Spaces:
Running
on
Zero
Running
on
Zero
Update app.py
Browse files
app.py
CHANGED
@@ -47,6 +47,25 @@ model_x = Qwen2VLForConditionalGeneration.from_pretrained(
|
|
47 |
torch_dtype=torch.float16
|
48 |
).to(device).eval()
|
49 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
50 |
def downsample_video(video_path):
|
51 |
"""
|
52 |
Downsamples the video to evenly spaced frames.
|
@@ -84,6 +103,9 @@ def generate_image(model_name: str, text: str, image: Image.Image,
|
|
84 |
elif model_name == "coreOCR-7B-050325-preview":
|
85 |
processor = processor_x
|
86 |
model = model_x
|
|
|
|
|
|
|
87 |
else:
|
88 |
yield "Invalid model selected."
|
89 |
return
|
@@ -135,6 +157,9 @@ def generate_video(model_name: str, text: str, video_path: str,
|
|
135 |
elif model_name == "coreOCR-7B-050325-preview":
|
136 |
processor = processor_x
|
137 |
model = model_x
|
|
|
|
|
|
|
138 |
else:
|
139 |
yield "Invalid model selected."
|
140 |
return
|
@@ -233,7 +258,7 @@ with gr.Blocks(css=css, theme="bethecloud/storj_theme") as demo:
|
|
233 |
with gr.Column():
|
234 |
output = gr.Textbox(label="Output", interactive=False, lines=2, scale=2)
|
235 |
model_choice = gr.Radio(
|
236 |
-
choices=["docscopeOCR-7B-050425-exp", "coreOCR-7B-050325-preview"],
|
237 |
label="Select Model",
|
238 |
value="docscopeOCR-7B-050425-exp"
|
239 |
)
|
|
|
47 |
torch_dtype=torch.float16
|
48 |
).to(device).eval()
|
49 |
|
50 |
+
#--------------------------------------------------------------------------------------#
|
51 |
+
#Load MonkeyOCR
|
52 |
+
MODEL_ID_G = "echo840/MonkeyOCR"
|
53 |
+
SUBFOLDER = "Recognition"
|
54 |
+
|
55 |
+
processor_g = AutoProcessor.from_pretrained(
|
56 |
+
MODEL_ID_G,
|
57 |
+
trust_remote_code=True,
|
58 |
+
subfolder=SUBFOLDER
|
59 |
+
)
|
60 |
+
|
61 |
+
model_g = Qwen2VLForConditionalGeneration.from_pretrained(
|
62 |
+
MODEL_ID_G,
|
63 |
+
trust_remote_code=True,
|
64 |
+
subfolder=SUBFOLDER,
|
65 |
+
torch_dtype=torch.float16
|
66 |
+
).to(device).eval()
|
67 |
+
#--------------------------------------------------------------------------------------#
|
68 |
+
|
69 |
def downsample_video(video_path):
|
70 |
"""
|
71 |
Downsamples the video to evenly spaced frames.
|
|
|
103 |
elif model_name == "coreOCR-7B-050325-preview":
|
104 |
processor = processor_x
|
105 |
model = model_x
|
106 |
+
elif model_name == "MonkeyOCR":
|
107 |
+
processor = processor_g
|
108 |
+
model = model_g
|
109 |
else:
|
110 |
yield "Invalid model selected."
|
111 |
return
|
|
|
157 |
elif model_name == "coreOCR-7B-050325-preview":
|
158 |
processor = processor_x
|
159 |
model = model_x
|
160 |
+
elif model_name == "MonkeyOCR":
|
161 |
+
processor = processor_g
|
162 |
+
model = model_g
|
163 |
else:
|
164 |
yield "Invalid model selected."
|
165 |
return
|
|
|
258 |
with gr.Column():
|
259 |
output = gr.Textbox(label="Output", interactive=False, lines=2, scale=2)
|
260 |
model_choice = gr.Radio(
|
261 |
+
choices=["docscopeOCR-7B-050425-exp", "coreOCR-7B-050325-preview", "MonkeyOCR"],
|
262 |
label="Select Model",
|
263 |
value="docscopeOCR-7B-050425-exp"
|
264 |
)
|