Daankular commited on
Commit
499128d
·
1 Parent(s): ceb42ce

Pre-download Gemma at startup; bump guidance to 5.0 for prompt adherence

Browse files
Files changed (2) hide show
  1. app.py +16 -8
  2. generate.py +1 -1
app.py CHANGED
@@ -91,13 +91,21 @@ def setup():
91
  for repo, fname, dest in SULPHUR_ASSETS + LTX_ASSETS:
92
  _download(repo, fname, dest)
93
 
94
- te_dir = CKPTS_DIR / "umt5-xxl"
95
- if not te_dir.exists():
96
- from huggingface_hub import snapshot_download
97
- print("[download] umt5-xxl...")
98
- snapshot_download(repo_id="DeepBeepMeep/Wan2.1",
99
- allow_patterns=["umt5-xxl/*"],
100
- local_dir=str(CKPTS_DIR), token=_HF_TOKEN)
 
 
 
 
 
 
 
 
101
 
102
  FINETUNES_DIR.mkdir(parents=True, exist_ok=True)
103
  (FINETUNES_DIR / "sulphur_2_base.json").write_text(json.dumps(SULPHUR_FINETUNE, indent=2))
@@ -178,7 +186,7 @@ with gr.Blocks(title="Sulphur — Image to Video") as demo:
178
  with gr.Accordion("Advanced", open=False):
179
  resolution_dd = gr.Dropdown(RESOLUTIONS, value="832x480", label="Resolution")
180
  steps_sl = gr.Slider(1, 50, value=8, step=1, label="Steps")
181
- guidance_sl = gr.Slider(1.0, 10.0, value=3.5, step=0.5, label="Guidance Scale")
182
  frames_sl = gr.Slider(17, 257, value=81, step=8, label="Frames")
183
  seed_num = gr.Number(value=-1, label="Seed (-1 = random)", precision=0)
184
  run_btn = gr.Button("Generate", variant="primary")
 
91
  for repo, fname, dest in SULPHUR_ASSETS + LTX_ASSETS:
92
  _download(repo, fname, dest)
93
 
94
+ # Gemma text encoder — must stay in its subfolder (Wan2GP looks there by name)
95
+ _gemma_folder = "gemma-3-12b-it-qat-q4_0-unquantized"
96
+ _gemma_file = f"{_gemma_folder}_quanto_bf16_int8.safetensors"
97
+ gemma_dest = CKPTS_DIR / _gemma_folder / _gemma_file
98
+ if not gemma_dest.exists():
99
+ from huggingface_hub import hf_hub_download
100
+ print("[download] Gemma text encoder...")
101
+ hf_hub_download(
102
+ repo_id="DeepBeepMeep/LTX-2",
103
+ filename=f"{_gemma_folder}/{_gemma_file}",
104
+ local_dir=str(CKPTS_DIR),
105
+ token=_HF_TOKEN,
106
+ )
107
+ else:
108
+ print("[download] cached: Gemma text encoder")
109
 
110
  FINETUNES_DIR.mkdir(parents=True, exist_ok=True)
111
  (FINETUNES_DIR / "sulphur_2_base.json").write_text(json.dumps(SULPHUR_FINETUNE, indent=2))
 
186
  with gr.Accordion("Advanced", open=False):
187
  resolution_dd = gr.Dropdown(RESOLUTIONS, value="832x480", label="Resolution")
188
  steps_sl = gr.Slider(1, 50, value=8, step=1, label="Steps")
189
+ guidance_sl = gr.Slider(1.0, 10.0, value=5.0, step=0.5, label="Guidance Scale")
190
  frames_sl = gr.Slider(17, 257, value=81, step=8, label="Frames")
191
  seed_num = gr.Number(value=-1, label="Seed (-1 = random)", precision=0)
192
  run_btn = gr.Button("Generate", variant="primary")
generate.py CHANGED
@@ -17,7 +17,7 @@ MODEL_SHORTHANDS = {
17
  DEFAULTS = {
18
  "sulphur_2_base": {
19
  "num_inference_steps": 8,
20
- "guidance_scale": 3.5,
21
  "resolution": "832x480",
22
  "video_length": 81,
23
  },
 
17
  DEFAULTS = {
18
  "sulphur_2_base": {
19
  "num_inference_steps": 8,
20
+ "guidance_scale": 5.0,
21
  "resolution": "832x480",
22
  "video_length": 81,
23
  },