[Bug] DMD Distillation for Wan2.1-I2V-14B collapses to static videos then blue flashing screens

[iHateTheWorld555]

Describe the bug

I am using FastVideo to perform a DMD2 distillation on a custom I2V model (based on Wan2.1-I2V-14B, but utilizing Wan2.2 VAE) on a single node with 8x NVIDIA H200 GPUs. I aim to distill the model to 3 steps. Instead of using the officially provided synthetic dataset, I used an internal dataset of real-world videos.

However, the training dynamics are highly unstable and the model completely collapses during the process.

Observed Symptoms Progression:

1. Without Training: If I directly infer the base model with 3 steps, it generates a gradually blurring video.
2. Steps 0 ~ 200: The model generates almost completely static videos (copying the first frame).

validation_step_200_inference_steps_3_video_4.mp4

3. Steps > 200: Only generates fixed, meaningless blue flashing screens/grids.

validation_step_7000_inference_steps_3_video_4.mp4

Loss Curves:

To isolate the issue, I have already verified the following:

• VAE is normal: The VAE vectors input during distillation are correct and can be normally decoded into videos offline. The VAE used during distillation perfectly matches the offline VAE.
• No immediate NaN/Inf: I wrote a script to check the transformer weights around step 200 (before the complete blue screen collapse). There are no NaN or Inf values.
• Weight magnitude anomaly: Although there are no NaNs, I noticed that the weight values of norm_k and norm_q in deeper blocks (e.g., blocks.33.attn1.norm_k.weight) have large max values (Max > 5.2).

Reproduction

#!/usr/bin/env bash

NNODES=${PET_NNODES:-1}
NPROC_PER_NODE=${NPROC_PER_NODE:-8}
NODE_RANK=${PET_NODE_RANK:-0}
MASTER_ADDR=${PET_MASTER_ADDR:-127.0.0.1}
MASTER_PORT=${PET_MASTER_PORT:-29500}
TOTAL_GPUS=$((NNODES * NPROC_PER_NODE))

MODEL_PATH="/path/to/your/wan2.1_i2v_14b_checkpoint"
DATA_DIR="/path/to/your/distill_dataset"
OUTPUT_DIR="checkpoints/wan_i2v_distill_dmd"
VALIDATION_DATASET_FILE="configs/validation_config.json"

NUM_HEIGHT=480
NUM_WIDTH=832
NUM_FRAMES=129
NUM_LATENT_T=33

SP_SIZE=1
TP_SIZE=1
HSDP_SHARD_DIM=8
HSDP_REPLICATE_DIM=$((TOTAL_GPUS / HSDP_SHARD_DIM))

LEARNING_RATE=2e-6
SEED=1000

training_args=(
--tracker_project_name wan_i2v_distill_dmd
--output_dir "$OUTPUT_DIR"
--max_train_steps 14000
--train_batch_size 1
--train_sp_batch_size 1
--gradient_accumulation_steps 1
--num_latent_t "$NUM_LATENT_T"
--num_height "$NUM_HEIGHT"
--num_width "$NUM_WIDTH"
--num_frames "$NUM_FRAMES"
--enable_gradient_checkpointing_type full
)

parallel_args=(
--num_gpus "$TOTAL_GPUS"
--sp_size "$SP_SIZE"
--tp_size "$TP_SIZE"
--hsdp_replicate_dim "$HSDP_REPLICATE_DIM"
--hsdp_shard_dim "$HSDP_SHARD_DIM"
)

model_args=(
--model_path "$MODEL_PATH"
--pretrained_model_name_or_path "$MODEL_PATH"
--real_score_model_path "$MODEL_PATH"
--fake_score_model_path "$MODEL_PATH"
)

dataset_args=(
--data_path "$DATA_DIR"
--dataloader_num_workers 4
)

optimizer_args=(
--learning_rate "$LEARNING_RATE"
--mixed_precision "bf16"
--lr_scheduler "linear"
--lr_warmup_steps 500
--weight_decay 0.01
--max_grad_norm 1.0
--training_state_checkpointing_steps 500
--weight_only_checkpointing_steps 500
)

