File size: 2,272 Bytes
bd499ae 92b64f3 bd499ae 92b64f3 bd499ae 92b64f3 bd499ae 92b64f3 bd499ae 92b64f3 bd499ae 92b64f3 bd499ae 92b64f3 |
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 |
import torch
from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline
class SimpleTransformersLLM:
"""A simple wrapper for Hugging Face Transformers models."""
def __init__(self, model_name="google/flan-t5-small"):
"""Initialize with a small model that works on CPU."""
try:
self.tokenizer = AutoTokenizer.from_pretrained(model_name)
self.pipe = pipeline(
"text-generation",
model=model_name,
tokenizer=self.tokenizer,
max_length=512,
device_map="auto"
)
except Exception as e:
print(f"Error initializing model: {e}")
self.pipe = None
def complete(self, prompt):
"""Complete a prompt with the model."""
class Response:
def __init__(self, text):
self.text = text
if self.pipe is None:
return Response("Model initialization failed.")
try:
result = self.pipe(prompt, max_length=len(prompt) + 200, do_sample=True)
generated_text = result[0]["generated_text"]
# Extract only the new text (not including the prompt)
response_text = generated_text[len(prompt):].strip()
if not response_text:
response_text = "I couldn't generate a proper response."
return Response(response_text)
except Exception as e:
print(f"Error generating response: {e}")
return Response(f"Error generating response: {str(e)}")
def setup_llm():
"""Set up a simple LLM that doesn't require API keys."""
try:
# Try with a very small model first
return SimpleTransformersLLM("google/flan-t5-small")
except Exception as e:
print(f"Error setting up LLM: {e}")
# Fallback to dummy LLM
class DummyLLM:
def complete(self, prompt):
class Response:
def __init__(self, text):
self.text = text
return Response("This is a dummy response. The actual model couldn't be loaded.")
return DummyLLM()
|