wiklif commited on
Commit
f090db9
·
1 Parent(s): a070f41
Files changed (2) hide show
  1. a.py +9 -0
  2. app.py +21 -28
a.py ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ from gradio_client import Client
2
+
3
+
4
+ client = Client("wiklif/mixtral-api")
5
+ result = client.predict(
6
+ name="Jak się dziś masz?",
7
+ api_name="/predict"
8
+ )
9
+ print(result)
app.py CHANGED
@@ -4,14 +4,16 @@ import gradio as gr
4
  client = InferenceClient("mistralai/Mixtral-8x7B-Instruct-v0.1")
5
 
6
  def format_prompt(message, history):
7
- prompt = "<s>"
8
- for user_prompt, bot_response in history:
9
- prompt += f"[INST] {user_prompt} [/INST]"
10
- prompt += f" {bot_response}</s> "
11
- prompt += f"[INST] {message} [/INST]"
12
- return prompt
13
-
14
- def generate(prompt, history, temperature=0.2, max_new_tokens=256, top_p=0.95, repetition_penalty=1.0):
 
 
15
  temperature = float(temperature)
16
  if temperature < 1e-2:
17
  temperature = 1e-2
@@ -36,24 +38,15 @@ def generate(prompt, history, temperature=0.2, max_new_tokens=256, top_p=0.95, r
36
  yield output
37
  return output
38
 
39
- # Definicja API Gradio
40
- def predict(prompt, temperature=0.2, max_new_tokens=256, top_p=0.95, repetition_penalty=1.0):
41
- history = [] # lub inny sposób na przechowywanie historii rozmowy
42
- output = generate(prompt, history, temperature, max_new_tokens, top_p, repetition_penalty)
43
- return output
44
 
45
- # Utworzenie interfejsu API w Gradio
46
- demo = gr.Interface(
47
- fn=predict,
48
- inputs=[
49
- gr.inputs.Textbox(lines=2, placeholder="Wpisz prompt tutaj..."),
50
- gr.inputs.Slider(0, 1, default=0.2, label="Temperature"),
51
- gr.inputs.Slider(1, 500, default=256, label="Max New Tokens"),
52
- gr.inputs.Slider(0, 1, default=0.95, label="Top P"),
53
- gr.inputs.Slider(0.5, 2, default=1.0, label="Repetition Penalty")
54
- ],
55
- outputs="text"
56
- )
57
-
58
- if __name__ == "__main__":
59
- demo.launch(server_name="0.0.0.0", server_port=7860)
 
4
  client = InferenceClient("mistralai/Mixtral-8x7B-Instruct-v0.1")
5
 
6
  def format_prompt(message, history):
7
+ prompt = "<s>"
8
+ for user_prompt, bot_response in history:
9
+ prompt += f"[INST] {user_prompt} [/INST]"
10
+ prompt += f" {bot_response}</s> "
11
+ prompt += f"[INST] {message} [/INST]"
12
+ return prompt
13
+
14
+ def generate(
15
+ prompt, history, temperature=0, max_new_tokens=3500, top_p=0.95, repetition_penalty=1.0,
16
+ ):
17
  temperature = float(temperature)
18
  if temperature < 1e-2:
19
  temperature = 1e-2
 
38
  yield output
39
  return output
40
 
 
 
 
 
 
41
 
42
+ mychatbot = gr.Chatbot(
43
+ avatar_images=["./user.png", "./botm.png"], bubble_full_width=False, show_label=False, show_copy_button=True, likeable=True,)
44
+
45
+ demo = gr.ChatInterface(fn=generate,
46
+ chatbot=mychatbot,
47
+ title="Tomoniai's Mixtral 8x7b Chat",
48
+ retry_btn=None,
49
+ undo_btn=None
50
+ )
51
+
52
+ demo.queue().launch(show_api=True)