ismailkattakath commited on
Commit
19f6b88
·
verified ·
1 Parent(s): 76f7d97

Switch base model to SulphurAI/Sulphur-2-base (fp8mixed transformer + Sulphur distill LoRA)

Browse files
Files changed (2) hide show
  1. README.md +3 -3
  2. pipeline.py +5 -5
README.md CHANGED
@@ -47,10 +47,10 @@ holds a coffee cup in their right hand, glances to the left.
47
 
48
  | Component | Source |
49
  |-----------|--------|
50
- | Transformer (22B, FP8) | [Kijai/LTX2.3_comfy](https://huggingface.co/Kijai/LTX2.3_comfy) |
51
  | Video VAE | [Kijai/LTX2.3_comfy](https://huggingface.co/Kijai/LTX2.3_comfy) |
52
- | Text encoder (Gemma 3 12B, FP8) | [Comfy-Org/ltx-2](https://huggingface.co/Comfy-Org/ltx-2) |
53
- | Motion LoRA | [Kijai/LTX2.3_comfy](https://huggingface.co/Kijai/LTX2.3_comfy) |
54
  | BFS head-swap LoRA | [Alissonerdx/BFS-Best-Face-Swap-Video](https://huggingface.co/Alissonerdx/BFS-Best-Face-Swap-Video) |
55
 
56
  ## Logs
 
47
 
48
  | Component | Source |
49
  |-----------|--------|
50
+ | Transformer (22B, FP8 mixed) | [SulphurAI/Sulphur-2-base](https://huggingface.co/SulphurAI/Sulphur-2-base) |
51
  | Video VAE | [Kijai/LTX2.3_comfy](https://huggingface.co/Kijai/LTX2.3_comfy) |
52
+ | Text encoder + scheduler | [Lightricks/LTX-Video](https://huggingface.co/Lightricks/LTX-Video) |
53
+ | Distill LoRA | [SulphurAI/Sulphur-2-base](https://huggingface.co/SulphurAI/Sulphur-2-base) |
54
  | BFS head-swap LoRA | [Alissonerdx/BFS-Best-Face-Swap-Video](https://huggingface.co/Alissonerdx/BFS-Best-Face-Swap-Video) |
55
 
56
  ## Logs
pipeline.py CHANGED
@@ -2,10 +2,10 @@
2
  LTX-2.3 inference pipeline for BFS head-swap.
3
 
4
  Model sources:
5
- Transformer : Kijai/LTX2.3_comfy — diffusion_models/ltx-2.3-22b-distilled_transformer_only_fp8_input_scaled_v3.safetensors
6
- Video VAE : Kijai/LTX2.3_comfy — vae/LTX23_video_vae_bf16.safetensors
7
  Pipeline cfg: Lightricks/LTX-Video (text encoder, scheduler, tokenizer via from_pretrained)
8
- LoRA motion : Kijai/LTX2.3_comfy loras/ltx-2.3-22b-distilled-lora-dynamic_fro09_avg_rank_105_bf16.safetensors
9
  LoRA BFS : Alissonerdx/BFS-Best-Face-Swap-Video — ltx-2.3/head_swap_v3_rank_adaptive_fro_098.safetensors
10
  """
11
 
@@ -29,9 +29,9 @@ from PIL import Image
29
  _HF_CACHE = os.environ.get("HF_HOME", os.path.expanduser("~/.cache/huggingface"))
30
 
31
  MODELS = {
32
- "transformer": ("Kijai/LTX2.3_comfy", "diffusion_models/ltx-2.3-22b-distilled_transformer_only_fp8_input_scaled_v3.safetensors"),
33
  "video_vae": ("Kijai/LTX2.3_comfy", "vae/LTX23_video_vae_bf16.safetensors"),
34
- "lora_motion": ("Kijai/LTX2.3_comfy", "loras/ltx-2.3-22b-distilled-lora-dynamic_fro09_avg_rank_105_bf16.safetensors"),
35
  "lora_bfs": ("Alissonerdx/BFS-Best-Face-Swap-Video", "ltx-2.3/head_swap_v3_rank_adaptive_fro_098.safetensors"),
36
  }
37
 
 
2
  LTX-2.3 inference pipeline for BFS head-swap.
3
 
4
  Model sources:
5
+ Transformer : SulphurAI/Sulphur-2-base — sulphur_dev_fp8mixed.safetensors
6
+ Video VAE : Kijai/LTX2.3_comfy — vae/LTX23_video_vae_bf16.safetensors
7
  Pipeline cfg: Lightricks/LTX-Video (text encoder, scheduler, tokenizer via from_pretrained)
8
+ LoRA distill: SulphurAI/Sulphur-2-base distill_loras/ltx-2.3-22b-distilled-lora-1.1_fro90_ceil72_condsafe.safetensors
9
  LoRA BFS : Alissonerdx/BFS-Best-Face-Swap-Video — ltx-2.3/head_swap_v3_rank_adaptive_fro_098.safetensors
10
  """
11
 
 
29
  _HF_CACHE = os.environ.get("HF_HOME", os.path.expanduser("~/.cache/huggingface"))
30
 
31
  MODELS = {
32
+ "transformer": ("SulphurAI/Sulphur-2-base", "sulphur_dev_fp8mixed.safetensors"),
33
  "video_vae": ("Kijai/LTX2.3_comfy", "vae/LTX23_video_vae_bf16.safetensors"),
34
+ "lora_motion": ("SulphurAI/Sulphur-2-base", "distill_loras/ltx-2.3-22b-distilled-lora-1.1_fro90_ceil72_condsafe.safetensors"),
35
  "lora_bfs": ("Alissonerdx/BFS-Best-Face-Swap-Video", "ltx-2.3/head_swap_v3_rank_adaptive_fro_098.safetensors"),
36
  }
37