File size: 781 Bytes
6081d20 f8f1df1 6081d20 3fc441b 6081d20 f8f1df1 6081d20 f8f1df1 6081d20 f8f1df1 6081d20 3fc441b 6081d20 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 |
import torch
from transformers import AutoModelForCausalLM, AutoTokenizer
from fastapi import FastAPI
from pydantic import BaseModel
# Load the model and tokenizer
model_name = "databricks/dolly-v2-3b"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
# FastAPI app
app = FastAPI()
# Input format
class ChatInput(BaseModel):
user_input: str
@app.post("/chat")
async def chat(chat_input: ChatInput):
inputs = tokenizer(chat_input.user_input, return_tensors="pt").to("cuda")
outputs = model.generate(**inputs, max_length=200, do_sample=True)
response_text = tokenizer.decode(outputs[0], skip_special_tokens=True)
return {"response": response_text}
|