File size: 3,297 Bytes
7bd6e67
 
 
 
 
 
 
a7cccd7
7bd6e67
 
 
 
 
 
a6ab0c4
1cc47b4
76a6d8e
79a49f5
205ebd7
1cc47b4
4759f07
e4cd356
4759f07
e4cd356
7b061fd
205ebd7
 
1cc47b4
4759f07
e4cd356
205ebd7
67fdcc7
 
205ebd7
7bd6e67
 
 
 
 
 
7b061fd
7bd6e67
 
 
 
 
 
e8081aa
7bd6e67
 
 
 
 
e4cd356
 
 
 
 
 
 
76a6d8e
7bd6e67
 
 
 
e4cd356
7bd6e67
 
 
 
e4cd356
7bd6e67
 
 
 
 
 
 
 
 
e4cd356
7bd6e67
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
7b061fd
df914b7
7bd6e67
84facec
06dfd60
7bd6e67
 
df914b7
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
113
114
115
116
117
import os
from threading import Thread
from typing import Iterator

import gradio as gr
import spaces
import torch
from transformers import AutoModelForCausalLM, AutoTokenizer, TextIteratorStreamer

MAX_MAX_NEW_TOKENS = 2048
DEFAULT_MAX_NEW_TOKENS = 1024
MAX_INPUT_TOKEN_LENGTH = int(os.getenv("MAX_INPUT_TOKEN_LENGTH", "4096"))
ACCESS_TOKEN = os.getenv("HF_TOKEN", "")

model_id = "mistralai/Mistral-Nemo-Instruct-2407"
#filename = "Mistral-Nemo-Instruct-2407-Q6_K_L.gguf"

model = AutoModelForCausalLM.from_pretrained(
    model_id,
    #gguf_file=filename,
    torch_dtype=torch.float16,
    device_map="auto",
    trust_remote_code=False,
    token=ACCESS_TOKEN)

tokenizer = AutoTokenizer.from_pretrained(
    model_id,
    #gguf_file=filename,
    trust_remote_code=False,
    token=ACCESS_TOKEN)

tokenizer.use_default_system_prompt = False
model.config.gradient_checkpointing = True

@spaces.GPU
def generate(
    message: str,
    system_prompt: str,
    max_new_tokens: int = 1024,
    temperature: float = 0.01,
    top_p: float = 0.7,
) -> Iterator[str]:
    conversation = []
    if system_prompt:
        conversation.append({"role": "system", "content": system_prompt})
    conversation.append({"role": "user", "content": message})

    input_ids = tokenizer.apply_chat_template(conversation, return_tensors="pt")
    if input_ids.shape[1] > MAX_INPUT_TOKEN_LENGTH:
        input_ids = input_ids[:, -MAX_INPUT_TOKEN_LENGTH:]
        gr.Warning(f"Trimmed input from conversation as it was longer than {MAX_INPUT_TOKEN_LENGTH} tokens.")
    input_ids = input_ids.to(model.device)

    '''
    terminators = [
        tokenizer.eos_token_id,
        tokenizer.convert_tokens_to_ids("<|eot_id|>")
    ]
    '''
    
    streamer = TextIteratorStreamer(tokenizer, timeout=600.0, skip_prompt=True, skip_special_tokens=True)
    generate_kwargs = dict(
        {"input_ids": input_ids},
        streamer=streamer,
        max_new_tokens=max_new_tokens,
        #eos_token_id=terminators,
        do_sample=True,
        top_p=top_p,
        temperature=temperature,
        num_beams=1,
        pad_token_id=tokenizer.eos_token_id, 
    )
    t = Thread(target=model.generate, kwargs=generate_kwargs)
    t.start()

    outputs = []
    for text in streamer:
        outputs.append(text)
        yield "".join(outputs)


chat_interface = gr.Interface(
    fn=generate,
    inputs=[
        gr.Textbox(lines=2, placeholder="Prompt", label="Prompt"),
    ],
    outputs="text",
    additional_inputs=[
        gr.Textbox(label="System prompt", lines=6),
        gr.Slider(
            label="Max new tokens",
            minimum=1,
            maximum=MAX_MAX_NEW_TOKENS,
            step=1,
            value=DEFAULT_MAX_NEW_TOKENS,
        ),
        gr.Slider(
            label="Temperature",
            minimum=0.1,
            maximum=4.0,
            step=0.01,
            value=0.01,
        ),
        gr.Slider(
            label="Top-p (nucleus sampling)",
            minimum=0.05,
            maximum=1.0,
            step=0.01,
            value=0.7,
        ),          
    ],
    title="Model testing - mistralai/Mistral-Nemo-Instruct-2407",
    description="Provide system settings and a prompt to interact with the model.",
)

chat_interface.queue(max_size=20).launch()