Spaces:
Runtime error
Runtime error
File size: 9,850 Bytes
edf3f62 58c54a2 186fd60 58c54a2 ff9681b 59edac2 be4764f 343dd2c f6fe3e1 58c54a2 186fd60 58c54a2 59edac2 58c54a2 343dd2c 59edac2 343dd2c 58c54a2 343dd2c 58c54a2 343dd2c 58c54a2 343dd2c 58c54a2 343dd2c 58c54a2 343dd2c 58c54a2 343dd2c 58c54a2 f6fe3e1 6302661 186fd60 58c54a2 6302661 58c54a2 186fd60 58c54a2 fc0862c 58c54a2 4d25f8e 58c54a2 186fd60 58c54a2 186fd60 58c54a2 893b675 58c54a2 6302661 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 |
import gradio as gr
import torch
from PIL import Image
import logging
from typing import Optional, Union
import os
import spaces
from dotenv import load_dotenv
load_dotenv()
# Disable torch compilation to avoid dynamo issues
torch._dynamo.config.disable = True
torch.backends.cudnn.allow_tf32 = True
# Configure logging
logging.basicConfig(level=logging.INFO)
logger = logging.getLogger(__name__)
class AtlasOCR:
def __init__(self, model_name: str = "atlasia/AtlasOCR", max_tokens: int = 2000):
"""Initialize the AtlasOCR model with proper error handling."""
try:
from unsloth import FastVisionModel
logger.info(f"Loading model: {model_name}")
# Disable compilation for the model
with torch._dynamo.config.patch(disable=True):
self.model, self.processor = FastVisionModel.from_pretrained(
model_name,
device_map="auto",
load_in_4bit=True,
use_gradient_checkpointing="unsloth",
token=os.environ["HF_API_KEY"]
)
# Ensure model is not compiled
if hasattr(self.model, '_dynamo_compile'):
self.model._dynamo_compile = False
self.max_tokens = max_tokens
self.prompt = ""
self.device = next(self.model.parameters()).device
logger.info(f"Model loaded successfully on device: {self.device}")
except ImportError:
logger.error("unsloth not found. Please install it: pip install unsloth")
raise
except Exception as e:
logger.error(f"Error loading model: {e}")
raise
def prepare_inputs(self, image: Image.Image) -> dict:
"""Prepare inputs for the model with proper error handling."""
try:
messages = [
{
"role": "user",
"content": [
{
"type": "image",
},
{"type": "text", "text": self.prompt},
],
}
]
text = self.processor.apply_chat_template(
messages, tokenize=False, add_generation_prompt=True
)
inputs = self.processor(
image,
text,
add_special_tokens=False,
return_tensors="pt",
)
return inputs
except Exception as e:
logger.error(f"Error preparing inputs: {e}")
raise
def predict(self, image: Image.Image) -> str:
"""Predict text from image with comprehensive error handling."""
try:
if image is None:
return "Please upload an image."
# Convert numpy array to PIL Image if needed
if hasattr(image, 'shape'): # numpy array
image = Image.fromarray(image)
inputs = self.prepare_inputs(image)
# Move inputs to the same device as model with explicit device handling
device = self.device
logger.info(f"Moving inputs to device: {device}")
# Manually move each tensor to device
for key in inputs:
if hasattr(inputs[key], 'to'):
inputs[key] = inputs[key].to(device)
# Ensure attention_mask is float32 and on correct device
if 'attention_mask' in inputs:
inputs['attention_mask'] = inputs['attention_mask'].to(dtype=torch.float32, device=device)
logger.info(f"Generating text with max_tokens={self.max_tokens}")
# Disable compilation during generation
with torch.no_grad(), torch._dynamo.config.patch(disable=True):
generated_ids = self.model.generate(
**inputs,
max_new_tokens=self.max_tokens,
use_cache=True,
do_sample=False,
temperature=0.1,
pad_token_id=self.processor.tokenizer.eos_token_id
)
generated_ids_trimmed = [
out_ids[len(in_ids):] for in_ids, out_ids in zip(inputs['input_ids'], generated_ids)
]
output_text = self.processor.batch_decode(
generated_ids_trimmed,
skip_special_tokens=True,
clean_up_tokenization_spaces=False
)
result = output_text[0].strip()
logger.info(f"Generated text: {result[:100]}...")
