Spaces:
Sleeping
Sleeping
File size: 10,609 Bytes
a77376b |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 |
"""
LLM utilities for handling different LLM providers.
Supports OpenAI and Hugging Face models.
"""
import os
from typing import Optional, List
from dotenv import load_dotenv
# Load environment variables
load_dotenv(override=True)
class LLMHandler:
"""Handler for different LLM providers."""
def __init__(
self,
provider: str = "openai",
model_name: Optional[str] = None,
temperature: float = 0.7,
max_tokens: int = 500
):
"""
Initialize LLM handler.
Args:
provider: LLM provider ("openai" or "huggingface")
model_name: Model name (optional, uses default if not provided)
temperature: Temperature for generation
max_tokens: Maximum tokens to generate
"""
self.provider = provider.lower()
self.temperature = temperature
self.max_tokens = max_tokens
self.model = None
self.tokenizer = None
self.embedding_model = None
if self.provider == "openai":
self._initialize_openai(model_name)
elif self.provider == "huggingface":
self._initialize_huggingface(model_name)
else:
raise ValueError(f"Unsupported provider: {provider}")
def _initialize_openai(self, model_name: Optional[str] = None):
"""Initialize OpenAI client."""
try:
from openai import OpenAI
api_key = os.getenv("OPENAI_API_KEY")
if not api_key:
raise ValueError("OPENAI_API_KEY not found in environment variables")
self.client = OpenAI(api_key=api_key)
self.model_name = model_name or os.getenv("OPENAI_MODEL", "gpt-3.5-turbo")
print(f"✓ OpenAI client initialized with model: {self.model_name}")
except ImportError:
raise ImportError("OpenAI package not installed. Run: pip install openai")
except Exception as e:
raise Exception(f"Failed to initialize OpenAI: {e}")
def _initialize_huggingface(self, model_name: Optional[str] = None):
"""Initialize Hugging Face model."""
try:
from transformers import AutoModelForSeq2SeqLM, AutoTokenizer, AutoModelForCausalLM
import torch
# Get model name from parameter or environment
if model_name is None:
model_name = os.getenv(
"HUGGINGFACE_MODEL",
"google/flan-t5-large"
)
self.model_name = model_name
print(f"Initializing LLM: huggingface - {self.model_name}")
# Get HF token
hf_token = os.getenv("HUGGINGFACE_API_TOKEN")
if not hf_token:
print("⚠️ Warning: HUGGINGFACE_API_TOKEN not found. Some models may not be accessible.")
# Determine device
self.device = "cuda" if torch.cuda.is_available() else "cpu"
print(f"Using device: {self.device}")
# Load tokenizer
print(f"Loading tokenizer for {self.model_name}...")
self.tokenizer = AutoTokenizer.from_pretrained(
self.model_name,
token=hf_token,
trust_remote_code=True
)
# Load model based on type
print(f"Loading model {self.model_name}...")
# Detect model type
if "t5" in self.model_name.lower() or "flan" in self.model_name.lower():
# Seq2Seq models (T5, Flan-T5)
self.model = AutoModelForSeq2SeqLM.from_pretrained(
self.model_name,
token=hf_token,
trust_remote_code=True,
torch_dtype=torch.float16 if self.device == "cuda" else torch.float32
)
else:
# Causal LM models (Mistral, Llama, etc.)
self.model = AutoModelForCausalLM.from_pretrained(
self.model_name,
token=hf_token,
trust_remote_code=True,
torch_dtype=torch.float16 if self.device == "cuda" else torch.float32
)
self.model.to(self.device)
self.model.eval()
print(f"✓ LLM initialized successfully")
except ImportError as e:
raise ImportError(f"Required packages not installed: {e}")
except Exception as e:
raise Exception(f"Failed to initialize Hugging Face model: {e}")
def generate(
self,
prompt: str,
system_message: Optional[str] = None,
temperature: Optional[float] = None,
max_tokens: Optional[int] = None
) -> str:
"""
Generate text using the LLM.
