Spaces:
Running
on
Zero
Running
on
Zero
add sliders for variables , add cpu support
Browse files
app.py
CHANGED
@@ -20,10 +20,12 @@ import re
|
|
20 |
|
21 |
model_name = 'ucaslcl/GOT-OCR2_0'
|
22 |
|
|
|
|
|
23 |
tokenizer = AutoTokenizer.from_pretrained('ucaslcl/GOT-OCR2_0', trust_remote_code=True)
|
24 |
config = AutoConfig.from_pretrained(model_name, trust_remote_code=True)
|
25 |
-
model = AutoModel.from_pretrained('ucaslcl/GOT-OCR2_0', trust_remote_code=True, low_cpu_mem_usage=True,
|
26 |
-
model = model.eval().
|
27 |
model.config.pad_token_id = tokenizer.eos_token_id
|
28 |
|
29 |
UPLOAD_FOLDER = "./uploads"
|
@@ -40,7 +42,7 @@ def image_to_base64(image):
|
|
40 |
|
41 |
|
42 |
@spaces.GPU()
|
43 |
-
def process_image(image, task, ocr_type=None, ocr_box=None, ocr_color=None):
|
44 |
if image is None:
|
45 |
return "Error: No image provided", None, None
|
46 |
|
@@ -49,7 +51,7 @@ def process_image(image, task, ocr_type=None, ocr_box=None, ocr_color=None):
|
|
49 |
result_path = os.path.join(RESULTS_FOLDER, f"{unique_id}.html")
|
50 |
|
51 |
try:
|
52 |
-
if isinstance(image, dict):
|
53 |
composite_image = image.get("composite")
|
54 |
if composite_image is not None:
|
55 |
if isinstance(composite_image, np.ndarray):
|
@@ -68,19 +70,19 @@ def process_image(image, task, ocr_type=None, ocr_box=None, ocr_color=None):
|
|
68 |
return "Error: Unsupported image format", None, None
|
69 |
|
70 |
if task == "Plain Text OCR":
|
71 |
-
res = model.chat(tokenizer, image_path, ocr_type='ocr')
|
72 |
return res, None, unique_id
|
73 |
else:
|
74 |
if task == "Format Text OCR":
|
75 |
-
res = model.chat(tokenizer, image_path, ocr_type='format', render=True, save_render_file=result_path)
|
76 |
elif task == "Fine-grained OCR (Box)":
|
77 |
-
res = model.chat(tokenizer, image_path, ocr_type=ocr_type, ocr_box=ocr_box, render=True, save_render_file=result_path)
|
78 |
elif task == "Fine-grained OCR (Color)":
|
79 |
-
res = model.chat(tokenizer, image_path, ocr_type=ocr_type, ocr_color=ocr_color, render=True, save_render_file=result_path)
|
80 |
elif task == "Multi-crop OCR":
|
81 |
-
res = model.chat_crop(tokenizer, image_path, ocr_type='format', render=True, save_render_file=result_path)
|
82 |
elif task == "Render Formatted OCR":
|
83 |
-
res = model.chat(tokenizer, image_path, ocr_type='format', render=True, save_render_file=result_path)
|
84 |
|
85 |
if os.path.exists(result_path):
|
86 |
with open(result_path, 'r') as f:
|
@@ -249,6 +251,10 @@ with gr.Blocks(theme=gr.themes.Base()) as demo:
|
|
249 |
label="OCR Color",
|
250 |
visible=False
|
251 |
)
|
|
|
|
|
|
|
|
|
252 |
submit_button = gr.Button("Process")
|
253 |
editor_submit_button = gr.Button("Process Edited Image", visible=False)
|
254 |
|
|
|
20 |
|
21 |
model_name = 'ucaslcl/GOT-OCR2_0'
|
22 |
|
23 |
+
device = 'cuda' if torch.cuda.is_available() else 'cpu'
|
24 |
+
|
25 |
tokenizer = AutoTokenizer.from_pretrained('ucaslcl/GOT-OCR2_0', trust_remote_code=True)