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
)

miscellaneous_args=(
--inference_mode False
--checkpoints_total_limit 3
--training_cfg_rate 0.0
--dit_precision "fp32"
--flow_shift 3
--seed "$SEED"
)

validation_args=(
--log_validation
--validation_dataset_file "$VALIDATION_DATASET_FILE"
--validation_steps 200
--validation_sampling_steps 3
--validation_guidance_scale 5.0
)

export TOKENIZERS_PARALLELISM=false
export FASTVIDEO_ATTENTION_BACKEND=FLASH_ATTN

torchrun
--nnodes "$NNODES"
--nproc_per_node "$NPROC_PER_NODE"
--node_rank "$NODE_RANK"
--master_addr "$MASTER_ADDR"
--master_port "$MASTER_PORT"
fastvideo/training/wan_i2v_distillation_pipeline.py
"${parallel_args[@]}"
"${model_args[@]}"
"${dataset_args[@]}"
"${training_args[@]}"
"${optimizer_args[@]}"
"${miscellaneous_args[@]}"
"${dmd_args[@]}"
"${validation_args[@]}"

Environment

INFO 04-06 02:05:03.015 [init.py:106] ROCm platform is unavailable: No module named 'amdsmi'
INFO 04-06 02:05:03.021 [init.py:47] CUDA is available
Collecting environment information...
PyTorch version: 2.11.0+cu128
Is debug build: False
CUDA used to build PyTorch: 12.8
ROCM used to build PyTorch: N/A

OS: Ubuntu 24.04.2 LTS (x86_64)
GCC version: (Ubuntu 13.3.0-6ubuntu2~24.04) 13.3.0
Clang version: Could not collect
CMake version: version 4.3.0
Libc version: glibc-2.39

Python version: 3.12.3 (main, Feb 4 2025, 14:48:35) [GCC 13.3.0] (64-bit runtime)
Python platform: Linux-5.15.0-119-generic-x86_64-with-glibc2.39
Is CUDA available: True
CUDA runtime version: 12.9.86
CUDA_MODULE_LOADING set to: LAZY
GPU models and configuration:
GPU 0: NVIDIA H200
GPU 1: NVIDIA H200
GPU 2: NVIDIA H200
GPU 3: NVIDIA H200
GPU 4: NVIDIA H200
GPU 5: NVIDIA H200
GPU 6: NVIDIA H200
GPU 7: NVIDIA H200

Nvidia driver version: 570.124.06
cuDNN version: Probably one of the following:
/usr/lib/x86_64-linux-gnu/libcudnn.so.9.10.2
/usr/lib/x86_64-linux-gnu/libcudnn_adv.so.9.10.2
/usr/lib/x86_64-linux-gnu/libcudnn_cnn.so.9.10.2
/usr/lib/x86_64-linux-gnu/libcudnn_engines_precompiled.so.9.10.2
/usr/lib/x86_64-linux-gnu/libcudnn_engines_runtime_compiled.so.9.10.2
/usr/lib/x86_64-linux-gnu/libcudnn_graph.so.9.10.2
/usr/lib/x86_64-linux-gnu/libcudnn_heuristic.so.9.10.2
/usr/lib/x86_64-linux-gnu/libcudnn_ops.so.9.10.2
HIP runtime version: N/A
MIOpen runtime version: N/A
Is XNNPACK available: True

