| 1 | # SPDX-License-Identifier: Apache-2.0 |
| 2 | # SPDX-FileCopyrightText: Copyright contributors to the vLLM project |
| 3 | """Example for starting a Gradio Webserver |
| 4 | Start vLLM API server: |
| 5 | python examples/applications/api_server/server.py \ |
| 6 | --model meta-llama/Llama-2-7b-chat-hf |
| 7 | |
| 8 | Start Webserver: |
| 9 | python examples/applications/chatbot/gradio_webserver.py |
| 10 | |
| 11 | Note that `pip install --upgrade gradio` is needed to run this example. |
| 12 | More details: https://github.com/gradio-app/gradio |
| 13 | |
| 14 | If your antivirus software blocks the download of frpc for gradio, |
| 15 | you can install it manually by following these steps: |
| 16 | |
| 17 | 1. Download this file: https://cdn-media.huggingface.co/frpc-gradio-0.3/frpc_linux_amd64 |
| 18 | 2. Rename the downloaded file to: frpc_linux_amd64_v0.3 |
| 19 | 3. Move the file to this location: /home/user/.cache/huggingface/gradio/frpc |
| 20 | """ |
| 21 | |
| 22 | import argparse |
| 23 | import json |
| 24 | |
| 25 | import gradio as gr |
| 26 | import requests |
| 27 | |
| 28 | |
| 29 | def http_bot(prompt): |
| 30 | headers = {"User-Agent": "vLLM Client"} |
| 31 | pload = { |
| 32 | "prompt": prompt, |
| 33 | "stream": True, |
| 34 | "max_tokens": 128, |
| 35 | } |
| 36 | response = requests.post(args.model_url, headers=headers, json=pload, stream=True) |
| 37 | |
| 38 | for chunk in response.iter_lines( |
| 39 | chunk_size=8192, decode_unicode=False, delimiter=b"\n" |
| 40 | ): |
| 41 | if chunk: |
| 42 | data = json.loads(chunk.decode("utf-8")) |
| 43 | output = data["text"][0] |
| 44 | yield output |
| 45 | |
| 46 | |
| 47 | def build_demo(): |
| 48 | with gr.Blocks() as demo: |
| 49 | gr.Markdown("# vLLM text completion demo\n") |
| 50 | inputbox = gr.Textbox(label="Input", placeholder="Enter text and press ENTER") |
| 51 | outputbox = gr.Textbox( |
| 52 | label="Output", placeholder="Generated result from the model" |
| 53 | ) |
| 54 | inputbox.submit(http_bot, [inputbox], [outputbox]) |
| 55 | return demo |
| 56 | |
| 57 | |
| 58 | def parse_args(): |
| 59 | parser = argparse.ArgumentParser() |
| 60 | parser.add_argument("--host", type=str, default=None) |
| 61 | parser.add_argument("--port", type=int, default=8001) |
| 62 | parser.add_argument( |
| 63 | "--model-url", type=str, default="http://localhost:8000/generate" |
| 64 | ) |
| 65 | return parser.parse_args() |
| 66 | |
| 67 | |
| 68 | def main(args): |
| 69 | demo = build_demo() |
| 70 | demo.queue().launch(server_name=args.host, server_port=args.port, share=True) |
| 71 | |
| 72 | |
| 73 | if __name__ == "__main__": |
| 74 | args = parse_args() |
| 75 | main(args) |