axteam/Motif-Video-2B-FP8

Motif-Video-2B · FP8 (transformer)

FP8 (float8_e4m3fn) quantization of the Motif Video DiT (MotifVideoTransformer3DModel) transformer from Motif-Technologies/Motif-Video-2B. Quantized by axteam.

File	Size
Motif-Video-2B-transformer-fp8_e4m3fn.safetensors	2.6 GB

What this is

• Only the transformer (DiT) is quantized to float8_e4m3fn (2-D weight tensors; norms/biases kept in BF16). ~half the size of the BF16 transformer.
• VAE and text encoder(s) are NOT included — use the ones from the original Motif-Technologies/Motif-Video-2B.

Usage

• ComfyUI / Forge: place the file in models/unet/ (or models/diffusion_models/) and load with the UNET/diffusion-model loader; pair with the original VAE + text encoder(s).
• Produced per the common ComfyUI fp8_e4m3fn convention (plain cast, no scaling).

License

Quantized derivative of Motif-Technologies/Motif-Video-2B. Original weights belong to the upstream authors; the base model's license applies to this derivative.

---

한국어 요약

Motif-Technologies/Motif-Video-2B 의 transformer(DiT)만 FP8(float8_e4m3fn) 로 양자화(2D 가중치만, norm/bias는 BF16). BF16 대비 약 절반 크기. VAE·텍스트인코더는 미포함(원본 사용). ComfyUI models/unet/ 에 넣어 사용. 작성 axteam.