Spaces:
Runtime error
Runtime error
Upload debug_model_loading.py with huggingface_hub
Browse files- debug_model_loading.py +108 -0
debug_model_loading.py
ADDED
@@ -0,0 +1,108 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import torch
|
2 |
+
import os
|
3 |
+
import sys
|
4 |
+
import traceback
|
5 |
+
import requests
|
6 |
+
import json
|
7 |
+
import platform
|
8 |
+
|
9 |
+
print("=" * 50)
|
10 |
+
print("DETAILED MODEL LOADING DIAGNOSTIC")
|
11 |
+
print("=" * 50)
|
12 |
+
|
13 |
+
# System information
|
14 |
+
print("\n1. SYSTEM INFORMATION:")
|
15 |
+
print(f"Python version: {sys.version}")
|
16 |
+
print(f"PyTorch version: {torch.__version__}")
|
17 |
+
print(f"Platform: {platform.platform()}")
|
18 |
+
print(f"Processor: {platform.processor()}")
|
19 |
+
|
20 |
+
# Environment variables
|
21 |
+
print("\n2. ENVIRONMENT VARIABLES:")
|
22 |
+
relevant_vars = ["CUDA_VISIBLE_DEVICES", "NVIDIA_VISIBLE_DEVICES", "TRANSFORMERS_CACHE", "HF_HOME"]
|
23 |
+
for var in relevant_vars:
|
24 |
+
print(f"{var}: {os.environ.get(var, 'Not set')}")
|
25 |
+
|
26 |
+
# GPU information
|
27 |
+
print("\n3. GPU DETECTION:")
|
28 |
+
print(f"CUDA available: {torch.cuda.is_available()}")
|
29 |
+
if torch.cuda.is_available():
|
30 |
+
try:
|
31 |
+
print(f"CUDA version: {torch.version.cuda}")
|
32 |
+
print(f"GPU count: {torch.cuda.device_count()}")
|
33 |
+
for i in range(torch.cuda.device_count()):
|
34 |
+
print(f"GPU {i}: {torch.cuda.get_device_name(i)}")
|
35 |
+
|
36 |
+
# Test GPU with a simple operation
|
37 |
+
print("\nTesting GPU with tensor operations...")
|
38 |
+
test_tensor = torch.rand(1000, 1000, device="cuda")
|
39 |
+
start = torch.cuda.Event(enable_timing=True)
|
40 |
+
end = torch.cuda.Event(enable_timing=True)
|
41 |
+
|
42 |
+
start.record()
|
43 |
+
result = torch.matmul(test_tensor, test_tensor)
|
44 |
+
end.record()
|
45 |
+
|
46 |
+
torch.cuda.synchronize()
|
47 |
+
print(f"GPU tensor operation completed in {start.elapsed_time(end):.2f} ms")
|
48 |
+
|
49 |
+
# Memory info
|
50 |
+
print(f"\nTotal GPU memory: {torch.cuda.get_device_properties(0).total_memory / 1e9:.2f} GB")
|
51 |
+
print(f"Allocated GPU memory: {torch.cuda.memory_allocated() / 1e9:.2f} GB")
|
52 |
+
print(f"Reserved GPU memory: {torch.cuda.memory_reserved() / 1e9:.2f} GB")
|
53 |
+
|
54 |
+
except Exception as e:
|
55 |
+
print(f"Error testing GPU: {str(e)}")
|
56 |
+
traceback.print_exc()
|
57 |
+
else:
|
58 |
+
print("CUDA is not available. This is a critical issue for model loading.")
|
59 |
+
|
60 |
+
# HuggingFace hub connectivity
|
61 |
+
print("\n4. HUGGINGFACE HUB CONNECTIVITY:")
|
62 |
+
try:
|
63 |
+
print("Testing connection to HuggingFace Hub...")
|
64 |
+
response = requests.get("https://huggingface.co/api/models/OpenGVLab/InternViT-6B-224px")
|
65 |
+
if response.status_code == 200:
|
66 |
+
print("Successfully connected to HuggingFace Hub")
|
67 |
+
model_info = response.json()
|
68 |
+
print(f"Model exists: OpenGVLab/InternViT-6B-224px")
|
69 |
+
if 'downloads' in model_info:
|
70 |
+
print(f"Downloads: {model_info['downloads']}")
|
71 |
+
else:
|
72 |
+
print(f"Failed to connect to HuggingFace Hub: Status code {response.status_code}")
|
73 |
+
print(response.text)
|
74 |
+
except Exception as e:
|
75 |
+
print(f"Error connecting to HuggingFace Hub: {str(e)}")
|
76 |
+
traceback.print_exc()
|
77 |
+
|
78 |
+
# Attempt model loading with detailed error capture
|
79 |
+
print("\n5. ATTEMPTING MODEL LOADING:")
|
80 |
+
try:
|
81 |
+
print("Importing transformers...")
|
82 |
+
from transformers import AutoModel, AutoProcessor
|
83 |
+
print("✓ Transformers imported successfully")
|
84 |
+
|
85 |
+
print("\nLoading AutoProcessor...")
|
86 |
+
processor = AutoProcessor.from_pretrained("OpenGVLab/InternViT-6B-224px")
|
87 |
+
print("✓ AutoProcessor loaded successfully")
|
88 |
+
|
89 |
+
print("\nLoading AutoModel...")
|
90 |
+
model = AutoModel.from_pretrained("OpenGVLab/InternViT-6B-224px")
|
91 |
+
print("✓ AutoModel loaded successfully")
|
92 |
+
|
93 |
+
if torch.cuda.is_available():
|
94 |
+
print("\nMoving model to CUDA...")
|
95 |
+
model = model.to("cuda")
|
96 |
+
print("✓ Model moved to CUDA successfully")
|
97 |
+
|
98 |
+
print("\nModel loading SUCCESSFUL")
|
99 |
+
print(f"Model parameters: {sum(p.numel() for p in model.parameters()):,}")
|
100 |
+
|
101 |
+
except Exception as e:
|
102 |
+
print(f"\n❌ ERROR LOADING MODEL: {str(e)}")
|
103 |
+
print("\nDetailed traceback:")
|
104 |
+
traceback.print_exc()
|
105 |
+
|
106 |
+
print("\n" + "=" * 50)
|
107 |
+
print("DIAGNOSTIC COMPLETE")
|
108 |
+
print("=" * 50)
|