models.dits.wanvideo import WanVideoConfig

-__all__ = [
-    "HunyuanVideoConfig", "HunyuanVideo15Config", "WanVideoConfig",
-    "StepVideoConfig", "CosmosVideoConfig", "Cosmos25VideoConfig"
-]
+__all__ = ["HunyuanVideoConfig", "WanVideoConfig", "StepVideoConfig"]
--- a/Show More
+++ b/Show More
Author	SHA1	Message	Date
SolitaryThinker	c4a0f789da	wip	2025-10-31 20:54:49 +00:00
SolitaryThinker	5ac14938d2	add example	2025-10-31 20:54:00 +00:00
SolitaryThinker	9d239e9f8b	prepare for wan2.2 self-forcing	2025-10-31 20:52:31 +00:00