File size: 4,012 Bytes
f1fef64
d9faa8c
 
317e409
d9faa8c
 
 
 
 
5146aa4
d9faa8c
 
317e409
d9faa8c
 
 
 
 
 
539566d
f70fc29
 
 
d9faa8c
 
 
 
 
 
 
 
 
 
 
 
 
 
9122113
d9faa8c
 
9122113
d9faa8c
 
 
 
f70fc29
 
d9faa8c
f70fc29
 
 
 
 
 
 
 
 
 
c2dfdca
5146aa4
 
f70fc29
 
 
 
c2dfdca
 
381d2e1
f70fc29
d9faa8c
f70fc29
c2dfdca
a26f5ee
f70fc29
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
d9faa8c
f70fc29
 
 
 
d9faa8c
317e409
 
5146aa4
 
 
d9faa8c
f1fef64
5146aa4
 
 
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
import gradio as gr
import spaces
from functools import lru_cache

# Cache model loading to optimize performance
@lru_cache(maxsize=3)
def load_hf_model(model_name):
    return gr.load(
        name=f"deepseek-ai/{model_name}",
        src="huggingface",
        api_name="/chat"
    )

# Load all models at startup
MODELS = {
    "DeepSeek-R1-Distill-Qwen-32B": load_hf_model("DeepSeek-R1-Distill-Qwen-32B"),
    "DeepSeek-R1": load_hf_model("DeepSeek-R1"),
    "DeepSeek-R1-Zero": load_hf_model("DeepSeek-R1-Zero")
}

# --- Chatbot function ---
def chatbot(input_text, history, model_choice, system_message, max_new_tokens, temperature, top_p):
    history = history or []
    
    # Get the selected model component
    model_component = MODELS[model_choice]
    
    # Create payload for the model
    payload = {
        "messages": [{"role": "user", "content": input_text}],
        "system": system_message,
        "max_tokens": max_new_tokens,
        "temperature": temperature,
        "top_p": top_p
    }
    
    # Run inference using the selected model
    try:
        response = model_component(payload)
        assistant_response = response[-1]["content"]
    except Exception as e:
        assistant_response = f"Error: {str(e)}"
    
    history.append((input_text, assistant_response))
    return history, history, ""

# --- Gradio Interface ---
with gr.Blocks(theme=gr.themes.Soft(), title="DeepSeek Chatbot") as demo:
    gr.Markdown(
        """
        # DeepSeek Chatbot
        Created by [ruslanmv.com](https://ruslanmv.com/)
        This is a demo of different DeepSeek models. Select a model, type your message, and click "Submit".
        You can also adjust optional parameters like system message, max new tokens, temperature, and top-p.
        """
    )

    with gr.Row():
        with gr.Column():
            # Specify type='messages' to avoid deprecation warning
            chatbot_output = gr.Chatbot(label="DeepSeek Chatbot", height=500, type="messages")
            msg = gr.Textbox(label="Your Message", placeholder="Type your message here...")
            with gr.Row():
                submit_btn = gr.Button("Submit", variant="primary")
                clear_btn = gr.ClearButton([msg, chatbot_output])

    with gr.Row():
        with gr.Accordion("Options", open=True):
            model_choice = gr.Radio(
                choices=list(MODELS.keys()),
                label="Choose a Model",
                value="DeepSeek-R1"
            )
            with gr.Accordion("Optional Parameters", open=False):
                system_message = gr.Textbox(
                    label="System Message",
                    value="You are a friendly Chatbot created by ruslanmv.com",
                    lines=2,
                )
                max_new_tokens = gr.Slider(
                    minimum=1, maximum=4000, value=200, label="Max New Tokens"
                )
                temperature = gr.Slider(
                    minimum=0.10, maximum=4.00, value=0.70, label="Temperature"
                )
                top_p = gr.Slider(
                    minimum=0.10, maximum=1.00, value=0.90, label="Top-p (nucleus sampling)"
                )

    chat_history = gr.State([])

    # Event handling
    submit_btn.click(
        chatbot,
        [msg, chat_history, model_choice, system_message, max_new_tokens, temperature, top_p],
        [chatbot_output, chat_history, msg]
    )
    msg.submit(
        chatbot,
        [msg, chat_history, model_choice, system_message, max_new_tokens, temperature, top_p],
        [chatbot_output, chat_history, msg]
    )

# Remove (or replace) the old `demo.fn = spaces.GPU()(demo.fn)` line
# If you need GPU acceleration in Spaces, wrap demo.launch() with spaces.GPU()
# as shown below. Otherwise, remove the spaces.GPU() call.

if __name__ == "__main__":
    # Launch the Gradio app with GPU from Hugging Face Spaces
    # If you DON'T need GPU from `spaces`, just do: demo.launch()
    spaces.GPU()(demo.launch)()