fix tensor type hint

.buildkite/pipeline.yml

@@ -1,224 +0,0 @@
-env:
-  IMAGE_VERSION: "py3.12-latest"
-  BUILDKITE_CLEAN_CHECKOUT: true
-
-steps:
-  - label: "pre-commit"
-    command: ".buildkite/scripts/pre_commit.sh"
-    agents:
-      queue: "default"
-
-  - wait
-
-  - label: "Trigger Tests"
-    plugins:
-      - monorepo-diff#v1.4.0:
-          diff: 'git fetch origin "$BUILDKITE_PULL_REQUEST_BASE_BRANCH" && git diff --name-only origin/"$BUILDKITE_PULL_REQUEST_BASE_BRANCH"...HEAD'
-          watch:
-            - path: 
-                - "fastvideo/models/encoders/**"
-                - "fastvideo/models/loader/**"
-                - "fastvideo/tests/encoders/**"
-                - "pyproject.toml"
-                - "docker/Dockerfile.python3.12"
-              config:
-                command: "timeout 15m .buildkite/scripts/pr_test.sh"
-                label: "Encoder Tests"
-                env:
-                  - TEST_TYPE=encoder
-                agents:
-                  queue: "default"
-            - path:
-                - "fastvideo/models/vaes/**"
-                - "fastvideo/models/loader/**"
-                - "fastvideo/tests/vaes/**"
-                - "pyproject.toml"
-                - "docker/Dockerfile.python3.12"
-              config:
-                command: "timeout 15m .buildkite/scripts/pr_test.sh"
-                label: "VAE Tests"
-                env:
-                  - TEST_TYPE=vae
-                agents:
-                  queue: "default"
-            - path:
-                - "fastvideo/models/dits/**"
-                - "fastvideo/models/loader/**"
-                - "fastvideo/tests/transformers/**"
-                - "fastvideo/layers/**"
-                - "fastvideo/attention/**"
-                - "pyproject.toml"
-                - "docker/Dockerfile.python3.12"
-              config:
-                command: "timeout 15m .buildkite/scripts/pr_test.sh"
-                label: "Transformer Tests"
-                env:
-                  - TEST_TYPE=transformer
-                agents:
-                  queue: "default"
-            - path: 
-                - "fastvideo/**/*.py"
-                - "pyproject.toml"
-                - "docker/Dockerfile.python3.12"
-              config:
-                command: "timeout 45m .buildkite/scripts/pr_test.sh"
-                label: "SSIM Tests"
-                env:
-                  - TEST_TYPE=ssim
-                agents:
-                  queue: "default"
-            - path:
-                - "fastvideo/tests/lora/**"
-                - "fastvideo/models/loader/**"
-                - "fastvideo/tests/transformers/**"
-                - "fastvideo/pipelines/**"
-                - "fastvideo/layers/lora/**"
-                - "pyproject.toml"
-                - "docker/Dockerfile.python3.12"
-              config:
-                command: "timeout 15m .buildkite/scripts/pr_test.sh"
-                label: "LoRA Inference Tests"
-                env:
-                  - TEST_TYPE=inference_lora
-                agents:
-                  queue: "default"
-            - path:
-                - "fastvideo/**"
-                - "pyproject.toml"
-                - "docker/Dockerfile.python3.12"
-              config:
-                command: "timeout 15m .buildkite/scripts/pr_test.sh"
-                label: "Training Tests"
-                env:
-                  - TEST_TYPE=training
-                agents:
-                  queue: "default"
-            - path:
-                - "fastvideo/training/*distillation_pipeline.py"
-                - "pyproject.toml"
-                - "docker/Dockerfile.python3.12"
-              config:
-                command: "timeout 15m .buildkite/scripts/pr_test.sh"
-                label: "Distillation DMDTests"
-                env:
-                  - TEST_TYPE=distillation_dmd
-                agents:
-                  queue: "default"
-            - path:
-                - "fastvideo/training/*self_forcing_distillation_pipeline.py"
-                - "fastvideo/tests/training/self-forcing/**"
-                - "pyproject.toml"
-                - "docker/Dockerfile.python3.12"
-              config:
-                command: "timeout 15m .buildkite/scripts/pr_test.sh"
-                label: "Self-Forcing Tests"
-                env:
-                  - TEST_TYPE=self_forcing
-                agents:
-                  queue: "default"
-            - path:
-                - "fastvideo/**"
-                - "pyproject.toml"
-                - "docker/Dockerfile.python3.12"
-              config:
-                command: "timeout 15m .buildkite/scripts/pr_test.sh"
-                label: "LoRA Training Tests"
-                env:
-                  - TEST_TYPE=training_lora
-                agents:
-                  queue: "default"
-            - path:
-                - "fastvideo/**"
-                - "csrc/attn/video_sparse_attn/**"
-                - "csrc/attn/video_sparse_attn/tk/**"
-                - "csrc/attn/video_sparse_attn/setup.py"
-                - "csrc/attn/video_sparse_attn/config_vsa.py"
-                - "csrc/attn/video_sparse_attn/vsa.cpp"
-                - "pyproject.toml"
-                - "docker/Dockerfile.python3.12"
-              config:
-                command: "timeout 15m .buildkite/scripts/pr_test.sh"
-                label: "Training Tests VSA"
-                env:
-                  - TEST_TYPE=training_vsa
-                agents:
-                  queue: "default"
-            - path:
-                - "fastvideo/**"
-                - "csrc/attn/sliding_tile_attn/**"
-                - "csrc/attn/sliding_tile_attn/setup.py"
-                - "csrc/attn/sliding_tile_attn/config_sta.py"
-                - "csrc/attn/sliding_tile_attn/st_attn.cpp"
-                - "pyproject.toml"
-                - "docker/Dockerfile.python3.12"
-              config:
-                command: "timeout 15m .buildkite/scripts/pr_test.sh"
-                label: "Inference Tests STA"
-                env:
-                  - TEST_TYPE=inference_sta
-                agents:
-                  queue: "default"
-            - path:
-                - "csrc/attn/sliding_tile_attn/**"
-                - "csrc/attn/sliding_tile_attn/setup.py"
-                - "csrc/attn/sliding_tile_attn/config_sta.py"
-                - "csrc/attn/sliding_tile_attn/st_attn.cpp"
-                - "pyproject.toml"
-                - "docker/Dockerfile.python3.12"
-              config:
-                command: "timeout 15m .buildkite/scripts/pr_test.sh"
-                label: "Precision Tests STA"
-                env:
-                  - TEST_TYPE=precision_sta
-                agents:
-                  queue: "default"
-            - path:
-                - "csrc/attn/video_sparse_attn/**"
-                - "csrc/attn/video_sparse_attn/tk/**"
-                - "csrc/attn/tests/test_vsa.py"
-                - "csrc/attn/video_sparse_attn/setup.py"
-                - "csrc/attn/video_sparse_attn/config_vsa.py"
-                - "csrc/attn/video_sparse_attn/vsa.cpp"
-                - "pyproject.toml"
-                - "docker/Dockerfile.python3.12"
-              config:
-                command: "timeout 15m .buildkite/scripts/pr_test.sh"
-                label: "Precision Tests VSA"
-                env:
-                  - TEST_TYPE=precision_vsa
-                agents:
-                  queue: "default"
-            - path:
-                - "csrc/attn/vmoba_attn/**"
-                - "pyproject.toml"
-                - "docker/Dockerfile.python3.12"
-              config:
-                command: "timeout 15m .buildkite/scripts/pr_test.sh"
-                label: "Precision Tests VMoBA"
-                env:
-                  - TEST_TYPE=precision_vmoba
-                agents:
-                  queue: "default"
-            - path:
-                - "csrc/attn/vmoba_attn/vmoba/**"
-                - "fastvideo/attention/backends/vmoba.py"
-                - "pyproject.toml"
-                - "docker/Dockerfile.python3.12"
-              config:
-                command: "timeout 15m .buildkite/scripts/pr_test.sh"
-                label: "Inference Tests VMoBA"
-                env:  
-                  - TEST_TYPE=inference_vmoba
-                agents:
-                  queue: "default"
-            - path:
-                - "fastvideo/**"
-                - "pyproject.toml"
-                - "docker/Dockerfile.python3.12"
-              config:
-                command: "timeout 15m .buildkite/scripts/pr_test.sh"
-                label: "Unit Tests"
-                env:
-                  - TEST_TYPE=unit_test
-                agents:
-                  queue: "default"

.buildkite/scripts/pr_test.sh

@@ -1,146 +0,0 @@
-#!/bin/bash
-set -uo pipefail
-
-log() {
-  echo "[$(date '+%Y-%m-%d %H:%M:%S')] $1"
-}
-
-log "=== Starting Modal test execution ==="
-
-# Change to the project directory
-cd "$(dirname "$0")/../.."
-PROJECT_ROOT=$(pwd)
-log "Project root: $PROJECT_ROOT"
-
-# Install Modal if not available
-if ! python3 -m modal --version &> /dev/null; then
-    log "Modal not found, installing..."
-    python3 -m pip install modal
-    
-    # Verify installation
-    if ! python3 -m modal --version &> /dev/null; then
-        log "Error: Failed to install modal. Please install it manually."
-        exit 1
-    fi
-fi
-
-log "modal version: $(python3 -m modal --version)"
-
-# Set up Modal authentication using Buildkite secrets
-log "Setting up Modal authentication from Buildkite secrets..."
-MODAL_TOKEN_ID=$(buildkite-agent secret get modal_token_id)
-MODAL_TOKEN_SECRET=$(buildkite-agent secret get modal_token_secret)
-
-WANDB_API_KEY=$(buildkite-agent secret get wandb_api_key)
-
-WANDB_API_KEY=$(buildkite-agent secret get wandb_api_key)
-
-if [ -n "$MODAL_TOKEN_ID" ] && [ -n "$MODAL_TOKEN_SECRET" ]; then
-    log "Retrieved Modal credentials from Buildkite secrets"
-    python3 -m modal token set --token-id "$MODAL_TOKEN_ID" --token-secret "$MODAL_TOKEN_SECRET" --profile buildkite-ci --activate --verify
-    if [ $? -eq 0 ]; then
-        log "Modal authentication successful"
-    else
-        log "Error: Failed to set Modal credentials"
-        exit 1
-    fi
-else
-    log "Error: Could not retrieve Modal credentials from Buildkite secrets."
-    log "Please ensure 'modal_token_id' and 'modal_token_secret' secrets are set in Buildkite."
-    exit 1
-fi
-
-MODAL_TEST_FILE="fastvideo/tests/modal/pr_test.py"
-
-if [ -z "${TEST_TYPE:-}" ]; then
-    log "Error: TEST_TYPE environment variable is not set"
-    exit 1
-fi
-log "Test type: $TEST_TYPE"
-
-MODAL_ENV="BUILDKITE_REPO=$BUILDKITE_REPO BUILDKITE_COMMIT=$BUILDKITE_COMMIT BUILDKITE_PULL_REQUEST=$BUILDKITE_PULL_REQUEST IMAGE_VERSION=$IMAGE_VERSION"
-
-case "$TEST_TYPE" in
-    "encoder")
-        log "Running 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 python3 -m modal run $MODAL_TEST_FILE::run_vae_tests"
-        ;;
-    "transformer")
-        log "Running transformer tests..."
-        MODAL_COMMAND="$MODAL_ENV python3 -m modal run $MODAL_TEST_FILE::run_transformer_tests"
-        ;;
-    "ssim")
-        log "Running SSIM tests..."
-        MODAL_COMMAND="$MODAL_ENV python3 -m modal run $MODAL_TEST_FILE::run_ssim_tests"
-        ;;
-    "training")
-        log "Running training tests..."
-        MODAL_COMMAND="$MODAL_ENV WANDB_API_KEY=$WANDB_API_KEY python3 -m modal run $MODAL_TEST_FILE::run_training_tests"
-        ;;
-    "training_lora")
-        log "Running LoRA training tests..."
-        MODAL_COMMAND="$MODAL_ENV WANDB_API_KEY=$WANDB_API_KEY python3 -m modal run $MODAL_TEST_FILE::run_training_lora_tests"
-        ;;
-    "training_vsa")
-        log "Running training VSA tests..."
-        MODAL_COMMAND="$MODAL_ENV WANDB_API_KEY=$WANDB_API_KEY python3 -m modal run $MODAL_TEST_FILE::run_training_tests_VSA"
-        ;;
-    "inference_sta")
-        log "Running inference STA tests..."
-        MODAL_COMMAND="$MODAL_ENV python3 -m modal run $MODAL_TEST_FILE::run_inference_tests_STA"
-        ;;
-    "precision_sta")
-        log "Running precision STA tests..."
-        MODAL_COMMAND="$MODAL_ENV python3 -m modal run $MODAL_TEST_FILE::run_precision_tests_STA"
-        ;;
-    "precision_vsa")
-        log "Running precision VSA tests..."
-        MODAL_COMMAND="$MODAL_ENV python3 -m modal run $MODAL_TEST_FILE::run_precision_tests_VSA"
-        ;;
-    "inference_lora")
-        log "Running LoRA tests..."
-        MODAL_COMMAND="$MODAL_ENV python3 -m modal run $MODAL_TEST_FILE::run_inference_lora_tests"
-        ;;
-    "distillation_dmd")
-        log "Running distillation DMD tests..."
-        MODAL_COMMAND="$MODAL_ENV WANDB_API_KEY=$WANDB_API_KEY python3 -m modal run $MODAL_TEST_FILE::run_distill_dmd_tests"
-        ;;
-        # run_inference_tests_vmoba
-    "self_forcing")
-        log "Running self-forcing tests..."
-        MODAL_COMMAND="$MODAL_ENV WANDB_API_KEY=$WANDB_API_KEY python3 -m modal run $MODAL_TEST_FILE::run_self_forcing_tests"
-        ;;
-    "inference_vmoba")
-        log "Running V-MoBA inference tests..."
-        MODAL_COMMAND="$MODAL_ENV python3 -m modal run $MODAL_TEST_FILE::run_inference_tests_vmoba"
-        ;;
-    "precision_vmoba")
-        log "Running V-MoBA precision tests..."
-        MODAL_COMMAND="$MODAL_ENV python3 -m modal run $MODAL_TEST_FILE::run_precision_tests_vmoba"
-        ;;
-    "unit_test")
-        log "Running unit tests..."
-        MODAL_COMMAND="$MODAL_ENV python3 -m modal run $MODAL_TEST_FILE::run_unit_test"
-        ;;
-    *)
-        log "Error: Unknown test type: $TEST_TYPE"
-        exit 1
-        ;;
-esac
-
-log "Executing: $MODAL_COMMAND"
-eval "$MODAL_COMMAND"
-TEST_EXIT_CODE=$?
-
-if [ $TEST_EXIT_CODE -eq 0 ]; then
-    log "Modal test completed successfully"
-else
-    log "Error: Modal test failed with exit code: $TEST_EXIT_CODE"
-fi
-
-log "=== Test execution completed with exit code: $TEST_EXIT_CODE ==="
-exit $TEST_EXIT_CODE

.buildkite/scripts/pre_commit.sh

@@ -1,40 +0,0 @@
-#!/bin/bash
-set -uo pipefail
-
-log() {
-  echo "[$(date '+%Y-%m-%d %H:%M:%S')] $1"
-}
-
-log "=== Starting pre-commit checks ==="
-
-cd "$(dirname "$0")/../.."
-PROJECT_ROOT=$(pwd)
-log "Project root: $PROJECT_ROOT"
-
-if ! python3 -m pre_commit --version &> /dev/null; then
-    log "pre-commit not found, installing..."
-    python3 -m pip install --user pre-commit==4.0.1
-    
-    if ! python3 -m pre_commit --version &> /dev/null; then
-        log "Error: Failed to install pre-commit."
-        exit 1
-    fi
-fi
-
-log "Pre-commit version: $(python3 -m pre_commit --version)"
-
-log "Installing/updating pre-commit hooks..."
-python3 -m pre_commit install --install-hooks
-
-log "Running pre-commit checks on all files..."
-python3 -m pre_commit run --all-files
-PRE_COMMIT_EXIT_CODE=$?
-
-if [ $PRE_COMMIT_EXIT_CODE -eq 0 ]; then
-    log "Pre-commit checks completed successfully"
-else
-    log "Error: Pre-commit checks failed with exit code: $PRE_COMMIT_EXIT_CODE"
-fi
-
-log "=== Pre-commit checks completed with exit code: $PRE_COMMIT_EXIT_CODE ==="
-exit $PRE_COMMIT_EXIT_CODE 

.github/ISSUE_TEMPLATE/1-bug-report.yml

@@ -4,6 +4,14 @@ title: "[Bug] "
 labels: ['Bug']
 
 body:
+- type: textarea
+  attributes:
+    label: Environment
+    description: |
+      Please share your environment with us. You can run the command **python fastvideo/utils/collect_env.py** and copy-paste its output below.
+    placeholder: FastVideo version, platform, python version, cuda version...
+  validations:
+    required: true
 - type: textarea
   attributes:
     label: Describe the bug
@@ -17,13 +25,5 @@ body:
       What command or script did you run? Which **model** are you using?
     placeholder: |
       A placeholder for the command.
-  validations:
-    required: true
-- type: textarea
-  attributes:
-    label: Environment
-    description: |
-      Please share your environment with us. You can run the command **python collect_env.py** and copy-paste its output below.
-    placeholder: FastVideo version, platform, python version, cuda version...
   validations:
     required: true

.github/ISSUE_TEMPLATE/3-RFC.yml

@@ -1,56 +0,0 @@
-name: 💬 Request for comments (RFC).
-description: Ask for feedback on major architectural changes or design choices.
-title: "[RFC]: "
-labels: ["RFC"]
-
-body:
-- type: markdown
-  attributes:
-    value: >
-      #### Please take a look at previous [RFCs](https://github.com/hao-ai-lab/FastVideo/issues?q=label%3ARFC+sort%3Aupdated-desc) for reference.
-- type: textarea
-  attributes:
-    label: Motivation.
-    description: >
-      The motivation of the RFC.
-  validations:
-    required: true
-- type: textarea
-  attributes:
-    label: Proposed Change.
-    description: >
-      The proposed change of the RFC.
-  validations:
-    required: true
-- type: textarea
-  attributes:
-    label: Feedback Period.
-    description: >
-      The feedback period of the RFC. Usually at least one week.
-  validations:
-    required: false
-- type: textarea
-  attributes:
-    label: CC List.
-    description: >
-      The list of people you want to CC.
-  validations:
-    required: false
-- type: textarea
-  attributes:
-    label: Any Other Things.
-    description: >
-      Any other things you would like to mention.
-  validations:
-    required: false
-- type: markdown
-  attributes:
-    value: >
-      Thanks for contributing 🎉!
-- type: checkboxes
-  id: askllm
-  attributes:
-    label: Before submitting a new issue...
-    options:
-      - label: Make sure you already searched for relevant issues.
-        required: true

.github/scripts/runpod_api.py

@@ -160,7 +160,8 @@ def execute_command(pod_id):
     setup_steps = [
         "tar -xzf /tmp/repo.tar.gz --no-same-owner -C /workspace/",
         f"cd /workspace/{repo_name}",
-        "source $HOME/.local/bin/env && source /opt/venv/bin/activate",
+        "source /opt/conda/etc/profile.d/conda.sh",
+        "conda activate fastvideo-dev",
         args.test_command
     ]
     

.github/workflows/build-image.yml

@@ -18,12 +18,6 @@ on:
         required: false
         default: false
         type: boolean
-      python_3_12_cuda_12_9:
-        description: 'Build Python 3.12 image Cuda 12.9'
-        required: false
-        default: false
-        type: boolean
-
 
 permissions:
   contents: read
@@ -55,13 +49,4 @@ jobs:
       python_version: '3.12'
       dockerfile_path: docker/Dockerfile.python3.12
       tag_suffix: py3.12
-    secrets: inherit
-
-  build-python-3-12-cuda-12-9:
-    if: ${{ github.event.inputs.python_3_12_cuda_12_9 == 'true' }}
-    uses: ./.github/workflows/build-image-template.yml
-    with:
-      python_version: '3.12'
-      dockerfile_path: docker/Dockerfile.python3.12.cuda12.9.1
-      tag_suffix: py3.12-cuda12.9.1
     secrets: inherit

.github/workflows/docs.yml

@@ -8,14 +8,14 @@ on:
       - main
     paths:
       - "docs/**/*.md"
-      - "fastvideo/examples/**/*.py"
+      - "fastvideo/v1/examples/**/*.py"
   pull_request:
     branches: 
       - main
     types: [opened, ready_for_review, synchronize, reopened]
     paths:
       - "docs/**/*.md"
-      - "fastvideo/examples/**/*.py"
+      - "fastvideo/v1/examples/**/*.py"
 
   # Allows you to run this workflow manually from the Actions tab
   workflow_dispatch:

.github/workflows/matchers/mypy.json

@@ -13,4 +13,4 @@
       ]
     }
   ]
-}
+}

.github/workflows/pr-test.yml

@@ -12,11 +12,13 @@ on:
     paths:
       - "fastvideo/**/*.py"
       - ".github/workflows/pr-test.yml"
-      - "pyproject.toml"
-      - "docker/Dockerfile.python3.12"
-      - "csrc/**"
   workflow_dispatch:
     inputs:
+      custom_image:
+        description: "Custom image from this repository (default: fastvideo-dev:latest)"
+        required: false
+        default: "fastvideo-dev:latest"
+        type: string
       run_encoder_test:
         description: "Run encoder-test"
         required: false
@@ -37,41 +39,10 @@ on:
         required: false
         default: false
         type: boolean
-      run_training_test:
-        description: "Run training-test"
-        required: false
-        default: false
-        type: boolean
-      run_training_test_VSA:
-        description: "Run training-test-VSA"
-        required: false
-        default: false
-        type: boolean
-      run_inference_test_STA:
-        description: "Run inference-test-STA"
-        required: false
-        default: false
-        type: boolean
-      run_precision_test_STA:
-        description: "Run precision-test-STA"
-        required: false
-        default: false
-        type: boolean
-      run_precision_test_VSA:
-        description: "Run precision-test-VSA"
-        required: false
-        default: false
-        type: boolean
-      run_unit_test:
-        description: "Run unit-test"
-        required: false
-        default: false
-        type: boolean
 
 env:
   PYTHONUNBUFFERED: "1"
 
