app.py

from fastapi import FastAPI
from transformers import AutoModelForCausalLM, AutoTokenizer

# Wrap problematic imports in try-except blocks
try:
    from peft import PeftModel, PeftConfig
except ImportError as e:
    print(f"Error importing from peft: {e}")
    raise

try:
    from mistral_common.tokenizer import AutoMistralTokenizer
except ImportError as e:
    print(f"Error importing from mistral_common: {e}")
    raise

# Initialize FastAPI app
app = FastAPI()

# Load PEFT model configuration and base model
try:
    config = PeftConfig.from_pretrained("frankmorales2020/Mistral-7B-text-to-sql-flash-attention-2-dataeval")
    base_model = AutoModelForCausalLM.from_pretrained("mistralai/Mistral-7B-Instruct-v0.3")
    model = PeftModel.from_pretrained(base_model, "frankmorales2020/Mistral-7B-text-to-sql-flash-attention-2-dataeval")

    # Load recommended tokenizer
    tokenizer = AutoMistralTokenizer.from_pretrained("mistralai/Mistral-7B-Instruct-v0.3")

    # Create the pipeline
    from transformers import pipeline
    pipe = pipeline("text2text-generation", model=model, tokenizer=tokenizer)
except Exception as e:
    print(f"Error loading model or creating pipeline: {e}")
    raise

@app.get("/")
def home():
    return {"message": "Hello World"}

@app.get("/generate")
def generate(text: str):
    try:
        output = pipe(text)
        return {"output": output[0]['generated_text']}
    except Exception as e:
        return {"error": str(e)}