Spaces:
Running
Running
| 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 | |
| llm = AutoModelForCausalLM.from_pretrained('TheBloke/MPT-7B-Storywriter-GGML', | |
| model_file='mpt-7b-storywriter.ggmlv3.q4_0.bin', | |
| model_type='mpt') | |
| app = fastapi.FastAPI() | |
| app.add_middleware( | |
| CORSMiddleware, | |
| allow_origins=["*"], | |
| allow_credentials=True, | |
| allow_methods=["*"], | |
| allow_headers=["*"], | |
| ) | |
| 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 | |
| async def demo(): | |
| html_content = """ | |
| <!DOCTYPE html> | |
| <html> | |
| <head> | |
| <style> | |
| body { | |
| align-items: center; | |
| background-color: #d9b99b; | |
| display: flex; | |
| height: 100vh; | |
| justify-content: center; | |
| } | |
| #content { | |
| align-items: center; | |
| background-color: #fff0db; | |
| box-shadow: | |
| 12px 12px 16px 0 rgba(0, 0, 0, 0.25), | |
| -8px -8px 12px 0 rgba(255, 255, 255, 0.3); | |
| border-radius: 50px; | |
| display: flex; | |
| padding: 50px; | |
| justify-content: center; | |
| margin-right: 4rem; | |
| font-size: 16px; | |
| } | |
| </style> | |
| </head> | |
| <body> | |
| <div id="content"></div> | |
| <script> | |
| var source = new EventSource("https://matthoffner-storywriter.hf.space/stream"); | |
| source.onmessage = function(event) { | |
| document.getElementById("content").innerHTML += event.data | |
| }; | |
| </script> | |
| </body> | |
| </html> | |
| """ | |
| return HTMLResponse(content=html_content, status_code=200) | |
| async def flow(prompt = ""): | |
| 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)) | |
| async def chat(prompt = "Once upon a time there was a "): | |
| tokens = llm.tokenize(prompt) | |
| async def server_sent_events(chat_chunks, llm): | |
| yield prompt | |
| for chat_chunk in llm.generate(chat_chunks): | |
| yield llm.detokenize(chat_chunk) | |
| yield "" | |
| return EventSourceResponse(server_sent_events(tokens, llm)) | |
| 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) | |