-
 concurrency:
   group: pr-test-${{ github.ref }}
   cancel-in-progress: true
@@ -88,79 +59,24 @@ jobs:
       encoder-test: ${{ steps.filter.outputs.encoder-test }}
       vae-test: ${{ steps.filter.outputs.vae-test }}
       transformer-test: ${{ steps.filter.outputs.transformer-test }}
-      training-test: ${{ steps.filter.outputs.training-test }}
-      training-test-VSA: ${{ steps.filter.outputs.training-test-VSA }}
-      inference-test-STA: ${{ steps.filter.outputs.inference-test-STA }}
-      precision-test-STA: ${{ steps.filter.outputs.precision-test-STA }}
-      precision-test-VSA: ${{ steps.filter.outputs.precision-test-VSA }}
-      unit-test: ${{ steps.filter.outputs.unit-test }}
     steps:
       - uses: actions/checkout@v4
       - uses: dorny/paths-filter@v3
         id: filter
         with:
           filters: |
-            # Define reusable path patterns
-            common-paths: &common-paths
-              - 'pyproject.toml'
-              - 'docker/Dockerfile.python3.10'
-              - 'docker/Dockerfile.python3.11'
-              - 'docker/Dockerfile.python3.12'
-            sta-kernel-paths: &sta-kernel-paths
-              - 'csrc/attn/sliding_tile_attn/**'
-              - 'csrc/attn/sliding_tile_attn/tk/**'
-              - 'csrc/attn/sliding_tile_attn/setup.py'
-              - 'csrc/attn/sliding_tile_attn/config_sta.py'
-              - 'csrc/attn/sliding_tile_attn/st_attn.cpp'
-            vsa-kernel-paths: &vsa-kernel-paths
-              - 'csrc/attn/video_sparse_attn/**'
-              - 'csrc/attn/video_sparse_attn/tk/**'
-              - 'csrc/attn/video_sparse_attn/setup.py'
-              - 'csrc/attn/video_sparse_attn/config_vsa.py'
-              - 'csrc/attn/video_sparse_attn/vsa.cpp'
-            vsa-paths: &vsa-paths
-              - 'fastvideo/**'
-              - *common-paths
-              - *vsa-kernel-paths
-            
-            # Actual tests
             encoder-test:
-              - 'fastvideo/models/encoders/**'
-              - 'fastvideo/models/loader/**'
-              - 'fastvideo/tests/encoders/**'
-              - *common-paths
+              - 'fastvideo/v1/models/encoders/**'
+              - 'fastvideo/v1/models/loaders/**'
+              - 'fastvideo/v1/tests/encoders/**'
             vae-test:
-              - 'fastvideo/models/vaes/**'
-              - 'fastvideo/models/loader/**'
-              - 'fastvideo/tests/vaes/**'
-              - *common-paths
+              - 'fastvideo/v1/models/vaes/**'
+              - 'fastvideo/v1/models/loaders/**'
+              - 'fastvideo/v1/tests/vaes/**'
             transformer-test:
-              - 'fastvideo/models/dits/**'
-              - 'fastvideo/models/loader/**'
-              - 'fastvideo/tests/transformers/**'
-              - 'fastvideo/layers/**'
-              - 'fastvideo/attention/**'
-              - *common-paths
-            training-test:
-              - 'fastvideo/**'
-              - *common-paths
-            training-test-VSA:
-              - 'fastvideo/**'
-              - *common-paths
-              - *vsa-kernel-paths
-            inference-test-STA:
-              - 'fastvideo/**'
-              - *common-paths
-              - *sta-kernel-paths
-            precision-test-STA:
-              - *common-paths
-              - *sta-kernel-paths
-            precision-test-VSA:
-              - *common-paths
-              - *vsa-kernel-paths
-            unit-test:
-              - 'fastvideo/**'
-              - *common-paths
+              - 'fastvideo/v1/models/dits/**'
+              - 'fastvideo/v1/models/loaders/**'
+              - 'fastvideo/v1/tests/transformers/**'
 
   encoder-test:
     needs: change-filter
@@ -173,8 +89,8 @@ jobs:
       gpu_type: "NVIDIA A40"
       gpu_count: 1
       volume_size: 100
-      image: "ghcr.io/${{ github.repository }}/fastvideo-dev:py3.12-latest"
-      test_command: "uv pip install -e .[test] && pytest ./fastvideo/tests/encoders -s"
+      image: "ghcr.io/${{ github.repository }}/${{ github.event.inputs.custom_image || 'fastvideo-dev:latest' }}"
+      test_command: "pip install -e .[test] && pytest ./fastvideo/v1/tests/encoders -s"
       timeout_minutes: 30
     secrets:
       RUNPOD_API_KEY: ${{ secrets.RUNPOD_API_KEY }}
@@ -191,8 +107,8 @@ jobs:
       gpu_type: "NVIDIA A40"
       gpu_count: 1
       volume_size: 100
-      image: "ghcr.io/${{ github.repository }}/fastvideo-dev:py3.12-latest"
-      test_command: "uv pip install -e .[test] && pytest ./fastvideo/tests/vaes -s"
+      image: "ghcr.io/${{ github.repository }}/${{ github.event.inputs.custom_image || 'fastvideo-dev:latest' }}"
+      test_command: "pip install -e .[test] && pytest ./fastvideo/v1/tests/vaes -s"
       timeout_minutes: 30
     secrets:
       RUNPOD_API_KEY: ${{ secrets.RUNPOD_API_KEY }}
@@ -209,8 +125,8 @@ jobs:
       gpu_type: "NVIDIA L40S"
       gpu_count: 1
       volume_size: 100
-      image: "ghcr.io/${{ github.repository }}/fastvideo-dev:py3.12-latest"
-      test_command: "uv pip install -e .[test] && pytest ./fastvideo/tests/transformers -s"
+      image: "ghcr.io/${{ github.repository }}/${{ github.event.inputs.custom_image || 'fastvideo-dev:latest' }}"
+      test_command: "pip install -e .[test] && pytest ./fastvideo/v1/tests/transformers -s"
       timeout_minutes: 30
     secrets:
       RUNPOD_API_KEY: ${{ secrets.RUNPOD_API_KEY }}
@@ -219,7 +135,8 @@ jobs:
   ssim-test:
     needs: change-filter
     if: >-
-      github.event_name != 'workflow_dispatch' || (github.event_name == 'workflow_dispatch' && github.event.inputs.run_ssim_test == 'true')
+      (github.event_name != 'workflow_dispatch' && github.event.pull_request.draft == false) || 
+      (github.event_name == 'workflow_dispatch' && github.event.inputs.run_ssim_test == 'true')
     strategy:
       fail-fast: false
       matrix:
@@ -236,149 +153,14 @@ jobs:
       volume_size: 200
       disk_size: 200
       image: "ghcr.io/${{ github.repository }}/fastvideo-dev:${{ matrix.python-version.tag }}"
-      test_command: "uv pip install -e .[test] && pytest ./fastvideo/tests/ssim -vs"
+      test_command: "pip install -e .[test] && pytest ./fastvideo/v1/tests/ssim -vs"
       timeout_minutes: 60
     secrets:
       RUNPOD_API_KEY: ${{ secrets.RUNPOD_API_KEY }}
       RUNPOD_PRIVATE_KEY: ${{ secrets.RUNPOD_PRIVATE_KEY }}
 
-  training-test:
-    needs: change-filter
-    if: >-
-      (github.event_name != 'workflow_dispatch' && needs.change-filter.outputs.training-test == 'true') || 
-      (github.event_name == 'workflow_dispatch' && github.event.inputs.run_training_test == 'true')
-    uses: ./.github/workflows/runpod-test.yml
-    with:
-      job_id: "training-test"
-      gpu_type: "NVIDIA A40"
-      gpu_count: 4
-      volume_size: 100
-      disk_size: 100
-      image: "ghcr.io/${{ github.repository }}/fastvideo-dev:py3.12-latest"
-      test_command: "wandb login $WANDB_API_KEY && uv pip install -e .[test] && pytest ./fastvideo/tests/training/Vanilla -srP"
-      timeout_minutes: 30
-    secrets:
-      RUNPOD_API_KEY: ${{ secrets.RUNPOD_API_KEY }}
-      RUNPOD_PRIVATE_KEY: ${{ secrets.RUNPOD_PRIVATE_KEY }}
-      WANDB_API_KEY: ${{ secrets.WANDB_API_KEY }}
-
-  training-test-VSA:
-    needs: change-filter
-    if: >-
-      (github.event_name != 'workflow_dispatch' && needs.change-filter.outputs.training-test-VSA == 'true') || 
-      (github.event_name == 'workflow_dispatch' && github.event.inputs.run_training_test_VSA == 'true')
-    uses: ./.github/workflows/runpod-test.yml
-    with:
-      job_id: "training-test-VSA"
-      gpu_type: "NVIDIA H100 NVL"
-      gpu_count: 2
-      volume_size: 100
-      disk_size: 100
-      image: "ghcr.io/${{ github.repository }}/fastvideo-dev:py3.12-latest"
-      test_command: "wandb login $WANDB_API_KEY && uv pip install -e .[test] && pytest ./fastvideo/tests/training/VSA -srP"
-      timeout_minutes: 30
-    secrets:
-      RUNPOD_API_KEY: ${{ secrets.RUNPOD_API_KEY }}
-      RUNPOD_PRIVATE_KEY: ${{ secrets.RUNPOD_PRIVATE_KEY }}
-      WANDB_API_KEY: ${{ secrets.WANDB_API_KEY }}
-
-  inference-test-STA:
-    needs: change-filter
-    if: >-
-      (github.event_name != 'workflow_dispatch' && needs.change-filter.outputs.inference-test-STA == 'true') || 
-      (github.event_name == 'workflow_dispatch' && github.event.inputs.run_inference_test_STA == 'true')
-    uses: ./.github/workflows/runpod-test.yml
-    with:
-      job_id: "inference-test-STA"
-      gpu_type: "NVIDIA H100 NVL"
-      gpu_count: 2
-      volume_size: 100
-      disk_size: 100
-      image: "ghcr.io/${{ github.repository }}/fastvideo-dev:py3.12-latest"
-      test_command: "uv pip install -e .[test] && pytest ./fastvideo/tests/inference/STA -srP"
-      timeout_minutes: 30
-    secrets:
-      RUNPOD_API_KEY: ${{ secrets.RUNPOD_API_KEY }}
-      RUNPOD_PRIVATE_KEY: ${{ secrets.RUNPOD_PRIVATE_KEY }}
-
-  precision-test-STA:
-    needs: change-filter
-    if: >-
-      (github.event_name != 'workflow_dispatch' && needs.change-filter.outputs.precision-test-STA == 'true') || 
-      (github.event_name == 'workflow_dispatch' && github.event.inputs.run_precision_test_STA == 'true')
-    uses: ./.github/workflows/runpod-test.yml
-    with:
-      job_id: "precision-test-STA"
-      gpu_type: "NVIDIA H100 NVL"
-      gpu_count: 1
-      volume_size: 100
-      disk_size: 100
-      image: "ghcr.io/${{ github.repository }}/fastvideo-dev:py3.12-latest"
-      test_command: "uv pip install -e .[test] && python csrc/attn/tests/test_sta.py"
-      timeout_minutes: 30
-    secrets:
-      RUNPOD_API_KEY: ${{ secrets.RUNPOD_API_KEY }}
-      RUNPOD_PRIVATE_KEY: ${{ secrets.RUNPOD_PRIVATE_KEY }}
-
-  precision-test-VSA:
-    needs: change-filter
-    if: >-
-      (github.event_name != 'workflow_dispatch' && needs.change-filter.outputs.precision-test-VSA == 'true') || 
-      (github.event_name == 'workflow_dispatch' && github.event.inputs.run_precision_test_VSA == 'true')
-    uses: ./.github/workflows/runpod-test.yml
-    with:
-      job_id: "precision-test-VSA"
-      gpu_type: "NVIDIA H100 NVL"
-      gpu_count: 1
-      volume_size: 100
-      disk_size: 100
-      image: "ghcr.io/${{ github.repository }}/fastvideo-dev:py3.12-latest"
-      test_command: "uv pip install -e .[test] && python csrc/attn/tests/test_vsa.py"
-      timeout_minutes: 30
-    secrets:
-      RUNPOD_API_KEY: ${{ secrets.RUNPOD_API_KEY }}
-      RUNPOD_PRIVATE_KEY: ${{ secrets.RUNPOD_PRIVATE_KEY }}
-
-  unit-test:
-    needs: change-filter
-    if: >-
-      (github.event_name != 'workflow_dispatch' && needs.change-filter.outputs.unit-test == 'true') || 
-      (github.event_name == 'workflow_dispatch' && github.event.inputs.run_unit_test == 'true')
-    uses: ./.github/workflows/runpod-test.yml
-    with:
-      job_id: "unit-test"
-      gpu_type: "NVIDIA L40S"
-      gpu_count: 1
-      volume_size: 100
-      disk_size: 100
-      image: "ghcr.io/${{ github.repository }}/fastvideo-dev:py3.12-latest"
-      test_command: "uv pip install -e .[test] && pytest ./fastvideo/dataset/ -vs && pytest ./fastvideo/workflow/ -vs"
-      timeout_minutes: 30
-    secrets:
-      RUNPOD_API_KEY: ${{ secrets.RUNPOD_API_KEY }}
-      RUNPOD_PRIVATE_KEY: ${{ secrets.RUNPOD_PRIVATE_KEY }}
-
-  # nightly-test:
-  #   if: >-
-  #     (github.event_name == 'workflow_dispatch' && github.event.inputs.run_nightly_test == 'true')
-  #   uses: ./.github/workflows/runpod-test.yml
-  #   with:
-  #     job_id: "nightly-test"
-  #     gpu_type: "NVIDIA A40"
-  #     gpu_count: 4
-  #     volume_size: 100
-  #     disk_size: 100
-  #     image: "ghcr.io/${{ github.repository }}/fastvideo-dev:py3.12-latest"
-  #     test_command: "wandb login $WANDB_API_KEY && uv pip install -e .[test] && pytest ./fastvideo/tests/nightly/test_e2e_overfit_single_sample.py -vs"
-  #     timeout_minutes: 30
-  #   secrets:
-  #     RUNPOD_API_KEY: ${{ secrets.RUNPOD_API_KEY }}
-  #     RUNPOD_PRIVATE_KEY: ${{ secrets.RUNPOD_PRIVATE_KEY }}
-  #     WANDB_API_KEY: ${{ secrets.WANDB_API_KEY }}
-
   runpod-cleanup:
-    # Add other jobs to this list as you create them
-    needs: [encoder-test, vae-test, transformer-test, ssim-test, training-test, training-test-VSA, inference-test-STA, precision-test-STA, precision-test-VSA] 
+    needs: [encoder-test, vae-test, transformer-test, ssim-test] # Add other jobs to this list as you create them
     if: ${{ always() && ((github.event_name != 'workflow_dispatch' && github.event.pull_request.draft == false) || github.event_name == 'workflow_dispatch') }}
     runs-on: ubuntu-latest
     steps:
@@ -395,7 +177,7 @@ jobs:
 
       - name: Cleanup all RunPod instances
         env:
-          JOB_IDS: '["encoder-test", "vae-test", "transformer-test", "ssim-test-py3.10", "ssim-test-py3.11", "ssim-test-py3.12", "training-test", "training-test-VSA", "inference-test-STA", "precision-test-STA", "precision-test-VSA"]'
+          JOB_IDS: '["encoder-test", "vae-test", "transformer-test", "ssim-test-py3.10", "ssim-test-py3.11", "ssim-test-py3.12"]'
           RUNPOD_API_KEY: ${{ secrets.RUNPOD_API_KEY }}
           GITHUB_RUN_ID: ${{ github.run_id }}
         run: python .github/scripts/runpod_cleanup.py

.github/workflows/pre-commit.yml

@@ -10,7 +10,7 @@ jobs:
     - uses: actions/checkout@v4
     - uses: actions/setup-python@v5
       with:
-        python-version: "3.12"
+        python-version: "3.10"
     - run: echo "::add-matcher::.github/workflows/matchers/actionlint.json"
     - run: echo "::add-matcher::.github/workflows/matchers/mypy.json"
     - uses: pre-commit/action@v3.0.1

.github/workflows/publish-comfyui.yml

@@ -1,28 +0,0 @@
-name: Publish to Comfy registry
-on:
-  workflow_dispatch:
-  push:
-    branches:
-      - main
-      - master
-    paths:
-      - "pyproject.toml"
-
-permissions:
-  issues: write
-
-jobs:
-  publish-node:
-    name: Publish Custom Node to registry
-    runs-on: ubuntu-latest
-    if: ${{ github.repository_owner == 'hao-ai-lab' }}
-    steps:
-      - name: Check out code
-        uses: actions/checkout@v4
-        with:
-          submodules: true
-      - name: Publish Custom Node
-        uses: Comfy-Org/publish-node-action@v1
-        with:
-          ## Add your own personal access token to your Github Repository secrets and reference it here.
-          personal_access_token: ${{ secrets.REGISTRY_ACCESS_TOKEN }}

.github/workflows/runpod-test.yml

@@ -43,8 +43,6 @@ on:
         required: true
       RUNPOD_PRIVATE_KEY:
         required: true
-      WANDB_API_KEY:
-        required: false
 
 jobs:
   run-test:
@@ -57,7 +55,7 @@ jobs:
       - name: Set up Python
         uses: actions/setup-python@v5
         with:
-          python-version: "3.12"
+          python-version: "3.10"
 
       - name: Set up SSH key
         run: |
@@ -74,7 +72,6 @@ jobs:
           JOB_ID: ${{ inputs.job_id }}
           RUNPOD_API_KEY: ${{ secrets.RUNPOD_API_KEY }}
           GITHUB_RUN_ID: ${{ github.run_id }}
-          WANDB_API_KEY: ${{ secrets.WANDB_API_KEY }}
         timeout-minutes: ${{ inputs.timeout_minutes }}
         run: >-
           python .github/scripts/runpod_api.py 

.github/workflows/sta-publish.yml

@@ -5,7 +5,7 @@ on:
     branches:
       - main
     paths:
-      - "csrc/attn/sliding_tile_attn/setup.py"
+      - "csrc/sliding_tile_attention/setup.py"
   workflow_dispatch:
 
 jobs:
@@ -23,7 +23,7 @@ jobs:
       - name: Check if version changed
         id: check-version
         run: |
-          cd csrc/attn/sliding_tile_attn
+          cd csrc/sliding_tile_attention
           # Get current commit's version
           NEW_VERSION=$(grep -oP 'VERSION\s*=\s*"\K[^"]+' setup.py)
           echo "New version: $NEW_VERSION"
@@ -136,21 +136,19 @@ jobs:
 
       - name: Build wheel
         run: |
-          export PYTHONPATH=$GITHUB_WORKSPACE:$PYTHONPATH
-          
           # We want setuptools >= 49.6.0 otherwise we can't compile the extension if system CUDA version is 11.7 and pytorch cuda version is 11.6
           # https://github.com/pytorch/pytorch/blob/664058fa83f1d8eede5d66418abff6e20bd76ca8/torch/utils/cpp_extension.py#L810
           # However this still fails so I'm using a newer version of setuptools
           pip install setuptools
           pip install ninja packaging wheel
 
-          cd csrc/attn/sliding_tile_attn  # Move into the correct folder
-          git submodule update --init --recursive # Ensure ThunderKittens submodule is initialized
+          cd csrc/sliding_tile_attention  # Move into the correct folder
+          git submodule update --init --recursive tk # Ensure ThunderKittens submodule is initialized
           python setup.py bdist_wheel --dist-dir=dist
 
       - name: Rename wheel file
         run: |
-          cd csrc/attn/sliding_tile_attn
+          cd csrc/sliding_tile_attention
 
           CUDA_SHORT_VERSION=$(echo ${{ matrix.cuda-version }} | cut -d. -f1,2 | sed 's/\.//g')
           TORCH_SHORT_VERSION=$(echo ${{ matrix.torch-version }} | cut -d. -f1,2)
@@ -165,7 +163,7 @@ jobs:
         uses: actions/upload-artifact@v4
         with:
           name: ${{ env.wheel_name }}
-          path: csrc/attn/sliding_tile_attn/dist/*.whl
+          path: csrc/sliding_tile_attention/dist/*.whl
           retention-days: 90
 
   publish_package:
@@ -231,19 +229,17 @@ jobs:
 
       - name: Build source distribution
         run: |
-          export PYTHONPATH=$GITHUB_WORKSPACE:$PYTHONPATH
-          
           # We want setuptools >= 49.6.0 otherwise we can't compile the extension if system CUDA version is 11.7 and pytorch cuda version is 11.6
           # https://github.com/pytorch/pytorch/blob/664058fa83f1d8eede5d66418abff6e20bd76ca8/torch/utils/cpp_extension.py#L810
           # However this still fails so I'm using a newer version of setuptools
           pip install setuptools
           pip install ninja packaging wheel
 
-          cd csrc/attn/sliding_tile_attn  # Move into the correct folder
-          git submodule update --init --recursive # Ensure ThunderKittens submodule is initialized
+          cd csrc/sliding_tile_attention  # Move into the correct folder
+          git submodule update --init --recursive tk # Ensure ThunderKittens submodule is initialized
           python setup.py sdist --dist-dir=dist
 
       - name: Publish release distributions to PyPI
         uses: pypa/gh-action-pypi-publish@release/v1
         with:
-          packages-dir: csrc/attn/sliding_tile_attn/dist/
+          packages-dir: csrc/sliding_tile_attention/dist/

.github/workflows/test.yml

@@ -28,4 +28,4 @@ jobs:
 
       - name: Run Pytest
         run: |
-          pytest --ignore csrc/attn/test
+          pytest --ignore csrc/sliding_tile_attention/test

