import fastapi
import json
import markdown
import uvicorn
from fastapi.responses import HTMLResponse
from fastapi.middleware.cors import CORSMiddleware
from langchain.callbacks.streaming_stdout import StreamingStdOutCallbackHandler
from ctransformers import AutoModelForCausalLM
from pydantic import BaseModel
from sse_starlette.sse import EventSourceResponse
config = {
"temperature": 0.8,
"top_p": 0.95,
"top_k": 50,
"max_new_tokens": 1024,
"use_cache": True,
"do_sample": True,
"repetition_penalty": 1.02,
"max_seq_len": 4096
}
llm = AutoModelForCausalLM.from_pretrained('TheBloke/MPT-7B-Storywriter-GGML',
model_file='mpt-7b-storywriter.ggmlv3.q4_0.bin',
model_type='mpt', config=config)
app = fastapi.FastAPI()
app.add_middleware(
CORSMiddleware,
allow_origins=["*"],
allow_credentials=True,
allow_methods=["*"],
allow_headers=["*"],
)
@app.get("/")
async def index():
with open("README.md", "r", encoding="utf-8") as readme_file:
md_template_string = readme_file.read()
html_content = markdown.markdown(md_template_string)
return HTMLResponse(content=html_content, status_code=200)
class ChatCompletionRequest(BaseModel):
prompt: str
@app.get("/demo")
async def demo():
html_content = """
StoryWriter Demo
"""
return HTMLResponse(content=html_content, status_code=200)
@app.get("/stream")
async def chat(prompt = "Once upon a time there was a "):
completion = llm(prompt)
async def server_sent_events(chat_chunks):
yield prompt
for chat_chunk in chat_chunks:
yield chat_chunk
yield ""
return EventSourceResponse(server_sent_events(completion))
@app.post("/v1/chat/completions")
async def chat(request: ChatCompletionRequest, response_mode=None):
completion = llm(request.prompt)
async def server_sent_events(
chat_chunks,
):
for chat_chunk in chat_chunks:
yield dict(data=json.dumps(chat_chunk))
yield dict(data="[DONE]")
chunks = completion
return EventSourceResponse(
server_sent_events(chunks),
)
if __name__ == "__main__":
uvicorn.run(app, host="0.0.0.0", port=8000)