Update app.py
Browse files
app.py
CHANGED
|
@@ -1,19 +1,32 @@
|
|
| 1 |
import gradio as gr
|
| 2 |
-
from transformers import AutoModelForCausalLM, AutoTokenizer
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 3 |
|
| 4 |
# Загрузка модели и токенизатора
|
| 5 |
model_name = "IlyaGusev/saiga_yandexgpt_8b"
|
| 6 |
tokenizer = AutoTokenizer.from_pretrained(model_name)
|
| 7 |
-
model = AutoModelForCausalLM.from_pretrained(
|
|
|
|
|
|
|
|
|
|
|
|
|
| 8 |
|
| 9 |
# Функция для генерации текста
|
| 10 |
def generate_text(input_text):
|
| 11 |
-
inputs = tokenizer(input_text, return_tensors="pt")
|
| 12 |
outputs = model.generate(
|
| 13 |
**inputs,
|
| 14 |
-
max_new_tokens=300,
|
| 15 |
-
do_sample=True,
|
| 16 |
-
temperature=0.7
|
| 17 |
)
|
| 18 |
return tokenizer.decode(outputs[0], skip_special_tokens=True)
|
| 19 |
|
|
@@ -22,8 +35,8 @@ interface = gr.Interface(
|
|
| 22 |
fn=generate_text,
|
| 23 |
inputs=gr.Textbox(lines=2, placeholder="Введите ваш запрос..."),
|
| 24 |
outputs="text",
|
| 25 |
-
title="Saiga YandexGPT 8B Demo",
|
| 26 |
-
description="Задайте вопрос модели Saiga YandexGPT 8B
|
| 27 |
)
|
| 28 |
|
| 29 |
# Запуск приложения
|
|
|
|
| 1 |
import gradio as gr
|
| 2 |
+
from transformers import AutoModelForCausalLM, AutoTokenizer, BitsAndBytesConfig
|
| 3 |
+
import torch
|
| 4 |
+
|
| 5 |
+
# Настройка 4-битной квантизации
|
| 6 |
+
quantization_config = BitsAndBytesConfig(
|
| 7 |
+
load_in_4bit=True,
|
| 8 |
+
bnb_4bit_compute_dtype=torch.float16, #
|
| 9 |
+
bnb_4bit_quant_type="nf4",
|
| 10 |
+
bnb_4bit_use_double_quant=True
|
| 11 |
+
)
|
| 12 |
|
| 13 |
# Загрузка модели и токенизатора
|
| 14 |
model_name = "IlyaGusev/saiga_yandexgpt_8b"
|
| 15 |
tokenizer = AutoTokenizer.from_pretrained(model_name)
|
| 16 |
+
model = AutoModelForCausalLM.from_pretrained(
|
| 17 |
+
model_name,
|
| 18 |
+
quantization_config=quantization_config,
|
| 19 |
+
device_map="auto"
|
| 20 |
+
)
|
| 21 |
|
| 22 |
# Функция для генерации текста
|
| 23 |
def generate_text(input_text):
|
| 24 |
+
inputs = tokenizer(input_text, return_tensors="pt") # Перенос на GPU
|
| 25 |
outputs = model.generate(
|
| 26 |
**inputs,
|
| 27 |
+
max_new_tokens=300,
|
| 28 |
+
do_sample=True,
|
| 29 |
+
temperature=0.7
|
| 30 |
)
|
| 31 |
return tokenizer.decode(outputs[0], skip_special_tokens=True)
|
| 32 |
|
|
|
|
| 35 |
fn=generate_text,
|
| 36 |
inputs=gr.Textbox(lines=2, placeholder="Введите ваш запрос..."),
|
| 37 |
outputs="text",
|
| 38 |
+
title="Saiga YandexGPT 8B Demo (4-bit)",
|
| 39 |
+
description="Задайте вопрос модели Saiga YandexGPT 8B в 4-битной квантизации!"
|
| 40 |
)
|
| 41 |
|
| 42 |
# Запуск приложения
|