File size: 1,586 Bytes
f1c69e7
a0e0f92
 
 
 
 
 
 
d42d32c
a0e0f92
 
 
 
 
 
f586823
a0e0f92
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2804f3c
a0e0f92
 
 
2804f3c
869a9b8
 
 
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
from fastapi import FastAPI, Query, Request
from pydantic import BaseModel
from huggingface_hub import InferenceClient
import uvicorn


app = FastAPI()

client = InferenceClient("mistralai/Mixtral-8x7B-Instruct-v0.1", headers={"X-use-cache": "false"})

class Item(BaseModel):
    prompt: str
    history: list
    system_prompt: str
    temperature: float = 0.0
    max_new_tokens: int = 16384
    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

def generate(item: Item):
    temperature = float(item.temperature)
    if temperature < 1e-2:
        temperature = 1e-2
    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 = client.text_generation(formatted_prompt, **generate_kwargs, stream=True, details=True, return_full_text=False)
    output = ""

    for response in stream:
        output += response.token.text
    return output

@app.post("/chat/completions")
async def generate_text(item: Item):
    return {"response": generate(item)}

@app.get("/ping")
async def ping(request: Request):
    return "pong"