Test502 commited on
Commit
d7e6590
·
verified ·
1 Parent(s): 8042b79

Update app.py

Browse files
Files changed (1) hide show
  1. app.py +50 -54
app.py CHANGED
@@ -1,69 +1,65 @@
1
  import gradio as gr
2
- from huggingface_hub import InferenceClient
 
3
 
 
4
 
5
- def respond(
6
- message,
7
- history: list[dict[str, str]],
8
- system_message,
9
- max_tokens,
10
- temperature,
11
- top_p,
12
- hf_token: gr.OAuthToken,
13
- ):
14
- """
15
- For more information on `huggingface_hub` Inference API support, please check the docs: https://huggingface.co/docs/huggingface_hub/v0.22.2/en/guides/inference
16
- """
17
- client = InferenceClient(token=hf_token.token, model="openai/gpt-oss-20b")
18
 
19
- messages = [{"role": "system", "content": system_message}]
 
 
 
 
 
20
 
21
- messages.extend(history)
 
22
 
23
- messages.append({"role": "user", "content": message})
 
 
24
 
25
- response = ""
 
 
 
26
 
27
- for message in client.chat_completion(
28
  messages,
29
- max_tokens=max_tokens,
30
- stream=True,
31
- temperature=temperature,
32
- top_p=top_p,
33
- ):
34
- choices = message.choices
35
- token = ""
36
- if len(choices) and choices[0].delta.content:
37
- token = choices[0].delta.content
38
 
39
- response += token
40
- yield response
 
 
41
 
 
 
 
 
 
 
 
 
42
 
43
- """
44
- For information on how to customize the ChatInterface, peruse the gradio docs: https://www.gradio.app/docs/chatinterface
45
- """
46
- chatbot = gr.ChatInterface(
47
- respond,
48
- additional_inputs=[
49
- gr.Textbox(value="You are a friendly Chatbot.", label="System message"),
50
- gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"),
51
- gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
52
- gr.Slider(
53
- minimum=0.1,
54
- maximum=1.0,
55
- value=0.95,
56
- step=0.05,
57
- label="Top-p (nucleus sampling)",
58
- ),
59
- ],
60
- )
61
 
62
- with gr.Blocks() as demo:
63
- with gr.Sidebar():
64
- gr.LoginButton()
65
- chatbot.render()
66
 
67
 
68
- if __name__ == "__main__":
69
- demo.launch()
 
 
 
 
 
 
1
  import gradio as gr
2
+ from transformers import AutoTokenizer, AutoModelForCausalLM
3
+ import torch
4
 
5
+ MODEL_ID = "saai-sa/ASL-4B-v1"
6
 
7
+ tokenizer = AutoTokenizer.from_pretrained(
8
+ MODEL_ID,
9
+ trust_remote_code=True
10
+ )
 
 
 
 
 
 
 
 
 
11
 
12
+ model = AutoModelForCausalLM.from_pretrained(
13
+ MODEL_ID,
14
+ torch_dtype=torch.float16 if torch.cuda.is_available() else torch.float32,
15
+ device_map="auto",
16
+ trust_remote_code=True
17
+ )
18
 
19
+ def respond(message, history):
20
+ messages = []
21
 
22
+ for item in history:
23
+ if isinstance(item, dict):
24
+ messages.append(item)
25
 
26
+ messages.append({
27
+ "role": "user",
28
+ "content": message
29
+ })
30
 
31
+ prompt = tokenizer.apply_chat_template(
32
  messages,
33
+ tokenize=False,
34
+ add_generation_prompt=True
35
+ )
 
 
 
 
 
 
36
 
37
+ inputs = tokenizer(
38
+ prompt,
39
+ return_tensors="pt"
40
+ ).to(model.device)
41
 
42
+ with torch.no_grad():
43
+ outputs = model.generate(
44
+ **inputs,
45
+ max_new_tokens=300,
46
+ temperature=0.7,
47
+ top_p=0.8,
48
+ do_sample=True
49
+ )
50
 
51
+ new_tokens = outputs[0][inputs.input_ids.shape[-1]:]
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
52
 
53
+ return tokenizer.decode(
54
+ new_tokens,
55
+ skip_special_tokens=True
56
+ )
57
 
58
 
59
+ demo = gr.ChatInterface(
60
+ fn=respond,
61
+ title="ASL-4B-v1",
62
+ description="Saudi Arabic Language Model"
63
+ )
64
+
65
+ demo.launch()