Spaces:
Running
Running
continues to remove cache parameter
Browse files
app.py
CHANGED
@@ -248,113 +248,25 @@ def image_to_base64(image):
|
|
248 |
image.save(buffered, format="PNG")
|
249 |
return base64.b64encode(buffered.getvalue()).decode()
|
250 |
|
251 |
-
def
|
252 |
"""
|
253 |
-
|
254 |
-
Optimized for ZeroGPU environments
|
255 |
"""
|
256 |
-
|
257 |
-
|
258 |
-
|
259 |
-
|
260 |
-
|
261 |
-
|
262 |
-
|
263 |
-
|
264 |
-
|
265 |
-
model.clear_cache()
|
266 |
-
elif hasattr(model, '_clear_cache'):
|
267 |
-
model._clear_cache()
|
268 |
-
|
269 |
-
# Try to clear cache from transformers
|
270 |
-
try:
|
271 |
-
from transformers import GenerationConfig
|
272 |
-
if hasattr(GenerationConfig, 'clear_cache'):
|
273 |
-
GenerationConfig.clear_cache()
|
274 |
-
except:
|
275 |
-
pass
|
276 |
-
|
277 |
-
# Retry the call
|
278 |
-
return model.chat(tokenizer, image_path, **kwargs)
|
279 |
-
except:
|
280 |
-
# If still failing, try with different parameters
|
281 |
-
try:
|
282 |
-
# Remove any cache-related parameters
|
283 |
-
kwargs_copy = kwargs.copy()
|
284 |
-
for key in list(kwargs_copy.keys()):
|
285 |
-
if 'cache' in key.lower():
|
286 |
-
del kwargs_copy[key]
|
287 |
-
return model.chat(tokenizer, image_path, **kwargs_copy)
|
288 |
-
except:
|
289 |
-
raise Exception("Model compatibility issue: DynamicCache error. Please try again.")
|
290 |
-
else:
|
291 |
-
raise e
|
292 |
-
except Exception as e:
|
293 |
-
# Handle other potential issues
|
294 |
-
if "attention_mask" in str(e).lower():
|
295 |
-
# Try to handle attention mask issues
|
296 |
-
try:
|
297 |
-
return model.chat(tokenizer, image_path, **kwargs)
|
298 |
-
except:
|
299 |
-
raise Exception(f"Attention mask error: {str(e)}")
|
300 |
-
else:
|
301 |
-
raise e
|
302 |
-
|
303 |
-
def safe_model_chat_crop(model, tokenizer, image_path, **kwargs):
|
304 |
-
"""
|
305 |
-
Safe wrapper for model.chat_crop to handle DynamicCache and other compatibility issues
|
306 |
-
Optimized for ZeroGPU environments
|
307 |
-
"""
|
308 |
-
try:
|
309 |
-
# First attempt: normal call
|
310 |
-
return model.chat_crop(tokenizer, image_path, **kwargs)
|
311 |
-
except AttributeError as e:
|
312 |
-
if "get_max_length" in str(e):
|
313 |
-
# Try to fix the cache issue by clearing it
|
314 |
-
try:
|
315 |
-
# Clear any existing cache
|
316 |
-
if hasattr(model, 'clear_cache'):
|
317 |
-
model.clear_cache()
|
318 |
-
elif hasattr(model, '_clear_cache'):
|
319 |
-
model._clear_cache()
|
320 |
-
|
321 |
-
# Try to clear cache from transformers
|
322 |
-
try:
|
323 |
-
from transformers import GenerationConfig
|
324 |
-
if hasattr(GenerationConfig, 'clear_cache'):
|
325 |
-
GenerationConfig.clear_cache()
|
326 |
-
except:
|
327 |
-
pass
|
328 |
-
|
329 |
-
# Retry the call
|
330 |
-
return model.chat_crop(tokenizer, image_path, **kwargs)
|
331 |
-
except:
|
332 |
-
# If still failing, try with different parameters
|
333 |
-
try:
|
334 |
-
# Remove any cache-related parameters
|
335 |
-
kwargs_copy = kwargs.copy()
|
336 |
-
for key in list(kwargs_copy.keys()):
|
337 |
-
if 'cache' in key.lower():
|
338 |
-
del kwargs_copy[key]
|
339 |
-
return model.chat_crop(tokenizer, image_path, **kwargs_copy)
|
340 |
-
except:
|
341 |
-
raise Exception("Model compatibility issue: DynamicCache error. Please try again.")