.github/workflows/vsa-publish.yml

@@ -1,257 +0,0 @@
-name: Publish Video Sparse Attention Kernel to PyPI on Version Change
-
-on:
-  push:
-    branches:
-      - main
-    paths:
-      - "csrc/attn/video_sparse_attn/setup.py"
-  workflow_dispatch:
-
-jobs:
-  check-version-change:
-    runs-on: ubuntu-latest
-    outputs:
-      version-changed: ${{ steps.check-version.outputs.changed }}
-      new-version: ${{ steps.check-version.outputs.new-version }}
-    steps:
-      - name: Checkout code
-        uses: actions/checkout@v4
-        with:
-          fetch-depth: 2
-
-      - name: Check if version changed
-        id: check-version
-        run: |
-          cd csrc/attn/video_sparse_attn
-          # Get current commit's version
-          NEW_VERSION=$(grep -oP 'VERSION\s*=\s*"\K[^"]+' setup.py)
-          echo "New version: $NEW_VERSION"
-          
-          # Get previous version from git history
-          OLD_VERSION=$(git show HEAD~1:./setup.py | grep -oP 'VERSION\s*=\s*"\K[^"]+' || echo "0.0.0")
-          echo "Old version: $OLD_VERSION"
-          
-          if [ "$NEW_VERSION" != "$OLD_VERSION" ]; then
-            echo "Version changed from $OLD_VERSION to $NEW_VERSION"
-            echo "changed=true" >> $GITHUB_OUTPUT
-            echo "new-version=$NEW_VERSION" >> $GITHUB_OUTPUT
-          else
-            echo "Version did not change"
-            echo "changed=false" >> $GITHUB_OUTPUT
-          fi
-
-  build_wheels:
-    name: Build Wheel
-    needs: check-version-change
-    if: ${{ needs.check-version-change.outputs.version-changed == 'true' || github.event_name == 'workflow_dispatch' }}
-    runs-on: ${{ matrix.os }}
-
-    strategy:
-      fail-fast: false
-      matrix:
-          # Using ubuntu-20.04 instead of 22.04 for more compatibility (glibc). Ideally we'd use the
-          # manylinux docker image, but I haven't figured out how to install CUDA on manylinux.
-          os: [ubuntu-22.04]
-          python-version: ['3.10', '3.11', '3.12', '3.13']
-          # For version reference https://pytorch.org/get-started/previous-versions/
-          torch-cuda:
-            - torch-version: '2.5.1'
-              cuda-version: '12.4.1'
-              torch-cuda-short: 'cu124'
-            - torch-version: '2.6.0'
-              cuda-version: '12.6.3'
-              torch-cuda-short: 'cu126'
-            - torch-version: '2.7.1'
-              cuda-version: '12.8.0'
-              torch-cuda-short: 'cu128'
-
-    steps:
-      - name: Free up disk space
-        run: |
-          echo "Initial disk space:"
-          df -h
-          
-          # Remove large directories
-          sudo rm -rf /usr/share/dotnet
-          sudo rm -rf /usr/local/lib/android
-          sudo rm -rf /opt/ghc
-          sudo rm -rf /usr/local/share/boost
-          sudo rm -rf /usr/share/swift
-          sudo rm -rf /usr/local/lib/node_modules
-          sudo rm -rf /usr/local/share/powershell
-          sudo rm -rf /usr/share/rust
-          sudo rm -rf /usr/local/.ghcup
-          
-          # Remove cached files
-          sudo rm -rf /var/lib/apt/lists/*
-          sudo rm -rf /var/cache/apt/archives/*
-          
-          echo "Disk space after cleanup:"
-          df -h
-
-      - name: Checkout
-        uses: actions/checkout@v4
-
-      - name: Set up Python
-        uses: actions/setup-python@v5
-        with:
-          python-version: ${{ matrix.python-version }}
-
-      - name: Install CUDA ${{ matrix.torch-cuda.cuda-version }}
-        uses: Jimver/cuda-toolkit@v0.2.21
-        id: cuda-toolkit
-        with:
-          cuda: ${{ matrix.torch-cuda.cuda-version }}
-          linux-local-args: '["--toolkit"]'
-          method: 'network'
-
-      - name: Install dependencies (GCC, Clang, CUDA Paths, Git)
-        run: |
-          sudo apt update
-          sudo apt install -y git patchelf gcc-11 g++-11 clang-11
-          sudo update-alternatives --install /usr/bin/gcc gcc /usr/bin/gcc-11 100 --slave /usr/bin/g++ g++ /usr/bin/g++-11
-
-          # Allow Git to Access Safe Directory
-          git config --global --add safe.directory /__w/FastVideo/FastVideo
-
-          # Set CUDA environment variables
-          export CUDA_HOME=/usr/local/cuda-${{ matrix.torch-cuda.cuda-version }}
-          export PATH=${CUDA_HOME}/bin:${PATH}
-          export LD_LIBRARY_PATH=${CUDA_HOME}/lib64:$LD_LIBRARY_PATH
-          
-          # Verify installation
-          gcc --version
-          g++ --version
-          clang-11 --version
-          nvcc --version
-
-      - name: Install PyTorch ${{ matrix.torch-cuda.torch-version }}+cu${{ matrix.torch-cuda.cuda-version }}
-        run: |
-          pip install --upgrade pip
-          # With python 3.13 and torch 2.5.1, unless we update typing-extensions, we get error
-          # AttributeError: attribute '__default__' of 'typing.ParamSpec' objects is not writable
-          pip install typing-extensions==4.12.2
-          # We want to figure out the CUDA version to download pytorch
-          # e.g. we can have system CUDA version being 11.7 but if torch==1.12 then we need to download the wheel from cu116
-          # see https://github.com/pytorch/pytorch/blob/main/RELEASE.md#release-compatibility-matrix
-          pip install --no-cache-dir torch==${{ matrix.torch-cuda.torch-version }} --index-url https://download.pytorch.org/whl/${{matrix.torch-cuda.torch-cuda-short}}
-          nvcc --version
-          python --version
-          python -c "import torch; print('PyTorch:', torch.__version__)"
-          python -c "import torch; print('CUDA:', torch.version.cuda)"
-          python -c "from torch.utils import cpp_extension; print (cpp_extension.CUDA_HOME)"
-
-      - name: Build wheel
-        run: |
-          export PYTHONPATH=$GITHUB_WORKSPACE:$PYTHONPATH
-          
-          # We want setuptools >= 49.6.0 otherwise we can't compile the extension if system CUDA version is 11.7 and pytorch cuda version is 11.6
-          # https://github.com/pytorch/pytorch/blob/664058fa83f1d8eede5d66418abff6e20bd76ca8/torch/utils/cpp_extension.py#L810
-          # However this still fails so I'm using a newer version of setuptools
-          pip install setuptools
-          pip install ninja packaging wheel
-
-          cd csrc/attn/video_sparse_attn  # Move into the correct folder
-          git submodule update --init --recursive # Ensure ThunderKittens submodule is initialized
-          python setup.py bdist_wheel --dist-dir=dist
-
-      - name: Rename wheel file
-        run: |
-          cd csrc/attn/video_sparse_attn
-
-          CUDA_SHORT_VERSION=$(echo ${{ matrix.torch-cuda.cuda-version }} | cut -d. -f1,2 | sed 's/\.//g')
-          TORCH_SHORT_VERSION=$(echo ${{ matrix.torch-cuda.torch-version }} | cut -d. -f1,2)
-          # Get the correct version format
-          tmpname=cu${CUDA_SHORT_VERSION}torch${TORCH_SHORT_VERSION}
-          wheel_name=$(ls dist/*whl | xargs -n 1 basename | sed "s/-/+$tmpname-/2")
-          # Rename with version information
-          ls dist/*whl |xargs -I {} mv {} dist/${wheel_name}
-          echo "wheel_name=${wheel_name}" >> $GITHUB_ENV
-
-      - name: Upload wheel artifact
-        uses: actions/upload-artifact@v4
-        with:
-          name: ${{ env.wheel_name }}
-          path: csrc/attn/video_sparse_attn/dist/*.whl
-          retention-days: 90
-
-  publish_package:
-    name: Publish package
-    needs: [build_wheels, check-version-change]
-    if: ${{ needs.check-version-change.outputs.version-changed == 'true' || github.event_name == 'workflow_dispatch' }}
-    runs-on: ubuntu-22.04
-    permissions:
-      id-token: write  # Needed for OIDC Trusted Publishing
-
-    steps:
-      - uses: actions/checkout@v4
-
-      - uses: actions/setup-python@v5
-        with:
-          python-version: '3.10'
-
-      - name: Install CUDA 12.4.1
-        uses: Jimver/cuda-toolkit@v0.2.21
-        id: cuda-toolkit
-        with:
-          cuda: 12.4.1
-          linux-local-args: '["--toolkit"]'
-          method: 'network'
-          sub-packages: '["nvcc"]'
-
-      - name: Install dependencies (GCC, Clang, CUDA Paths, Git)
-        run: |
-          sudo apt update
-          sudo apt install -y git patchelf gcc-11 g++-11 clang-11
-          sudo update-alternatives --install /usr/bin/gcc gcc /usr/bin/gcc-11 100 --slave /usr/bin/g++ g++ /usr/bin/g++-11
-
-          # Allow Git to Access Safe Directory
-          git config --global --add safe.directory /__w/FastVideo/FastVideo
-
-          # Set CUDA environment variables
-          export CUDA_HOME=/usr/local/cuda-12.4.1
-          export PATH=${CUDA_HOME}/bin:${PATH}
-          export LD_LIBRARY_PATH=${CUDA_HOME}/lib64:$LD_LIBRARY_PATH
-          
-          # Verify installation
-          gcc --version
-          g++ --version
-          clang-11 --version
-          nvcc --version
-
-      - name: Install PyTorch 2.5.1+cu12.4.1
-        run: |
-          pip install --upgrade pip
-          # With python 3.13 and torch 2.5.1, unless we update typing-extensions, we get error
-          # AttributeError: attribute '__default__' of 'typing.ParamSpec' objects is not writable
-          pip install typing-extensions==4.12.2
-          # We want to figure out the CUDA version to download pytorch
-          # e.g. we can have system CUDA version being 11.7 but if torch==1.12 then we need to download the wheel from cu116
-          # see https://github.com/pytorch/pytorch/blob/main/RELEASE.md#release-compatibility-matrix
-          export TORCH_CUDA_VERSION=124
-          pip install --no-cache-dir torch==2.5.1 --index-url https://download.pytorch.org/whl/cu${TORCH_CUDA_VERSION}
-          nvcc --version
-          python --version
-          python -c "import torch; print('PyTorch:', torch.__version__)"
-          python -c "import torch; print('CUDA:', torch.version.cuda)"
-          python -c "from torch.utils import cpp_extension; print (cpp_extension.CUDA_HOME)"
-
-      - name: Build source distribution
-        run: |
-          export PYTHONPATH=$GITHUB_WORKSPACE:$PYTHONPATH
-          
-          # We want setuptools >= 49.6.0 otherwise we can't compile the extension if system CUDA version is 11.7 and pytorch cuda version is 11.6
-          # https://github.com/pytorch/pytorch/blob/664058fa83f1d8eede5d66418abff6e20bd76ca8/torch/utils/cpp_extension.py#L810
-          # However this still fails so I'm using a newer version of setuptools
-          pip install setuptools
-          pip install ninja packaging wheel
-
-          cd csrc/attn/video_sparse_attn  # Move into the correct folder
-          git submodule update --init --recursive # Ensure ThunderKittens submodule is initialized
-          python setup.py sdist --dist-dir=dist
-
-      - name: Publish release distributions to PyPI
-        uses: pypa/gh-action-pypi-publish@release/v1
-        with:
-          packages-dir: csrc/attn/video_sparse_attn/dist/

.gitignore

@@ -20,7 +20,6 @@ samples/
 data/
 outputs/
 outputs_video
-checkpoints/
 sbatch.sh
 *.out
 env
@@ -28,6 +27,7 @@ env
 **/build/
 **.pyc
 **.txt
+**.json
 
 # Distribution / packaging
 build/
@@ -41,8 +41,6 @@ eggs/
 docs/_build/
 docs/source/getting_started/examples/
 docs/source/inference/examples/
-docs/source/training/examples/
-docs/source/distillation/examples/
 
 # VSCode
 .vscode/
@@ -58,12 +56,7 @@ docs/source/distillation/examples/
 *.pkl
 
 # Reference videos
-!fastvideo/tests/ssim/reference_videos/**/*.mp4
+!fastvideo/v1/tests/ssim/reference_videos/**/*.mp4
 
 # Static images
 !docs/source/_static/images/**/*.png
-!comfyui/assets/**/*.png
-!comfyui/assets/**/*.gif
-
-dmd_t2v_output/
-preprocess_output_text/

.gitmodules

@@ -1,7 +1,3 @@
-[submodule "csrc/attn/video_sparse_attn/tk"]
-	path = csrc/attn/video_sparse_attn/tk
-	url = https://github.com/HazyResearch/ThunderKittens.git
-
-[submodule "csrc/attn/sliding_tile_attn/tk"]
-	path = csrc/attn/sliding_tile_attn/tk
+[submodule "csrc/sliding_tile_attention/tk"]
+	path = csrc/sliding_tile_attention/tk
 	url = https://github.com/HazyResearch/ThunderKittens.git

.pre-commit-config.yaml

@@ -3,7 +3,7 @@ default_stages:
   - manual # Run in CI
 exclude: |
   (?x)(
-    fastvideo/third_party/.*|
+    fastvideo/v1/third_party/.*|
     csrc/.*|
     assets/.*|
     tests/.*|
@@ -12,6 +12,9 @@ exclude: |
     scripts/.*|
     fastvideo/data_preprocess/.*|
     fastvideo/dataset/.*|
+    fastvideo/distill/.*|
+    fastvideo/distill\.py|
+    fastvideo/distill_adv\.py|
     fastvideo/models/.*|
     fastvideo/sample/.*|
     fastvideo/train\.py|
@@ -19,7 +22,6 @@ exclude: |
     examples/.*|
     .github/workflows/fastvideo-publish.yml|
     .github/workflows/sta-publish.yml|
-    .github/workflows/vsa-publish.yml|
     .github/workflows/build-image-template.yml|
     docs/source/inference/support_matrix.md
   )
@@ -31,7 +33,7 @@ repos:
     args: [--in-place, --verbose]
     additional_dependencies: [toml] # TODO: Remove when yapf is upgraded
 - repo: https://github.com/astral-sh/ruff-pre-commit
-  rev: v0.11.12
+  rev: v0.11.4
   hooks:
   - id: ruff
     args: [--output-format, github, --fix]
@@ -41,12 +43,12 @@ repos:
   - id: codespell
     additional_dependencies: ['tomli']
     args: ['--toml', 'pyproject.toml']
-# - repo: https://github.com/PyCQA/isort
-#   rev: 6.0.1
-#   hooks:
-#   - id: isort
+- repo: https://github.com/PyCQA/isort
+  rev: 6.0.1
+  hooks:
+  - id: isort
 - repo: https://github.com/jackdewinter/pymarkdown
-  rev: v0.9.30
+  rev: v0.9.29
   hooks:
   - id: pymarkdown
     args: [fix]
@@ -58,7 +60,7 @@ repos:
   rev: v1.15.0
   hooks:
   - id: mypy
-    args: [--python-version, '3.10', --follow-imports, "skip", "--disable-error-code", "union-attr", "--disable-error-code", "override" ]
+    args: [--python-version, '3.10', --follow-imports, "skip", ]
     additional_dependencies: [types-cachetools, types-setuptools, types-PyYAML, types-requests]
 - repo: local
   hooks:
@@ -67,7 +69,7 @@ repos:
     entry: bash
     args:
       - -c
-      - 'git ls-files | grep -v "^fastvideo/tests/ssim/" | grep -v "^fastvideo/tests/inference/lora/L40S_reference_videos/" | grep " " && echo "Filenames should not contain spaces!" && exit 1 || exit 0'
+      - 'git ls-files | grep -v "^fastvideo/v1/tests/ssim/" | grep " " && echo "Filenames should not contain spaces!" && exit 1 || exit 0'
     language: system
     always_run: true
     pass_filenames: false

README.md

@@ -1,41 +1,37 @@
 <div align="center">
-<img src=assets/logos/logo.svg width="30%"/>
+<img src=assets/logo.jpg width="30%"/>
 </div>
 
-**FastVideo is a unified post-training and inference framework for accelerated video generation.**
+**FastVideo is a unified 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.
+It features a clean, consistent API that works across popular video models, making it easier for developers to author new models and incorporate system- or kernel-level optimizations.
+With FastVideo's optimizations, you can achieve more than 3x inference improvement compared to other systems.
 
 <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.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/q46BbX6" target="_blank"> <b> WeChat </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/FastVideo/FastHunyuan"  target="_blank"><b>FastHunyuan</b></a>  | 🤗 <a href="https://huggingface.co/FastVideo/FastMochi-diffusers" target="_blank"><b>FastMochi</b></a> | 🟣💬 <a href="https://join.slack.com/t/fastvideo/shared_invite/zt-2zf6ru791-sRwI9lPIUJQq1mIeB_yjJg" target="_blank"> <b>Slack</b> </a> |
 </p>
 
 <div align="center">
-<img src=assets/fastwan.png width="90%"/>
+<img src=assets/perf.png width="90%"/>
 </div>
 
-## NEWS
-- ```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/).
-
 ## Key Features
 
 FastVideo has the following features:
-- End-to-end post-training support:
-  - [Sparse distillation](https://hao-ai-lab.github.io/blogs/fastvideo_post_training/) for Wan2.1 and Wan2.2 to achineve >50x denoising speedup
-  - Data preprocessing pipeline for video data
-  - 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
 - State-of-the-art performance optimizations for inference
-  - [Video Sparse Attention](https://arxiv.org/pdf/2505.13389)
   - [Sliding Tile Attention](https://arxiv.org/pdf/2502.04507)
   - [TeaCache](https://arxiv.org/pdf/2411.19108)
   - [Sage Attention](https://arxiv.org/abs/2410.02367)
-- Diverse hardware and OS support
-  - Support H100, A100, 4090
-  - Support Linux, Windows, MacOS
+- Cutting edge models
+  - Wan2.1 T2V, I2V
+  - HunyuanVideo
+  - FastHunyuan: consistency distilled video diffusion models for 8x inference speedup.
+  - StepVideo T2V
+- Distillation support
+  - Recipes for video DiT, based on [PCM](https://github.com/G-U-N/Phased-Consistency-Model).
+  - Support distilling/finetuning/inferencing state-of-the-art open video DiTs: 1. Mochi 2. Hunyuan.
+- Scalable training with FSDP, sequence parallelism, and selective activation checkpointing, with near linear scaling to 64 GPUs.
+- Memory efficient finetuning with LoRA, precomputed latent, and precomputed text embeddings.
 
 ## Getting Started
 We recommend using an environment manager such as `Conda` to create a clean environment:
@@ -51,31 +47,17 @@ pip install fastvideo
 
 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.html) and check out our [blog](https://hao-ai-lab.github.io/blogs/fastvideo_post_training/).
-
-See below for recipes and datasets:
-
-|                                            Model                                              |                                               Sparse Distillation                                                 |                                                  Dataset                                                  |
-|:-------------------------------------------------------------------------------------------:  |:---------------------------------------------------------------------------------------------------------------:  |:--------------------------------------------------------------------------------------------------------: |
-| [FastWan2.1-T2V-1.3B](https://huggingface.co/FastVideo/FastWan2.1-T2V-1.3B-Diffusers)         |    [Recipe](https://github.com/hao-ai-lab/FastVideo/tree/main/examples/distill/Wan2.1-T2V/Wan-Syn-Data-480P)      | [FastVideo Synthetic Wan2.1 480P](https://huggingface.co/datasets/FastVideo/Wan-Syn_77x448x832_600k)      |
-| [FastWan2.1-T2V-14B-Preview](https://huggingface.co/FastVideo/FastWan2.1-T2V-14B-Diffusers)   |                                                   Coming soon!                                                    |   [FastVideo Synthetic Wan2.1 720P](https://huggingface.co/datasets/FastVideo/Wan-Syn_77x768x1280_250k)   |
-| [FastWan2.2-TI2V-5B](https://huggingface.co/FastVideo/FastWan2.2-TI2V-5B-Diffusers)           | [Recipe](https://github.com/hao-ai-lab/FastVideo/tree/main/examples/distill/Wan2.2-TI2V-5B-Diffusers/Data-free)   | [FastVideo Synthetic Wan2.2 720P](https://huggingface.co/datasets/FastVideo/Wan2.2-Syn-121x704x1280_32k)  |
-
 ## 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.html). Create a file called `example.py` with the following code:
+Here's a minimal example to generate a video using the default settings. Create a file called `example.py` with the following code:
 
 ```python
-import os
 from fastvideo import VideoGenerator
 
 def main():
-    os.environ["FASTVIDEO_ATTENTION_BACKEND"] = "VIDEO_SPARSE_ATTN"
-
     # Create a video generator with a pre-trained model
     generator = VideoGenerator.from_pretrained(
-        "FastVideo/FastWan2.1-T2V-1.3B-Diffusers",
+        "Wan-AI/Wan2.1-T2V-1.3B-Diffusers",
         num_gpus=1,  # Adjust based on your hardware
     )
 
@@ -108,63 +90,60 @@ For a more detailed guide, please see our [inference quick start](https://hao-ai
 - [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.html)
-<!-- - [Finetuning Guide](https://hao-ai-lab.github.io/FastVideo/training/finetune.html) -->
+- [Distillation Guide](https://hao-ai-lab.github.io/FastVideo/training/distillation.html)
+- [Finetuning Guide](https://hao-ai-lab.github.io/FastVideo/training/finetuning.html)
 
 ## 📑 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 -->
+- More models support
+  <!-- - [ ] Add CogvideoX model -->
+  - [x] Add StepVideo to V1
+- Optimization features
+  - [x] Teacache in V1
+  - [x] SageAttention in V1
+- Code updates
+  - [x] V1 Configuration API
+  - [ ] Support Training in V1
   <!-- - [ ] fp8 support -->
   <!-- - [ ] faster load model and save model support -->
 
-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.html)
+We welcome all contributions. Please check out our guide [here](https://hao-ai-lab.github.io/FastVideo/developer_guide/overview.html)
 
 ## Acknowledgement
 We learned and reused code from the following projects:
-- [Wan-Video](https://github.com/Wan-Video)
-- [ThunderKittens](https://github.com/HazyResearch/ThunderKittens)
-- [Triton](https://github.com/triton-lang/triton)
-- [DMD2](https://github.com/tianweiy/DMD2)
+- [PCM](https://github.com/G-U-N/Phased-Consistency-Model)
 - [diffusers](https://github.com/huggingface/diffusers)
+- [OpenSoraPlan](https://github.com/PKU-YuanGroup/Open-Sora-Plan)
 - [xDiT](https://github.com/xdit-project/xDiT)
 - [vLLM](https://github.com/vllm-project/vllm)
 - [SGLang](https://github.com/sgl-project/sglang)
 
-We thank [MBZUAI](https://ifm.mbzuai.ac.ae/), [Anyscale](https://www.anyscale.com/), and [GMI Cloud](https://www.gmicloud.ai/) for their support throughout this project.
+We thank MBZUAI and [Anyscale](https://www.anyscale.com/) for their support throughout this project.
 
 ## Citation
-If you find FastVideo useful, please considering citing our work:
+If you use FastVideo for your research, please cite our paper:
 
