Hugging Face's logo

PetraAI
/
Nashmi

Transformers
4 languages
llama
chemistry
biology
finance
legal
music
code
art
climate
medical
text-generation-inference
Inference Endpoints
Model card Files Community
Nashmi
File size: 4,606 Bytes 
8eddc47
 
 
 
ec0f90b
 
8eddc47
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
ec0f90b
8eddc47
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
ec0f90b
 
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
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
 
import os
import sys
from pathlib import Path
from setuptools import setup, find_packages


common_setup_kwargs = {
    "version": "0.4.1",
    "name": "auto_gptq",
    "author": "PanQiWei",
    "description": "An easy-to-use LLMs quantization package with user-friendly apis, based on GPTQ algorithm.",
    "long_description": (Path(__file__).parent / "README.md").read_text(encoding="UTF-8"),
    "long_description_content_type": "text/markdown",
    "url": "https://github.com/PanQiWei/AutoGPTQ",
    "keywords": ["gptq", "quantization", "large-language-models", "transformers"],
    "platforms": ["windows", "linux"],
    "classifiers": [
        "Environment :: GPU :: NVIDIA CUDA :: 11.7",
        "Environment :: GPU :: NVIDIA CUDA :: 11.8",
        "Environment :: GPU :: NVIDIA CUDA :: 12.0",
        "License :: OSI Approved :: MIT License",
        "Natural Language :: Chinese (Simplified)",
        "Natural Language :: English",
        "Programming Language :: Python :: 3.8",
        "Programming Language :: Python :: 3.9",
        "Programming Language :: Python :: 3.10",
        "Programming Language :: Python :: 3.11",
        "Programming Language :: C++",
    ]
}


BUILD_CUDA_EXT = int(os.environ.get('BUILD_CUDA_EXT', '1')) == 1
if BUILD_CUDA_EXT:
    try:
        import torch
    except:
        print("Building cuda extension requires PyTorch(>=1.13.0) been installed, please install PyTorch first!")
        sys.exit(-1)

    CUDA_VERSION = None
    ROCM_VERSION = os.environ.get('ROCM_VERSION', None)
    if ROCM_VERSION and not torch.version.hip:
        print(
            f"Trying to compile auto-gptq for RoCm, but PyTorch {torch.__version__} "
            "is installed without RoCm support."
        )
        sys.exit(-1)

    if not ROCM_VERSION:
        default_cuda_version = torch.version.cuda
        CUDA_VERSION = "".join(os.environ.get("CUDA_VERSION", default_cuda_version).split("."))

    if ROCM_VERSION:
        common_setup_kwargs['version'] += f"+rocm{ROCM_VERSION}"
    else:
        if not CUDA_VERSION:
            print(
                f"Trying to compile auto-gptq for CUDA, byt Pytorch {torch.__version__} "
                "is installed without CUDA support."
            )
            sys.exit(-1)
        common_setup_kwargs['version'] += f"+cu{CUDA_VERSION}"


requirements = [
    "accelerate>=0.19.0",
    "datasets",
    "numpy",
    "rouge",
    "torch>=1.13.0",
    "safetensors",
    "transformers>=4.31.0",
    "peft"
]

extras_require = {
    "triton": ["triton==2.0.0"],
    "test": ["parameterized"]
}

include_dirs = ["autogptq_cuda"]

additional_setup_kwargs = dict()
if BUILD_CUDA_EXT:
    from torch.utils import cpp_extension

    if not ROCM_VERSION:
        from distutils.sysconfig import get_python_lib
        conda_cuda_include_dir = os.path.join(get_python_lib(), "nvidia/cuda_runtime/include")

        print("conda_cuda_include_dir", conda_cuda_include_dir)
        if os.path.isdir(conda_cuda_include_dir):
            include_dirs.append(conda_cuda_include_dir)
            print(f"appending conda cuda include dir {conda_cuda_include_dir}")
    extensions = [
        cpp_extension.CUDAExtension(
            "autogptq_cuda_64",
            [
                "autogptq_cuda/autogptq_cuda_64.cpp",
                "autogptq_cuda/autogptq_cuda_kernel_64.cu"
            ]
        ),
        cpp_extension.CUDAExtension(
            "autogptq_cuda_256",
            [
                "autogptq_cuda/autogptq_cuda_256.cpp",
                "autogptq_cuda/autogptq_cuda_kernel_256.cu"
            ]
        )
    ]

    if os.environ.get("INCLUDE_EXLLAMA_KERNELS", "1") == "1":  # TODO: improve github action to always compile exllama_kernels
        extensions.append(
            cpp_extension.CUDAExtension(
                "exllama_kernels",
                [
                    "autogptq_cuda/exllama/exllama_ext.cpp",
                    "autogptq_cuda/exllama/cuda_buffers.cu",
                    "autogptq_cuda/exllama/cuda_func/column_remap.cu",
                    "autogptq_cuda/exllama/cuda_func/q4_matmul.cu",
                    "autogptq_cuda/exllama/cuda_func/q4_matrix.cu"
                ]
            )
        )

    additional_setup_kwargs = {
        "ext_modules": extensions,
        "cmdclass": {'build_ext': cpp_extension.BuildExtension}
    }
common_setup_kwargs.update(additional_setup_kwargs)
setup(
    packages=find_packages(),
    install_requires=requirements,
    extras_require=extras_require,
    include_dirs=include_dirs,
    python_requires=">=3.8.0",
    **common_setup_kwargs
)