|
342 |
-
else:
|
343 |
-
raise e
|
344 |
-
except Exception as e:
|
345 |
-
# Handle other potential issues
|
346 |
-
if "attention_mask" in str(e).lower():
|
347 |
-
# Try to handle attention mask issues
|
348 |
-
try:
|
349 |
-
return model.chat_crop(tokenizer, image_path, **kwargs)
|
350 |
-
except:
|
351 |
-
raise Exception(f"Attention mask error: {str(e)}")
|
352 |
-
else:
|
353 |
-
raise e
|
354 |
-
|
355 |
|
356 |
@spaces.GPU()
|
357 |
def process_image(image, task, ocr_type=None, ocr_box=None, ocr_color=None):
|
|
|
|
|
|
|
358 |
if image is None:
|
359 |
return "Error: No image provided", None, None
|
360 |
|
@@ -381,61 +293,22 @@ def process_image(image, task, ocr_type=None, ocr_box=None, ocr_color=None):
|
|
381 |
else:
|
382 |
return "Error: Unsupported image format", None, None
|
383 |
|
384 |
-
#
|
385 |
try:
|
386 |
if task == "Plain Text OCR":
|
387 |
-
|
388 |
-
try:
|
389 |
-
res = cache_manager.safe_call('chat', tokenizer, image_path, ocr_type='ocr')
|
390 |
-
except:
|
391 |
-
try:
|
392 |
-
# Fallback to direct call
|
393 |
-
res = cache_manager.direct_call('chat', tokenizer, image_path, ocr_type='ocr')
|
394 |
-
except:
|
395 |
-
# Final fallback to legacy call
|
396 |
-
res = cache_manager.legacy_call('chat', tokenizer, image_path, ocr_type='ocr')
|
397 |
return res, None, unique_id
|
398 |
else:
|
399 |
if task == "Format Text OCR":
|
400 |
-
|
401 |
-
res = cache_manager.safe_call('chat', tokenizer, image_path, ocr_type='format', render=True, save_render_file=result_path)
|
402 |
-
except:
|
403 |
-
try:
|
404 |
-
res = cache_manager.direct_call('chat', tokenizer, image_path, ocr_type='format', render=True, save_render_file=result_path)
|
405 |
-
except:
|
406 |
-
res = cache_manager.legacy_call('chat', tokenizer, image_path, ocr_type='format', render=True, save_render_file=result_path)
|
407 |
elif task == "Fine-grained OCR (Box)":
|
408 |
-
|
409 |
-
res = cache_manager.safe_call('chat', tokenizer, image_path, ocr_type=ocr_type, ocr_box=ocr_box, render=True, save_render_file=result_path)
|
410 |
-
except:
|
411 |
-
try:
|
412 |
-
res = cache_manager.direct_call('chat', tokenizer, image_path, ocr_type=ocr_type, ocr_box=ocr_box, render=True, save_render_file=result_path)
|
413 |
-
except:
|
414 |
-
res = cache_manager.legacy_call('chat', tokenizer, image_path, ocr_type=ocr_type, ocr_box=ocr_box, render=True, save_render_file=result_path)
|
415 |
elif task == "Fine-grained OCR (Color)":
|
416 |
-
|
417 |
-
res = cache_manager.safe_call('chat', tokenizer, image_path, ocr_type=ocr_type, ocr_color=ocr_color, render=True, save_render_file=result_path)
|
418 |
-
except:
|
419 |
-
try:
|
420 |
-
res = cache_manager.direct_call('chat', tokenizer, image_path, ocr_type=ocr_type, ocr_color=ocr_color, render=True, save_render_file=result_path)
|
421 |
-
except:
|
422 |
-
res = cache_manager.legacy_call('chat', tokenizer, image_path, ocr_type=ocr_type, ocr_color=ocr_color, render=True, save_render_file=result_path)
|
423 |
elif task == "Multi-crop OCR":
|
424 |
-
|
425 |
-
res = cache_manager.safe_call('chat_crop', tokenizer, image_path, ocr_type='format', render=True, save_render_file=result_path)
