prithivMLmods commited on
Commit
cccc9b3
·
verified ·
1 Parent(s): 09a1add

Update app.py

Browse files
Files changed (1) hide show
  1. app.py +9 -1
app.py CHANGED
@@ -130,6 +130,13 @@ model_f = Qwen2_5_VLForConditionalGeneration.from_pretrained(
130
  MODEL_ID_F, trust_remote_code=True, torch_dtype=torch.float16
131
  ).to(device).eval()
132
 
 
 
 
 
 
 
 
133
  # --- PDF Generation and Preview Utility Function ---
134
  def generate_and_preview_pdf(image: Image.Image, text_content: str, font_size: int, line_spacing: float, alignment: str, image_size: str):
135
  """
@@ -240,6 +247,7 @@ def process_document_stream(
240
  elif model_name == "Megalodon-OCR-Sync-0713(ocr)": processor, model = processor_x, model_x
241
  elif model_name == "Qwen2.5-VL-3B-Abliterated-Caption-it(caption)": processor, model = processor_n, model_n
242
  elif model_name == "LMM-R1-MGT-PerceReason(reason)": processor, model = processor_f, model_f
 
243
  else:
244
  yield "Invalid model selected.", ""
245
  return
@@ -298,7 +306,7 @@ def create_gradio_interface():
298
  model_choice = gr.Dropdown(
299
  choices=["LFM2-VL-450M(fast)", "LFM2-VL-1.6B(fast)", "SmolVLM-Instruct-250M(smol)", "Moondream2(vision)", "ShotVL-3B(cinematic)", "Megalodon-OCR-Sync-0713(ocr)",
300
  "VLAA-Thinker-Qwen2VL-2B(reason)", "MonkeyOCR-pro-1.2B(ocr)", "Qwen2.5-VL-3B-Abliterated-Caption-it(caption)", "Nanonets-OCR-s(ocr)",
301
- "LMM-R1-MGT-PerceReason(reason)"],
302
  label="Select Model", value= "LFM2-VL-450M(fast)"
303
  )
304
  prompt_input = gr.Textbox(label="Query Input", placeholder="✦︎ Enter your query", value="Describe the image!")
 
130
  MODEL_ID_F, trust_remote_code=True, torch_dtype=torch.float16
131
  ).to(device).eval()
132
 
133
+ # TencentBAC/TBAC-VLR1-3B
134
+ MODEL_ID_G = "TencentBAC/TBAC-VLR1-3B"
135
+ processor_g = AutoProcessor.from_pretrained(MODEL_ID_G, trust_remote_code=True)
136
+ model_g = Qwen2_5_VLForConditionalGeneration.from_pretrained(
137
+ MODEL_ID_G, trust_remote_code=True, torch_dtype=torch.float16
138
+ ).to(device).eval()
139
+
140
  # --- PDF Generation and Preview Utility Function ---
141
  def generate_and_preview_pdf(image: Image.Image, text_content: str, font_size: int, line_spacing: float, alignment: str, image_size: str):
142
  """
 
247
  elif model_name == "Megalodon-OCR-Sync-0713(ocr)": processor, model = processor_x, model_x
248
  elif model_name == "Qwen2.5-VL-3B-Abliterated-Caption-it(caption)": processor, model = processor_n, model_n
249
  elif model_name == "LMM-R1-MGT-PerceReason(reason)": processor, model = processor_f, model_f
250
+ elif model_name == "TBAC-VLR1-3B(open-r1)": processor, model = processor_g, model_g
251
  else:
252
  yield "Invalid model selected.", ""
253
  return
 
306
  model_choice = gr.Dropdown(
307
  choices=["LFM2-VL-450M(fast)", "LFM2-VL-1.6B(fast)", "SmolVLM-Instruct-250M(smol)", "Moondream2(vision)", "ShotVL-3B(cinematic)", "Megalodon-OCR-Sync-0713(ocr)",
308
  "VLAA-Thinker-Qwen2VL-2B(reason)", "MonkeyOCR-pro-1.2B(ocr)", "Qwen2.5-VL-3B-Abliterated-Caption-it(caption)", "Nanonets-OCR-s(ocr)",
309
+ "LMM-R1-MGT-PerceReason(reason)", "TBAC-VLR1-3B(open-r1)"],
310
  label="Select Model", value= "LFM2-VL-450M(fast)"
311
  )
312
  prompt_input = gr.Textbox(label="Query Input", placeholder="✦︎ Enter your query", value="Describe the image!")