SMD-Wan2.1-1.3B

4-step SMD for video generation. The student adapters, standalone inference code and reproducible examples are included. Upstream base components are downloaded separately.

Training code · SMD collection

Example

Download the video

Prompt: A small white delivery robot labeled "SMD" in large blue capital letters drives slowly along a bright hallway. The camera tracks alongside the robot, realistic motion, soft daylight.

Seed: 1234. Size: 832 × 480. Frames: 81. FPS: 16.

Case settings · Generation receipt

Models

Variant	Weights	Steps	Size
default	LoRA	4	0.70 GB

Run

Base: Wan-AI/Wan2.1-T2V-1.3B-Diffusers.

Use Linux, Python 3.12 and an NVIDIA GPU with CUDA support. The included inference code does not import the SMD training repository.

python3.12 -m venv .venv
source .venv/bin/activate
python -m pip install huggingface-hub
# Sign in to Hugging Face.
hf auth login
hf download zjwfufu/SMD-Wan2.1-1.3B --local-dir SMD-Wan2.1-1.3B --exclude 'lora/lora.safetensors'
cd SMD-Wan2.1-1.3B
python -m pip install -r requirements.txt
python infer.py --output outputs

The example was run on an A100 80 GB. Model CPU offload reduces GPU memory usage; enough host RAM is still needed to load the upstream components. Use --prompt, --seed and --output to generate another case. --base accepts a local upstream snapshot.

The student weights are published as smd_lora.safetensors at the repository root. The legacy lora/lora.safetensors remains available for existing integrations; the command above excludes that duplicate. The supplied loader prefers the root weight file and falls back to the legacy path. The adapter configuration, merge precision and sampling schedule are unchanged.

The first run downloads the upstream base components to the Hugging Face cache. Plan disk space for those components as well as the released weights; subsequent runs reuse the cache.

Sampling

The released sampler uses raw final-segment scale 0.3. The raw grid is built before the model-specific time shift. The included runner follows sampling.json; changing only an upstream pipeline's step count does not reproduce the student.

External classifier-free guidance is disabled for the students. Any embedded guidance remains specified in sampling.json.

The supplied loader validates and applies all LoRA matrices and saved replacement tensors, with the recorded merge precision. Omitting replacement tensors or using a different base can change the result.

License and attribution

Base components retain the terms described in the Wan-AI/Wan2.1-T2V-1.3B-Diffusers model card. Original code notices and bundled licenses are retained. Adapter use remains subject to the applicable base-model terms.