CPU:
Architecture: x86_64
CPU op-mode(s): 32-bit, 64-bit
Address sizes: 52 bits physical, 57 bits virtual
Byte Order: Little Endian
CPU(s): 192
On-line CPU(s) list: 0-191
Vendor ID: GenuineIntel
Model name: INTEL(R) XEON(R) PLATINUM 8558
CPU family: 6
Model: 207
Thread(s) per core: 2
Core(s) per socket: 48
Socket(s): 2
Stepping: 2
CPU(s) scaling MHz: 75%
CPU max MHz: 4000.0000
CPU min MHz: 800.0000
BogoMIPS: 4200.00
Flags: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx pdpe1gb rdtscp lm constant_tsc art arch_perfmon pebs bts rep_good nopl xtopology nonstop_tsc cpuid aperfmperf tsc_known_freq pni pclmulqdq dtes64 ds_cpl vmx smx est tm2 ssse3 sdbg fma cx16 xtpr pdcm pcid dca sse4_1 sse4_2 x2apic movbe popcnt tsc_deadline_timer aes xsave avx f16c rdrand lahf_lm abm 3dnowprefetch cpuid_fault epb cat_l3 cat_l2 cdp_l3 invpcid_single intel_ppin cdp_l2 ssbd mba ibrs ibpb stibp ibrs_enhanced tpr_shadow vnmi flexpriority ept vpid ept_ad fsgsbase tsc_adjust bmi1 avx2 smep bmi2 erms invpcid cqm rdt_a avx512f avx512dq rdseed adx smap avx512ifma clflushopt clwb intel_pt avx512cd sha_ni avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local split_lock_detect avx_vnni avx512_bf16 wbnoinvd dtherm ida arat pln pts hwp hwp_act_window hwp_epp hwp_pkg_req avx512vbmi umip pku ospke waitpkg avx512_vbmi2 gfni vaes vpclmulqdq avx512_vnni avx512_bitalg tme avx512_vpopcntdq la57 rdpid bus_lock_detect cldemote movdiri movdir64b enqcmd fsrm md_clear serialize tsxldtrk pconfig arch_lbr amx_bf16 avx512_fp16 amx_tile amx_int8 flush_l1d arch_capabilities
Virtualization: VT-x
L1d cache: 4.5 MiB (96 instances)
L1i cache: 3 MiB (96 instances)
L2 cache: 192 MiB (96 instances)
L3 cache: 520 MiB (2 instances)
NUMA node(s): 2
NUMA node0 CPU(s): 0-47,96-143
NUMA node1 CPU(s): 48-95,144-191
Vulnerability Gather data sampling: Not affected
Vulnerability Itlb multihit: Not affected
Vulnerability L1tf: Not affected
Vulnerability Mds: Not affected
Vulnerability Meltdown: Not affected
Vulnerability Mmio stale data: Not affected
Vulnerability Reg file data sampling: Not affected
Vulnerability Retbleed: Not affected
Vulnerability Spec rstack overflow: Not affected
Vulnerability Spec store bypass: Mitigation; Speculative Store Bypass disabled via prctl and seccomp
Vulnerability Spectre v1: Mitigation; usercopy/swapgs barriers and __user pointer sanitization
Vulnerability Spectre v2: Mitigation; Enhanced / Automatic IBRS; IBPB conditional; RSB filling; PBRSB-eIBRS SW sequence; BHI BHI_DIS_S
Vulnerability Srbds: Not affected
Vulnerability Tsx async abort: Not affected