 ```bibtex
-@software{fastvideo2024,
-  title        = {FastVideo: A Unified Framework for Accelerated Video Generation},
-  author       = {The FastVideo Team},
-  url          = {https://github.com/hao-ai-lab/FastVideo},
-  month        = apr,
-  year         = {2024},
+@misc{zhang2025fastvideogenerationsliding,
+      title={Fast Video Generation with Sliding Tile Attention},
+      author={Peiyuan Zhang and Yongqi Chen and Runlong Su and Hangliang Ding and Ion Stoica and Zhenghong Liu and Hao Zhang},
+      year={2025},
+      eprint={2502.04507},
+      archivePrefix={arXiv},
+      primaryClass={cs.CV},
+      url={https://arxiv.org/abs/2502.04507},
 }
-
-@article{zhang2025vsa,
-  title={Vsa: Faster video diffusion with trainable sparse attention},
-  author={Zhang, Peiyuan and Chen, Yongqi and Huang, Haofeng and Lin, Will and Liu, Zhengzhong and Stoica, Ion and Xing, Eric and Zhang, Hao},
-  journal={arXiv preprint arXiv:2505.13389},
-  year={2025}
-}
-
-@article{zhang2025fast,
-  title={Fast video generation with sliding tile attention},
-  author={Zhang, Peiyuan and Chen, Yongqi and Su, Runlong and Ding, Hangliang and Stoica, Ion and Liu, Zhengzhong and Zhang, Hao},
-  journal={arXiv preprint arXiv:2502.04507},
-  year={2025}
+@misc{ding2025efficientvditefficientvideodiffusion,
+      title={Efficient-vDiT: Efficient Video Diffusion Transformers With Attention Tile},
+      author={Hangliang Ding and Dacheng Li and Runlong Su and Peiyuan Zhang and Zhijie Deng and Ion Stoica and Hao Zhang},
+      year={2025},
+      eprint={2502.06155},
+      archivePrefix={arXiv},
+      primaryClass={cs.CV},
+      url={https://arxiv.org/abs/2502.06155},
 }
 ```

__init__.py

@@ -1,15 +0,0 @@
-try:
-    from .comfyui.video_generator.nodes import (NODE_CLASS_MAPPINGS,
-                                                NODE_DISPLAY_NAME_MAPPINGS)
-    WEB_DIRECTORY = "./web"
-    __all__ = [
-        'NODE_CLASS_MAPPINGS', 'NODE_DISPLAY_NAME_MAPPINGS', 'WEB_DIRECTORY'
-    ]
-except ImportError:
-    # ComfyUI environment not available, skip comfyui imports
-    NODE_CLASS_MAPPINGS = {}
-    NODE_DISPLAY_NAME_MAPPINGS = {}
-    WEB_DIRECTORY = "./web"
-    __all__ = [
-        'NODE_CLASS_MAPPINGS', 'NODE_DISPLAY_NAME_MAPPINGS', 'WEB_DIRECTORY'
-    ]

assets/logos/icon_simple.svg

@@ -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>

assets/logos/logo.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>

cog.yaml

@@ -0,0 +1,24 @@
+# Configuration for Cog ⚙️
+# Reference: https://cog.run/yaml
+
+build:
+  gpu: true
+  cuda: "12.1"
+  python_version: "3.10"
+  python_packages:
+    - "torch==2.4.0"
+    - "torchvision"
+    - "ninja==1.11.1.3"
+    - "transformers==4.46.1"
+    - "git+https://github.com/huggingface/diffusers.git@bf64b32652a63a1865a0528a73a13652b201698b"
+    - "accelerate==1.0.1"
+    - "safetensors==0.4.5"
+    - "peft==0.13.2"
+    - "packaging==24.2"
+    - "git+https://github.com/hao-ai-lab/FastVideo"
+
+  run:
+    - FLASH_ATTENTION_SKIP_CUDA_BUILD=TRUE pip install flash-attn --no-build-isolation
+    - curl -o /usr/local/bin/pget -L "https://github.com/replicate/pget/releases/latest/download/pget_$(uname -s)_$(uname -m)" && chmod +x /usr/local/bin/pget
+
+predict: "predict.py:Predictor"

comfyui/README.md

@@ -1,138 +0,0 @@
-# ComfyUI-FastVideo
-
-A custom node suite for ComfyUI that provides accelerated video generation using [FastVideo](https://github.com/hao-ai-labs/FastVideo). See the [blog post](https://hao-ai-lab.github.io/blogs/fastvideo/) about FastVideo V1 to learn more.
-
-## Multi-GPU Parallel Inference
-
-One of the key features ComfyUI-FastVideo brings to ComfyUI is its ability to distribute the generation workload across multiple GPUs, resulting in significantly faster inference times.
-
-![Wan2.1-I2V-14B-480P-Diffusers](./assets/wani2v.gif)
-
-Example of Wan2.1-I2V-14B-480P-Diffusers model running on 4 GPUs.
-## Features
-
-- Generate high-quality videos from text prompts and images
-- Configurable video parameters (prompt, resolution, frame count, FPS)
-- Support for multiple GPUs with tensor and sequence parallelism
-- Advanced configuration options for VAE, Text Encoder, and DIT components
-- Interruption/cancellation support for long-running generations
-
-## Installation
-
-### Requirements
-
-- [ComfyUI](https://github.com/comfyanonymous/ComfyUI)
-- CUDA-capable GPU(s) with sufficient VRAM
-
-### Install using ComfyUI Manager
-
-Coming soon!
-
-### Manual Installation
-
-#### Copy the FastVideo `comfyui` directory into your ComfyUI custom_nodes directory:
-
-```bash
-cp -r /path/to/FastVideo/comfyui /path/to/ComfyUI/custom_nodes/FastVideo
-```
-
-#### Install dependencies:
-
-Currently, the only dependency is `fastvideo`, which can be installed using pip.
-
-```bash
-pip install fastvideo
-```
-
-#### Install missing custom nodes:
-
-`ComfyUI-VideoHelperSuite`:
-
-```bash
-cd /path/to/ComfyUI/custom_nodes
-git clone https://github.com/Kosinkadink/ComfyUI-VideoHelperSuite.git
-```
-
-If you're seeing `ImportError: libGL.so.1: cannot open shared object file: No such file or directory`,
-you may need to install ffmpeg
-
-```bash
-apt-get update && apt-get install ffmpeg
-```
-
-## Usage
-
-After installation, the following nodes will be available in the ComfyUI interface under the "fastvideo" category:
-
-- **Video Generator**: The main node for generating videos from prompts
-- **Inference Args**: Configure video generation parameters
-- **VAE Config**
-- **Text Encoder Config**
-- **DIT Config**
-- **Load Image Path**: Load images for potential conditioning
-
-You may have noticed many arguments on the nodes have 'auto' as the default value. This is because FastVideo will automatically detect the best values for these parameters based on the model and the hardware. However, you can also manually configure these parameters to get the best performance for your specific use case. We plan on releasing more optimized workflow files for different models and hardware configurations in the future.
-
-You can see what some of the default configurations are by looking at the FastVideo repo:
-- [Wan2.1-I2V-14B-480P-Diffusers](https://github.com/hao-ai-lab/FastVideo/blob/main/fastvideo/configs/wan_14B_i2v_480p_pipeline.json)
-- [FastHunyuan-diffusers](https://github.com/hao-ai-lab/FastVideo/blob/main/fastvideo/configs/fasthunyuan_t2v.json)
-
-### Node Configuration
-
-#### Video Generator
-
-- **prompt**: Text description of the video to generate
-- **output_path**: Directory where generated videos will be saved
-- **num_gpus**: Number of GPUs to use for generation
-- **model_path**: Path to the FastVideo model
-- **embedded_cfg_scale**: Classifier-free guidance scale
-- **sp_size**: Sequence parallelism size (usually should match num_gpus)
-- **tp_size**: Tensor parallelism size (usually should match num_gpus)
-- **precision**: Model precision (fp16 or bf16)
-
-`model_path takes either a model id from huggingface or a local path to a model. Models by default will be downloaded to ~/.cache/huggingface/hub/ and cached for subsequent runs.`
-
-#### Inference Args
-
-- **height/width**: Resolution of the output video
-- **num_frames**: Number of frames to generate
-- **num_inference_steps**: Number of diffusion steps per frame
-- **guidance_scale**: Classifier-free guidance scale
-- **flow_shift**: Frame flow shift parameter
-- **seed**: Random seed for reproducible generation
-- **fps**: Frames per second of the output video
-- **image_path**: Optional path to input image for conditioning (for i2v models)
-
-## Memory Management
-
-Models will remain loaded in GPU memory between runs when you only change inference arguments (such as prompt, resolution, frame count, FPS, guidance scale, etc.) or the prompt text. This allows for faster subsequent generations since the model doesn't need to be reloaded.
-
-However, if you need to change the following parameters, you will need to restart the ComfyUI server:
-- **Number of GPUs** (`num_gpus`)
-- **Model path** (`model_path`)
-- **Tensor parallelism size** (`tp_size`)
-- **Sequence parallelism size** (`sp_size`)
-
-These parameters affect the model's distribution across GPUs and require a complete reinitialization of the model pipeline.
-
-## Example workflows
-
-### Text to Video
-
-FastVideo-FastHunyuan-diffusers
-
-![FastVideo-FastHunyuan-diffusers](./assets/fasthunyuan.png)
-
-- [FastHunyuan-diffusers.json](./examples/FastHunyuan-diffusers.json)
-
-### Image to Video
-
-Wan2.1-I2V-14B-480P-Diffusers
-
-![Wan2.1-I2V-14B-480P-Diffusers](./assets/wani2v.png)
-
-- [Wan2.1-I2V-14B-480P-Diffusers.json](./examples/Wan2.1-I2V-14B-480P-Diffusers.json)
-
-## License
-
-This project is licensed under Apache 2.0.

comfyui/__init__.py

@@ -1,5 +0,0 @@
-from .video_generator.nodes import (NODE_CLASS_MAPPINGS,
-                                    NODE_DISPLAY_NAME_MAPPINGS)
-
-WEB_DIRECTORY = "./web"
-__all__ = ['NODE_CLASS_MAPPINGS', 'NODE_DISPLAY_NAME_MAPPINGS', 'WEB_DIRECTORY']

comfyui/assets/icon_simple.svg

@@ -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>

comfyui/examples/FastHunyuan-diffusers.json

@@ -1,645 +0,0 @@
-{
-  "id": "23a1f065-bbba-4a8f-b144-944e1318fcbf",
-  "revision": 0,
-  "last_node_id": 8,
-  "last_link_id": 7,
-  "nodes": [
-    {
-      "id": 4,
-      "type": "VAEConfig",
-      "pos": [
-        374.2159423828125,
-        554.85888671875
-      ],
-      "size": [
-        334.080078125,
-        322
-      ],
-      "flags": {},
-      "order": 0,
-      "mode": 0,
-      "inputs": [],
-      "outputs": [
-        {
-          "name": "vae_config",
-          "type": "VAE_CONFIG",
-          "links": [
-            2
-          ]
-        }
-      ],
-      "properties": {
-        "Node name for S&R": "VAEConfig"
-      },
-      "widgets_values": [
-        -99999,
-        -99999,
-        -99999,
-        -99999,
-        -99999,
-        -99999,
-        -99999,
-        -99999,
-        -99999,
-        -99999,
-        -99999,
-        -99999
-      ],
-      "auto_widget_states": {
-        "load_encoder": {
-          "isAuto": true,
-          "value": -99999,
-          "cachedValue": true
-        },
-        "load_decoder": {
-          "isAuto": true,
-          "value": -99999,
-          "cachedValue": true
-        },
-        "tile_sample_min_height": {
-          "isAuto": true,
-          "value": -99999,
-          "cachedValue": 256
-        },
-        "tile_sample_min_width": {
-          "isAuto": true,
-          "value": -99999,
-          "cachedValue": 256
-        },
-        "tile_sample_min_num_frames": {
-          "isAuto": true,
-          "value": -99999,
-          "cachedValue": 16
-        },
-        "tile_sample_stride_height": {
-          "isAuto": true,
-          "value": -99999,
-          "cachedValue": 192
-        },
-        "tile_sample_stride_width": {
-          "isAuto": true,
-          "value": -99999,
-          "cachedValue": 192
-        },
-        "tile_sample_stride_num_frames": {
-          "isAuto": true,
-          "value": -99999,
-          "cachedValue": 12
-        },
-        "blend_num_frames": {
-          "isAuto": true,
-          "value": -99999,
-          "cachedValue": 0
-        },
-        "use_tiling": {
-          "isAuto": true,
-          "value": -99999,
-          "cachedValue": true
-        },
-        "use_temporal_tiling": {
-          "isAuto": true,
-          "value": -99999,
-          "cachedValue": true
-        },
-        "use_parallel_tiling": {
-          "isAuto": true,
-          "value": -99999,
-          "cachedValue": true
-        }
-      }
-    },
-    {
-      "id": 5,
-      "type": "TextEncoderConfig",
-      "pos": [
-        416.4937744140625,
-        953.6171875
-      ],
-      "size": [
-        270,
-        106
-      ],
-      "flags": {},
-      "order": 1,
-      "mode": 0,
-      "inputs": [],
-      "outputs": [
-        {
-          "name": "text_encoder_config",
-          "type": "TEXT_ENCODER_CONFIG",
-          "links": [
-            7
-          ]
-        }
-      ],
-      "properties": {
-        "Node name for S&R": "TextEncoderConfig"
-      },
-      "widgets_values": [
-        -99999,
-        -99999,
-        -99999
-      ],
-      "auto_widget_states": {
-        "prefix": {
-          "isAuto": true,
-          "value": -99999,
-          "cachedValue": ""
-        },
-        "quant_config": {
-          "isAuto": true,
-          "value": -99999,
-          "cachedValue": ""
-        },
-        "lora_config": {
-          "isAuto": true,
-          "value": -99999,
-          "cachedValue": ""
-        }
-      }
-    },
-    {
-      "id": 6,
-      "type": "DITConfig",
-      "pos": [
-        415.1928405761719,
-        1154.1573486328125
-      ],
-      "size": [
-        270,
-        82
-      ],
-      "flags": {},
-      "order": 2,
-      "mode": 0,
-      "inputs": [],
-      "outputs": [
-        {
-          "name": "dit_config",
-          "type": "DIT_CONFIG",
-          "links": [
-            6
-          ]
-        }
-      ],
-      "properties": {
-        "Node name for S&R": "DITConfig"
-      },
-      "widgets_values": [
-        -99999,
-        -99999
-      ],
-      "auto_widget_states": {
-        "prefix": {
-          "isAuto": true,
-          "value": -99999,
-          "cachedValue": ""
-        },
-        "quant_config": {
-          "isAuto": true,
-          "value": -99999,
-          "cachedValue": ""
-        }
-      }
-    },
-    {
-      "id": 1,
-      "type": "VideoGenerator",
-      "pos": [
-        818.804931640625,
-        348.9299621582031
-      ],
-      "size": [
-        400,
-        436
-      ],
-      "flags": {},
-      "order": 4,
-      "mode": 0,
-      "inputs": [
-        {
-          "name": "inference_args",
-          "shape": 7,
-          "type": "INFERENCE_ARGS",
-          "link": 3
-        },
-        {
-          "name": "vae_config",
-          "shape": 7,
-          "type": "VAE_CONFIG",
-          "link": 2
-        },
-        {
-          "name": "text_encoder_config",
-          "shape": 7,
-          "type": "TEXT_ENCODER_CONFIG",
-          "link": 7
-        },
-        {
-          "name": "dit_config",
-          "shape": 7,
-          "type": "DIT_CONFIG",
-          "link": 6
-        }
-      ],
-      "outputs": [
-        {
-          "name": "video_path",
-          "type": "STRING",
-          "links": [
-            4
-          ]
-        }
-      ],
-      "properties": {
-        "Node name for S&R": "VideoGenerator"
-      },
-      "widgets_values": [
-        "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.",
-        "/workspace/ComfyUI/outputs_video/",
-        2,
-        "FastVideo/FastHunyuan-diffusers",
-        -99999,
-        -99999,
-        -99999,
-        -99999,
-        -99999,
-        -99999,
-        -99999,
-        -99999,
-        -99999
-      ],
-      "auto_widget_states": {
-        "embedded_cfg_scale": {
-          "isAuto": true,
-          "value": -99999,
-          "cachedValue": 6
-        },
-        "sp_size": {
-          "isAuto": true,
-          "value": -99999,
-          "cachedValue": 2
-        },
-        "tp_size": {
-          "isAuto": true,
-          "value": -99999,
-          "cachedValue": 2
-        },
-        "vae_precision": {
-          "isAuto": true,
-          "value": -99999,
-          "cachedValue": "fp16"
-        },
-        "vae_tiling": {
-          "isAuto": true,
-          "value": -99999,
-          "cachedValue": true
-        },
-        "vae_sp": {
-          "isAuto": true,
-          "value": -99999,
-          "cachedValue": true
-        },
-        "text_encoder_precision": {
-          "isAuto": true,
-          "value": -99999,
-          "cachedValue": "fp16"
-        },
-        "precision": {
-          "isAuto": true,
-          "value": -99999,
-          "cachedValue": "fp16"
-        },
-        "dit_cpu_offload": {
-          "isAuto": true,
-          "value": -99999,
-          "cachedValue": true
-        }
-      }
-    },
-    {
-      "id": 3,
-      "type": "VHS_LoadVideoPath",
-      "pos": [
-        1350.136962890625,
-        331.20361328125
-      ],
-      "size": [
-        231.8896484375,
-        286
-      ],
-      "flags": {},
-      "order": 5,
-      "mode": 0,
-      "inputs": [
-        {
-          "name": "meta_batch",
-          "shape": 7,
-          "type": "VHS_BatchManager",
-          "link": null
-        },
-        {
-          "name": "vae",
-          "shape": 7,
-          "type": "VAE",
-          "link": null
-        },
-        {
-          "name": "video",
-          "type": "STRING",
-          "widget": {
-            "name": "video"
-          },
-          "link": 4
-        }
-      ],
-      "outputs": [
-        {
-          "name": "IMAGE",
-          "type": "IMAGE",
-          "links": [
-            5
-          ]
-        },
-        {
-          "name": "frame_count",
-          "type": "INT",
-          "links": null
-        },
-        {
-          "name": "audio",
-          "type": "AUDIO",
-          "links": null
-        },
-        {
-          "name": "video_info",
-          "type": "VHS_VIDEOINFO",
-          "links": null
-        }
-      ],
-      "properties": {
-        "Node name for S&R": "VHS_LoadVideoPath"
-      },
-      "widgets_values": {
-        "video": "",
-        "force_rate": 0,
-        "custom_width": 0,
-        "custom_height": 0,
-        "frame_load_cap": 0,
-        "skip_first_frames": 0,
-        "select_every_nth": 1,
-        "format": "Wan",
-        "videopreview": {
-          "hidden": false,
-          "paused": false,
-          "params": {
-            "filename": "",
-            "type": "path",
-            "format": "video/",
-            "force_rate": 0,
-            "custom_width": 0,
-            "custom_height": 0,
-            "frame_load_cap": 0,
-            "skip_first_frames": 0,
-            "select_every_nth": 1
-          }
-        }
-      }
-    },
-    {
-      "id": 2,
-      "type": "InferenceArgs",
-      "pos": [
-        411.46307373046875,
-        178.18182373046875
-      ],
-      "size": [
-        278.73828125,
-        298
-      ],
-      "flags": {},
-      "order": 3,
-      "mode": 0,
-      "inputs": [],
-      "outputs": [
-        {
-          "name": "inference_args",
-          "type": "INFERENCE_ARGS",
-          "links": [
-            3
-          ]
-        }
-      ],
-      "properties": {
-        "Node name for S&R": "InferenceArgs"
-      },
-      "widgets_values": [
-        720,
-        1280,
-        45,
-        6,
-        -99999,
-        -99999,
-        1025,
-        "fixed",
-        24,
-        -99999,
-        -99999
-      ],
-      "auto_widget_states": {
-        "height": {
-          "isAuto": false,
-          "value": 720,
-          "cachedValue": 720
-        },
-        "width": {
-          "isAuto": false,
-          "value": 1280,
-          "cachedValue": 1280
-        },
-        "num_frames": {
-          "isAuto": false,
-          "value": 45,
-          "cachedValue": 45
-        },
-        "num_inference_steps": {
-          "isAuto": false,
-          "value": 6,
-          "cachedValue": 6
-        },
-        "guidance_scale": {
-          "isAuto": true,
-          "value": -99999,
-          "cachedValue": 1
-        },
-        "flow_shift": {
-          "isAuto": true,
-          "value": -99999,
-          "cachedValue": 17
-        },
-        "seed": {
-          "isAuto": false,
-          "value": 1025,
-          "cachedValue": 1024
-        },
-        "fps": {
-          "isAuto": false,
-          "value": 24,
-          "cachedValue": 24
-        },
-        "image_path": {
-          "isAuto": true,
-          "value": -99999,
-          "cachedValue": "X://insert/path/here.mp4"
-        },
-        "enable_teacache": {
-          "isAuto": true,
-          "value": -99999,
-          "cachedValue": true
-        }
-      }
-    },
-    {
-      "id": 8,
-      "type": "VHS_VideoCombine",
-      "pos": [
-        1668.3499755859375,
-        328.22625732421875
-      ],
-      "size": [
-        507.507080078125,
-        622.2227172851562
-      ],
-      "flags": {},
-      "order": 6,
-      "mode": 0,
-      "inputs": [
-        {
-          "name": "images",
-          "type": "IMAGE",
-          "link": 5
-        },
-        {
-          "name": "audio",
-          "shape": 7,
-          "type": "AUDIO",
-          "link": null
-        },
-        {
-          "name": "meta_batch",
-          "shape": 7,
-          "type": "VHS_BatchManager",
-          "link": null
-        },
-        {
-          "name": "vae",
-          "shape": 7,
-          "type": "VAE",
-          "link": null
-        }
-      ],
-      "outputs": [
-        {
-          "name": "Filenames",
-          "type": "VHS_FILENAMES",
-          "links": null
-        }
-      ],
-      "properties": {
-        "Node name for S&R": "VHS_VideoCombine"
-      },
-      "widgets_values": {
-        "frame_rate": 24,
-        "loop_count": 0,
-        "filename_prefix": "",
-        "format": "video/h264-mp4",
-        "pix_fmt": "yuv420p",
-        "crf": 19,
-        "save_metadata": true,
-        "trim_to_audio": false,
-        "pingpong": false,
-        "save_output": false,
-        "videopreview": {
-          "hidden": false,
-          "paused": false,
-          "params": {
-            "filename": "._00003.mp4",
-            "subfolder": "",
-            "type": "temp",
-            "format": "video/h264-mp4",
-            "frame_rate": 24,
-            "workflow": "._00003.png",
-            "fullpath": "/workspace/ComfyUI/temp/._00003.mp4"
-          }
-        }
-      }
-    }
-  ],
-  "links": [
-    [
-      2,
-      4,
-      0,
-      1,
-      1,
-      "VAE_CONFIG"
-    ],
-    [
-      3,
-      2,
-      0,
-      1,
-      0,
-      "INFERENCE_ARGS"
-    ],
-    [
-      4,
-      1,
-      0,
-      3,
-      2,
-      "STRING"
-    ],
-    [
-      5,
-      3,
-      0,
-      8,
-      0,
-      "IMAGE"
-    ],
-    [
-      6,
-      6,
-      0,
-      1,
-      3,
-      "DIT_CONFIG"
-    ],
-    [
-      7,
-      5,
-      0,
-      1,
-      2,
-      "TEXT_ENCODER_CONFIG"
-    ]
-  ],
-  "groups": [],
-  "config": {},
-  "extra": {
-    "ds": {
-      "scale": 0.9090909090909091,
-      "offset": [
-        112.86678372727341,