Args:
prompt: Input prompt
system_message: Optional system message
temperature: Optional temperature override
max_tokens: Optional max tokens override
Returns:
Generated text
"""
temp = temperature if temperature is not None else self.temperature
max_tok = max_tokens if max_tokens is not None else self.max_tokens
if self.provider == "openai":
return self._generate_openai(prompt, system_message, temp, max_tok)
elif self.provider == "huggingface":
return self._generate_huggingface(prompt, system_message, temp, max_tok)
def _generate_openai(
self,
prompt: str,
system_message: Optional[str],
temperature: float,
max_tokens: int
) -> str:
"""Generate using OpenAI."""
messages = []
if system_message:
messages.append({"role": "system", "content": system_message})
messages.append({"role": "user", "content": prompt})
response = self.client.chat.completions.create(
model=self.model_name,
messages=messages,
temperature=temperature,
max_tokens=max_tokens
)
return response.choices[0].message.content
def _generate_huggingface(
self,
prompt: str,
system_message: Optional[str],
temperature: float,
max_tokens: int
) -> str:
"""Generate using Hugging Face."""
import torch
# Construct full prompt
if system_message:
full_prompt = f"{system_message}\n\n{prompt}"
else:
full_prompt = prompt
# Tokenize
inputs = self.tokenizer(
full_prompt,
return_tensors="pt",
truncation=True,
max_length=512
).to(self.device)
# Generate
with torch.no_grad():
outputs = self.model.generate(
**inputs,
max_new_tokens=max_tokens,
temperature=temperature,
do_sample=temperature > 0,
top_p=0.9,
num_return_sequences=1,
pad_token_id=self.tokenizer.eos_token_id
)
# Decode
generated_text = self.tokenizer.decode(outputs[0], skip_special_tokens=True)
# For seq2seq models, return as-is
# For causal models, remove the prompt
if "t5" in self.model_name.lower() or "flan" in self.model_name.lower():
return generated_text.strip()
else:
# Remove the input prompt from output
return generated_text[len(full_prompt):].strip()
def generate_with_context(
self,
query: str,
context: str,
system_message: Optional[str] = None
) -> str:
"""
Generate answer using query and context.
Args:
query: User query
context: Retrieved context
system_message: Optional system message
Returns:
Generated answer
"""
prompt = f"""Context:
{context}
Question: {query}
Answer the question based on the context provided above. Be concise and accurate."""
return self.generate(prompt, system_message)
class EmbeddingHandler:
"""Handler for embedding models."""
def __init__(self, model_name: Optional[str] = None):
"""
Initialize embedding handler.
Args:
model_name: Embedding model name
"""
from sentence_transformers import SentenceTransformer
self.model_name = model_name or os.getenv(
"EMBEDDING_MODEL",
"sentence-transformers/all-MiniLM-L6-v2"
)
print(f"Loading embedding model: {self.model_name}")
self.model = SentenceTransformer(self.model_name)
print(f"✓ Embedding model loaded successfully")
def embed_documents(self, texts: List[str]) -> List[List[float]]:
"""
Embed a list of documents.
Args:
texts: List of text documents
Returns:
List of embeddings
"""
embeddings = self.model.encode(texts, convert_to_numpy=True)
return embeddings.tolist()
def embed_query(self, text: str) -> List[float]:
"""
Embed a single query.
Args:
text: Query text
Returns:
Embedding vector
"""
embedding = self.model.encode(text, convert_to_numpy=True)
return embedding.tolist()
def create_llm_handler(
provider: str = "openai",
model_name: Optional[str] = None,
temperature: float = 0.7,
max_tokens: int = 500
) -> LLMHandler:
"""
Create and return an LLM handler.
Args:
provider: LLM provider
model_name: Model name
temperature: Temperature
max_tokens: Max tokens
Returns:
LLMHandler instance
"""
return LLMHandler(provider, model_name, temperature, max_tokens)
def create_embedding_handler(model_name: Optional[str] = None) -> EmbeddingHandler:
"""
Create and return an embedding handler.
Args:
model_name: Embedding model name
Returns:
EmbeddingHandler instance
"""
return EmbeddingHandler(model_name) |