|
26 |
config = AutoConfig.from_pretrained(model_name, trust_remote_code=True)
|
27 |
+
model = AutoModel.from_pretrained('ucaslcl/GOT-OCR2_0', trust_remote_code=True, low_cpu_mem_usage=True, device_map=device, use_safetensors=True, pad_token_id=tokenizer.eos_token_id)
|
28 |
+
model = model.eval().to(device)
|
29 |
model.config.pad_token_id = tokenizer.eos_token_id
|
30 |
|
31 |
UPLOAD_FOLDER = "./uploads"
|
|
|
42 |
|
43 |
|
44 |
@spaces.GPU()
|
45 |
+
def process_image(image, task, max_new_tokens, no_repeat_ngram_size, ocr_type=None, ocr_box=None, ocr_color=None):
|
46 |
if image is None:
|
47 |
return "Error: No image provided", None, None
|
48 |
|
|
|
51 |
result_path = os.path.join(RESULTS_FOLDER, f"{unique_id}.html")
|
52 |
|
53 |
try:
|
54 |
+
if isinstance(image, dict):
|
55 |
composite_image = image.get("composite")
|
56 |
if composite_image is not None:
|
57 |
if isinstance(composite_image, np.ndarray):
|
|
|
70 |
return "Error: Unsupported image format", None, None
|
71 |
|
72 |
if task == "Plain Text OCR":
|
73 |
+
res = model.chat(tokenizer, image_path, ocr_type='ocr', max_new_tokens=max_new_tokens, no_repeat_ngram_size=no_repeat_ngram_size)
|
74 |
return res, None, unique_id
|
75 |
else:
|
76 |
if task == "Format Text OCR":
|
77 |
+
res = model.chat(tokenizer, image_path, ocr_type='format', render=True, save_render_file=result_path, max_new_tokens=max_new_tokens, no_repeat_ngram_size=no_repeat_ngram_size)
|
78 |
elif task == "Fine-grained OCR (Box)":
|
79 |
+
res = model.chat(tokenizer, image_path, ocr_type=ocr_type, ocr_box=ocr_box, render=True, save_render_file=result_path, max_new_tokens=max_new_tokens, no_repeat_ngram_size=no_repeat_ngram_size)
|
80 |
elif task == "Fine-grained OCR (Color)":
|
81 |
+
res = model.chat(tokenizer, image_path, ocr_type=ocr_type, ocr_color=ocr_color, render=True, save_render_file=result_path, max_new_tokens=max_new_tokens, no_repeat_ngram_size=no_repeat_ngram_size)
|
82 |
elif task == "Multi-crop OCR":
|
83 |
+
res = model.chat_crop(tokenizer, image_path, ocr_type='format', render=True, save_render_file=result_path, max_new_tokens=max_new_tokens, no_repeat_ngram_size=no_repeat_ngram_size)
|
84 |
elif task == "Render Formatted OCR":
|
85 |
+
res = model.chat(tokenizer, image_path, ocr_type='format', render=True, save_render_file=result_path, max_new_tokens=max_new_tokens, no_repeat_ngram_size=no_repeat_ngram_size)
|
86 |
|
87 |
if os.path.exists(result_path):
|
88 |
with open(result_path, 'r') as f:
|
|
|
251 |
label="OCR Color",
|
252 |
visible=False
|
253 |
)
|
254 |
+
with gr.Row():
|
255 |
+
max_new_tokens_slider = gr.Slider(50, 500, step=10, value=150, label="Max New Tokens")
|
256 |
+
no_repeat_ngram_size_slider = gr.Slider(1, 10, step=1, value=2, label="No Repeat N-gram Size")
|
257 |
+
|
258 |
submit_button = gr.Button("Process")
|
259 |
editor_submit_button = gr.Button("Process Edited Image", visible=False)
|
260 |
|