-        -71.45635903989245
-      ]
-    },
-    "frontendVersion": "1.20.4",
-    "VHS_latentpreview": false,
-    "VHS_latentpreviewrate": 0,
-    "VHS_MetadataImage": true,
-    "VHS_KeepIntermediate": true
-  },
-  "version": 0.4
-}

comfyui/examples/Wan2.1-I2V-14B-480P-Diffusers.json

@@ -1,697 +0,0 @@
-{
-  "id": "23a1f065-bbba-4a8f-b144-944e1318fcbf",
-  "revision": 0,
-  "last_node_id": 8,
-  "last_link_id": 7,
-  "nodes": [
-    {
-      "id": 7,
-      "type": "LoadImagePath",
-      "pos": [
-        33.15385437011719,
-        191.2037353515625
-      ],
-      "size": [
-        270,
-        334
-      ],
-      "flags": {},
-      "order": 0,
-      "mode": 0,
-      "inputs": [],
-      "outputs": [
-        {
-          "name": "image_path",
-          "type": "STRING",
-          "links": [
-            1
-          ]
-        },
-        {
-          "name": "IMAGE",
-          "type": "IMAGE",
-          "links": null
-        },
-        {
-          "name": "MASK",
-          "type": "MASK",
-          "links": null
-        }
-      ],
-      "properties": {
-        "Node name for S&R": "LoadImagePath"
-      },
-      "widgets_values": [
-        "woman.jpg",
-        "image"
-      ]
-    },
-    {
-      "id": 3,
-      "type": "VHS_LoadVideoPath",
-      "pos": [
-        1350.136962890625,
-        331.20361328125
-      ],
-      "size": [
-        231.8896484375,
-        286
-      ],
-      "flags": {},
-      "order": 6,
-      "mode": 0,
-      "inputs": [
-        {
-          "name": "meta_batch",
-          "shape": 7,
-          "type": "VHS_BatchManager",
-          "link": null
-        },
-        {
-          "name": "vae",
-          "shape": 7,
-          "type": "VAE",
-          "link": null
-        },
-        {
-          "name": "video",
-          "type": "STRING",
-          "widget": {
-            "name": "video"
-          },
-          "link": 4
-        }
-      ],
-      "outputs": [
-        {
-          "name": "IMAGE",
-          "type": "IMAGE",
-          "links": [
-            5
-          ]
-        },
-        {
-          "name": "frame_count",
-          "type": "INT",
-          "links": null
-        },
-        {
-          "name": "audio",
-          "type": "AUDIO",
-          "links": null
-        },
-        {
-          "name": "video_info",
-          "type": "VHS_VIDEOINFO",
-          "links": null
-        }
-      ],
-      "properties": {
-        "Node name for S&R": "VHS_LoadVideoPath"
-      },
-      "widgets_values": {
-        "video": "",
-        "force_rate": 0,
-        "custom_width": 0,
-        "custom_height": 0,
-        "frame_load_cap": 0,
-        "skip_first_frames": 0,
-        "select_every_nth": 1,
-        "format": "Wan",
-        "videopreview": {
-          "hidden": false,
-          "paused": false,
-          "params": {
-            "filename": "",
-            "type": "path",
-            "format": "video/",
-            "force_rate": 0,
-            "custom_width": 0,
-            "custom_height": 0,
-            "frame_load_cap": 0,
-            "skip_first_frames": 0,
-            "select_every_nth": 1
-          }
-        }
-      }
-    },
-    {
-      "id": 8,
-      "type": "VHS_VideoCombine",
-      "pos": [
-        1668.3499755859375,
-        328.22625732421875
-      ],
-      "size": [
-        214.7587890625,
-        334
-      ],
-      "flags": {},
-      "order": 7,
-      "mode": 0,
-      "inputs": [
-        {
-          "name": "images",
-          "type": "IMAGE",
-          "link": 5
-        },
-        {
-          "name": "audio",
-          "shape": 7,
-          "type": "AUDIO",
-          "link": null
-        },
-        {
-          "name": "meta_batch",
-          "shape": 7,
-          "type": "VHS_BatchManager",
-          "link": null
-        },
-        {
-          "name": "vae",
-          "shape": 7,
-          "type": "VAE",
-          "link": null
-        }
-      ],
-      "outputs": [
-        {
-          "name": "Filenames",
-          "type": "VHS_FILENAMES",
-          "links": null
-        }
-      ],
-      "properties": {
-        "Node name for S&R": "VHS_VideoCombine"
-      },
-      "widgets_values": {
-        "frame_rate": 24,
-        "loop_count": 0,
-        "filename_prefix": "",
-        "format": "video/h264-mp4",
-        "pix_fmt": "yuv420p",
-        "crf": 19,
-        "save_metadata": true,
-        "trim_to_audio": false,
-        "pingpong": false,
-        "save_output": false,
-        "videopreview": {
-          "hidden": false,
-          "paused": false,
-          "params": {}
-        }
-      }
-    },
-    {
-      "id": 4,
-      "type": "VAEConfig",
-      "pos": [
-        374.2159423828125,
-        554.85888671875
-      ],
-      "size": [
-        334.080078125,
-        322
-      ],
-      "flags": {},
-      "order": 1,
-      "mode": 0,
-      "inputs": [],
-      "outputs": [
-        {
-          "name": "vae_config",
-          "type": "VAE_CONFIG",
-          "links": [
-            2
-          ]
-        }
-      ],
-      "properties": {
-        "Node name for S&R": "VAEConfig"
-      },
-      "widgets_values": [
-        true,
-        true,
-        256,
-        256,
-        16,
-        192,
-        192,
-        12,
-        0,
-        true,
-        true,
-        true
-      ],
-      "auto_widget_states": {
-        "load_encoder": {
-          "isAuto": true,
-          "value": true,
-          "cachedValue": true
-        },
-        "load_decoder": {
-          "isAuto": true,
-          "value": true,
-          "cachedValue": true
-        },
-        "tile_sample_min_height": {
-          "isAuto": true,
-          "value": 256,
-          "cachedValue": 256
-        },
-        "tile_sample_min_width": {
-          "isAuto": true,
-          "value": 256,
-          "cachedValue": 256
-        },
-        "tile_sample_min_num_frames": {
-          "isAuto": true,
-          "value": 16,
-          "cachedValue": 16
-        },
-        "tile_sample_stride_height": {
-          "isAuto": true,
-          "value": 192,
-          "cachedValue": 192
-        },
-        "tile_sample_stride_width": {
-          "isAuto": true,
-          "value": 192,
-          "cachedValue": 192
-        },
-        "tile_sample_stride_num_frames": {
-          "isAuto": true,
-          "value": 12,
-          "cachedValue": 12
-        },
-        "blend_num_frames": {
-          "isAuto": true,
-          "value": 0,
-          "cachedValue": 0
-        },
-        "use_tiling": {
-          "isAuto": true,
-          "value": true,
-          "cachedValue": true
-        },
-        "use_temporal_tiling": {
-          "isAuto": true,
-          "value": true,
-          "cachedValue": true
-        },
-        "use_parallel_tiling": {
-          "isAuto": true,
-          "value": true,
-          "cachedValue": true
-        }
-      }
-    },
-    {
-      "id": 2,
-      "type": "InferenceArgs",
-      "pos": [
-        411.46307373046875,
-        178.18182373046875
-      ],
-      "size": [
-        278.73828125,
-        298
-      ],
-      "flags": {},
-      "order": 4,
-      "mode": 0,
-      "inputs": [
-        {
-          "name": "image_path",
-          "shape": 7,
-          "type": "STRING",
-          "widget": {
-            "name": "image_path"
-          },
-          "link": 1
-        }
-      ],
-      "outputs": [
-        {
-          "name": "inference_args",
-          "type": "INFERENCE_ARGS",
-          "links": [
-            3
-          ]
-        }
-      ],
-      "properties": {
-        "Node name for S&R": "InferenceArgs"
-      },
-      "widgets_values": [
-        832,
-        480,
-        45,
-        20,
-        1,
-        17,
-        1024,
-        "fixed",
-        24,
-        "X://insert/path/here.mp4",
-        true
-      ],
-      "auto_widget_states": {
-        "height": {
-          "isAuto": false,
-          "value": 832,
-          "cachedValue": 720
-        },
-        "width": {
-          "isAuto": false,
-          "value": 480,
-          "cachedValue": 1280
-        },
-        "num_frames": {
-          "isAuto": false,
-          "value": 45,
-          "cachedValue": 45
-        },
-        "num_inference_steps": {
-          "isAuto": false,
-          "value": 20,
-          "cachedValue": 6
-        },
-        "guidance_scale": {
-          "isAuto": true,
-          "value": 1,
-          "cachedValue": 1
-        },
-        "flow_shift": {
-          "isAuto": true,
-          "value": 17,
-          "cachedValue": 17
-        },
-        "seed": {
-          "isAuto": false,
-          "value": 1024,
-          "cachedValue": 1024
-        },
-        "fps": {
-          "isAuto": false,
-          "value": 24,
-          "cachedValue": 24
-        },
-        "image_path": {
-          "isAuto": true,
-          "value": "X://insert/path/here.mp4",
-          "cachedValue": "X://insert/path/here.mp4"
-        },
-        "enable_teacache": {
-          "isAuto": true,
-          "value": true,
-          "cachedValue": true
-        }
-      }
-    },
-    {
-      "id": 1,
-      "type": "VideoGenerator",
-      "pos": [
-        818.804931640625,
-        348.9299621582031
-      ],
-      "size": [
-        400,
-        436
-      ],
-      "flags": {},
-      "order": 5,
-      "mode": 0,
-      "inputs": [
-        {
-          "name": "inference_args",
-          "shape": 7,
-          "type": "INFERENCE_ARGS",
-          "link": 3
-        },
-        {
-          "name": "vae_config",
-          "shape": 7,
-          "type": "VAE_CONFIG",
-          "link": 2
-        },
-        {
-          "name": "text_encoder_config",
-          "shape": 7,
-          "type": "TEXT_ENCODER_CONFIG",
-          "link": 7
-        },
-        {
-          "name": "dit_config",
-          "shape": 7,
-          "type": "DIT_CONFIG",
-          "link": 6
-        }
-      ],
-      "outputs": [
-        {
-          "name": "video_path",
-          "type": "STRING",
-          "links": [
-            4
-          ]
-        }
-      ],
-      "properties": {
-        "Node name for S&R": "VideoGenerator"
-      },
-      "widgets_values": [
-        "A woman crying from laughter.",
-        "/workspace/ComfyUI/outputs_video/",
-        4,
-        "Wan-AI/Wan2.1-I2V-14B-480P-Diffusers",
-        6,
-        2,
-        2,
-        "fp16",
-        true,
-        true,
-        "fp16",
-        "fp16",
-        true
-      ],
-      "auto_widget_states": {
-        "embedded_cfg_scale": {
-          "isAuto": true,
-          "value": 6,
-          "cachedValue": 6
-        },
-        "sp_size": {
-          "isAuto": true,
-          "value": 2,
-          "cachedValue": 2
-        },
-        "tp_size": {
-          "isAuto": true,
-          "value": 2,
-          "cachedValue": 2
-        },
-        "vae_precision": {
-          "isAuto": true,
-          "value": "fp16",
-          "cachedValue": "fp16"
-        },
-        "vae_tiling": {
-          "isAuto": true,
-          "value": true,
-          "cachedValue": true
-        },
-        "vae_sp": {
-          "isAuto": true,
-          "value": true,
-          "cachedValue": true
-        },
-        "text_encoder_precision": {
-          "isAuto": true,
-          "value": "fp16",
-          "cachedValue": "fp16"
-        },
-        "precision": {
-          "isAuto": true,
-          "value": "fp16",
-          "cachedValue": "fp16"
-        },
-        "dit_cpu_offload": {
-          "isAuto": true,
-          "value": true,
-          "cachedValue": true
-        }
-      }
-    },
-    {
-      "id": 5,
-      "type": "TextEncoderConfig",
-      "pos": [
-        416.4937744140625,
-        953.6171875
-      ],
-      "size": [
-        270,
-        106
-      ],
-      "flags": {},
-      "order": 2,
-      "mode": 0,
-      "inputs": [],
-      "outputs": [
-        {
-          "name": "text_encoder_config",
-          "type": "TEXT_ENCODER_CONFIG",
-          "links": [
-            7
-          ]
-        }
-      ],
-      "properties": {
-        "Node name for S&R": "TextEncoderConfig"
-      },
-      "widgets_values": [
-        "",
-        "",
-        ""
-      ],
-      "auto_widget_states": {
-        "prefix": {
-          "isAuto": true,
-          "value": "",
-          "cachedValue": ""
-        },
-        "quant_config": {
-          "isAuto": true,
-          "value": "",
-          "cachedValue": ""
-        },
-        "lora_config": {
-          "isAuto": true,
-          "value": "",
-          "cachedValue": ""
-        }
-      }
-    },
-    {
-      "id": 6,
-      "type": "DITConfig",
-      "pos": [
-        415.1928405761719,
-        1154.1573486328125
-      ],
-      "size": [
-        270,
-        82
-      ],
-      "flags": {},
-      "order": 3,
-      "mode": 0,
-      "inputs": [],
-      "outputs": [
-        {
-          "name": "dit_config",
-          "type": "DIT_CONFIG",
-          "links": [
-            6
-          ]
-        }
-      ],
-      "properties": {
-        "Node name for S&R": "DITConfig"
-      },
-      "widgets_values": [
-        "",
-        ""
-      ],
-      "auto_widget_states": {
-        "prefix": {
-          "isAuto": true,
-          "value": "",
-          "cachedValue": ""
-        },
-        "quant_config": {
-          "isAuto": true,
-          "value": "",
-          "cachedValue": ""
-        }
-      }
-    }
-  ],
-  "links": [
-    [
-      1,
-      7,
-      0,
-      2,
-      0,
-      "STRING"
-    ],
-    [
-      2,
-      4,
-      0,
-      1,
-      1,
-      "VAE_CONFIG"
-    ],
-    [
-      3,
-      2,
-      0,
-      1,
-      0,
-      "INFERENCE_ARGS"
-    ],
-    [
-      4,
-      1,
-      0,
-      3,
-      2,
-      "STRING"
-    ],
-    [
-      5,
-      3,
-      0,
-      8,
-      0,
-      "IMAGE"
-    ],
-    [
-      6,
-      6,
-      0,
-      1,
-      3,
-      "DIT_CONFIG"
-    ],
-    [
-      7,
-      5,
-      0,
-      1,
-      2,
-      "TEXT_ENCODER_CONFIG"
-    ]
-  ],
-  "groups": [],
-  "config": {},
-  "extra": {
-    "ds": {
-      "scale": 0.8264462809917354,
-      "offset": [
-        646.7950212991898,
-        66.17259910028655
-      ]
-    },
-    "frontendVersion": "1.20.4",
-    "VHS_latentpreview": false,
-    "VHS_latentpreviewrate": 0,
-    "VHS_MetadataImage": true,
-    "VHS_KeepIntermediate": true
-  },
-  "version": 0.4
-}

comfyui/video_generator/dit_config.py

@@ -1,31 +0,0 @@
-class DITConfig:
-
-    @classmethod
-    def INPUT_TYPES(cls):
-        return {
-            "optional": {
-                "prefix": ("STRING", {
-                    "default": ""
-                }),
-                "quant_config": ("STRING", {
-                    "default": ""
-                }),
-            }
-        }
-
-    @classmethod
-    def VALIDATE_INPUTS(cls, **kwargs):
-        return True
-
-    RETURN_TYPES = ("DIT_CONFIG", )
-    RETURN_NAMES = ("dit_config", )
-    FUNCTION = "set_args"
-    CATEGORY = "fastvideo"
-
-    def set_args(self, prefix, quant_config):
-        raw_args = {"prefix": prefix, "quant_config": quant_config}
-
-        # Filter out keys where value is -99999
-        args = {k: v for k, v in raw_args.items() if str(int(v)) != str(-99999)}
-
-        return (args, )

comfyui/video_generator/inference_args.py

@@ -1,89 +0,0 @@
-class InferenceArgs:
-
-    @classmethod
-    def INPUT_TYPES(cls):
-        return {
-            "optional": {
-                "height": ("INT", {
-                    "default": 720
-                }),
-                "width": ("INT", {
-                    "default": 1280
-                }),
-                "num_frames": ("INT", {
-                    "default": 45
-                }),
-                "num_inference_steps": ("INT", {
-                    "default": 6
-                }),
-                "guidance_scale": ("FLOAT", {
-                    "default": 1.0
-                }),
-                "flow_shift": ("INT", {
-                    "default": 17
-                }),
-                "seed": ("INT", {
-                    "default": 1024
-                }),
-                "fps": ("INT", {
-                    "default": 24
-                }),
-                "image_path": ("STRING", {
-                    "default": "X://insert/path/here.mp4"
-                }),
-                "enable_teacache": ([True, False], {
-                    "default": False
-                }),
-            }
-        }
-
-    @classmethod
-    def VALIDATE_INPUTS(cls, **kwargs):
-        return True
-
-    RETURN_TYPES = ("INFERENCE_ARGS", )
-    RETURN_NAMES = ("inference_args", )
-    FUNCTION = "set_args"
-    CATEGORY = "fastvideo"
-
-    def set_args(
-        self,
-        height,
-        width,
-        num_frames,
-        num_inference_steps,
-        guidance_scale,
-        flow_shift,
-        seed,
-        fps,
-        image_path,
-        enable_teacache,
-    ):
-        raw_args = {
-            "height": height,
-            "width": width,
-            "num_frames": num_frames,
-            "num_inference_steps": num_inference_steps,
-            "guidance_scale": guidance_scale,
-            "flow_shift": flow_shift,
-            "seed": seed,
-            "fps": fps,
-            "image_path": image_path,
-            "enable_teacache": enable_teacache,
-        }
-
-        # Filter out keys where value is -99999, handling different types properly
-        args = {}
-        for k, v in raw_args.items():
-            try:
-                if isinstance(v, str):
-                    if v != "-99999":
-                        args[k] = v
-                elif v != -99999:
-                    # If it's not a string, compare directly
-                    args[k] = v
-            except (ValueError, TypeError):
-                # Include any value that causes an error in comparison
-                args[k] = v
-
-        return (args, )

comfyui/video_generator/load_image.py

@@ -1,103 +0,0 @@
-import hashlib
-import os
-
-import folder_paths
-import numpy as np
-import torch
-from PIL import Image, ImageOps, ImageSequence
-
-from .node_helpers import pillow
-
-
-class LoadImagePath:
-
-    @classmethod
-    def INPUT_TYPES(s):
-        input_dir = folder_paths.get_input_directory()
-        files = [
-            f for f in os.listdir(input_dir)
-            if os.path.isfile(os.path.join(input_dir, f))
-        ]
-        files = folder_paths.filter_files_content_types(files, ["image"])
-        return {
-            "required": {
-                "image": (sorted(files), {
-                    "image_upload": True
-                })
-            },
-        }
-
-    CATEGORY = "fastvideo"
-
-    RETURN_TYPES = ("STRING", "IMAGE", "MASK")
-    RETURN_NAMES = ("image_path", "IMAGE", "MASK")
-    FUNCTION = "load_image"
-
-    def load_image(self, image):
-        image_path = folder_paths.get_annotated_filepath(image)
-
-        img = pillow(Image.open, image_path)
-
-        output_images: list[torch.Tensor] = []
-        output_masks: list[torch.Tensor] = []
-        w, h = None, None
-
-        excluded_formats = ['MPO']
-
-        for i in ImageSequence.Iterator(img):
-            processed_image = pillow(ImageOps.exif_transpose, i)
-            if processed_image is None:
-                continue
-
-            if processed_image.mode == 'I':
-                processed_image = processed_image.point(lambda i: i * (1 / 255))
-            image = processed_image.convert("RGB")
-
-            if len(output_images) == 0:
-                w = image.size[0]
-                h = image.size[1]
-
-            if image.size[0] != w or image.size[1] != h:
-                continue
-
-            image = np.array(image).astype(np.float32) / 255.0
-            image = torch.from_numpy(image)[
-                None,
-            ]
-            if 'A' in processed_image.getbands():
-                mask = np.array(processed_image.getchannel('A')).astype(
-                    np.float32) / 255.0
-                mask = 1. - torch.from_numpy(mask)
-            elif processed_image.mode == 'P' and 'transparency' in processed_image.info:
-                mask = np.array(
-                    processed_image.convert('RGBA').getchannel('A')).astype(
-                        np.float32) / 255.0
-                mask = 1. - torch.from_numpy(mask)
-            else:
-                mask = torch.zeros((64, 64), dtype=torch.float32, device="cpu")
-            output_images.append(image)
-            output_masks.append(mask.unsqueeze(0))
-
-        if len(output_images) > 1 and img.format not in excluded_formats:
-            output_image = torch.cat(output_images, dim=0)
-            output_mask = torch.cat(output_masks, dim=0)
-        else:
-            output_image = output_images[0]
-            output_mask = output_masks[0]
-
-        return (image_path, output_image, output_mask)
-
-    @classmethod
-    def IS_CHANGED(s, image):
-        image_path = folder_paths.get_annotated_filepath(image)
-        m = hashlib.sha256()
-        with open(image_path, 'rb') as f:
-            m.update(f.read())
-        return m.digest().hex()
-
-    @classmethod
-    def VALIDATE_INPUTS(s, image):
-        if not folder_paths.exists_annotated_filepath(image):
-            return "Invalid image file: {}".format(image)
-
-        return True

comfyui/video_generator/node_helpers.py

@@ -1,68 +0,0 @@
-import hashlib
-from collections.abc import Callable
-from typing import Any, TypeVar
-
-import torch
-from comfy.cli_args import args
-from PIL import ImageFile, UnidentifiedImageError
-
-T = TypeVar('T')
-
-
-def conditioning_set_values(conditioning: list[Any],
-                            values: dict[str, Any] | None = None) -> list[Any]:
-    if values is None:
-        values = {}
-    c = []
-    for t in conditioning:
-        n = [t[0], t[1].copy()]
-        for k in values:
-            n[1][k] = values[k]
-        c.append(n)
-
-    return c
-
-
-def pillow(fn: Callable[[Any], T], arg: Any) -> T:
-    prev_value = None
-    try:
-        x = fn(arg)
-    except (OSError, UnidentifiedImageError, ValueError
-            ):  #PIL issues #4472 and #2445, also fixes ComfyUI issue #3416
-        prev_value = ImageFile.LOAD_TRUNCATED_IMAGES
-        ImageFile.LOAD_TRUNCATED_IMAGES = True
-        x = fn(arg)
-    finally:
