Spaces:
Running
Running
File size: 5,825 Bytes
3ddd5b6 f9e8a03 3ddd5b6 9b54a18 3ddd5b6 f023e65 f9e8a03 3ddd5b6 f023e65 3ddd5b6 f023e65 3ddd5b6 f023e65 f9e8a03 f023e65 3ddd5b6 f9e8a03 3ddd5b6 f9e8a03 f023e65 3ddd5b6 f9e8a03 3ddd5b6 f9e8a03 f023e65 f9e8a03 f023e65 f9e8a03 f023e65 f9e8a03 f023e65 f9e8a03 f023e65 f9e8a03 9b54a18 3ddd5b6 9b54a18 3ddd5b6 9b54a18 3ddd5b6 f9e8a03 3ddd5b6 f9e8a03 3ddd5b6 f9e8a03 3ddd5b6 f9e8a03 3ddd5b6 f9e8a03 |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 |
import os
import requests
from fastapi import FastAPI, HTTPException
from fastapi.middleware.cors import CORSMiddleware
from fastapi.responses import HTMLResponse
from llama_cpp import Llama
from pydantic import BaseModel
import uvicorn
import json
# Configuration
MODEL_URL = "https://huggingface.co/unsloth/DeepSeek-R1-Distill-Qwen-1.5B-GGUF/resolve/main/DeepSeek-R1-Distill-Qwen-1.5B-Q5_K_M.gguf"
MODEL_NAME = "DeepSeek-R1-Distill-Qwen-1.5B-Q5_K_M.gguf"
MODEL_DIR = "model"
MODEL_PATH = os.path.join(MODEL_DIR, MODEL_NAME)
# Create model directory if it doesn't exist
os.makedirs(MODEL_DIR, exist_ok=True)
# Download the model if it doesn't exist
if not os.path.exists(MODEL_PATH):
print(f"Downloading model from {MODEL_URL}...")
response = requests.get(MODEL_URL, stream=True)
if response.status_code == 200:
with open(MODEL_PATH, "wb") as f:
for chunk in response.iter_content(chunk_size=8192):
f.write(chunk)
print("Model downloaded successfully!")
else:
raise RuntimeError(f"Failed to download model: HTTP {response.status_code}")
else:
print("Model already exists. Skipping download.")
# Initialize FastAPI
app = FastAPI(
title="DeepSeek-R1 OpenAI-Compatible API",
description="OpenAI-compatible API for DeepSeek-R1-Distill-Qwen-1.5B",
version="1.0.0"
)
# CORS Configuration
app.add_middleware(
CORSMiddleware,
allow_origins=["*"],
allow_methods=["*"],
allow_headers=["*"],
)
# Load the model
print("Loading model...")
try:
llm = Llama(
model_path=MODEL_PATH,
n_ctx=2048,
n_threads=4,
n_gpu_layers=0,
verbose=False
)
print("Model loaded successfully!")
except Exception as e:
raise RuntimeError(f"Failed to load model: {str(e)}")
# Root endpoint with documentation
@app.get("/", response_class=HTMLResponse)
async def root():
return f"""
<html>
<head>
<title>DeepSeek-R1 OpenAI API</title>
<style>
body {{ font-family: Arial, sans-serif; max-width: 800px; margin: 20px auto; padding: 0 20px; }}
.warning {{ color: #dc3545; background: #ffeef0; padding: 15px; border-radius: 5px; }}
a {{ color: #007bff; text-decoration: none; }}
code {{ background: #f8f9fa; padding: 2px 4px; border-radius: 4px; }}
</style>
</head>
<body>
<h1>DeepSeek-R1 OpenAI-Compatible API</h1>
<div class="warning">
<h3>⚠️ Important Notice</h3>
<p>For private use, please duplicate this space:<br>
1. Click your profile picture in the top-right<br>
2. Select "Duplicate Space"<br>
3. Set visibility to Private</p>
</div>
<h2>API Documentation</h2>
<ul>
<li><a href="/docs">Interactive Swagger Documentation</a></li>
<li><a href="/redoc">ReDoc Documentation</a></li>
</ul>
<h2>API Endpoints</h2>
<h3>Chat Completion</h3>
<p><code>POST /v1/chat/completions</code></p>
<p>Parameters:</p>
<ul>
<li><strong>messages</strong>: List of message objects</li>
<li><strong>max_tokens</strong>: Maximum response length (default: 128)</li>
<li><strong>temperature</strong>: Sampling temperature (default: 0.7)</li>
<li><strong>top_p</strong>: Nucleus sampling threshold (default: 0.9)</li>
</ul>
<h2>Example Request</h2>
<pre>
curl -X POST "{os.environ.get('SPACE_HOST', 'http://localhost:7860')}/v1/chat/completions" \\
-H "Content-Type: application/json" \\
-d '{{
"messages": [{{"role": "user", "content": "Explain quantum computing"}}],
"max_tokens": 150
}}'
</pre>
</body>
</html>
"""
# OpenAI-Compatible Request Schema
class ChatCompletionRequest(BaseModel):
model: str = "DeepSeek-R1-Distill-Qwen-1.5B"
messages: list[dict]
max_tokens: int = 128
temperature: float = 0.7
top_p: float = 0.9
stream: bool = False
# OpenAI-Compatible Response Schema
class ChatCompletionResponse(BaseModel):
id: str = "chatcmpl-12345"
object: str = "chat.completion"
created: int = 1693161600
model: str = "DeepSeek-R1-Distill-Qwen-1.5B"
choices: list[dict]
usage: dict
@app.get("/v1/chat/completions")
async def chat_completion(
messages: str,
max_tokens: int = 128,
temperature: float = 0.7,
top_p: float = 0.9,
stream: bool = False
):
try:
messages_list = json.loads(messages)
prompt = "\n".join([f"{msg['role']}: {msg['content']}" for msg in messages_list])
prompt += "\nassistant:"
response = llm(
prompt=prompt,
max_tokens=max_tokens,
temperature=temperature,
top_p=top_p,
stop=["</s>"]
)
return ChatCompletionResponse(
choices=[{
"index": 0,
"message": {
"role": "assistant",
"content": response['choices'][0]['text'].strip()
},
"finish_reason": "stop"
}],
usage={
"prompt_tokens": len(prompt),
"completion_tokens": len(response['choices'][0]['text']),
"total_tokens": len(prompt) + len(response['choices'][0]['text'])
}
)
except Exception as e:
raise HTTPException(status_code=500, detail=str(e))
@app.get("/health")
def health_check():
return {"status": "healthy"}
if __name__ == "__main__":
uvicorn.run(app, host="0.0.0.0", port=7860) |