Hugging Face's logo

barelymining
/
ComfyUI-MiniMax-H3-FastVideo

Diffusers
Safetensors
MiniMax H3
comfyui
lora
video
fasth3
vsa
sparse-attention
Model card Files
xet
 Community
barelymining commited on
Commit
aa55405
·
verified ·
1 Parent(s): 5e30cd9

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +0 -2
README.md CHANGED
@@ -59,8 +59,6 @@ Converted from FastVideo's [VSA-DataFree checkpoint](https://huggingface.co/Fast
59
  - Projecting `adaln_proj.linear.lora_A` from full BF16 space (`[96768, 2688]`) into the pruned model's 8-dim compressed space via least-squares fit against the base's `adaln_t_table` (residual ≈ 0 — mathematically lossless)
60
  - Extracting `to_gate_compress.set_weight` tensors into a separate file for runtime injection
61
 
62
- Conversion script (`convert_fastvideo_vsa_to_comfy.py`) is in the [node repo](https://github.com/barelymining/ComfyUI-MiniMax-H3-FastVideo) if you want to re-run it yourself.
63
-
64
  ## Attribution
65
 
66
  - Original weights: [FastVideo/FastVideo-FastH3-4-step-Preview-v1-VSA-DataFree](https://huggingface.co/FastVideo/FastVideo-FastH3-4-step-Preview-v1-VSA-DataFree)
 
59
  - Projecting `adaln_proj.linear.lora_A` from full BF16 space (`[96768, 2688]`) into the pruned model's 8-dim compressed space via least-squares fit against the base's `adaln_t_table` (residual ≈ 0 — mathematically lossless)
60
  - Extracting `to_gate_compress.set_weight` tensors into a separate file for runtime injection
61
 
 
 
62
  ## Attribution
63
 
64
  - Original weights: [FastVideo/FastVideo-FastH3-4-step-Preview-v1-VSA-DataFree](https://huggingface.co/FastVideo/FastVideo-FastH3-4-step-Preview-v1-VSA-DataFree)