-        if prev_value is not None:
-            ImageFile.LOAD_TRUNCATED_IMAGES = prev_value
-    return x
-
-
-def hasher() -> Callable[[], Any]:
-    hashfuncs = {
-        "md5": hashlib.md5,
-        "sha1": hashlib.sha1,
-        "sha256": hashlib.sha256,
-        "sha512": hashlib.sha512
-    }
-    return hashfuncs[args.default_hashing_function]
-
-
-def string_to_torch_dtype(string: str) -> torch.dtype | None:
-    if string == "fp32":
-        return torch.float32
-    if string == "fp16":
-        return torch.float16
-    if string == "bf16":
-        return torch.bfloat16
-    return None
-
-
-def image_alpha_fix(destination: torch.Tensor,
-                    source: torch.Tensor) -> tuple[torch.Tensor, torch.Tensor]:
-    if destination.shape[-1] < source.shape[-1]:
-        source = source[..., :destination.shape[-1]]
-    elif destination.shape[-1] > source.shape[-1]:
-        destination = torch.nn.functional.pad(destination, (0, 1))
-        destination[..., -1] = 1.0
-    return destination, source

comfyui/video_generator/nodes.py

@@ -1,24 +0,0 @@
-from .dit_config import DITConfig
-from .inference_args import InferenceArgs
-from .load_image import LoadImagePath
-from .text_encoder_config import TextEncoderConfig
-from .vae_config import VAEConfig
-from .video_generator import VideoGenerator
-
-NODE_CLASS_MAPPINGS = {
-    "VideoGenerator": VideoGenerator,
-    "InferenceArgs": InferenceArgs,
-    "VAEConfig": VAEConfig,
-    "TextEncoderConfig": TextEncoderConfig,
-    "DITConfig": DITConfig,
-    "LoadImagePath": LoadImagePath
-}
-
-NODE_DISPLAY_NAME_MAPPINGS = {
-    "VideoGenerator": "Video Generator",
-    "InferenceArgs": "Inference Args",
-    "VAEConfig": "VAE Config",
-    "TextEncoderConfig": "Text Encoder Config",
-    "DITConfig": "DIT Config",
-    "LoadImagePath": "Load Image Path"
-}

comfyui/video_generator/text_encoder_config.py

@@ -1,38 +0,0 @@
-class TextEncoderConfig:
-
-    @classmethod
-    def INPUT_TYPES(cls):
-        return {
-            "optional": {
-                "prefix": ("STRING", {
-                    "default": ""
-                }),
-                "quant_config": ("STRING", {
-                    "default": ""
-                }),
-                "lora_config": ("STRING", {
-                    "default": ""
-                }),
-            }
-        }
-
-    @classmethod
-    def VALIDATE_INPUTS(cls, **kwargs):
-        return True
-
-    RETURN_TYPES = ("TEXT_ENCODER_CONFIG", )
-    RETURN_NAMES = ("text_encoder_config", )
-    FUNCTION = "set_args"
-    CATEGORY = "fastvideo"
-
-    def set_args(self, prefix, quant_config, lora_config):
-        raw_args = {
-            "prefix": prefix,
-            "quant_config": quant_config,
-            "lora_config": lora_config
-        }
-
-        # Filter out keys where value is -99999
-        args = {k: v for k, v in raw_args.items() if str(int(v)) != str(-99999)}
-
-        return (args, )

comfyui/video_generator/vae_config.py

@@ -1,88 +0,0 @@
-class VAEConfig:
-
-    @classmethod
-    def INPUT_TYPES(cls):
-        return {
-            "optional": {
-                "load_encoder": ([True, False], {
-                    "default": True
-                }),
-                "load_decoder": ([True, False], {
-                    "default": True
-                }),
-                "tile_sample_min_height": ("INT", {
-                    "default": 256
-                }),
-                "tile_sample_min_width": ("INT", {
-                    "default": 256
-                }),
-                "tile_sample_min_num_frames": ("INT", {
-                    "default": 16
-                }),
-                "tile_sample_stride_height": ("INT", {
-                    "default": 192
-                }),
-                "tile_sample_stride_width": ("INT", {
-                    "default": 192
-                }),
-                "tile_sample_stride_num_frames": ("INT", {
-                    "default": 12
-                }),
-                "blend_num_frames": ("INT", {
-                    "default": 0
-                }),
-                "use_tiling": ([True, False], {
-                    "default": True
-                }),
-                "use_temporal_tiling": ([True, False], {
-                    "default": True
-                }),
-                "use_parallel_tiling": ([True, False], {
-                    "default": True
-                }),
-            }
-        }
-
-    @classmethod
-    def VALIDATE_INPUTS(cls, **kwargs):
-        return True
-
-    RETURN_TYPES = ("VAE_CONFIG", )
-    RETURN_NAMES = ("vae_config", )
-    FUNCTION = "set_args"
-    CATEGORY = "fastvideo"
-
-    def set_args(
-        self,
-        load_encoder,
-        load_decoder,
-        tile_sample_min_height,
-        tile_sample_min_width,
-        tile_sample_min_num_frames,
-        tile_sample_stride_height,
-        tile_sample_stride_width,
-        tile_sample_stride_num_frames,
-        blend_num_frames,
-        use_tiling,
-        use_temporal_tiling,
-        use_parallel_tiling,
-    ):
-        raw_args = {
-            "load_encoder": load_encoder,
-            "load_decoder": load_decoder,
-            "tile_sample_min_height": tile_sample_min_height,
-            "tile_sample_min_width": tile_sample_min_width,
-            "tile_sample_min_num_frames": tile_sample_min_num_frames,
-            "tile_sample_stride_height": tile_sample_stride_height,
-            "tile_sample_stride_width": tile_sample_stride_width,
-            "tile_sample_stride_num_frames": tile_sample_stride_num_frames,
-            "blend_num_frames": blend_num_frames,
-            "use_tiling": use_tiling,
-            "use_temporal_tiling": use_temporal_tiling,
-            "use_parallel_tiling": use_parallel_tiling,
-        }
-
-        # Filter out any value explicitly set to -99999
-        args = {k: v for k, v in raw_args.items() if str(int(v)) != str(-99999)}
-
-        return (args, )

comfyui/video_generator/video_generator.py

@@ -1,315 +0,0 @@
-from __future__ import annotations
-
-import glob
-import os
-import signal
-import sys
-import threading
-import time
-from typing import Any
-
-from comfy.model_management import processing_interrupted
-
-from fastvideo import PipelineConfig
-from fastvideo import VideoGenerator as FastVideoGenerator
-
-sys.path.insert(
-    0,
-    os.path.dirname(
-        os.path.dirname(
-            os.path.dirname(os.path.dirname(os.path.abspath(__file__))))))
-
-
-# Custom exception for interruption
-class GenerationInterruptedException(Exception):
-    pass
-
-
-# Custom exception for interruption that ComfyUI will recognize
-class GenerationCancelledException(Exception):
-
-    def __init__(self,
-                 message: str = "Generation was cancelled by user") -> None:
-        self.message = message
-        super().__init__(self.message)
-
-
-def update_config_from_args(config: Any, args_dict: dict[str, Any]) -> None:
-    """
-    Update configuration object from arguments dictionary.
-    
-    Args:
-        config: The configuration object to update
-        args_dict: Dictionary containing arguments
-    """
-    for key, value in args_dict.items():
-        if hasattr(config, key) and value is not None:
-            if key == "text_encoder_precisions" and isinstance(value, list):
-                setattr(config, key, tuple(value))
-            else:
-                setattr(config, key, value)
-
-
-class VideoGenerator:
-
-    @classmethod
-    def INPUT_TYPES(s):
-        return {
-            "required": {
-                "prompt": ("STRING", {
-                    "multiline":
-                    True,
-                    "default":
-                    "A ripe orange tumbles gently from a tree and lands on the head of a lounging capybara, "
-                    "who blinks slowly in response. The moment is quietly humorous and oddly serene, framed by "
-                    "lush green foliage and dappled sunlight. Mid-shot, warm and whimsical tones."
-                }),
-                "output_path": ("STRING", {
-                    "default": "/workspace/ComfyUI/outputs_video/"
-                }),
-                "num_gpus": ("INT", {
-                    "default": 2,
-                    "min": 1,
-                    "max": 16
-                }),
-                "model_path": ("STRING", {
-                    "default": "FastVideo/FastHunyuan-diffusers"
-                })
-            },
-            "optional": {
-                "inference_args": ("INFERENCE_ARGS", ),
-                "embedded_cfg_scale": ("FLOAT", {
-                    "default": 6.0
-                }),
-                "sp_size": ("INT", {
-                    "default": 2
-                }),
-                "tp_size": ("INT", {
-                    "default": 2
-                }),
-                "vae_config": ("VAE_CONFIG", ),
-                "vae_precision": (["fp16", "bf16"], {
-                    "default": "fp16"
-                }),
-                "vae_tiling": ([True, False], {
-                    "default": True
-                }),
-                "vae_sp": ([True, False], {
-                    "default": False
-                }),
-                "text_encoder_config": ("TEXT_ENCODER_CONFIG", ),
-                "text_encoder_precision": (["fp16", "bf16"], {
-                    "default": "fp16"
-                }),
-                "dit_config": ("DIT_CONFIG", ),
-                "precision": (["fp16", "bf16"], {
-                    "default": "fp16"
-                }),
-                "dit_cpu_offload": ([True, False], {
-                    "default": False
-                }),
-            }
-        }
-
-    @classmethod
-    def VALIDATE_INPUTS(cls, **kwargs):
-        return True
-
-    RETURN_TYPES = ("STRING", )
-    RETURN_NAMES = ("video_path", )
-    FUNCTION = "launch_inference"
-    CATEGORY = "fastvideo"
-
-    generator: FastVideoGenerator | None = None
-    _interrupt_thread: threading.Thread | None = None
-    _generation_active: bool = False
-    _generation_interrupted: bool = False
-    _interrupt_event: threading.Event = threading.Event()
-    _generation_thread: threading.Thread | None = None
-    _generation_result: str | None = None
-    _generation_exception: Exception | None = None
-
-    def _monitor_for_interruption(self):
-        """Background thread that monitors for interruption requests"""
-        time.sleep(2)  # Give the generation thread time to send execute_forward
-
-        while self._generation_active and not self._interrupt_event.is_set():
-            if processing_interrupted():
-                print("Video generation interrupted by user")
-                self._generation_interrupted = True
-
-                # Try to send interrupt signal to worker processes
-                if self.generator is not None and hasattr(
-                        self.generator, 'executor'):
-                    try:
-                        # The MultiprocExecutor has a workers attribute
-                        if hasattr(self.generator.executor, 'workers'):
-                            for worker in self.generator.executor.workers:
-                                if worker.is_alive():
-                                    os.kill(worker.pid, signal.SIGINT)
-                            print("Interrupt signal sent to worker processes")
-                    except Exception as e:
-                        print(f"Error sending interrupt signal: {e}")
-
-                # Set the interrupt event to notify other threads
-                self._interrupt_event.set()
-                break
-            time.sleep(0.5)
-
-    def _run_generation(self, prompt: str, output_path: str,
-                        inference_args: dict[str, Any]) -> None:
-        """Thread function to run the generation"""
-        try:
-            if self.generator is not None:
-                self.generator.generate_video(prompt=prompt,
-                                              output_path=output_path,
-                                              **inference_args)
-                self._generation_result = os.path.join(output_path,
-                                                       f"{prompt[:100]}.mp4")
-            else:
-                raise RuntimeError("Generator is not initialized")
-        except Exception as e:
-            self._generation_exception = e
-            self._interrupt_event.set()
-
-    def load_output_video(self, output_dir):
-        video_extensions = ["*.mp4", "*.avi", "*.mov", "*.mkv"]
-        video_files = []
-
-        for ext in video_extensions:
-            video_files.extend(glob.glob(os.path.join(output_dir, ext)))
-
-        if not video_files:
-            print("No video files found in output directory: %s", output_dir)
-            return ""
-
-        video_files.sort()
-        return video_files[0]
-
-    def launch_inference(
-        self,
-        prompt,
-        output_path,
-        num_gpus,
-        model_path,
-        embedded_cfg_scale,
-        sp_size,
-        tp_size,
-        vae_precision,
-        vae_tiling,
-        vae_sp,
-        text_encoder_precision,
-        precision,
-        inference_args=None,
-        vae_config=None,
-        text_encoder_config=None,
-        dit_config=None,
-        dit_cpu_offload=None,
-    ):
-        print('Running FastVideo inference')
-
-        # Reset interruption flag and event
-        self._generation_interrupted = False
-        self._interrupt_event.clear()
-        self._generation_result = None
-        self._generation_exception = None
-
-        # Load pipeline config from model path
-        pipeline_config = PipelineConfig.from_pretrained(model_path)
-        print('pipeline_config', pipeline_config)
-
-        # Update configs with provided config dictionaries
-        if dit_config is not None:
-            update_config_from_args(pipeline_config.dit_config, dit_config)
-
-        if vae_config is not None:
-            update_config_from_args(pipeline_config.vae_config, vae_config)
-
-        if text_encoder_config is not None:
-            update_config_from_args(pipeline_config.text_encoder_configs,
-                                    text_encoder_config)
-
-        # Update top-level pipeline config with remaining arguments
-        raw_pipeline_args = {}
-        if embedded_cfg_scale is not None:
-            raw_pipeline_args['embedded_cfg_scale'] = embedded_cfg_scale
-        if precision is not None:
-            raw_pipeline_args['precision'] = precision
-        if vae_precision is not None:
-            raw_pipeline_args['vae_precision'] = vae_precision
-        if vae_tiling is not None:
-            raw_pipeline_args['vae_tiling'] = vae_tiling
-        if vae_sp is not None:
-            raw_pipeline_args['vae_sp'] = vae_sp
-        if text_encoder_precision is not None:
-            raw_pipeline_args['text_encoder_precision'] = text_encoder_precision
-
-        # Filter out any value explicitly set to -99999 (auto values)
-        pipeline_args = {
-            k: v
-            for k, v in raw_pipeline_args.items() if str(int(v)) != str(-99999)
-        }
-
-        update_config_from_args(pipeline_config, pipeline_args)
-
-        raw_generation_args = {}
-        if num_gpus is not None:
-            raw_generation_args['num_gpus'] = num_gpus
-        if tp_size is not None:
-            raw_generation_args['tp_size'] = tp_size
-        if sp_size is not None:
-            raw_generation_args['sp_size'] = sp_size
-        if dit_cpu_offload is not None:
-            raw_generation_args['dit_cpu_offload'] = dit_cpu_offload
-
-        generation_args = {
-            k: v
-            for k, v in raw_generation_args.items()
-            if str(int(v)) != str(-99999)
-        }
-
-        if self.generator is None:
-            print('generation_args', generation_args)
-            print('pipeline_config', pipeline_config)
-            self.generator = FastVideoGenerator.from_pretrained(
-                model_path=model_path,
-                **generation_args,
-                pipeline_config=pipeline_config)
-
-        print('inference_args', inference_args)
-
-        # Start a thread to run the generation
-        self._generation_thread = threading.Thread(target=self._run_generation,
-                                                   args=(prompt, output_path,
-                                                         inference_args),
-                                                   daemon=True)
-        self._generation_thread.start()
-
-        # Start a background thread to monitor for interruptions
-        self._generation_active = True
-        self._interrupt_thread = threading.Thread(
-            target=self._monitor_for_interruption, daemon=True)
-        self._interrupt_thread.start()
-
-        # Wait for either completion or interruption
-        while self._generation_thread.is_alive(
-        ) and not self._interrupt_event.is_set():
-            self._generation_thread.join(timeout=0.5)
-
-        self._generation_active = False
-        if self._interrupt_thread:
-            self._interrupt_thread.join(timeout=1.0)
-            self._interrupt_thread = None
-
-        if self._generation_interrupted:
-            print("Video generation was cancelled by user")
-            raise GenerationCancelledException()
-        elif self._generation_exception:
-            # Re-raise the exception from the generation thread
-            raise self._generation_exception
-        elif self._generation_result:
-            return (self._generation_result, )
-        else:
-            # This shouldn't happen, but just in case
-            print("Generation completed but no result was produced")
-            raise Exception("Generation failed to produce a result")

comfyui/web/js/FastVideo.core.js

@@ -1,593 +0,0 @@
-import { app } from '../../../scripts/app.js'
-
-function chainCallback(object, property, callback) {
-    if (object == undefined) {
-        console.error("Tried to add callback to non-existent object");
-        return;
-    }
-    if (property in object && object[property]) {
-        const callback_orig = object[property];
-        object[property] = function () {
-            const r = callback_orig.apply(this, arguments);
-            return callback.apply(this, arguments) ?? r;
-        };
-    } else {
-        object[property] = callback;
-    }
-}
-
-function drawAutoAnnotated(ctx, node, widget_width, y, H) {
-    const litegraph_base = LiteGraph;
-    const show_text = app.canvas.ds.scale >= 0.5;
-    const margin = 15;
-
-    const autoTextWidth = 30;
-    const autoTextRightMargin = 5;
-
-    ctx.textAlign = 'left';
-    ctx.strokeStyle = litegraph_base.WIDGET_OUTLINE_COLOR;
-    ctx.fillStyle = litegraph_base.WIDGET_BGCOLOR;
-
-    ctx.beginPath();
-    if (show_text && ctx.roundRect) {
-        ctx.roundRect(margin, y, widget_width - margin * 2, H, [H * 0.5]);
-    } else {
-        ctx.rect(margin, y, widget_width - margin * 2, H);
-    }
-    ctx.fill();
-
-    if (show_text) {
-        if (!this.disabled) ctx.stroke();
-        const isAuto = this.isAuto === true;
-
-        ctx.save();
-        if (isAuto) {
-            ctx.fillStyle = litegraph_base.WIDGET_TEXT_COLOR;
-            ctx.strokeStyle = litegraph_base.WIDGET_TEXT_COLOR;
-        } else {
-            ctx.fillStyle = litegraph_base.WIDGET_SECONDARY_TEXT_COLOR;
-            ctx.strokeStyle = litegraph_base.WIDGET_SECONDARY_TEXT_COLOR;
-        }
-
-        // Position for the cog
-        const cogX = widget_width - autoTextRightMargin - autoTextWidth - 6;
-        const cogY = y + H * 0.5;
-        const cogRadius = 6;
-        const toothLength = 2;
-        const numTeeth = 8;
-        const holeRadius = 2; // Radius of the center hole
-
-        // Draw the cog
-        ctx.beginPath();
-        ctx.arc(cogX, cogY, cogRadius - toothLength, 0, Math.PI * 2);
-        ctx.fill();
-
-        // Draw the center hole (by clearing it)
-        ctx.beginPath();
-        ctx.arc(cogX, cogY, holeRadius, 0, Math.PI * 2);
-        ctx.fillStyle = litegraph_base.WIDGET_BGCOLOR;
-        ctx.fill();
-
-        // Reset fill style for the teeth
-        if (isAuto) {
-            ctx.fillStyle = litegraph_base.WIDGET_TEXT_COLOR;
-        } else {
-            ctx.fillStyle = litegraph_base.WIDGET_SECONDARY_TEXT_COLOR;
-        }
-
-        // Draw teeth
-        ctx.beginPath();
-        for (let i = 0; i < numTeeth; i++) {
-            const angle = (i / numTeeth) * Math.PI * 2;
-            const innerX = cogX + (cogRadius - toothLength) * Math.cos(angle);
-            const innerY = cogY + (cogRadius - toothLength) * Math.sin(angle);
-            const outerX = cogX + cogRadius * Math.cos(angle);
-            const outerY = cogY + cogRadius * Math.sin(angle);
-
-            ctx.moveTo(innerX, innerY);
-            ctx.lineTo(outerX, outerY);
-        }
-        ctx.lineWidth = 2;
-        ctx.stroke();
-
-        ctx.restore();
-
-        // Draw label
-        ctx.fillStyle = litegraph_base.WIDGET_SECONDARY_TEXT_COLOR;
-        const label = this.label || this.name;
-        if (label != null) {
-            ctx.fillText(label, margin * 2 + 5, y + H * 0.7);
-        }
-
-        // Draw value
-        ctx.textAlign = 'right';
-        const text = isAuto ? "auto" : this.displayValue();
-        ctx.fillStyle = isAuto ? litegraph_base.WIDGET_SECONDARY_TEXT_COLOR : litegraph_base.WIDGET_TEXT_COLOR;
-        ctx.fillText(text, widget_width - autoTextRightMargin - autoTextWidth - 15, y + H * 0.7);
-
-        // Draw increment/decrement buttons if not in AUTO mode and not a string widget
-        if (!isAuto && !this.disabled && this.config[0] !== "FVAUTOSTRING") {
-            // Draw decrement button (left triangle)
-            ctx.fillStyle = litegraph_base.WIDGET_TEXT_COLOR;
-            ctx.beginPath();
-            ctx.moveTo(margin + 16, y + 5);
-            ctx.lineTo(margin + 6, y + H * 0.5);
-            ctx.lineTo(margin + 16, y + H - 5);
-            ctx.fill();
-
-            // Draw increment button (right triangle)
-            ctx.beginPath();
-            ctx.moveTo(widget_width - margin - 16, y + 5);
-            ctx.lineTo(widget_width - margin - 6, y + H * 0.5);
-            ctx.lineTo(widget_width - margin - 16, y + H - 5);
-            ctx.fill();
-        }
-    }
-}
-
-function mouseAutoAnnotated(event, [x, y], node) {
-    const widget_width = node.size[0];
