File size: 4,606 Bytes
a57f72f
 
7fdf7e0
2b6fd3b
 
 
a57f72f
2b6fd3b
 
a57f72f
 
bf31376
 
463d78b
a57f72f
2b6fd3b
3c54391
0247d00
341885c
08d233d
210500b
a57f72f
 
 
 
 
 
 
 
 
 
82c09a3
 
 
 
0b2af73
 
 
 
b11f460
82c09a3
 
 
a57f72f
 
46ac909
a57f72f
 
0d49ac1
 
 
 
46ac909
0d49ac1
4efb547
0d49ac1
7982958
65c273f
7982958
 
 
c796178
 
 
 
 
 
 
 
 
8abd12f
c796178
 
 
 
7904b82
c796178
210500b
94d3ebe
4f37acf
 
0d49ac1
 
ebbbded
0d49ac1
 
 
e177246
0d49ac1
865b816
3d19e0f
 
 
 
 
 
 
 
 
 
3afb5fe
3d19e0f
 
 
 
 
 
 
 
9e9fcb0
2b6fd3b
 
 
 
865b816
9e9fcb0
 
d7501b1
9e9fcb0
2b6fd3b
d7501b1
210500b
46ac909
f7d8687
3c54391
 
3650102
40717b9
a57f72f
3c54391
a57f72f
 
865b816
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
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
import json
import markdown
from typing import Callable, List, Dict, Any, Generator
from functools import partial

import fastapi
import uvicorn
from fastapi import HTTPException, Depends, Request
from fastapi.responses import HTMLResponse, StreamingResponse
from fastapi.middleware.cors import CORSMiddleware
from sse_starlette.sse import EventSourceResponse
from anyio import create_memory_object_stream
from anyio.to_thread import run_sync
from ctransformers import AutoModelForCausalLM
from pydantic import BaseModel

llm = AutoModelForCausalLM.from_pretrained("TheBloke/WizardCoder-15B-1.0-GGML",
                                           model_file="WizardCoder-15B-1.0.ggmlv3.q5_0.bin",
                                           model_type="starcoder",
                                           threads=8)
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="font-family:system-ui">
            <h2><a href="https://huggingface.co./TheBloke/WizardCoder-15B-1.0-GGML">wizardcoder-ggml</a></h2>
            <h2><a href="https://matthoffner-wizardcoder-ggml.hf.space/docs">FastAPI Docs</a></h2>
            <h2><a href="https://wizardcoder-sandbox.netlify.app">Wizardcoder Sandbox</a></h2>
            <h2><a href="https://matthoffner-monacopilot.hf.space">monacopilot</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: ChatCompletionRequestV0, response_mode=None):
    response = llm(request.prompt)
    return response

async def generate_response(chat_chunks, llm):
    for chat_chunk in chat_chunks:
        response = {
            'choices': [
                {
                    'message': {
                        'role': 'system',
                        'content': llm.detokenize(chat_chunk)
                    },
                    'finish_reason': 'stop' if chat_chunk == "<|end|>" else 'unknown'
                }
            ]
        }
        yield f"data: {json.dumps(response)}\n\n"
    yield dict(data="[DONE]")

@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)
    except Exception as e:
        raise HTTPException(status_code=500, detail=str(e))

    return StreamingResponse(generate_response(chat_chunks, llm), media_type="text/event-stream")

async def stream_response(tokens: Any) -> None:
    try:
        iterator: Generator = llm.generate(tokens)
        for chat_chunk in iterator:
            response = {
                'choices': [
                    {
                        'message': {
                            'role': 'system',
                            'content': llm.detokenize(chat_chunk)
                        },
                        'finish_reason': 'stop' if llm.is_eos_token(chat_chunk) else 'unknown'
                    }
                ]
            }
            yield f"data: {json.dumps(response)}\n\n"
        yield b"event: done\ndata: {}\n\n"
    except Exception as e:
        print(f"Exception in event publisher: {str(e)}")


async def chatV2(request: Request, body: ChatCompletionRequest):
    combined_messages = ' '.join([message.content for message in body.messages])
    tokens = llm.tokenize(combined_messages)

    return StreamingResponse(stream_response(tokens))

@app.post("/v2/chat/completions")
async def chatV2_endpoint(request: Request, body: ChatCompletionRequest):
    return await chatV2(request, body)


@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)