Spaces:
Runtime error
Runtime error
Error: Error processing document with GOT-OCR (ZeroGPU): Numpy is not available
Browse files- src/parsers/got_ocr_parser.py +88 -12
src/parsers/got_ocr_parser.py
CHANGED
@@ -4,6 +4,8 @@ import json
|
|
4 |
import os
|
5 |
import tempfile
|
6 |
import logging
|
|
|
|
|
7 |
|
8 |
from src.parsers.parser_interface import DocumentParser
|
9 |
from src.parsers.parser_registry import ParserRegistry
|
@@ -11,17 +13,22 @@ from src.parsers.parser_registry import ParserRegistry
|
|
11 |
# Configure logging
|
12 |
logger = logging.getLogger(__name__)
|
13 |
|
14 |
-
#
|
|
|
|
|
|
|
|
|
15 |
try:
|
16 |
-
|
17 |
-
|
18 |
-
|
19 |
-
|
20 |
-
|
21 |
-
|
22 |
-
|
23 |
-
|
24 |
-
|
|
|
25 |
import torch
|
26 |
import transformers
|
27 |
from transformers import AutoModel, AutoTokenizer
|
@@ -37,7 +44,6 @@ try:
|
|
37 |
GOT_AVAILABLE = True and NUMPY_AVAILABLE
|
38 |
except ImportError:
|
39 |
GOT_AVAILABLE = False
|
40 |
-
NUMPY_AVAILABLE = False
|
41 |
logger.warning("GOT-OCR dependencies not installed. The parser will not be available.")
|
42 |
|
43 |
class GotOcrParser(DocumentParser):
|
@@ -72,6 +78,8 @@ class GotOcrParser(DocumentParser):
|
|
72 |
@classmethod
|
73 |
def _load_model(cls):
|
74 |
"""Load the GOT-OCR model and tokenizer if not already loaded."""
|
|
|
|
|
75 |
if not NUMPY_AVAILABLE:
|
76 |
raise ImportError("NumPy is not available. This is required for GOT-OCR.")
|
77 |
|
@@ -127,15 +135,73 @@ class GotOcrParser(DocumentParser):
|
|
127 |
|
128 |
logger.info("GOT-OCR model released from memory")
|
129 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
130 |
def parse(self, file_path: Union[str, Path], ocr_method: Optional[str] = None, **kwargs) -> str:
|
131 |
"""Parse a document using GOT-OCR 2.0."""
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
132 |
if not GOT_AVAILABLE:
|
133 |
if not NUMPY_AVAILABLE:
|
|
|
134 |
raise ImportError(
|
135 |
"NumPy is not available. This is required for GOT-OCR. "
|
136 |
-
"Please ensure NumPy is installed in your environment."
|
|
|
137 |
)
|
138 |
else:
|
|
|
139 |
raise ImportError(
|
140 |
"GOT-OCR dependencies not installed. Please install required packages: "
|
141 |
"torch, transformers, tiktoken, verovio, accelerate"
|
@@ -157,6 +223,16 @@ class GotOcrParser(DocumentParser):
|
|
157 |
ocr_type = "format" if ocr_method == "format" else "ocr"
|
158 |
|
159 |
try:
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
160 |
# Load the model
|
161 |
self._load_model()
|
162 |
|
|
|
4 |
import os
|
5 |
import tempfile
|
6 |
import logging
|
7 |
+
import sys
|
8 |
+
import importlib
|
9 |
|
10 |
from src.parsers.parser_interface import DocumentParser
|
11 |
from src.parsers.parser_registry import ParserRegistry
|
|
|
13 |
# Configure logging
|
14 |
logger = logging.getLogger(__name__)
|
15 |
|
16 |
+
# Global flag for NumPy availability
|
17 |
+
NUMPY_AVAILABLE = False
|
18 |
+
NUMPY_VERSION = None
|
19 |
+
|
20 |
+
# Try to import NumPy
|
21 |
try:
|
22 |
+
import numpy as np
|
23 |
+
NUMPY_AVAILABLE = True
|
24 |
+
NUMPY_VERSION = np.__version__
|
25 |
+
logger.info(f"NumPy version {NUMPY_VERSION} is available")
|
26 |
+
except ImportError:
|
27 |
+
NUMPY_AVAILABLE = False
|
28 |
+
logger.error("NumPy is not available. This is required for GOT-OCR.")
|
29 |
+
|
30 |
+
# Check if required packages are installed
|
31 |
+
try:
|
32 |
import torch
|
33 |
import transformers
|
34 |
from transformers import AutoModel, AutoTokenizer
|
|
|
44 |
GOT_AVAILABLE = True and NUMPY_AVAILABLE
|
45 |
except ImportError:
|
46 |
GOT_AVAILABLE = False
|
|
|
47 |
logger.warning("GOT-OCR dependencies not installed. The parser will not be available.")
|
48 |
|
49 |
class GotOcrParser(DocumentParser):
|
|
|
78 |
@classmethod
|
79 |
def _load_model(cls):
|
80 |
"""Load the GOT-OCR model and tokenizer if not already loaded."""