|
426 |
-
except:
|
427 |
-
try:
|
428 |
-
res = cache_manager.direct_call('chat_crop', tokenizer, image_path, ocr_type='format', render=True, save_render_file=result_path)
|
429 |
-
except:
|
430 |
-
res = cache_manager.legacy_call('chat_crop', tokenizer, image_path, ocr_type='format', render=True, save_render_file=result_path)
|
431 |
elif task == "Render Formatted OCR":
|
432 |
-
|
433 |
-
res = cache_manager.safe_call('chat', tokenizer, image_path, ocr_type='format', render=True, save_render_file=result_path)
|
434 |
-
except:
|
435 |
-
try:
|
436 |
-
res = cache_manager.direct_call('chat', tokenizer, image_path, ocr_type='format', render=True, save_render_file=result_path)
|
437 |
-
except:
|
438 |
-
res = cache_manager.legacy_call('chat', tokenizer, image_path, ocr_type='format', render=True, save_render_file=result_path)
|
439 |
|
440 |
if os.path.exists(result_path):
|
441 |
with open(result_path, 'r') as f:
|
@@ -443,19 +316,39 @@ def process_image(image, task, ocr_type=None, ocr_box=None, ocr_color=None):
|
|
443 |
return res, html_content, unique_id
|
444 |
else:
|
445 |
return res, None, unique_id
|
446 |
-
except
|
447 |
-
|
448 |
-
|
449 |
-
|
450 |
-
|
451 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
452 |
|
453 |
except Exception as e:
|
454 |
return f"Error: {str(e)}", None, None
|
455 |
finally:
|
456 |
if os.path.exists(image_path):
|
457 |
os.remove(image_path)
|
458 |
-
|
459 |
def update_image_input(task):
|
460 |
if task == "Fine-grained OCR (Color)":
|
461 |
return gr.update(visible=False), gr.update(visible=True), gr.update(visible=True)
|
|
|
248 |
image.save(buffered, format="PNG")
|
249 |
return base64.b64encode(buffered.getvalue()).decode()
|
250 |
|
251 |
+
def direct_model_call(model, method_name, *args, **kwargs):
|
252 |
"""
|
253 |
+
Direct model call without any cache-related parameters
|
|
|
254 |
"""
|
255 |
+
# Create a clean kwargs dict without any cache-related parameters
|
256 |
+
clean_kwargs = {}
|
257 |
+
for key, value in kwargs.items():
|
258 |
+
if 'cache' not in key.lower():
|
259 |
+
clean_kwargs[key] = value
|
260 |
+
|
261 |
+
# Get the method and call it directly
|
262 |
+
method = getattr(model, method_name)
|
263 |
+
return method(*args, **clean_kwargs)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
264 |
|
265 |
@spaces.GPU()
|
266 |
def process_image(image, task, ocr_type=None, ocr_box=None, ocr_color=None):
|
267 |
+
"""
|
268 |
+
Process image with OCR using ZeroGPU-compatible approach
|
269 |
+
"""
|
270 |
if image is None:
|
271 |
return "Error: No image provided", None, None
|
272 |
|
|
|
293 |
else:
|
294 |
return "Error: Unsupported image format", None, None
|
295 |
|
296 |
+
# Use direct model calls without any cache management
|
297 |
try:
|
298 |
if task == "Plain Text OCR":
|
299 |
+
res = direct_model_call(model, 'chat', tokenizer, image_path, ocr_type='ocr')
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
300 |
return res, None, unique_id
|
301 |
else:
|
302 |
if task == "Format Text OCR":
|
303 |
+