Versions of relevant libraries:
[pip3] accelerate==1.0.1
[pip3] numpy==2.2.6
[pip3] nvidia-cublas==13.1.0.3
[pip3] nvidia-cublas-cu12==12.8.4.1
[pip3] nvidia-cuda-cupti==13.0.85
[pip3] nvidia-cuda-cupti-cu12==12.8.90
[pip3] nvidia-cuda-nvrtc==13.0.88
[pip3] nvidia-cuda-nvrtc-cu12==12.8.93
[pip3] nvidia-cuda-runtime==13.0.96
[pip3] nvidia-cuda-runtime-cu12==12.8.90
[pip3] nvidia-cudnn-cu12==9.19.0.56
[pip3] nvidia-cudnn-cu13==9.19.0.56
[pip3] nvidia-cufft==12.0.0.61
[pip3] nvidia-cufft-cu12==11.3.3.83
[pip3] nvidia-cufile==1.15.1.6
[pip3] nvidia-cufile-cu12==1.13.1.3
[pip3] nvidia-curand==10.4.0.35
[pip3] nvidia-curand-cu12==10.3.9.90
[pip3] nvidia-cusolver==12.0.4.66
[pip3] nvidia-cusolver-cu12==11.7.3.90
[pip3] nvidia-cusparse==12.6.3.3
[pip3] nvidia-cusparse-cu12==12.5.8.93
[pip3] nvidia-cusparselt-cu12==0.7.1
[pip3] nvidia-cusparselt-cu13==0.8.0
[pip3] nvidia-cutlass-dsl==4.4.2
[pip3] nvidia-cutlass-dsl-libs-base==4.4.2
[pip3] nvidia-ml-py==13.595.45
[pip3] nvidia-nccl-cu12==2.28.9
[pip3] nvidia-nccl-cu13==2.28.9
[pip3] nvidia-nvjitlink==13.0.88
[pip3] nvidia-nvjitlink-cu12==12.8.93
[pip3] nvidia-nvshmem-cu12==3.4.5
[pip3] nvidia-nvshmem-cu13==3.4.5
[pip3] nvidia-nvtx==13.0.85
[pip3] nvidia-nvtx-cu12==12.8.90
[pip3] peft==0.18.1
[pip3] torch==2.11.0+cu128
[pip3] torch_c_dlpack_ext==0.1.5
[pip3] torchaudio==2.11.0
[pip3] torchcodec==0.5
[pip3] torchdata==0.11.0
[pip3] torchvision==0.26.0+cu128
[pip3] transformers==4.57.3
[pip3] triton==3.6.0
[conda] Could not collect
FastVideo Version:
FastVideo Build Flags:
CUDA Archs: 7.5 8.0 8.6 9.0 10.0 12.0+PTX; ROCm: Disabled; Neuron: Disabled
GPU Topology:
GPU0 GPU1 GPU2 GPU3 GPU4 GPU5 GPU6 GPU7 NIC0 NIC1 NIC2 NIC3 NIC4 NIC5 NIC6 NIC7 NIC8 NIC9 NIC10 CPU Affinity NUMA Affinity GPU NUMA ID�[0m
GPU0 X NV18 NV18 NV18 NV18 NV18 NV18 NV18 PIX NODE NODE NODE SYS SYS SYS SYS NODE NODE SYS 0-47,96-143 0 N/A
GPU1 NV18 X NV18 NV18 NV18 NV18 NV18 NV18 NODE PIX NODE NODE SYS SYS SYS SYS NODE NODE SYS 0-47,96-143 0 N/A
GPU2 NV18 NV18 X NV18 NV18 NV18 NV18 NV18 NODE NODE PIX NODE SYS SYS SYS SYS NODE NODE SYS 0-47,96-143 0 N/A
GPU3 NV18 NV18 NV18 X NV18 NV18 NV18 NV18 NODE NODE NODE PIX SYS SYS SYS SYS NODE NODE SYS 0-47,96-143 0 N/A
GPU4 NV18 NV18 NV18 NV18 X NV18 NV18 NV18 SYS SYS SYS SYS PIX NODE NODE NODE SYS SYS NODE 48-95,144-191 1 N/A
GPU5 NV18 NV18 NV18 NV18 NV18 X NV18 NV18 SYS SYS SYS SYS NODE PIX NODE NODE SYS SYS NODE 48-95,144-191 1 N/A
GPU6 NV18 NV18 NV18 NV18 NV18 NV18 X NV18 SYS SYS SYS SYS NODE NODE PIX NODE SYS SYS NODE 48-95,144-191 1 N/A
GPU7 NV18 NV18 NV18 NV18 NV18 NV18 NV18 X SYS SYS SYS SYS NODE NODE NODE PIX SYS SYS NODE 48-95,144-191 1 N/A
NIC0 PIX NODE NODE NODE SYS SYS SYS SYS X NODE NODE NODE SYS SYS SYS SYS NODE NODE SYS
NIC1 NODE PIX NODE NODE SYS SYS SYS SYS NODE X NODE NODE SYS SYS SYS SYS NODE NODE SYS
NIC2 NODE NODE PIX NODE SYS SYS SYS SYS NODE NODE X NODE SYS SYS SYS SYS NODE NODE SYS
NIC3 NODE NODE NODE PIX SYS SYS SYS SYS NODE NODE NODE X SYS SYS SYS SYS NODE NODE SYS
NIC4 SYS SYS SYS SYS PIX NODE NODE NODE SYS SYS SYS SYS X NODE NODE NODE SYS SYS NODE
NIC5 SYS SYS SYS SYS NODE PIX NODE NODE SYS SYS SYS SYS NODE X NODE NODE SYS SYS NODE
NIC6 SYS SYS SYS SYS NODE NODE PIX NODE SYS SYS SYS SYS NODE NODE X NODE SYS SYS NODE
NIC7 SYS SYS SYS SYS NODE NODE NODE PIX SYS SYS SYS SYS NODE NODE NODE X SYS SYS NODE
NIC8 NODE NODE NODE NODE SYS SYS SYS SYS NODE NODE NODE NODE SYS SYS SYS SYS X PIX SYS
NIC9 NODE NODE NODE NODE SYS SYS SYS SYS NODE NODE NODE NODE SYS SYS SYS SYS PIX X SYS
NIC10 SYS SYS SYS SYS NODE NODE NODE NODE SYS SYS SYS SYS NODE NODE NODE NODE SYS SYS X

