File size: 1,190 Bytes
b166270
aa8cda8
53fc999
aa8cda8
 
b166270
 
b7cbc3b
53fc999
b166270
aa8cda8
 
 
b166270
 
 
 
 
 
aa8cda8
 
 
 
 
53fc999
 
b7cbc3b
 
53fc999
 
b7cbc3b
b166270
 
 
 
 
 
 
 
 
01c7904
b166270
b7cbc3b
b166270
 
 
 
 
 
 
b7cbc3b
b166270
 
 
 
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
import spaces

import torch
import torch._dynamo

import gradio as gr

from peft import AutoPeftModelForCausalLM
from transformers import AutoTokenizer, BitsAndBytesConfig

torch._dynamo.config.suppress_errors = True
torch._dynamo.disable()

max_seq_length = 2048
dtype = (
    None
)
load_in_4bit = True


quantization_config = BitsAndBytesConfig(
    load_in_4bit=True,
    # bnb_4bit_compute_dtype=torch.float16,
)

tokenizer = AutoTokenizer.from_pretrained("ua-l/gemma-2-9b-legal-uk")
model = AutoPeftModelForCausalLM.from_pretrained(
    "ua-l/gemma-2-9b-legal-uk",
    quantization_config=quantization_config,
    device_map='auto'
)


@spaces.GPU
def predict(question):
    inputs = tokenizer(
    [f'''### Question:
    {question}
    
    ### Answer:
'''], return_tensors = "pt").to("cuda")

    outputs = model.generate(**inputs, max_new_tokens = 128)
    
    results = tokenizer.batch_decode(outputs, skip_special_tokens=True)

    return results[0]

inputs = gr.Textbox(lines=2, label="Enter a question", value="Як отримати виплати ВПО?")

outputs = gr.Textbox(label="Answer")

demo = gr.Interface(fn=predict, inputs=inputs, outputs=outputs)
demo.launch()