Spaces:
Running
on
Zero
Running
on
Zero
Not use 8 bit quantization
Browse files
app.py
CHANGED
|
@@ -5,12 +5,7 @@ from typing import Iterator
|
|
| 5 |
import gradio as gr
|
| 6 |
import spaces
|
| 7 |
import torch
|
| 8 |
-
from transformers import
|
| 9 |
-
AutoModelForCausalLM,
|
| 10 |
-
BitsAndBytesConfig,
|
| 11 |
-
GemmaTokenizerFast,
|
| 12 |
-
TextIteratorStreamer,
|
| 13 |
-
)
|
| 14 |
|
| 15 |
DESCRIPTION = """\
|
| 16 |
# Gemma 2 2B IT
|
|
@@ -34,7 +29,6 @@ model = AutoModelForCausalLM.from_pretrained(
|
|
| 34 |
model_id,
|
| 35 |
device_map="auto",
|
| 36 |
torch_dtype=torch.bfloat16,
|
| 37 |
-
quantization_config=BitsAndBytesConfig(load_in_8bit=True),
|
| 38 |
)
|
| 39 |
model.config.sliding_window = 4096
|
| 40 |
model.eval()
|
|
|
|
| 5 |
import gradio as gr
|
| 6 |
import spaces
|
| 7 |
import torch
|
| 8 |
+
from transformers import AutoModelForCausalLM, GemmaTokenizerFast, TextIteratorStreamer
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 9 |
|
| 10 |
DESCRIPTION = """\
|
| 11 |
# Gemma 2 2B IT
|
|
|
|
| 29 |
model_id,
|
| 30 |
device_map="auto",
|
| 31 |
torch_dtype=torch.bfloat16,
|
|
|
|
| 32 |
)
|
| 33 |
model.config.sliding_window = 4096
|
| 34 |
model.eval()
|