return result
except Exception as e:
logger.error(f"Error during prediction: {e}")
return f"Error processing image: {str(e)}"
def __call__(self, image: Union[Image.Image, str]) -> str:
"""Callable interface for the model."""
if isinstance(image, str):
return "Please upload an image file."
return self.predict(image)
# Global model instance
atlas_ocr = None
def load_model():
"""Load the model globally to avoid reloading."""
global atlas_ocr
if atlas_ocr is None:
try:
atlas_ocr = AtlasOCR()
except Exception as e:
logger.error(f"Failed to load model: {e}")
return False
return True
@spaces.GPU
def perform_ocr(image):
"""Main OCR function with proper error handling."""
try:
if not load_model():
return "Error: Failed to load model. Please check the logs."
if image is None:
return "Please upload an image to extract text."
result = atlas_ocr(image)
return result
except Exception as e:
logger.error(f"Error in perform_ocr: {e}")
return f"An error occurred: {str(e)}"
def process_with_status(image):
"""Process image and return result with status - moved outside to avoid pickling issues."""
if image is None:
return "Please upload an image.", "No image provided"
try:
result = perform_ocr(image)
return result, "Processing completed successfully"
except Exception as e:
return f"Error: {str(e)}", f"Error occurred: {str(e)}"
def create_interface():
"""Create the Gradio interface with proper configuration."""
with gr.Blocks(
title="AtlasOCR - Darija Document OCR",
theme=gr.themes.Soft(),
css="""
.gradio-container {
max-width: 1200px !important;
}
"""
) as demo:
gr.Markdown("""
# AtlasOCR - Darija Document OCR
Upload an image to extract Darija text in real-time. This model is specialized for Darija document OCR.
""")
with gr.Row():
with gr.Column(scale=1):
# Input image
image_input = gr.Image(
type="pil",
label="Upload Image",
height=400
)
# Submit button
submit_btn = gr.Button(
"Extract Text",
variant="primary",
size="lg"
)
# Clear button
clear_btn = gr.Button("Clear", variant="secondary")
with gr.Column(scale=1):
# Output text
output = gr.Textbox(
label="Extracted Text",
lines=20,
show_copy_button=True,
placeholder="Extracted text will appear here..."
)
# Status indicator
status = gr.Textbox(
label="Status",
value="Ready to process images",
interactive=False
)
# Model details
with gr.Accordion("Model Information", open=False):
gr.Markdown("""
**Model:** AtlasOCR-v0
**Description:** Specialized Darija OCR model for Arabic dialect text extraction
**Size:** 3B parameters
**Context window:** Supports up to 2000 output tokens
**Optimization:** 4-bit quantization for efficient inference
""")
gr.Examples(
examples=[
["i3.png"],
["i6.png"]
],
inputs=image_input,
outputs=[output, status], # <-- required
fn=process_with_status, # <-- required
label="Example Images",
examples_per_page=4,
cache_examples=True
)
# Set up processing flow
submit_btn.click(
fn=process_with_status,
inputs=image_input,
outputs=[output, status]
)
image_input.change(
fn=process_with_status,
inputs=image_input,
outputs=[output, status]
)
clear_btn.click(
fn=lambda: (None, "", "Ready to process images"),
outputs=[image_input, output, status]
)
return demo
# Create and launch the interface
if __name__ == "__main__":
demo = create_interface()
demo.launch(
server_name="0.0.0.0",
server_port=7860,
share=False,
debug=True
) |