Spaces:
Runtime error
Runtime error
Update app.py
Browse files
app.py
CHANGED
@@ -25,6 +25,8 @@ from werkzeug.utils import secure_filename
|
|
25 |
from geopy.geocoders import Nominatim
|
26 |
import pickle
|
27 |
import numpy as np
|
|
|
|
|
28 |
|
29 |
# Configure logging
|
30 |
logging.basicConfig(level=logging.INFO)
|
@@ -50,24 +52,41 @@ model_dir = "./models/llm_model"
|
|
50 |
device = "cuda" if torch.cuda.is_available() else "cpu"
|
51 |
print(f"Using device: {device}")
|
52 |
|
|
|
|
|
|
|
|
|
|
|
53 |
# Load SentenceTransformer model
|
54 |
def load_sentence_transformer():
|
55 |
print("Loading SentenceTransformer model...")
|
56 |
try:
|
57 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
58 |
|
59 |
# Load and optimize model state dict
|
60 |
-
|
61 |
-
|
62 |
-
|
63 |
-
|
64 |
-
|
65 |
-
|
66 |
-
|
67 |
-
|
68 |
-
|
69 |
-
|
70 |
-
|
|
|
|
|
|
|
|
|
71 |
return model_embedding
|
72 |
except Exception as e:
|
73 |
print(f"Error loading model: {str(e)}")
|
|
|
25 |
from geopy.geocoders import Nominatim
|
26 |
import pickle
|
27 |
import numpy as np
|
28 |
+
import tempfile
|
29 |
+
from pathlib import Path
|
30 |
|
31 |
# Configure logging
|
32 |
logging.basicConfig(level=logging.INFO)
|
|
|
52 |
device = "cuda" if torch.cuda.is_available() else "cpu"
|
53 |
print(f"Using device: {device}")
|
54 |
|
55 |
+
# Configure cache directories
|
56 |
+
os.environ['TRANSFORMERS_CACHE'] = '/cache'
|
57 |
+
os.environ['HF_HOME'] = '/cache'
|
58 |
+
os.environ['XDG_CACHE_HOME'] = '/cache'
|
59 |
+
|
60 |
# Load SentenceTransformer model
|
61 |
def load_sentence_transformer():
|
62 |
print("Loading SentenceTransformer model...")
|
63 |
try:
|
64 |
+
# Create cache directory if it doesn't exist
|
65 |
+
cache_dir = Path('/cache')
|
66 |
+
cache_dir.mkdir(parents=True, exist_ok=True)
|
67 |
+
|
68 |
+
model_embedding = SentenceTransformer(
|
69 |
+
"jinaai/jina-embeddings-v3",
|
70 |
+
trust_remote_code=True,
|
71 |
+
cache_folder=str(cache_dir)
|
72 |
+
).to(device)
|
73 |
|
74 |
# Load and optimize model state dict
|
75 |
+
if os.path.exists(model_path):
|
76 |
+
state_dict = torch.load(model_path, map_location=device)
|
77 |
+
|
78 |
+
# Dequantize if needed
|
79 |
+
for key, tensor in state_dict.items():
|
80 |
+
if hasattr(tensor, 'dequantize'):
|
81 |
+
state_dict[key] = tensor.dequantize().to(dtype=torch.float32)
|
82 |
+
elif tensor.dtype == torch.bfloat16:
|
83 |
+
state_dict[key] = tensor.to(dtype=torch.float32)
|
84 |
+
|
85 |
+
model_embedding.load_state_dict(state_dict)
|
86 |
+
print("SentenceTransformer model loaded successfully.")
|
87 |
+
else:
|
88 |
+
print(f"Warning: Model file not found at {model_path}")
|
89 |
+
|
90 |
return model_embedding
|
91 |
except Exception as e:
|
92 |
print(f"Error loading model: {str(e)}")
|