|
81 |
+
global NUMPY_AVAILABLE
|
82 |
+
|
83 |
if not NUMPY_AVAILABLE:
|
84 |
raise ImportError("NumPy is not available. This is required for GOT-OCR.")
|
85 |
|
|
|
135 |
|
136 |
logger.info("GOT-OCR model released from memory")
|
137 |
|
138 |
+
def _try_install_numpy(self):
|
139 |
+
"""Attempt to install NumPy using pip."""
|
140 |
+
global NUMPY_AVAILABLE, NUMPY_VERSION
|
141 |
+
|
142 |
+
logger.warning("Attempting to install NumPy...")
|
143 |
+
try:
|
144 |
+
import subprocess
|
145 |
+
# Try to install numpy without specifying a version
|
146 |
+
result = subprocess.run(
|
147 |
+
[sys.executable, "-m", "pip", "install", "-q", "numpy", "--no-cache-dir"],
|
148 |
+
capture_output=True,
|
149 |
+
text=True,
|
150 |
+
check=True
|
151 |
+
)
|
152 |
+
logger.info(f"NumPy installation result: {result.stdout}")
|
153 |
+
|
154 |
+
# Try to import numpy again
|
155 |
+
importlib.invalidate_caches()
|
156 |
+
import numpy as np
|
157 |
+
importlib.reload(np)
|
158 |
+
|
159 |
+
NUMPY_AVAILABLE = True
|
160 |
+
NUMPY_VERSION = np.__version__
|
161 |
+
logger.info(f"NumPy installed successfully: version {NUMPY_VERSION}")
|
162 |
+
return True
|
163 |
+
except Exception as e:
|
164 |
+
logger.error(f"Failed to install NumPy: {str(e)}")
|
165 |
+
if hasattr(e, 'stderr'):
|
166 |
+
logger.error(f"Installation error output: {e.stderr}")
|
167 |
+
return False
|
168 |
+
|
169 |
def parse(self, file_path: Union[str, Path], ocr_method: Optional[str] = None, **kwargs) -> str:
|
170 |
"""Parse a document using GOT-OCR 2.0."""
|
171 |
+
global NUMPY_AVAILABLE, GOT_AVAILABLE
|
172 |
+
|
173 |
+
# Check NumPy availability and try to install if not available
|
174 |
+
if not NUMPY_AVAILABLE:
|
175 |
+
logger.warning("NumPy not available, attempting to install it...")
|
176 |
+
if self._try_install_numpy():
|
177 |
+
# Update GOT availability flag if NumPy is now available
|
178 |
+
try:
|
179 |
+
import torch
|
180 |
+
import transformers
|
181 |
+
GOT_AVAILABLE = True
|
182 |
+
logger.info("Updated GOT availability after installing NumPy: Available")
|
183 |
+
except ImportError:
|
184 |
+
GOT_AVAILABLE = False
|
185 |
+
logger.error("Updated GOT availability after installing NumPy: Not Available (missing other dependencies)")
|
186 |
+
else:
|
187 |
+
logger.error("Failed to install NumPy. Cannot proceed with GOT-OCR.")
|
188 |
+
raise ImportError(
|
189 |
+
"NumPy is not available and could not be installed automatically. "
|
190 |
+
"Please ensure NumPy is installed in your environment. "
|
191 |
+
"Add the following to your logs for debugging: NUMPY_INSTALLATION_FAILED"
|
192 |
+
)
|
193 |
+
|
194 |
+
# Check overall GOT availability
|
195 |
if not GOT_AVAILABLE:
|
196 |
if not NUMPY_AVAILABLE:
|
197 |
+
logger.error("NumPy is still not available after installation attempt.")
|
198 |
raise ImportError(
|
199 |
"NumPy is not available. This is required for GOT-OCR. "
|
200 |
+
"Please ensure NumPy is installed in your environment. "
|
201 |
+
"Environment details: Python " + sys.version
|
202 |
)
|
203 |
else:
|
204 |
+
logger.error("Other GOT-OCR dependencies missing even though NumPy is available.")
|
205 |
raise ImportError(
|
206 |
"GOT-OCR dependencies not installed. Please install required packages: "
|
207 |
"torch, transformers, tiktoken, verovio, accelerate"
|
|
|
223 |
ocr_type = "format" if ocr_method == "format" else "ocr"
|
224 |
|
225 |
try:
|
226 |
+
# Check if numpy needs to be reloaded
|
227 |
+
if 'numpy' in sys.modules:
|
228 |
+
logger.info("NumPy module found in sys.modules, attempting to reload...")
|
229 |
+
try:
|
230 |
+
importlib.reload(sys.modules['numpy'])
|
231 |
+
import numpy as np
|
232 |
+
logger.info(f"NumPy reloaded successfully: version {np.__version__}")
|
233 |
+
except Exception as e:
|
234 |
+
logger.error(f"Error reloading NumPy: {str(e)}")
|
235 |
+
|
236 |
# Load the model
|
237 |
self._load_model()
|
238 |
|