|
import fastapi |
|
import json |
|
import markdown |
|
import uvicorn |
|
from fastapi import HTTPException |
|
from fastapi.responses import HTMLResponse |
|
from fastapi.middleware.cors import CORSMiddleware |
|
from sse_starlette.sse import EventSourceResponse |
|
from starlette.responses import StreamingResponse |
|
from ctransformers import AutoModelForCausalLM |
|
from pydantic import BaseModel |
|
from typing import List, Dict, Any, Generator |
|
|
|
|
|
llm = AutoModelForCausalLM.from_pretrained("TheBloke/WizardCoder-15B-1.0-GGML", |
|
model_file="WizardCoder-15B-1.0.ggmlv3.q4_0.bin", |
|
model_type="starcoder") |
|
app = fastapi.FastAPI(title="🪄WizardCoder💫") |
|
app.add_middleware( |
|
CORSMiddleware, |
|
allow_origins=["*"], |
|
allow_credentials=True, |
|
allow_methods=["*"], |
|
allow_headers=["*"], |
|
) |
|
|
|
@app.get("/") |
|
async def index(): |
|
html_content = """ |
|
<html> |
|
<head> |
|
</head> |
|
<body style="background-color:black"> |
|
<h2 style="font-family:system-ui"><a href="https://huggingface.co./TheBloke/WizardCoder-15B-1.0-GGML">wizardcoder-ggml</a></h2> |
|
<iframe |
|
src="https://matthoffner-monacopilot.hf.space" |
|
frameborder="0" |
|
width="95%" |
|
height="90%" |
|
></iframe> |
|
<h2 style="font-family:system-ui"><a href="https://matthoffner-wizardcoder-ggml.hf.space/docs">FastAPI Docs</a></h2> |
|
</body> |
|
</html> |
|
""" |
|
return HTMLResponse(content=html_content, status_code=200) |
|
|
|
class ChatCompletionRequestV0(BaseModel): |
|
prompt: str |
|
|
|
class Message(BaseModel): |
|
role: str |
|
content: str |
|
|
|
class ChatCompletionRequest(BaseModel): |
|
messages: List[Message] |
|
max_tokens: int = 250 |
|
|
|
@app.post("/v1/completions") |
|
async def completion(request: ChatCompletionRequest, response_mode=None): |
|
response = llm(request.prompt) |
|
return response |
|
|
|
@app.post("/v1/chat/completions") |
|
async def chat(request: ChatCompletionRequest): |
|
combined_messages = ' '.join([message.content for message in request.messages]) |
|
tokens = llm.tokenize(combined_messages) |
|
|
|
try: |
|
chat_chunks = llm.generate(tokens, max_new_tokens=request.max_tokens) |
|
except Exception as e: |
|
raise HTTPException(status_code=500, detail=str(e)) |
|
|
|
async def format_response(chat_chunks: Generator) -> Any: |
|
for chat_chunk in chat_chunks: |
|
response = { |
|
'choices': [ |
|
{ |
|
'message': { |
|
'role': 'system', |
|
'content': llm.detokenize(chat_chunk) |
|
}, |
|
'finish_reason': 'stop' if llm.detokenize(chat_chunk) == "[DONE]" else 'unknown' |
|
} |
|
] |
|
} |
|
yield f"data: {json.dumps(response)}\n\n" |
|
yield "event: done\ndata: {}\n\n" |
|
|
|
return StreamingResponse(format_response(chat_chunks), media_type="text/event-stream") |
|
|
|
@app.post("/v0/chat/completions") |
|
async def chat(request: ChatCompletionRequestV0, response_mode=None): |
|
tokens = llm.tokenize(request.prompt) |
|
async def server_sent_events(chat_chunks, llm): |
|
for chat_chunk in llm.generate(chat_chunks): |
|
yield dict(data=json.dumps(llm.detokenize(chat_chunk))) |
|
yield dict(data="[DONE]") |
|
|
|
return EventSourceResponse(server_sent_events(tokens, llm)) |
|
|
|
if __name__ == "__main__": |
|
uvicorn.run(app, host="0.0.0.0", port=8000) |
