Spaces:
Running
on
Zero
Running
on
Zero
Update app.py
Browse files
app.py
CHANGED
|
@@ -32,14 +32,12 @@ model_m = Qwen2_5_VLForConditionalGeneration.from_pretrained(
|
|
| 32 |
torch_dtype=torch.float16
|
| 33 |
).to(device).eval()
|
| 34 |
|
| 35 |
-
# Load
|
| 36 |
-
MODEL_ID_X = "
|
| 37 |
-
|
| 38 |
-
processor_x = AutoProcessor.from_pretrained(MODEL_ID_X, subfolder=SUBFOLDER, trust_remote_code=True)
|
| 39 |
model_x = Qwen2_5_VLForConditionalGeneration.from_pretrained(
|
| 40 |
MODEL_ID_X,
|
| 41 |
trust_remote_code=True,
|
| 42 |
-
subfolder=SUBFOLDER,
|
| 43 |
torch_dtype=torch.float16
|
| 44 |
).to(device).eval()
|
| 45 |
|
|
@@ -103,7 +101,7 @@ def generate_image(model_name: str, text: str, image: Image.Image,
|
|
| 103 |
if model_name == "Camel-Doc-OCR-062825":
|
| 104 |
processor = processor_m
|
| 105 |
model = model_m
|
| 106 |
-
elif model_name == "
|
| 107 |
processor = processor_x
|
| 108 |
model = model_x
|
| 109 |
elif model_name == "Megalodon-OCR-Sync-0713":
|
|
@@ -162,7 +160,7 @@ def generate_video(model_name: str, text: str, video_path: str,
|
|
| 162 |
if model_name == "Camel-Doc-OCR-062825":
|
| 163 |
processor = processor_m
|
| 164 |
model = model_m
|
| 165 |
-
elif model_name == "
|
| 166 |
processor = processor_x
|
| 167 |
model = model_x
|
| 168 |
elif model_name == "Megalodon-OCR-Sync-0713":
|
|
@@ -289,7 +287,7 @@ with gr.Blocks(css=css, theme="bethecloud/storj_theme") as demo:
|
|
| 289 |
markdown_output = gr.Markdown(label="(Result.md)")
|
| 290 |
|
| 291 |
model_choice = gr.Radio(
|
| 292 |
-
choices=["Camel-Doc-OCR-062825", "GLM-4.1V-9B-Thinking", "Megalodon-OCR-Sync-0713", "
|
| 293 |
label="Select Model",
|
| 294 |
value="Camel-Doc-OCR-062825"
|
| 295 |
)
|
|
|
|
| 32 |
torch_dtype=torch.float16
|
| 33 |
).to(device).eval()
|
| 34 |
|
| 35 |
+
# Load OVR-7B-RL
|
| 36 |
+
MODEL_ID_X = "Kangheng/OVR-7B-RL"
|
| 37 |
+
processor_x = AutoProcessor.from_pretrained(MODEL_ID_X, trust_remote_code=True)
|
|
|
|
| 38 |
model_x = Qwen2_5_VLForConditionalGeneration.from_pretrained(
|
| 39 |
MODEL_ID_X,
|
| 40 |
trust_remote_code=True,
|
|
|
|
| 41 |
torch_dtype=torch.float16
|
| 42 |
).to(device).eval()
|
| 43 |
|
|
|
|
| 101 |
if model_name == "Camel-Doc-OCR-062825":
|
| 102 |
processor = processor_m
|
| 103 |
model = model_m
|
| 104 |
+
elif model_name == "Open-Vision-Reasoner":
|
| 105 |
processor = processor_x
|
| 106 |
model = model_x
|
| 107 |
elif model_name == "Megalodon-OCR-Sync-0713":
|
|
|
|
| 160 |
if model_name == "Camel-Doc-OCR-062825":
|
| 161 |
processor = processor_m
|
| 162 |
model = model_m
|
| 163 |
+
elif model_name == "Open-Vision-Reasoner":
|
| 164 |
processor = processor_x
|
| 165 |
model = model_x
|
| 166 |
elif model_name == "Megalodon-OCR-Sync-0713":
|
|
|
|
| 287 |
markdown_output = gr.Markdown(label="(Result.md)")
|
| 288 |
|
| 289 |
model_choice = gr.Radio(
|
| 290 |
+
choices=["Camel-Doc-OCR-062825", "GLM-4.1V-9B-Thinking", "Megalodon-OCR-Sync-0713", "Open-Vision-Reasoner", "DeepEyes-7B-Thinking"],
|
| 291 |
label="Select Model",
|
| 292 |
value="Camel-Doc-OCR-062825"
|
| 293 |
)
|