prithivMLmods commited on
Commit
a11c817
·
verified ·
1 Parent(s): 85d529e

Update app.py

Browse files
Files changed (1) hide show
  1. app.py +18 -4
app.py CHANGED
@@ -35,6 +35,13 @@ model_m = Qwen2_5_VLForConditionalGeneration.from_pretrained(
35
  MODEL_ID_M, trust_remote_code=True,
36
  torch_dtype=torch.float16).to(device).eval()
37
 
 
 
 
 
 
 
 
38
  # Load OCRFlux-3B
39
  MODEL_ID_X = "ChatDOC/OCRFlux-3B"
40
  processor_x = AutoProcessor.from_pretrained(MODEL_ID_X, trust_remote_code=True)
@@ -105,7 +112,7 @@ def generate_image(model_name: str,
105
  elif model_name == "OCRFlux-3B":
106
  processor = processor_x
107
  model = model_x
108
- elif model_name == "owlet-safety-3b-1":
109
  processor = processor_o
110
  model = model_o
111
  elif model_name == "MonkeyOCR-pro-1.2B":
@@ -114,6 +121,9 @@ def generate_image(model_name: str,
114
  elif model_name == "OpenVLThinker-7B":
115
  processor = processor_a
116
  model = model_a
 
 
 
117
  else:
118
  yield "Invalid model selected.", "Invalid model selected."
119
  return
@@ -172,7 +182,7 @@ def generate_video(model_name: str,
172
  elif model_name == "OCRFlux-3B":
173
  processor = processor_x
174
  model = model_x
175
- elif model_name == "owlet-safety-3b-1":
176
  processor = processor_o
177
  model = model_o
178
  elif model_name == "MonkeyOCR-pro-1.2B":
@@ -181,6 +191,9 @@ def generate_video(model_name: str,
181
  elif model_name == "OpenVLThinker-7B":
182
  processor = processor_a
183
  model = model_a
 
 
 
184
  else:
185
  yield "Invalid model selected.", "Invalid model selected."
186
  return
@@ -326,8 +339,9 @@ with gr.Blocks(css=css, theme="bethecloud/storj_theme") as demo:
326
  #download_btn = gr.Button("Download Result.md")
327
 
328
  model_choice = gr.Radio(choices=[
329
- "Camel-Doc-OCR-080125(v2)", "OCRFlux-3B", "OpenVLThinker-7B",
330
- "owlet-safety-3b-1", "MonkeyOCR-pro-1.2B"
 
331
  ],
332
  label="Select Model",
333
  value="Camel-Doc-OCR-080125(v2)")
 
35
  MODEL_ID_M, trust_remote_code=True,
36
  torch_dtype=torch.float16).to(device).eval()
37
 
38
+ # Load ViGoRL-MCTS-SFT-3b-Spatial
39
+ MODEL_ID_P = "gsarch/ViGoRL-MCTS-SFT-3b-Spatial"
40
+ processor_p = AutoProcessor.from_pretrained(MODEL_ID_P, trust_remote_code=True)
41
+ model_p = Qwen2_5_VLForConditionalGeneration.from_pretrained(
42
+ MODEL_ID_P, trust_remote_code=True,
43
+ torch_dtype=torch.float16).to(device).eval()
44
+
45
  # Load OCRFlux-3B
46
  MODEL_ID_X = "ChatDOC/OCRFlux-3B"
47
  processor_x = AutoProcessor.from_pretrained(MODEL_ID_X, trust_remote_code=True)
 
112
  elif model_name == "OCRFlux-3B":
113
  processor = processor_x
114
  model = model_x
115
+ elif model_name == "Owlet-Safety-3B-1":
116
  processor = processor_o
117
  model = model_o
118
  elif model_name == "MonkeyOCR-pro-1.2B":
 
121
  elif model_name == "OpenVLThinker-7B":
122
  processor = processor_a
123
  model = model_a
124
+ elif model_name == "ViGoRL-MCTS-SFT-3B":
125
+ processor = processor_p
126
+ model = model_p
127
  else:
128
  yield "Invalid model selected.", "Invalid model selected."
129
  return
 
182
  elif model_name == "OCRFlux-3B":
183
  processor = processor_x
184
  model = model_x
185
+ elif model_name == "Owlet-Safety-3B-1":
186
  processor = processor_o
187
  model = model_o
188
  elif model_name == "MonkeyOCR-pro-1.2B":
 
191
  elif model_name == "OpenVLThinker-7B":
192
  processor = processor_a
193
  model = model_a
194
+ elif model_name == "ViGoRL-MCTS-SFT-3B":
195
+ processor = processor_p
196
+ model = model_p
197
  else:
198
  yield "Invalid model selected.", "Invalid model selected."
199
  return
 
339
  #download_btn = gr.Button("Download Result.md")
340
 
341
  model_choice = gr.Radio(choices=[
342
+ "Camel-Doc-OCR-080125(v2)", "OCRFlux-3B",
343
+ "OpenVLThinker-7B", "ViGoRL-MCTS-SFT-3B",
344
+ "Owlet-Safety-3B-1", "MonkeyOCR-pro-1.2B"
345
  ],
346
  label="Select Model",
347
  value="Camel-Doc-OCR-080125(v2)")