-    const margin = 15;
-    const H = 20; // Widget height
-
-    const autoTextWidth = 30;
-    const autoTextRightMargin = 5;
-
-    const cogRadius = 6;
-
-    if (this.isAuto) {
-        if (event.type === "pointerup" || event.type === "mouseup") {
-            const cogX = widget_width - autoTextRightMargin - autoTextWidth - 6;
-            const cogLeftEdge = cogX - cogRadius;
-            const cogRightEdge = cogX + cogRadius;
-
-            if (x > cogLeftEdge && x < cogRightEdge) {
-                this.isAuto = false;
-                this.value = this.cachedValue !== undefined ? this.cachedValue : (this.options.default || 0);
-
-                if (this.callback) {
-                    this.callback(this.value);
-                }
-                node.graph.setDirtyCanvas(true, false);
-            }
-        }
-
-        // Block ALL events in auto mode except cog clicks
-        event.preventDefault?.();
-        event.stopPropagation?.();
-        event.stopImmediatePropagation?.();
-        return true; // Always return true to indicate event was handled
-    }
-
-    // Determine if clicking on increment/decrement buttons
-    const delta = this.config[0] === "FVAUTOSTRING" ? 0 :
-        (x < 40 ? -1 : x > widget_width - 48 ? 1 : 0);
-
-    if (event.type === "pointerdown" || event.type === "mousedown") {
-        // ComfyUI appears to intercept pointerdown events, so this code path is never reached
-        console.log("pointerdown received (unexpected)");
-        return false;
-    } else if (event.type === "pointerup" || event.type === "mouseup") {
-        // Stop event propagation to prevent double handling
-        event.preventDefault?.();
-        event.stopPropagation?.();
-        event.stopImmediatePropagation?.();
-
-        const cogX = widget_width - autoTextRightMargin - autoTextWidth - 6;
-        const cogLeftEdge = widget_width - autoTextRightMargin - autoTextWidth - 6 - cogRadius;
-        const cogRightEdge = widget_width - autoTextRightMargin - autoTextWidth - 6 + cogRadius;
-
-        if (x > cogLeftEdge && x < cogRightEdge) {
-            this.isAuto = !this.isAuto;
-
-            if (this.isAuto) {
-                this.cachedValue = this.value;
-                this.value = -99999;
-            } else {
-                this.value = this.cachedValue !== undefined ? this.cachedValue : (this.options.default || 0);
-            }
-
-            if (this.callback) {
-                this.callback(this.value);
-            }
-
-            node.graph.setDirtyCanvas(true, false);
-            return true;
-        }
-
-        // If in auto mode and NOT clicking the cog, block all other interactions
-        if (this.isAuto) {
-            return true;
-        }
-
-        // Handle increment/decrement buttons if not in auto mode
-        if (delta !== 0 && !this.isAuto) {
-            if (this.config[0] === "FVAUTOCOMBO") {
-                const options = this.options.values || [];
-                if (options.length === 0) return true;
-
-                let currentIndex = -1;
-                for (let i = 0; i < options.length; i++) {
-                    const optValue = typeof options[i] === 'object' ? options[i].value : options[i];
-                    if (optValue == this.value || String(optValue) === String(this.value)) {
-                        currentIndex = i;
-                        break;
-                    }
-                }
-
-                if (currentIndex === -1) {
-                    currentIndex = 0;
-                }
-
-                let newIndex = currentIndex + delta;
-                if (newIndex < 0) {
-                    newIndex = options.length - 1;
-                } else if (newIndex >= options.length) {
-                    newIndex = 0;
-                }
-
-                const newOption = options[newIndex];
-                this.value = typeof newOption === 'object' ? newOption.value : newOption;
-
-                if (this.callback) {
-                    this.callback(this.value);
-                }
-
-                node.graph.setDirtyCanvas(true, false);
-                return true;
-            } else {
-                let v = parseFloat(this.value);
-                const increment = delta * 0.1 * (this.options.step || 1);
-
-                v += increment;
-
-                // Apply min/max constraints
-                if (this.options.min != null) {
-                    v = Math.max(this.options.min, v);
-                }
-                if (this.options.max != null) {
-                    v = Math.min(this.options.max, v);
-                }
-
-                // Round to precision or to integer
-                if (this.config[0] === "FVAUTOINT") {
-                    v = Math.round(v);
-                } else if (this.options.precision !== undefined) {
-                    const precision = Math.pow(10, this.options.precision);
-                    v = Math.round(v * precision) / precision;
-                }
-
-                this.value = v;
-
-                if (this.callback) {
-                    this.callback(this.value);
-                }
-
-                node.graph.setDirtyCanvas(true, false);
-                return true;
-            }
-        }
-
-        if (delta === 0 && !this.isAuto) {
-            if (this.config[0] === "FVAUTOCOMBO") {
-                const options = this.options.values || [];
-
-                // Create menu items
-                const menuItems = options.map(opt => {
-                    const value = typeof opt === 'object' ? opt.value : opt;
-                    const label = typeof opt === 'object' ? opt.label : opt.toString();
-
-                    return {
-                        content: label,
-                        callback: () => {
-                            this.value = value;
-
-                            if (this.callback) {
-                                this.callback(this.value);
-                            }
-
-                            node.graph.setDirtyCanvas(true, false);
-                        }
-                    };
-                });
-
-                new LiteGraph.ContextMenu(menuItems, {
-                    event: event,
-                    title: null,
-                    callback: null,
-                    extra: node
-                });
-
-                return true;
-            } else if (this.config[0] === "FVAUTOSTRING") {
-                const d_callback = (v) => {
-                    this.value = v;
-
-                    if (this.callback) {
-                        this.callback(this.value);
-                    }
-
-                    node.graph.setDirtyCanvas(true, false);
-                };
-
-                const dialog = app.canvas.prompt(
-                    'Value',
-                    this.value,
-                    d_callback,
-                    event
-                );
-
-                return true;
-            } else {
-                // For numeric widgets, show input dialog
-                const d_callback = (v) => {
-                    this.value = this.parseValue?.(v) ?? Number(v);
-
-                    // Apply min/max constraints
-                    if (this.options.min != null) {
-                        this.value = Math.max(this.options.min, this.value);
-                    }
-                    if (this.options.max != null) {
-                        this.value = Math.min(this.options.max, this.value);
-                    }
-
-                    // Round to precision or to integer
-                    if (this.config[0] === "FVAUTOINT") {
-                        this.value = Math.round(this.value);
-                    } else if (this.options.precision !== undefined) {
-                        const precision = Math.pow(10, this.options.precision);
-                        this.value = Math.round(this.value * precision) / precision;
-                    }
-
-                    if (this.callback) {
-                        this.callback(this.value);
-                    }
-
-                    node.graph.setDirtyCanvas(true, false);
-                };
-
-                const dialog = app.canvas.prompt(
-                    'Value',
-                    this.value,
-                    d_callback,
-                    event
-                );
-
-                return true;
-            }
-        }
-
-        return true;
-    }
-
-    return false;
-}
-
-function makeAutoAnnotated(widget, inputData) {
-    const original = {
-        callback: widget.callback,
-        type: widget.type,
-        value: widget.value
-    };
-
-    // Add AUTO properties to the widget
-    Object.assign(widget, {
-        type: "BOOLEAN",
-        draw: drawAutoAnnotated,
-        mouse: mouseAutoAnnotated,
-        onMouse: null, // Explicitly disable original onMouse handler
-
-        isAuto: true,
-        cachedValue: widget.value,
-        config: inputData,
-        options: Object.assign({}, inputData[1], widget.options),
-        original: original,  // Store original properties for reference
-
-        // Disable other potential mouse handlers with no-op functions
-        onClick: function () {
-            return false;
-        },
-        onPointerUp: function () {
-            return false;
-        },
-        onPointerDown: function () {
-            return false;
-        },
-        onMouseUp: function () {
-            return false;
-        },
-        onMouseDown: function () {
-            return false;
-        },
-
-        computeSize(width) {
-            return [width, 20];
-        },
-        displayValue: function () {
-            if (this.config[0] === "FVAUTOINT") {
-                return Math.round(this.value).toString();
-            }
-            if (this.config[0] === "FVAUTOCOMBO") {
-                return this.value;
-            }
-            if (this.config[0] === "FVAUTOSTRING") {
-                return this.value;
-            }
-            // For FLOAT values, check if it's actually an integer
-            if (Number.isInteger(this.value)) {
-                return this.value.toString();
-            }
-            return this.value.toFixed(this.options.precision || 2);
-        },
-        parseValue: function (v) {
-            if (this.config[0] === "FVAUTOSTRING") {
-                return v;
-            }
-            if (typeof v === "string") {
-                return parseFloat(v);
-            }
-            return v;
-        },
-        serializeValue: function () {
-            // Return special value for AUTO mode
-            return this.isAuto ? -99999 : this.value;
-        },
-        deserializeValue: function (data) {
-            if (data === -99999) {
-                this.isAuto = true;
-                this.value = -99999;
-            } else {
-                this.isAuto = false;
-                this.value = data;
-                this.cachedValue = data;
-            }
-        }
-    });
-
-    // Override callback to handle AUTO mode
-    widget.callback = function (v) {
-        if (this.isAuto) {
-            return; // Don't call the original callback in AUTO mode
-        }
-        const result = original.callback?.call(this, v);
-        return result;
-    };
-
-    // Override any potential click handlers
-    const originalOnClick = widget.onClick;
-    if (originalOnClick) {
-        widget.onClick = function (...args) {
-            if (this.isAuto) {
-                return false;
-            }
-            return originalOnClick.call(this, ...args);
-        };
-    }
-
-    return widget;
-}
-
-app.registerExtension({
-    name: "FastVideo.AutoWidgets",
-
-    async beforeRegisterNodeDef(nodeType, nodeData, app) {
-        if (nodeData?.name == "VideoGenerator" || nodeData?.name === "InferenceArgs" || nodeData?.name === "VAEConfig" ||
-            nodeData?.name === "TextEncoderConfig" || nodeData?.name === "DITConfig") {
-            // Add serialization support
-            chainCallback(nodeType.prototype, "onSerialize", function (info) {
-                if (!this.widgets) {
-                    return;
-                }
-                // Ensure widgets_values exists
-                if (!info.widgets_values) {
-                    info.widgets_values = {};
-                }
-
-                // Store AUTO widget states in a separate property
-                if (!info.auto_widget_states) {
-                    info.auto_widget_states = {};
-                }
-
-                // Handle AUTO widgets specially
-                for (const w of this.widgets) {
-                    if (w.type === "BOOLEAN" && w.isAuto !== undefined) {
-                        // Store the serialized value (for Python node)
-                        info.widgets_values[w.name] = w.serializeValue();
-
-                        // Store the full state (for UI restoration)
-                        info.auto_widget_states[w.name] = {
-                            isAuto: w.isAuto,
-                            value: w.value,
-                            cachedValue: w.cachedValue
-                        };
-                    }
-                }
-            });
-
-            // Add deserialization support
-            chainCallback(nodeType.prototype, "onConfigure", function (info) {
-                if (!this.widgets) {
-                    return;
-                }
-
-                // First, restore from widgets_values (for backward compatibility)
-                if (info.widgets_values && Array.isArray(info.widgets_values)) {
-                    for (let i = 0; i < this.widgets.length && i < info.widgets_values.length; i++) {
-                        const w = this.widgets[i];
-                        const value = info.widgets_values[i];
-
-                        if (w.type === "BOOLEAN" && w.isAuto !== undefined) {
-                            w.deserializeValue(value);
-                        }
-                    }
-                }
-
-                // Then, restore full state if available
-                if (info.auto_widget_states) {
-                    for (const w of this.widgets) {
-                        if (w.type === "BOOLEAN" && w.isAuto !== undefined && w.name in info.auto_widget_states) {
-                            const state = info.auto_widget_states[w.name];
-
-                            w.isAuto = state.isAuto;
-                            w.cachedValue = state.cachedValue;
-                            w.value = state.isAuto ? -99999 : state.value;
-
-                            w.callback?.(w.value);
-                        }
-                    }
-                }
-
-                // Force a redraw
-                this.graph?.setDirtyCanvas(true, true);
-            });
-
-
-            // Override addInput to handle AUTO widgets
-            chainCallback(nodeType.prototype, "onNodeCreated", function () {
-                // Convert any existing widgets to AUTO widgets if needed
-                let new_widgets = [];
-                const intWidgetNames = ["sp_size", "tp_size", "height", "width", "num_frames", "num_inference_steps", "flow_shift", "seed", "fps", "scale_factor",
-                    "tile_sample_min_height", "tile_sample_min_width", "tile_sample_min_num_frames", "tile_sample_stride_height", "tile_sample_stride_width",
-                    "tile_sample_stride_num_frames", "blend_num_frames"
-                ]
-                const floatWidgetNames = ["embedded_cfg_scale", "guidance_scale"]
-                const comboWidgetNames = ["vae_tiling", "vae_precision", "vae_sp", "text_encoder_precision", "precision",
-                    "load_encoder", "load_decoder", "use_tiling", "use_temporal_tiling", "use_parallel_tiling", "dit_cpu_offload", "enable_teacache"
-                ]
-                const stringWidgetNames = ["prefix", "quant_config", "lora_config", "image_path"]
-
-                if (this.widgets) {
-                    for (let w of this.widgets) {
-                        if (intWidgetNames.includes(w.name)) {
-                            new_widgets.push(makeAutoAnnotated(w, ["FVAUTOINT", { "default": 0 }]));
-                        } else if (floatWidgetNames.includes(w.name)) {
-                            new_widgets.push(makeAutoAnnotated(w, ["FVAUTOFLOAT", { "default": 0 }]));
-                        } else if (comboWidgetNames.includes(w.name)) {
-                            new_widgets.push(makeAutoAnnotated(w, ["FVAUTOCOMBO", { "default": 0 }]));
-                        } else if (stringWidgetNames.includes(w.name)) {
-                            new_widgets.push(makeAutoAnnotated(w, ["FVAUTOSTRING", { "default": "" }]));
-                        } else {
-                            new_widgets.push(w);
-                        }
-                    }
-                    this.widgets = new_widgets;
-
-                    const autoWidgets = this.widgets.filter(w => w.type === "BOOLEAN" && w.isAuto !== undefined);
-                }
-
-                this.graph?.setDirtyCanvas(true, true);
-            });
-        }
-    },
-
-    async init() {
-        // Force a redraw of all nodes when the extension initializes
-        if (app.graph) {
-            setTimeout(() => {
-                app.graph.setDirtyCanvas(true, true);
-            }, 1000);
-        }
-    }
-});
-
-console.log("FastVideo.core.js loaded");

csrc/attn/README.md

@@ -1,113 +0,0 @@
-
-
-# Attention Kernel Used in FastVideo
-
-
-
-## Video Sparse Attention (VSA)
-
-### Installation
-We support H100 (via TK) and any other GPU (via triton) for VSA.
-```bash
-git submodule update --init --recursive
-python setup_vsa.py install
-```
-
-
-If you encounter error during installation, try below:
-Install C++20 for ThunderKittens:
-```bash
-sudo apt update
-sudo apt install gcc-11 g++-11
-
-sudo update-alternatives --install /usr/bin/gcc gcc /usr/bin/gcc-11 100 --slave /usr/bin/g++ g++ /usr/bin/g++-11
-
-sudo apt update
-sudo apt install clang-11
-```
-(If you use CUDA12.8)
-```bash
-export CUDA_HOME=/usr/local/cuda-12.8
-export PATH=${CUDA_HOME}/bin:${PATH} 
-export LD_LIBRARY_PATH=${CUDA_HOME}/lib64:$LD_LIBRARY_PATH
-```
-
-### Verify if you have successfully installed
-
-```bash
-# test numerical
-python tests/test_vsa.py
-# (For H100) test speed
-python benchmarks/bench_vsa_hopper.py
-```
-bench_vsa_hopper.py should print something like this:
-```bash
-Using topk=76 kv blocks per q block (out of 768 total kv blocks)
-
-=== BLOCK SPARSE ATTENTION BENCHMARK ===
-Block Sparse Forward  - TFLOPS: 5622.26
-Block Sparse Backward - TFLOPS: 3865.68
-```
-
-
-## Sliding Tile Attention (STA)
-We only support H100 for STA.
-```bash
-git submodule update --init --recursive
-python setup_sta.py install
-```
-
-
-
-
-###  Usage
-End-2-end inference with FastVideo:
-```bash
-bash scripts/inference/v1_inference_wan_STA.sh
-```
-
-If you want to use sliding tile attention in your custom model:
-```python
-from st_attn import sliding_tile_attention
-# assuming video size (T, H, W) = (30, 48, 80), text tokens = 256 with padding. 
-# q, k, v: [batch_size, num_heads, seq_length, head_dim], seq_length = T*H*W + 256
-# a tile is a cube of size (6, 8, 8)
-# window_size in tiles: [(window_t, window_h, window_w), (..)...]. For example, window size (3, 3, 3) means a query can attend to (3x6, 3x8, 3x8) = (18, 24, 24) tokens out of the total 30x48x80 video.
-# text_length: int ranging from 0 to 256
-# If your attention contains text token (Hunyuan)
-out = sliding_tile_attention(q, k, v, window_size, text_length)
-# If your attention does not contain text token (StepVideo)
-out = sliding_tile_attention(q, k, v, window_size, 0, False)
-```
-
-
-### Test
-```bash
-python tests/test_sta.py # test STA
-python tests/test_vsa.py # test VSA
-```
-### Benchmark
-```bash
-python benchmarks/bench_sta.py
-```
-
-
-### How Does STA Work?
-We give a demo for 2D STA with window size (6,6) operating on a (10, 10) image. 
-
-
-https://github.com/user-attachments/assets/f3b6dd79-7b43-4b60-a0fa-3d6495ec5747
-
-## Why is STA Fast?
-2D/3D Sliding Window Attention (SWA) creates many mixed blocks in the attention map. Even though mixed blocks have less output value,a mixed block is significantly slower than a dense block due to the GPU-unfriendly masking operation. 
-
-STA removes mixed blocks.
-
-
-<div align="center">
-<img src=../../assets/sliding_tile_attn_map.png width="80%"/>
-</div>
-
-## Acknowledgement
-
-We learned or reuse code from FlexAtteniton, NATEN, and ThunderKittens.

csrc/attn/benchmarks/bench_vsa_hopper.py

@@ -1,224 +0,0 @@
-import torch
-import argparse
-from triton.testing import do_bench
-from vsa import block_sparse_fwd, block_sparse_bwd
-from vsa import BLOCK_M, BLOCK_N
-import  triton
-import numpy as np
-import random
-
-def set_seed(seed: int = 42):
-    # Python random module
-    random.seed(seed)
-
-    # NumPy
-    np.random.seed(seed)
-
-    # PyTorch
-    torch.manual_seed(seed)
-    torch.cuda.manual_seed(seed)
-    torch.cuda.manual_seed_all(seed)  # if using multi-GPU
-
-def parse_arguments():
-    parser = argparse.ArgumentParser(description='Benchmark Block Sparse Attention')
-    parser.add_argument('--batch_size', type=int, default=1, help='Batch size')
-    parser.add_argument('--num_heads', type=int, default=12, help='Number of heads')
-    parser.add_argument('--head_dim', type=int, default=128, help='Head dimension')
-    parser.add_argument('--topk', type=int, default=None, help='Number of kv blocks each q block attends to')
-    parser.add_argument('--seq_lengths', type=int, nargs='+', default=[49152], help='Sequence lengths to benchmark')
-    return parser.parse_args()
-
-def create_input_tensors(batch, head, seq_len, headdim):
-    """Create random input tensors for attention."""
-    q = torch.randn(batch, head, seq_len, headdim, dtype=torch.bfloat16, device="cuda")
-    k = torch.randn(batch, head, seq_len, headdim, dtype=torch.bfloat16, device="cuda")
-    v = torch.randn(batch, head, seq_len, headdim, dtype=torch.bfloat16, device="cuda")
-    return q, k, v
-
-def generate_block_sparse_pattern(bs, h, num_q_blocks, num_kv_blocks, k, device="cuda"):
-    """
-    Generate a block sparse pattern where each q block attends to exactly k kv blocks.
-    
-    Args:
-        bs: batch size
-        h: number of heads
-        num_q_blocks: number of query blocks
-        num_kv_blocks: number of key-value blocks
-        k: number of kv blocks each q block attends to
-        device: device to create tensors on
