File size: 2,798 Bytes
21b8ce0
d8a3c53
e2f7d5c
 
 
d8a3c53
 
 
 
e2f7d5c
d8a3c53
21b8ce0
d8a3c53
 
 
 
 
 
 
 
fa52871
d8a3c53
fa52871
 
 
 
 
 
 
 
 
 
 
d8a3c53
fa52871
 
 
 
 
 
 
 
 
d8a3c53
fa52871
 
 
 
 
 
 
 
 
 
 
 
d8a3c53
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
6d5f110
c83546c
6d5f110
d8a3c53
 
 
 
 
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
import spaces
import gradio as gr
import torch
from transformers import AutoModelForCausalLM, AutoTokenizer, TextIteratorStreamer
from threading import Thread

"""
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
"""
# client = InferenceClient("cognitivecomputations/dolphin-2.8-mistral-7b-v02")

@spaces.GPU
def respond(
    message,
    history: list[tuple[str, str]],
    system_message,
    max_tokens,
    temperature,
    top_p,
):
    torch.set_default_device("cuda")

    tokenizer = AutoTokenizer.from_pretrained(
        "cognitivecomputations/dolphin-2.8-mistral-7b-v02",
        trust_remote_code=True
    )
    model = AutoModelForCausalLM.from_pretrained(
        "cognitivecomputations/dolphin-2.8-mistral-7b-v02",
        torch_dtype="auto",
        load_in_4bit=True,
        trust_remote_code=True
    )
    history_transformer_format = history + [[message, ""]]

    system_prompt = f"<|im_start|>system\n{system_message}.<|im_end|>"
    messages = system_prompt + "".join(["".join(["\n<|im_start|>user\n" + item[0], "<|im_end|>\n<|im_start|>assistant\n" + item[1]]) for item in history_transformer_format])
    input_ids = tokenizer([messages], return_tensors="pt").to('cuda')
    streamer = TextIteratorStreamer(tokenizer, skip_prompt=True, skip_special_tokens=True)
    generate_kwargs = dict(
        input_ids,
        streamer=streamer,
        max_new_tokens=max_tokens,
        do_sample=True,
        top_p=top_p,
        top_k=50,
        temperature=temperature,
        num_beams=1
    )
    t = Thread(target=model.generate, kwargs=generate_kwargs)
    t.start()
    partial_message = ""
    for new_token in streamer:
        partial_message += new_token
        if '<|im_end|>' in partial_message:
            break
        yield partial_message

"""
For information on how to customize the ChatInterface, peruse the gradio docs: https://www.gradio.app/docs/chatinterface
"""
demo = gr.ChatInterface(
    respond,
    additional_inputs=[
        gr.Textbox(value="You are a friendly Chatbot.", label="System message"),
        gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"),
        gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
        gr.Slider(
            minimum=0.1,
            maximum=1.0,
            value=0.95,
            step=0.05,
            label="Top-p (nucleus sampling)",
        ),
    ],
    theme=gr.themes.Soft(primary_hue="green", secondary_hue="indigo", neutral_hue="zinc",font=[gr.themes.GoogleFont("Exo 2"), "ui-sans-serif", "system-ui", "sans-serif"]).set(
        block_background_fill_dark="*neutral_800"
    )
)


if __name__ == "__main__":
    demo.launch()