From e332dbb857875f1cce01ee9cdf608719aa129cd3 Mon Sep 17 00:00:00 2001
From: RandNMR73 <notomatthew31@gmail.com>
Date: Thu, 22 Jan 2026 11:51:12 +0000
Subject: [PATCH] more scripts

---
 .../distill_dmd_t2v_1.3B_vidprom.sh           | 122 ++++++++++++++++++
 .../crush_smol/finetune_t2v_nvfp4.sh          |  94 --------------
 2 files changed, 122 insertions(+), 94 deletions(-)
 create mode 100644 examples/distill/Wan2.1-T2V/Wan-Syn-Data-480P/distill_dmd_t2v_1.3B_vidprom.sh
 delete mode 100644 examples/training/finetune/wan_t2v_1.3B/crush_smol/finetune_t2v_nvfp4.sh

diff --git a/examples/distill/Wan2.1-T2V/Wan-Syn-Data-480P/distill_dmd_t2v_1.3B_vidprom.sh b/examples/distill/Wan2.1-T2V/Wan-Syn-Data-480P/distill_dmd_t2v_1.3B_vidprom.sh
new file mode 100644
index 000000000..bc258a031
--- /dev/null
+++ b/examples/distill/Wan2.1-T2V/Wan-Syn-Data-480P/distill_dmd_t2v_1.3B_vidprom.sh
@@ -0,0 +1,122 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+############################################
+# Single node, 6 GPUs
+############################################
+NUM_GPUS=1
+export WANDB_API_KEY=2f25ad37933894dbf0966c838c0b8494987f9f2f
+export WANDB_MODE="online"
+export WANDB_BASE_URL="https://api.wandb.ai"
+export TOKENIZERS_PARALLELISM=false
+export FASTVIDEO_ATTENTION_BACKEND=FLASH_ATTN   # or TORCH_SDPA
+export NCCL_P2P_DISABLE=1
+export TORCH_NCCL_ENABLE_MONITORING=0
+export TRITON_CACHE_DIR=/tmp/triton_cache
+
+echo "NUM_GPUS:    $NUM_GPUS"
+echo "CUDA_VISIBLE_DEVICES: $CUDA_VISIBLE_DEVICES"
+
+############################################
+# Configs
+############################################
+MODEL_PATH="Wan-AI/Wan2.1-T2V-1.3B-Diffusers"
+DATA_DIR=data/vidprom_16k_umt5_text_embed
+VALIDATION_DATASET_FILE="examples/training/finetune/wan_t2v_1.3B/crush_smol/validation.json"
+OUTPUT_DIR="checkpoints/wan_1.3B_t2v_distill_dmd_vidprom"
+REAL_SCORE_MODEL_PATH="Wan-AI/Wan2.1-T2V-1.3B-Diffusers"
+FAKE_SCORE_MODEL_PATH="Wan-AI/Wan2.1-T2V-1.3B-Diffusers"
+
+# Training arguments
+training_args=(
+  --tracker_project_name wan_t2v_distill_dmd_qat
+  --output_dir $OUTPUT_DIR
+  --max_train_steps 4000
+  --train_batch_size 1
+  --train_sp_batch_size 1
+  --gradient_accumulation_steps 1
+  --num_latent_t 21
+  --num_height 480
+  --num_width 832
+  --num_frames 81
+  --enable_gradient_checkpointing_type "full"
+)
+
+# Parallel arguments (adjusted to 6 GPUs)
+parallel_args=(
+  --num_gpus $NUM_GPUS
+  --sp_size 1
+  --tp_size 1
+  --hsdp_replicate_dim 1
+  --hsdp_shard_dim $NUM_GPUS
+)
+
+# Model arguments
+model_args=(
+  --model_path $MODEL_PATH
+  --pretrained_model_name_or_path $MODEL_PATH
+  --real_score_model_path $REAL_SCORE_MODEL_PATH
+  --fake_score_model_path $FAKE_SCORE_MODEL_PATH
+)
+
+# Dataset arguments
+dataset_args=(
+  --data_path "$DATA_DIR"
+  --dataloader_num_workers 4
+)
+
+# Validation arguments
+validation_args=(
+  --log_validation
+  --validation_dataset_file "$VALIDATION_DATASET_FILE"
+  --validation_steps 200
+  --validation_sampling_steps "3"
+  --validation_guidance_scale "6.0" # not used for dmd inference
+)
+
+# Optimizer arguments
+optimizer_args=(
+  --learning_rate 2e-6
+  --mixed_precision "bf16"
+  --training_state_checkpointing_steps 500
+  --weight_only_checkpointing_steps 500
+  --weight_decay 0.01
+  --max_grad_norm 1.0
+)
+
+# Miscellaneous arguments
+miscellaneous_args=(
+  --inference_mode False
+  --checkpoints_total_limit 3
+  --training_cfg_rate 0.0