-        
-    Returns:
-        q2k_block_sparse_index: [bs, h, num_q_blocks, k]
-            Contains the indices of kv blocks that each q block attends to.
-        q2k_block_sparse_num: [bs, h, num_q_blocks]
-            Contains the number of kv blocks that each q block attends to (all equal to k).
-        k2q_block_sparse_index: [bs, h, num_kv_blocks, num_q_blocks]
-            Contains the indices of q blocks that attend to each kv block.
-        k2q_block_sparse_num: [bs, h, num_kv_blocks]
-            Contains the number of q blocks that attend to each kv block.
-        block_sparse_mask: [bs, h, num_q_blocks, num_kv_blocks]
-            Binary mask where 1 indicates attention connection.
-    """
-    # Ensure k is not larger than num_kv_blocks
-    k = min(k, num_kv_blocks)
-    
-    # Create random scores for sampling
-    scores = torch.rand(bs, h, num_q_blocks, num_kv_blocks, device=device)
-    
-    # Get top-k indices for each q block
-    _, q2k_block_sparse_index = torch.topk(scores, k, dim=-1)
-    q2k_block_sparse_index = q2k_block_sparse_index.to(torch.int32)
-    
-    # sort q2k_block_sparse_index
-    q2k_block_sparse_index, _ = torch.sort(q2k_block_sparse_index, dim=-1)
-
-    # All q blocks attend to exactly k kv blocks
-    q2k_block_sparse_num = torch.full((bs, h, num_q_blocks), k, dtype=torch.int32, device=device)
-    
-    # Create the corresponding mask
-    block_sparse_mask = torch.zeros(bs, h, num_q_blocks, num_kv_blocks, dtype=torch.bool, device=device)
-    
-    # Fill in the mask based on the indices
-    for b in range(bs):
-        for head in range(h):
-            for q_idx in range(num_q_blocks):
-                kv_indices = q2k_block_sparse_index[b, head, q_idx]
-                block_sparse_mask[b, head, q_idx, kv_indices] = True
-    
-    # Create the reverse mapping (k2q)
-    # First, initialize lists to collect q indices for each kv block
-    k2q_indices_list = [[[] for _ in range(num_kv_blocks)] for _ in range(bs * h)]
-    
-    # Populate the lists based on q2k mapping
-    for b in range(bs):
-        for head in range(h):
-            flat_idx = b * h + head
-            for q_idx in range(num_q_blocks):
-                kv_indices = q2k_block_sparse_index[b, head, q_idx].tolist()
-                for kv_idx in kv_indices:
-                    k2q_indices_list[flat_idx][kv_idx].append(q_idx)
-    
-    # Find the maximum number of q blocks that attend to any kv block
-    max_q_per_kv = 0
-    for flat_idx in range(bs * h):
-        for kv_idx in range(num_kv_blocks):
-            max_q_per_kv = max(max_q_per_kv, len(k2q_indices_list[flat_idx][kv_idx]))
-    
-    # Create tensors for k2q mapping
-    k2q_block_sparse_index = torch.full((bs, h, num_kv_blocks, max_q_per_kv), -1, 
-                                        dtype=torch.int32, device=device)
-    k2q_block_sparse_num = torch.zeros((bs, h, num_kv_blocks), 
-                                       dtype=torch.int32, device=device)
-    
-    # Fill the tensors
-    for b in range(bs):
-        for head in range(h):
-            flat_idx = b * h + head
-            for kv_idx in range(num_kv_blocks):
-                q_indices = k2q_indices_list[flat_idx][kv_idx]
-                num_q = len(q_indices)
-                k2q_block_sparse_num[b, head, kv_idx] = num_q
-                if num_q > 0:
-                    k2q_block_sparse_index[b, head, kv_idx, :num_q] = torch.tensor(
-                        q_indices, dtype=torch.int32, device=device)
-                
-    return q2k_block_sparse_index, q2k_block_sparse_num, k2q_block_sparse_index, k2q_block_sparse_num, block_sparse_mask
-
-def benchmark_block_sparse_attention(q, k, v, q2k_block_sparse_index, q2k_block_sparse_num, k2q_block_sparse_index, k2q_block_sparse_num, flops):
-    """Benchmark block sparse attention forward and backward passes."""
-    print("\n=== BLOCK SPARSE ATTENTION BENCHMARK ===")
-    
-    # Forward pass
-    # Warm-up run
-    variable_block_sizes = torch.ones(q2k_block_sparse_index.shape[2], device=q.device).int() * BLOCK_M
-    o, l_vec = block_sparse_fwd(q, k, v, q2k_block_sparse_index, q2k_block_sparse_num, variable_block_sizes)
-    torch.cuda.synchronize()
-    
-    # Benchmark forward
-    fwd_time = do_bench(
-        lambda: block_sparse_fwd(q, k, v, q2k_block_sparse_index, q2k_block_sparse_num, variable_block_sizes),
-        warmup=5,
-        rep=20,
-        quantiles=None
-    )
-    
-    sparse_tflops = flops / fwd_time * 1e-12 * 1e3
-    print(f"Block Sparse Forward - TFLOPS: {sparse_tflops:.2f}")
-    
-    # Backward pass
-    grad_output = torch.randn_like(o)
-    
-    # Warm-up runs
-    for _ in range(5):
-        block_sparse_bwd(q, k, v, o, l_vec, grad_output, k2q_block_sparse_index, k2q_block_sparse_num, variable_block_sizes)
-    torch.cuda.synchronize()
-    
-    # Benchmark backward
-    bwd_time = do_bench(
-        lambda: block_sparse_bwd(q, k, v, o, l_vec, grad_output, k2q_block_sparse_index, k2q_block_sparse_num, variable_block_sizes),
-        warmup=5,
-        rep=20,
-        quantiles=None
-    )
-    bwd_flops = 2.5 * flops  # Approximation
-
-    sparse_bwd_tflops = bwd_flops / bwd_time * 1e-12 * 1e3
-    print(f"Block Sparse Backward - TFLOPS: {sparse_bwd_tflops:.2f}")
-    
-    return sparse_tflops, sparse_bwd_tflops
-
-def main():
-    args = parse_arguments()
-
-    set_seed(42)
-    
-    # Extract parameters
-    batch = args.batch_size
-    head = args.num_heads
-    headdim = args.head_dim
-    
-    print(f"Block Sparse Attention Benchmark")
-    print(f"batch: {batch}, head: {head}, headdim: {headdim}")
-    
-    # Test with different sequence lengths
-    for seq_len in args.seq_lengths:
-        # Skip very long sequences if they might cause OOM
-        if seq_len > 16384 and batch > 1:
-            continue
-            
-        print("="*100)
-        print(f"\nSequence length: {seq_len}")
-        
-        # Calculate theoretical FLOPs for attention
-        flops = 4 * batch * head * headdim * seq_len * seq_len
-        
-        # Create input tensors
-        q, k, v = create_input_tensors(batch, head, seq_len, headdim)
-        
-        # Setup block sparse parameters
-        num_q_blocks = seq_len // BLOCK_M
-        num_kv_blocks = seq_len // BLOCK_N
-        
-        # Determine k value (number of kv blocks per q block)
-        topk = args.topk
-        if topk is None:
-            topk = num_kv_blocks // 10  # Default to ~90% sparsity if k is not specified
-        topk = max(1, topk)       
-        print(f"Using topk={topk} kv blocks per q block (out of {num_kv_blocks} total kv blocks)")
-        
-        # Generate block sparse pattern
-        q2k_block_sparse_index, q2k_block_sparse_num, k2q_block_sparse_index, k2q_block_sparse_num, _ = generate_block_sparse_pattern(
-            batch, head, num_q_blocks, num_kv_blocks, topk, device="cuda")
-        
-        # Benchmark block sparse attention
-        sparse_fwd, sparse_bwd = benchmark_block_sparse_attention(
-            q, k, v, q2k_block_sparse_index, q2k_block_sparse_num, k2q_block_sparse_index, k2q_block_sparse_num, flops
-        )
-        
-        # Print results
-        print("\n=== PERFORMANCE RESULTS ===")
-        print(f"Block Sparse Forward - TFLOPS: {sparse_fwd:.2f}")
-        print(f"Block Sparse Backward - TFLOPS: {sparse_bwd:.2f}")
-
-if __name__ == "__main__":
-    main()

csrc/attn/benchmarks/bench_vsa_triton.py

@@ -1,217 +0,0 @@
-import torch
-import argparse
-import triton.testing
-from vsa import block_sparse_attn
-from vsa import BLOCK_M, BLOCK_N
-
-import numpy as np
-import random
-
-def set_seed(seed: int = 42):
-    # Python random module
-    random.seed(seed)
-
-    # NumPy
-    np.random.seed(seed)
-
-    # PyTorch
-    torch.manual_seed(seed)
-    torch.cuda.manual_seed(seed)
-    torch.cuda.manual_seed_all(seed)  # if using multi-GPU
-
-def parse_arguments():
-    parser = argparse.ArgumentParser(description='Benchmark Block Sparse Attention')
-    parser.add_argument('--batch_size', type=int, default=1, help='Batch size')
-    parser.add_argument('--num_heads', type=int, default=12, help='Number of heads')
-    parser.add_argument('--head_dim', type=int, default=64, help='Head dimension')
-    parser.add_argument('--topk', type=int, default=None, help='Number of kv blocks each q block attends to')
-    parser.add_argument('--seq_lengths', type=int, nargs='+', default=[49152], help='Sequence lengths to benchmark')
-    return parser.parse_args()
-
-def create_input_tensors(batch, head, seq_len, headdim):
-    """Create random input tensors for attention."""
-    q = torch.randn(batch, head, seq_len, headdim, dtype=torch.bfloat16, device="cuda")
-    k = torch.randn(batch, head, seq_len, headdim, dtype=torch.bfloat16, device="cuda")
-    v = torch.randn(batch, head, seq_len, headdim, dtype=torch.bfloat16, device="cuda")
-    return q, k, v
-
-def generate_block_sparse_pattern(bs, h, num_q_blocks, num_kv_blocks, k, device="cuda"):
-    """
-    Generate a block sparse pattern where each q block attends to exactly k kv blocks.
-    
-    Args:
-        bs: batch size
-        h: number of heads
-        num_q_blocks: number of query blocks
-        num_kv_blocks: number of key-value blocks
-        k: number of kv blocks each q block attends to
-        device: device to create tensors on
-        
-    Returns:
-        q2k_block_sparse_index: [bs, h, num_q_blocks, k]
-            Contains the indices of kv blocks that each q block attends to.
-        q2k_block_sparse_num: [bs, h, num_q_blocks]
-            Contains the number of kv blocks that each q block attends to (all equal to k).
-        k2q_block_sparse_index: [bs, h, num_kv_blocks, num_q_blocks]
-            Contains the indices of q blocks that attend to each kv block.
-        k2q_block_sparse_num: [bs, h, num_kv_blocks]
-            Contains the number of q blocks that attend to each kv block.
-        block_sparse_mask: [bs, h, num_q_blocks, num_kv_blocks]
-            Binary mask where 1 indicates attention connection.
-    """
-    # Ensure k is not larger than num_kv_blocks
-    k = min(k, num_kv_blocks)
-    
-    # Create random scores for sampling
-    scores = torch.rand(bs, h, num_q_blocks, num_kv_blocks, device=device)
-    
-    # Get top-k indices for each q block
-    _, q2k_block_sparse_index = torch.topk(scores, k, dim=-1)
-    q2k_block_sparse_index = q2k_block_sparse_index.to(torch.int32)
-    
-    # sort q2k_block_sparse_index
-    q2k_block_sparse_index, _ = torch.sort(q2k_block_sparse_index, dim=-1)
-
-    # All q blocks attend to exactly k kv blocks
-    q2k_block_sparse_num = torch.full((bs, h, num_q_blocks), k, dtype=torch.int32, device=device)
-    
-    # Create the corresponding mask
-    block_sparse_mask = torch.zeros(bs, h, num_q_blocks, num_kv_blocks, dtype=torch.bool, device=device)
-    
-    # Fill in the mask based on the indices
-    for b in range(bs):
-        for head in range(h):
-            for q_idx in range(num_q_blocks):
-                kv_indices = q2k_block_sparse_index[b, head, q_idx]
-                block_sparse_mask[b, head, q_idx, kv_indices] = True
-    
-    # Create the reverse mapping (k2q)
-    # First, initialize lists to collect q indices for each kv block
-    k2q_indices_list = [[[] for _ in range(num_kv_blocks)] for _ in range(bs * h)]
-    
-    # Populate the lists based on q2k mapping
-    for b in range(bs):
-        for head in range(h):
-            flat_idx = b * h + head
-            for q_idx in range(num_q_blocks):
-                kv_indices = q2k_block_sparse_index[b, head, q_idx].tolist()
-                for kv_idx in kv_indices:
-                    k2q_indices_list[flat_idx][kv_idx].append(q_idx)
-    
-    # Find the maximum number of q blocks that attend to any kv block
-    max_q_per_kv = 0
-    for flat_idx in range(bs * h):
-        for kv_idx in range(num_kv_blocks):
-            max_q_per_kv = max(max_q_per_kv, len(k2q_indices_list[flat_idx][kv_idx]))
-    
-    # Create tensors for k2q mapping
-    k2q_block_sparse_index = torch.full((bs, h, num_kv_blocks, max_q_per_kv), -1, 
-                                        dtype=torch.int32, device=device)
-    k2q_block_sparse_num = torch.zeros((bs, h, num_kv_blocks), 
-                                       dtype=torch.int32, device=device)
-    
-    # Fill the tensors
-    for b in range(bs):
-        for head in range(h):
-            flat_idx = b * h + head
-            for kv_idx in range(num_kv_blocks):
-                q_indices = k2q_indices_list[flat_idx][kv_idx]
-                num_q = len(q_indices)
-                k2q_block_sparse_num[b, head, kv_idx] = num_q
-                if num_q > 0:
-                    k2q_block_sparse_index[b, head, kv_idx, :num_q] = torch.tensor(
-                        q_indices, dtype=torch.int32, device=device)
-                
-    return q2k_block_sparse_index, q2k_block_sparse_num, k2q_block_sparse_index, k2q_block_sparse_num, block_sparse_mask
-
-def benchmark_block_sparse_attention(q, k, v, q2k_block_sparse_index, q2k_block_sparse_num, k2q_block_sparse_index, k2q_block_sparse_num, flops):
-    """Benchmark block sparse attention forward+backward pass."""
-    print("\n=== BLOCK SPARSE ATTENTION FORWARD+BACKWARD BENCHMARK ===")
-    
-    # Combined forward+backward pass
-    # Warm-up run
-    q_fwd = q.clone().requires_grad_(True)
-    k_fwd = k.clone().requires_grad_(True)
-    v_fwd = v.clone().requires_grad_(True)
-    o = block_sparse_attn(q_fwd, k_fwd, v_fwd, q2k_block_sparse_index, q2k_block_sparse_num, k2q_block_sparse_index, k2q_block_sparse_num)
-    grad_output = torch.randn_like(o)
-    o.backward(grad_output)
-    torch.cuda.synchronize()
-    
-    # Benchmark forward+backward
-    def forward_backward_fn():
-        q_fwd = q.clone().requires_grad_(True)
-        k_fwd = k.clone().requires_grad_(True)
-        v_fwd = v.clone().requires_grad_(True)
-        o = block_sparse_attn(q_fwd, k_fwd, v_fwd, q2k_block_sparse_index, q2k_block_sparse_num, k2q_block_sparse_index, k2q_block_sparse_num)
-        grad_output = torch.randn_like(o)
-        o.backward(grad_output)
-    
-    total_time = triton.testing.do_bench(
-        forward_backward_fn,
-        warmup=25,
-        rep=100,
-        return_mode='mean'
-    )
-    
-    # Total flops for forward + backward (forward + 2.5x backward approximation)
-    total_flops = flops + 2.5 * flops  # 3.5x the forward flops
-    sparse_tflops = total_flops / total_time * 1e-12 * 1e3
-    print(f"Block Sparse Forward+Backward - TFLOPS: {sparse_tflops:.2f}")
-    
-    return sparse_tflops
-
-def main():
-    args = parse_arguments()
-
-    set_seed(42)
-    
-    # Extract parameters
-    batch = args.batch_size
-    head = args.num_heads
-    headdim = args.head_dim
-    
-    print(f"Block Sparse Attention Benchmark")
-    print(f"batch: {batch}, head: {head}, headdim: {headdim}")
-    
-    # Test with different sequence lengths
-    for seq_len in args.seq_lengths:
-        # Skip very long sequences if they might cause OOM
-        if seq_len > 16384 and batch > 1:
-            continue
-            
-        print("="*100)
-        print(f"\nSequence length: {seq_len}")
-        
-        # Calculate theoretical FLOPs for attention
-        flops = 4 * batch * head * headdim * seq_len * seq_len
-        
-        # Create input tensors
-        q, k, v = create_input_tensors(batch, head, seq_len, headdim)
-        
-        # Setup block sparse parameters
-        num_q_blocks = seq_len // BLOCK_M
-        num_kv_blocks = seq_len // BLOCK_N
-        
-        # Determine k value (number of kv blocks per q block)
-        topk = args.topk
-        if topk is None:
-            topk = num_kv_blocks // 10  # Default to ~90% sparsity if k is not specified
-        topk = max(1, topk)       
-        print(f"Using topk={topk} kv blocks per q block (out of {num_kv_blocks} total kv blocks)")
-        
-        # Generate block sparse pattern
-        q2k_block_sparse_index, q2k_block_sparse_num, k2q_block_sparse_index, k2q_block_sparse_num, _ = generate_block_sparse_pattern(
-            batch, head, num_q_blocks, num_kv_blocks, topk, device="cuda")
-        
-        # Benchmark block sparse attention
-        sparse_fwd = benchmark_block_sparse_attention(
-            q, k, v, q2k_block_sparse_index, q2k_block_sparse_num, k2q_block_sparse_index, k2q_block_sparse_num, flops
-        )
-        
-        # Print results
-        print("\n=== PERFORMANCE RESULTS ===")
-        print(f"Block Sparse Forward+Backward - TFLOPS: {sparse_fwd:.2f}")
-
-if __name__ == "__main__":
-    main()

csrc/attn/tests/test_vsa.py

@@ -1,156 +0,0 @@
-import torch
-import sys
-import os
-import numpy as np
-from tqdm import tqdm
-
-# Add the parent directory to the path to import block_sparse_attn
-sys.path.append(os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
-from tests.utils import generate_block_sparse_mask_for_function, create_full_mask_from_block_mask
-from vsa import block_sparse_attn
-
-BLOCK_M = 64
-BLOCK_N = 64
-
-def pytorch_test(Q, K, V, block_sparse_mask, dO):
-    q_ = Q.clone().float().requires_grad_()
-    k_ = K.clone().float().requires_grad_()
-    v_ = V.clone().float().requires_grad_()
-
-    QK = torch.matmul(q_, k_.transpose(-2, -1))
-    QK /= (q_.size(-1) ** 0.5)
-    QK = QK.masked_fill(~block_sparse_mask.unsqueeze(0), float('-inf'))
-
-    QK = torch.nn.functional.softmax(QK, dim=-1)
-    output = torch.matmul(QK, v_)
-
-    dO_ = dO
-    output.backward(dO_)
-    return (
-        output.to(torch.bfloat16),
-        q_.grad.to(torch.bfloat16),
-        k_.grad.to(torch.bfloat16),
-        v_.grad.to(torch.bfloat16),
-    )
-
-
-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, 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[:, :, non_pad_index, :]
-    output.backward(dO)
-    return output, Q.grad, K.grad, V.grad
-
-
-def get_non_pad_index(
-    vid_len: torch.LongTensor,
-    n_win: int,
-    win_size: int,
-):
-    device = vid_len.device
-    starts_pad = torch.arange(n_win, device=device) * win_size
-    index_pad = starts_pad[:, None] + torch.arange(win_size, device=device)[None, :]
-    index_mask = torch.arange(win_size, device=device)[None, :] < vid_len[:, None]
-
-    return index_pad[index_mask]
-
-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=32, max_size=64, device="cuda"):
-    return torch.randint(min_size, max_size + 1, (num_blocks,), device=device, dtype=torch.int32)
-
-
-def vsa_pad(x, non_pad_index, num_blocks, block_size):
-    padded_x = torch.zeros((1, x.shape[1], num_blocks * BLOCK_M, x.shape[3]), device=x.device, dtype=x.dtype)
-    padded_x[:, :, non_pad_index, :] = x
-    return padded_x
-
-def check_correctness(h, d, num_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"
-    variable_block_sizes = generate_variable_block_sizes(num_blocks, device=device)
-    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, 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)
-
-        # 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, 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 = h * S * d * num_iterations
-    for name, data in results.items():
-        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"},
-        {"num_blocks": 32, "k": 4, "description": "Medium sequence"},
-        {"num_blocks": 53, "k": 6, "description": "Large sequence"},
-    ]
-    
-    print(f"\nError Analysis for h={h}, d={d}, mode={error_mode}")
-    print("=" * 150)
-    print(f"{'Config':<20} {'Blocks':<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_blocks = config["num_blocks"]
-        k = config["k"]
-        description = config["description"]
-        results = check_correctness(h, d, num_blocks, k, error_mode=error_mode)
-        print(f"{description:<20} {num_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)
-    print("\nAnalysis completed for all modes.")

csrc/attn/tests/utils.py

@@ -1,54 +0,0 @@
-import torch
-
-def generate_block_sparse_mask_for_function(h, num_blocks, k, device="cuda"):
-    """
-    Generate block sparse mask of shape [h, num_blocks, num_blocks].
-    
-    Args:
-        h: number of heads
-        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_blocks, num_blocks] bool tensor
-    """
-    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_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, variable_block_sizes, device="cuda"):
-    """
-    Convert block-level sparse mask to full attention mask.
-    
-    Args:
-        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, S] bool tensor where S = total sequence length
-    """
-    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_seq_len, total_seq_len, dtype=torch.bool, device=device)
-
-    for head in range(h):
-        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_blocks):
-                if block_sparse_mask[head, q_block, 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

csrc/attn/video_sparse_attn/MANIFEST.in

@@ -1,2 +0,0 @@
-recursive-include tk *
-include config_vsa.py

csrc/attn/video_sparse_attn/README.md

@@ -1,61 +0,0 @@
-
-
-# Attention Kernel Used in FastVideo
-
-## Video Sparse Attention (VSA)
-
-### Installation
-We support H100 (via TK) and any other GPU (via triton) for VSA.
-
-```bash
-pip install vsa
-``` 
-
-Install from source:
-
-```bash
-git submodule update --init --recursive
-python setup.py install
-```
-
-
-If you encounter error during installation, try below:
-Install C++20 for ThunderKittens:
-```bash
-sudo apt update
-sudo apt install gcc-11 g++-11
-
-sudo update-alternatives --install /usr/bin/gcc gcc /usr/bin/gcc-11 100 --slave /usr/bin/g++ g++ /usr/bin/g++-11
-
-sudo apt update
-sudo apt install clang-11
-```
-(If you use CUDA12.8)
-```bash
-export CUDA_HOME=/usr/local/cuda-12.8
-export PATH=${CUDA_HOME}/bin:${PATH} 
-export LD_LIBRARY_PATH=${CUDA_HOME}/lib64:$LD_LIBRARY_PATH
-```
-
-### Verify if you have successfully installed
-
-```bash
-# test numerical
-python ../tests/test_vsa.py
-# (For H100) test speed
-python ../benchmarks/bench_vsa_hopper.py
-```
-
-bench_vsa_hopper.py should print something like this:
-
-```bash
-Using topk=76 kv blocks per q block (out of 768 total kv blocks)
-
-=== BLOCK SPARSE ATTENTION BENCHMARK ===
-Block Sparse Forward  - TFLOPS: 5622.26
-Block Sparse Backward - TFLOPS: 3865.68
-```
-
-## Acknowledgement
-
-We learned or reuse code from FlexAtteniton, NATEN, and ThunderKittens.