seawolf2357 commited on
Commit
0096596
·
verified ·
1 Parent(s): 1aa7536

Create app.py

Browse files
Files changed (1) hide show
  1. app.py +40 -0
app.py ADDED
@@ -0,0 +1,40 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ import gradio as gr
2
+ import requests
3
+
4
+ def call_api(prompt, history, system_prompt, temperature, max_new_tokens, top_p, repetition_penalty):
5
+ url = "https://seawolf2357-nvidia-llama-fastapi.hf.space/generate/"
6
+ data = {
7
+ "prompt": prompt,
8
+ "history": history,
9
+ "system_prompt": system_prompt,
10
+ "temperature": temperature,
11
+ "max_new_tokens": max_new_tokens,
12
+ "top_p": top_p,
13
+ "repetition_penalty": repetition_penalty
14
+ }
15
+ response = requests.post(url, json=data)
16
+ return response.json()['response']
17
+
18
+ def main_interface():
19
+ with gr.Blocks() as demo:
20
+ with gr.Row():
21
+ prompt = gr.Textbox(label="Prompt", placeholder="Enter your prompt here...")
22
+ history = gr.Textbox(label="History", placeholder="Enter history here...")
23
+ system_prompt = gr.Textbox(label="System Prompt", placeholder="Enter system prompt here...")
24
+ temperature = gr.Number(label="Temperature", value=0.0)
25
+ max_new_tokens = gr.Number(label="Max New Tokens", value=1048)
26
+ top_p = gr.Number(label="Top P", value=0.15)
27
+ repetition_penalty = gr.Number(label="Repetition Penalty", value=1.0)
28
+ submit_button = gr.Button("Generate")
29
+
30
+ submit_button.click(
31
+ fn=call_api,
32
+ inputs=[prompt, history, system_prompt, temperature, max_new_tokens, top_p, repetition_penalty],
33
+ outputs=[gr.Textbox(label="Response")]
34
+ )
35
+
36
+ return demo
37
+
38
+ if __name__ == "__main__":
39
+ demo = main_interface()
40
+ demo.launch()