Upload app.py
Browse files
app.py
CHANGED
@@ -23,38 +23,16 @@ import gradio as gr
|
|
23 |
import os
|
24 |
import glob
|
25 |
import json
|
26 |
-
from setuptools import setup, find_packages
|
27 |
-
|
28 |
-
_PATH_ROOT = os.path.dirname(__file__)
|
29 |
-
|
30 |
-
with open(os.path.join(_PATH_ROOT, "README.md"), encoding="utf-8") as fo:
|
31 |
-
readme = fo.read()
|
32 |
-
|
33 |
-
setup(
|
34 |
-
name='lit-llama',
|
35 |
-
version='0.1.0',
|
36 |
-
description='Implementation of the LLaMA language model',
|
37 |
-
author='Lightning AI',
|
38 |
-
url='https://github.com/lightning-AI/lit-llama',
|
39 |
-
install_requires=[
|
40 |
-
"torch>=2.0.0",
|
41 |
-
"lightning @ git+https://github.com/Lightning-AI/lightning@master",
|
42 |
-
"sentencepiece",
|
43 |
-
"bitsandbytes",
|
44 |
-
],
|
45 |
-
packages=find_packages(),
|
46 |
-
long_description=readme,
|
47 |
-
long_description_content_type="text/markdown",
|
48 |
-
)
|
49 |
|
50 |
|
51 |
# os.environ['CUDA_LAUNCH_BLOCKING'] = '1'
|
52 |
torch.set_float32_matmul_precision("high")
|
|
|
53 |
|
54 |
def model_load(
|
55 |
adapter_path: Path = Path("out/adapter/alpaca/lit-llama-adapter-finetuned_15k.pth"),
|
56 |
pretrained_path: Path = Path("checkpoints/lit-llama/7B/lit-llama.pth"),
|
57 |
-
quantize: Optional[str] = "
|
58 |
):
|
59 |
|
60 |
fabric = L.Fabric(devices=1)
|
|
|
23 |
import os
|
24 |
import glob
|
25 |
import json
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
26 |
|
27 |
|
28 |
# os.environ['CUDA_LAUNCH_BLOCKING'] = '1'
|
29 |
torch.set_float32_matmul_precision("high")
|
30 |
+
# quantize: Optional[str] = "llm.int8",
|
31 |
|
32 |
def model_load(
|
33 |
adapter_path: Path = Path("out/adapter/alpaca/lit-llama-adapter-finetuned_15k.pth"),
|
34 |
pretrained_path: Path = Path("checkpoints/lit-llama/7B/lit-llama.pth"),
|
35 |
+
quantize: Optional[str] = "",
|
36 |
):
|
37 |
|
38 |
fabric = L.Fabric(devices=1)
|