Spaces:
Running
Running
File size: 5,677 Bytes
77daf9a |
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 |
# model_investigation.py - Run this to investigate your model
# Add this as a separate file to debug your model loading
from transformers import AutoConfig, AutoTokenizer
import logging
import traceback
logging.basicConfig(level=logging.INFO)
logger = logging.getLogger(__name__)
def investigate_model(model_name="Girinath11/aiml_code_debug_model"):
"""Investigate the model to understand its structure."""
print(f"π Investigating model: {model_name}")
print("=" * 60)
try:
# 1. Check config
print("π STEP 1: Checking model configuration...")
config = AutoConfig.from_pretrained(model_name)
print(f" β Config type: {type(config).__name__}")
print(f" β Model type: {getattr(config, 'model_type', 'Unknown')}")
print(f" β Architectures: {getattr(config, 'architectures', 'Unknown')}")
# Print all config attributes
print("\n π Full config attributes:")
for attr in sorted(dir(config)):
if not attr.startswith('_') and hasattr(config, attr):
try:
value = getattr(config, attr)
if not callable(value):
print(f" {attr}: {value}")
except:
pass
# 2. Check tokenizer
print(f"\nπ€ STEP 2: Checking tokenizer...")
try:
tokenizer = AutoTokenizer.from_pretrained(model_name)
print(f" β Tokenizer type: {type(tokenizer).__name__}")
print(f" β Vocab size: {len(tokenizer)}")
print(f" β Special tokens:")
print(f" pad_token: {tokenizer.pad_token}")
print(f" eos_token: {tokenizer.eos_token}")
print(f" unk_token: {tokenizer.unk_token}")
print(f" bos_token: {getattr(tokenizer, 'bos_token', 'None')}")
except Exception as e:
print(f" β Tokenizer error: {e}")
# 3. Try different loading approaches
print(f"\nπ€ STEP 3: Testing model loading approaches...")
from transformers import (
AutoModel,
AutoModelForSeq2SeqLM,
AutoModelForCausalLM,
pipeline
)
approaches = [
("AutoModel", lambda: AutoModel.from_pretrained(model_name, trust_remote_code=True)),
("AutoModelForCausalLM", lambda: AutoModelForCausalLM.from_pretrained(model_name, trust_remote_code=True)),
("AutoModelForSeq2SeqLM", lambda: AutoModelForSeq2SeqLM.from_pretrained(model_name, trust_remote_code=True)),
("Pipeline text-generation", lambda: pipeline("text-generation", model=model_name, trust_remote_code=True)),
("Pipeline text2text-generation", lambda: pipeline("text2text-generation", model=model_name, trust_remote_code=True)),
]
for approach_name, approach_func in approaches:
try:
print(f" π Trying {approach_name}...")
model = approach_func()
print(f" β
SUCCESS with {approach_name}!")
print(f" Model type: {type(model)}")
# Test a simple generation if possible
if hasattr(model, 'generate') or 'pipeline' in approach_name.lower():
print(f" β Supports text generation")
# Quick test
try:
test_input = "def hello(): print('world')"
if 'pipeline' in approach_name.lower():
result = model(f"Fix this code: {test_input}", max_length=100)
print(f" β Test generation successful: {str(result)[:100]}...")
else:
inputs = tokenizer(f"Fix: {test_input}", return_tensors="pt", max_length=100, truncation=True)
outputs = model.generate(**inputs, max_new_tokens=50)
result = tokenizer.decode(outputs[0], skip_special_tokens=True)
print(f" β Test generation successful: {result[:100]}...")
except Exception as gen_error:
print(f" β οΈ Generation test failed: {str(gen_error)[:100]}...")
# We found a working approach, let's break
print(f"\nπ RECOMMENDATION: Use {approach_name} for loading this model!")
break
except Exception as e:
print(f" β {approach_name} failed: {str(e)[:100]}...")
# 4. Model files inspection
print(f"\nπ STEP 4: Model files information...")
try:
from huggingface_hub import list_repo_files
files = list_repo_files(model_name)
model_files = [f for f in files if f.endswith(('.bin', '.safetensors', '.json'))]
print(f" π Key model files found:")
for file in sorted(model_files):
print(f" {file}")
except Exception as e:
print(f" β οΈ Could not list files: {e}")
except Exception as main_error:
print(f"\nβ CRITICAL ERROR: {main_error}")
print(f"Full traceback:\n{traceback.format_exc()}")
if __name__ == "__main__":
investigate_model()
print("\n" + "="*60)
print("π Investigation complete!")
print("Copy the successful loading approach to your model_wrapper.py") |