app.py

from fastapi import FastAPI, HTTPException
from pydantic import BaseModel
from transformers import pipeline, AutoModelForCausalLM, AutoTokenizer, PeftModel

class ModelInput(BaseModel):
    prompt: str
    max_new_tokens: int = 50

app = FastAPI()

# Load base model and tokenizer
base_model_path = "HuggingFaceTB/SmolLM2-135M-Instruct"
adapter_path = "khurrameycon/SmolLM-135M-Instruct-qa_pairs_converted.json-25epochs"

# Initialize tokenizer from base model
tokenizer = AutoTokenizer.from_pretrained(base_model_path)

# Load base model
base_model = AutoModelForCausalLM.from_pretrained(
    base_model_path,
    device_map="auto",
    trust_remote_code=True
)

# Load and merge adapter weights
model = PeftModel.from_pretrained(base_model, adapter_path)
model = model.merge_and_unload()

# Initialize pipeline
generator = pipeline("text-generation", model=model, tokenizer=tokenizer)

def generate_response(model, tokenizer, instruction, max_new_tokens=128):
    try:
        messages = [{"role": "user", "content": instruction}]
        input_text = tokenizer.apply_chat_template(
            messages, tokenize=False, add_generation_prompt=True
        )
        inputs = tokenizer.encode(input_text, return_tensors="pt").to(model.device)
        outputs = model.generate(
            inputs,
            max_new_tokens=max_new_tokens,
            temperature=0.2,
            top_p=0.9,
            do_sample=True,
        )
        response = tokenizer.decode(outputs[0], skip_special_tokens=True)
        return response
    except Exception as e:
        raise ValueError(f"Error generating response: {e}")

@app.post("/generate")
def generate_text(input: ModelInput):
    try:
        response = generate_response(
            model=model,
            tokenizer=tokenizer,
            instruction=input.prompt,
            max_new_tokens=input.max_new_tokens
        )
        return {"generated_text": response}
    except Exception as e:
        raise HTTPException(status_code=500, detail=str(e))

@app.get("/")
def root():
    return {"message": "Welcome to the Hugging Face Model API!"}