Spaces:
Running
on
Zero
Running
on
Zero
Update app.py
Browse files
app.py
CHANGED
|
@@ -137,6 +137,13 @@ 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 |
"""
|
|
@@ -248,6 +255,7 @@ def process_document_stream(
|
|
| 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
|
|
@@ -306,7 +314,7 @@ def create_gradio_interface():
|
|
| 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!")
|
|
|
|
| 137 |
MODEL_ID_G, trust_remote_code=True, torch_dtype=torch.float16
|
| 138 |
).to(device).eval()
|
| 139 |
|
| 140 |
+
# OCRFlux-3B
|
| 141 |
+
MODEL_ID_V = "ChatDOC/OCRFlux-3B"
|
| 142 |
+
processor_v = AutoProcessor.from_pretrained(MODEL_ID_V, trust_remote_code=True)
|
| 143 |
+
model_v = Qwen2_5_VLForConditionalGeneration.from_pretrained(
|
| 144 |
+
MODEL_ID_V, trust_remote_code=True, torch_dtype=torch.float16
|
| 145 |
+
).to(device).eval()
|
| 146 |
+
|
| 147 |
# --- PDF Generation and Preview Utility Function ---
|
| 148 |
def generate_and_preview_pdf(image: Image.Image, text_content: str, font_size: int, line_spacing: float, alignment: str, image_size: str):
|
| 149 |
"""
|
|
|
|
| 255 |
elif model_name == "Qwen2.5-VL-3B-Abliterated-Caption-it(caption)": processor, model = processor_n, model_n
|
| 256 |
elif model_name == "LMM-R1-MGT-PerceReason(reason)": processor, model = processor_f, model_f
|
| 257 |
elif model_name == "TBAC-VLR1-3B(open-r1)": processor, model = processor_g, model_g
|
| 258 |
+
elif model_name == "OCRFlux-3B(ocr)": processor, model = processor_v, modelv
|
| 259 |
else:
|
| 260 |
yield "Invalid model selected.", ""
|
| 261 |
return
|
|
|
|
| 314 |
model_choice = gr.Dropdown(
|
| 315 |
choices=["LFM2-VL-450M(fast)", "LFM2-VL-1.6B(fast)", "SmolVLM-Instruct-250M(smol)", "Moondream2(vision)", "ShotVL-3B(cinematic)", "Megalodon-OCR-Sync-0713(ocr)",
|
| 316 |
"VLAA-Thinker-Qwen2VL-2B(reason)", "MonkeyOCR-pro-1.2B(ocr)", "Qwen2.5-VL-3B-Abliterated-Caption-it(caption)", "Nanonets-OCR-s(ocr)",
|
| 317 |
+
"LMM-R1-MGT-PerceReason(reason)", "TBAC-VLR1-3B(open-r1)", "OCRFlux-3B(ocr)"],
|
| 318 |
label="Select Model", value= "LFM2-VL-450M(fast)"
|
| 319 |
)
|
| 320 |
prompt_input = gr.Textbox(label="Query Input", placeholder="✦︎ Enter your query", value="Describe the image!")
|