Legend:

X = Self
SYS = Connection traversing PCIe as well as the SMP interconnect between NUMA nodes (e.g., QPI/UPI)
NODE = Connection traversing PCIe as well as the interconnect between PCIe Host Bridges within a NUMA node
PHB = Connection traversing PCIe as well as a PCIe Host Bridge (typically the CPU)
PXB = Connection traversing multiple PCIe bridges (without traversing the PCIe Host Bridge)
PIX = Connection traversing at most a single PCIe bridge
NV# = Connection traversing a bonded set of # NVLinks

NIC Legend:

NIC0: mlx5_0
NIC1: mlx5_1
NIC2: mlx5_2
NIC3: mlx5_3
NIC4: mlx5_4
NIC5: mlx5_5
NIC6: mlx5_6
NIC7: mlx5_7
NIC8: mlx5_10
NIC9: mlx5_11
NIC10: mlx5_bond_0

NVIDIA_VISIBLE_DEVICES=GPU-3e2a24dd-3354-eadc-3b75-61c360649a25,GPU-d31ed0cd-0596-28c5-57d7-ecd9b3ccc0bb,GPU-c8a9c70c-1d43-8164-c47d-7580a36dfe42,GPU-87e1662d-fa25-3a74-a718-6dff167aabfb,GPU-2d9dfbc9-5422-035b-dc03-810694f464b8,GPU-593ed04d-b765-70ab-4862-f28a8f04150c,GPU-9679af87-937f-ce91-c39a-e295c69a1343,GPU-ad3f6227-2b39-418a-f217-58ce4c1f5d84
CUBLAS_VERSION=12.9.1.4
NVIDIA_REQUIRE_CUDA=cuda>=9.0
TORCH_CUDA_ARCH_LIST=7.5 8.0 8.6 9.0 10.0 12.0+PTX
NCCL_VERSION=2.27.3
NVIDIA_DRIVER_CAPABILITIES=all
TORCH_NCCL_USE_COMM_NONBLOCKING=0
CUDA_ARCH_LIST=7.5 8.0 8.6 9.0 10.0 12.0
NVIDIA_PRODUCT_NAME=PyTorch
CUDA_VERSION=12.9.1.010
PYTORCH_VERSION=2.8.0a0+5228986
PYTORCH_BUILD_NUMBER=0
CUBLASMP_VERSION=0.4.0.789
CUDNN_FRONTEND_VERSION=1.12.0
CUDNN_VERSION=9.10.2.21
PYTORCH_HOME=/opt/pytorch/pytorch
LD_LIBRARY_PATH=/usr/local/cuda/compat/lib.real:/usr/local/lib/python3.12/dist-packages/torch/lib:/usr/local/lib/python3.12/dist-packages/torch_tensorrt/lib:/usr/local/cuda/compat/lib:/usr/local/nvidia/lib:/usr/local/nvidia/lib64
NVIDIA_BUILD_ID=177567386
CUDA_DRIVER_VERSION=575.57.08
PYTORCH_BUILD_VERSION=2.8.0a0+5228986
CUDA_HOME=/usr/local/cuda
CUDA_MODULE_LOADING=LAZY
NVIDIA_REQUIRE_JETPACK_HOST_MOUNTS=
NVIDIA_PYTORCH_VERSION=25.06
TORCH_ALLOW_TF32_CUBLAS_OVERRIDE=1
TORCHINDUCTOR_CACHE_DIR=/tmp/torchinductor_root

[github-actions]

Welcome to FastVideo! Thanks for opening your first issue.

To help us investigate, please include:

• FastVideo version: pip show fastvideo
• GPU: nvidia-smi output (GPU model, driver, CUDA version)
• Python version: python --version
• OS: e.g., Ubuntu 22.04

If this is a bug, a minimal reproduction script helps us fix it faster.

Useful links:

• Documentation
• Contributing Guide
• Slack