res = direct_model_call(model, 'chat', tokenizer, image_path, ocr_type='format', render=True, save_render_file=result_path)
|
|
|
|
|
|
|
|
|
|
|
|
|
304 |
elif task == "Fine-grained OCR (Box)":
|
305 |
+
res = direct_model_call(model, 'chat', tokenizer, image_path, ocr_type=ocr_type, ocr_box=ocr_box, render=True, save_render_file=result_path)
|
|
|
|
|
|
|
|
|
|
|
|
|
306 |
elif task == "Fine-grained OCR (Color)":
|
307 |
+
res = direct_model_call(model, 'chat', tokenizer, image_path, ocr_type=ocr_type, ocr_color=ocr_color, render=True, save_render_file=result_path)
|
|
|
|
|
|
|
|
|
|
|
|
|
308 |
elif task == "Multi-crop OCR":
|
309 |
+
res = direct_model_call(model, 'chat_crop', tokenizer, image_path, ocr_type='format', render=True, save_render_file=result_path)
|
|
|
|
|
|
|
|
|
|
|
|
|
310 |
elif task == "Render Formatted OCR":
|
311 |
+
res = direct_model_call(model, 'chat', tokenizer, image_path, ocr_type='format', render=True, save_render_file=result_path)
|
|
|
|
|
|
|
|
|
|
|
|
|
312 |
|
313 |
if os.path.exists(result_path):
|
314 |
with open(result_path, 'r') as f:
|
|
|
316 |
return res, html_content, unique_id
|
317 |
else:
|
318 |
return res, None, unique_id
|
319 |
+
except Exception as e:
|
320 |
+
# If direct call fails, try with cache manager as fallback
|
321 |
+
try:
|
322 |
+
if task == "Plain Text OCR":
|
323 |
+
res = cache_manager.safe_call('chat', tokenizer, image_path, ocr_type='ocr')
|
324 |
+
return res, None, unique_id
|
325 |
+
else:
|
326 |
+
if task == "Format Text OCR":
|
327 |
+
res = cache_manager.safe_call('chat', tokenizer, image_path, ocr_type='format', render=True, save_render_file=result_path)
|
328 |
+
elif task == "Fine-grained OCR (Box)":
|
329 |
+
res = cache_manager.safe_call('chat', tokenizer, image_path, ocr_type=ocr_type, ocr_box=ocr_box, render=True, save_render_file=result_path)
|
330 |
+
elif task == "Fine-grained OCR (Color)":
|
331 |
+
res = cache_manager.safe_call('chat', tokenizer, image_path, ocr_type=ocr_type, ocr_color=ocr_color, render=True, save_render_file=result_path)
|
332 |
+
elif task == "Multi-crop OCR":
|
333 |
+
res = cache_manager.safe_call('chat_crop', tokenizer, image_path, ocr_type='format', render=True, save_render_file=result_path)
|
334 |
+
elif task == "Render Formatted OCR":
|
335 |
+
res = cache_manager.safe_call('chat', tokenizer, image_path, ocr_type='format', render=True, save_render_file=result_path)
|
336 |
+
|
337 |
+
if os.path.exists(result_path):
|
338 |
+
with open(result_path, 'r') as f:
|
339 |
+
html_content = f.read()
|
340 |
+
return res, html_content, unique_id
|
341 |
+
else:
|
342 |
+
return res, None, unique_id
|
343 |
+
except Exception as fallback_error:
|
344 |
+
return f"Error: {str(fallback_error)}", None, None
|
345 |
|
346 |
except Exception as e:
|
347 |
return f"Error: {str(e)}", None, None
|
348 |
finally:
|
349 |
if os.path.exists(image_path):
|
350 |
os.remove(image_path)
|
351 |
+
|
352 |
def update_image_input(task):
|
353 |
if task == "Fine-grained OCR (Color)":
|
354 |
return gr.update(visible=False), gr.update(visible=True), gr.update(visible=True)
|