Spaces:
Running
Running
File size: 1,814 Bytes
5fa76ab dad25ac 5fa76ab dad25ac 5fa76ab 68b47d7 5fa76ab e6fa3d8 473963a 5fa76ab dad25ac 5fa76ab dad25ac 5fa76ab dad25ac |
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 |
from fastapi import FastAPI
from fastapi.responses import StreamingResponse
from pydantic import BaseModel
from huggingface_hub import InferenceClient
import uvicorn
import asyncio
app = FastAPI()
client = InferenceClient("mistralai/Mistral-7B-Instruct-v0.2")
class Item(BaseModel):
prompt: str
history: list
system_prompt: str
temperature: float = 0.0
max_new_tokens: int = 1048
top_p: float = 0.15
repetition_penalty: float = 1.0
def format_prompt(message, history):
prompt = "<s>"
for user_prompt, bot_response in history:
prompt += f"[INST] {user_prompt} [/INST]"
prompt += f" {bot_response}</s> "
prompt += f"[INST] {message} [/INST]"
return prompt
async def generate(item: Item):
temperature = max(float(item.temperature), 1e-2) # Ensure temperature is not too low
top_p = float(item.top_p)
generate_kwargs = dict(
temperature=temperature,
max_new_tokens=item.max_new_tokens,
top_p=top_p,
repetition_penalty=item.repetition_penalty,
do_sample=True,
seed=42,
)
formatted_prompt = format_prompt(f"{item.system_prompt}, {item.prompt}", item.history)
# Stream the response from the model
async def event_stream():
stream = client.text_generation(formatted_prompt, **generate_kwargs, stream=True, details=True, return_full_text=False)
async for response in stream:
yield response.token.text # Yield each token as it is received
# Optional: Add a small delay to simulate streaming effect (if needed)
await asyncio.sleep(0.1)
return event_stream()
@app.post("/generate/")
async def generate_text(item: Item):
return StreamingResponse(generate(item), media_type="text/event-stream") |