+  --dit_precision "fp32"
+  --ema_start_step 0
+  --flow_shift 8
+  --seed 1000
+)
+
+# DMD arguments
+dmd_args=(
+  --dmd_denoising_steps '1000,757,522'
+  --min_timestep_ratio 0.02
+  --max_timestep_ratio 0.98
+  --generator_update_interval 5
+  --real_score_guidance_scale 3.5
+  --simulate_generator_forward
+)
+
+############################################
+# Launch (no rendezvous flags needed)
+############################################
+set -x
+torchrun \
+  --nproc_per_node="$NUM_GPUS" \
+  fastvideo/training/wan_distillation_pipeline.py \
+    "${parallel_args[@]}" \
+    "${model_args[@]}" \
+    "${dataset_args[@]}" \
+    "${training_args[@]}" \
+    "${optimizer_args[@]}" \
+    "${validation_args[@]}" \
+    "${miscellaneous_args[@]}" \
+    "${dmd_args[@]}"
\ No newline at end of file
diff --git a/examples/training/finetune/wan_t2v_1.3B/crush_smol/finetune_t2v_nvfp4.sh b/examples/training/finetune/wan_t2v_1.3B/crush_smol/finetune_t2v_nvfp4.sh
deleted file mode 100644
index 56c7da9fd..000000000
--- a/examples/training/finetune/wan_t2v_1.3B/crush_smol/finetune_t2v_nvfp4.sh
+++ /dev/null
@@ -1,94 +0,0 @@
-#!/bin/bash
-
-export WANDB_BASE_URL="https://api.wandb.ai"
-export WANDB_MODE=online
-export TOKENIZERS_PARALLELISM=false
-# export FASTVIDEO_ATTENTION_BACKEND=TORCH_SDPA
-
-MODEL_PATH="Wan-AI/Wan2.1-T2V-1.3B-Diffusers"
-DATA_DIR="data/Wan-Syn_77x448x832_600k/train/"
-VALIDATION_DATASET_FILE="data/Wan-Syn_77x448x832_600k/validation_6.json"
-NUM_GPUS=6
-# export CUDA_VISIBLE_DEVICES=4,5
-
-
-# Training arguments
-training_args=(
-  --tracker_project_name "wan_t2v_finetune"
-  --output_dir "checkpoints/wan_t2v_finetune"
-  --max_train_steps 5000
-  --train_batch_size 1
-  --train_sp_batch_size 1
-  --gradient_accumulation_steps 1
-  --num_latent_t 6
-  --num_height 480
-  --num_width 832
-  --num_frames 21
-  --enable_gradient_checkpointing_type "full"
-  --generator_4bit_linear True
-)
-
-# Parallel arguments
-parallel_args=(
-  --num_gpus $NUM_GPUS 
-  --sp_size 1
-  --tp_size 1
-  --hsdp_replicate_dim 1
-  --hsdp_shard_dim $NUM_GPUS
-)
-
-# Model arguments
-model_args=(
-  --model_path $MODEL_PATH
-  --pretrained_model_name_or_path $MODEL_PATH
-)
-
-# Dataset arguments
-dataset_args=(
-  --data_path $DATA_DIR
-  --dataloader_num_workers 1
-)
-
-# Validation arguments
-validation_args=(
-  --log_validation 
-  --validation_dataset_file $VALIDATION_DATASET_FILE
-  --validation_steps 50
-  --validation_sampling_steps "50" 
-  --validation_guidance_scale "6.0"
-)
-
-# Optimizer arguments
-optimizer_args=(
-  --learning_rate 1e-6
-  --mixed_precision "bf16"
-  --checkpointing_steps 2000
-  --weight_decay 1e-4
-  --max_grad_norm 1.0
-)
-
-# Miscellaneous arguments
-miscellaneous_args=(
-  --inference_mode False
-  --checkpoints_total_limit 3
-  --training_cfg_rate 0.1
-  --multi_phased_distill_schedule "4000-1"
-  --not_apply_cfg_solver
-  --dit_precision "fp32"
-  --num_euler_timesteps 50
-  --ema_start_step 0
-  --enable_gradient_checkpointing_type "full"
-  # --resume_from_checkpoint "checkpoints/wan_t2v_finetune/checkpoint-2500"
-)
-
-torchrun \
-  --nnodes 1 \
-  --nproc_per_node $NUM_GPUS \
-    fastvideo/training/wan_training_pipeline.py \
-    "${parallel_args[@]}" \
-    "${model_args[@]}" \
-    "${dataset_args[@]}" \
-    "${training_args[@]}" \
-    "${optimizer_args[@]}" \
-    "${validation_args[@]}" \
-    "${miscellaneous_args[@]}"