Spaces:
Sleeping
Sleeping
File size: 3,752 Bytes
087ce88 93aa8dc 087ce88 93aa8dc 087ce88 93aa8dc 087ce88 93aa8dc 087ce88 93aa8dc 087ce88 93aa8dc 087ce88 |
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 |
import logging
from transformers import AutoTokenizer, AutoModelForCausalLM
import torch
from huggingface_hub import login
from .config import Config
logger = logging.getLogger(__name__)
class ModelManager:
def __init__(self, model_name: str):
self.model_name = model_name
self.tokenizer = None
self.model = None
self.device = "cuda" if torch.cuda.is_available() else "cpu"
# Login to Hugging Face Hub
if Config.HUGGING_FACE_TOKEN:
logger.info("Logging in to Hugging Face Hub")
login(token=Config.HUGGING_FACE_TOKEN)
# Initialize tokenizer and model
self._init_tokenizer()
self._init_model()
def _init_tokenizer(self):
"""Initialize the tokenizer."""
try:
logger.info(f"Loading tokenizer: {self.model_name}")
self.tokenizer = AutoTokenizer.from_pretrained(
self.model_name,
token=Config.HUGGING_FACE_TOKEN
)
# Ensure we have the necessary special tokens
special_tokens = {
'pad_token': '[PAD]',
'eos_token': '</s>',
'bos_token': '<s>'
}
self.tokenizer.add_special_tokens(special_tokens)
logger.info("Tokenizer loaded successfully.")
except Exception as e:
logger.error(f"Error loading tokenizer: {str(e)}")
raise
def _init_model(self):
"""Initialize the model."""
try:
logger.info(f"Loading model: {self.model_name}")
# Load model with CPU configuration
self.model = AutoModelForCausalLM.from_pretrained(
self.model_name,
device_map={"": self.device},
torch_dtype=torch.float32, # Use float32 for CPU
token=Config.HUGGING_FACE_TOKEN,
low_cpu_mem_usage=True
)
# Resize embeddings to match tokenizer
self.model.resize_token_embeddings(len(self.tokenizer))
logger.info(f"Using device: {self.device}")
except Exception as e:
logger.error(f"Error loading model: {str(e)}")
raise
def generate_text(self, prompt: str, max_new_tokens: int = 1024) -> str:
"""Generate text from prompt."""
try:
# Encode the prompt
inputs = self.tokenizer(prompt, return_tensors="pt", truncation=True, max_length=2048)
inputs = {k: v.to(self.device) for k, v in inputs.items()}
# Generate response
with torch.no_grad():
outputs = self.model.generate(
**inputs,
max_new_tokens=max_new_tokens,
do_sample=True,
temperature=Config.TEMPERATURE,
top_p=Config.TOP_P,
pad_token_id=self.tokenizer.pad_token_id,
eos_token_id=self.tokenizer.eos_token_id,
)
# Decode and return the generated text
generated_text = self.tokenizer.decode(outputs[0], skip_special_tokens=True)
# Extract only the generated part (remove the prompt)
response = generated_text[len(prompt):].strip()
return response
except Exception as e:
logger.error(f"Error generating text: {str(e)}")
return """- Issues:
- Error generating code review
- Model inference failed
- Improvements:
- Please try again
- Check model configuration
- Best Practices:
- Ensure proper model setup
- Verify token permissions
- Security:
- No immediate concerns"""
|