AutoAWQ Kernels implements the AWQ kernels.
Project description
AutoAWQ Kernels
AutoAWQ Kernels is a new package that is split up from the main repository in order to avoid compilation times.
Requirements
-
Windows: Must use WSL2.
-
NVIDIA:
- GPU: Must be compute capability 7.5 or higher.
- CUDA Toolkit: Must be 11.8 or higher.
-
AMD:
- ROCm: Must be 5.6 or higher. Build from source
Install
Install from PyPi
The package is available on PyPi with CUDA 12.4.1 wheels:
pip install autoawq-kernels
Build from source
To build the kernels from source, you first need to setup an environment containing the necessary dependencies.
Build Requirements
- Python>=3.8.0
- Numpy
- Wheel
- PyTorch
- ROCm: You need to install the following packages
rocsparse-dev hipsparse-dev rocthrust-dev rocblas-dev hipblas-dev
.
Building process
pip install git+https://github.com/casper-hansen/AutoAWQ_kernels.git
Notes on environment variables:
TORCH_VERSION
: By default, we build using the current version of torch bytorch.__version__
. You can override it withTORCH_VERSION
.CUDA_VERSION
orROCM_VERSION
can also be used to build for a specific version of CUDA or ROCm.
CC
andCXX
: You can specify which build system to use for the C code, e.g.CC=g++-13 CXX=g++-13 pip install -e .
COMPUTE_CAPABILITIES
: You can specify specific compute capabilities to compile for:COMPUTE_CAPABILITIES="75,80,86,87,89,90" pip install -e .
Project details
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distributions
No source distribution files available for this release.See tutorial on generating distribution archives.
Built Distributions
Close
Hashes for autoawq_kernels-0.0.8-cp312-cp312-win_amd64.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | 11ae0d441ad30bc75099cbe7bc79d21b7969c73ecedfcfde1c0e691caf434cf5 |
|
MD5 | 49f6d52dc984cca65b3d7626a205758c |
|
BLAKE2b-256 | 95e8808214b8589f5629f033db9e22a55fd7813e9d9a7afce0e417c2a941df3b |
Close
Hashes for autoawq_kernels-0.0.8-cp312-cp312-manylinux2014_x86_64.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | 091e99424af55c1992eb460098051542d8611523eb14700d65bfbf19abcf2259 |
|
MD5 | e16fd5d6f0ca511008c6733389f1aa6a |
|
BLAKE2b-256 | d0ed6b22d32aa59388df713382128c3982b9147b6641c1037e2f2850c9f00c9c |
Close
Hashes for autoawq_kernels-0.0.8-cp311-cp311-win_amd64.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | be76051f63d08e52c2e0233a152d9b9c63eec9c53f657f264fd71ad574009266 |
|
MD5 | 85dbfc507757606a15d6f9d66e891ac8 |
|
BLAKE2b-256 | e2b9da156719223b289b83b2ce03f269a7612219e1d2dec00d56e318fe26f280 |
Close
Hashes for autoawq_kernels-0.0.8-cp311-cp311-manylinux2014_x86_64.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | 74800947b5f3c0c924c4c7c572ba2143d7436687b85603c160b4a3cdfc54b2ed |
|
MD5 | 2cc9bf767ac3cc271972dbc3d3485aa4 |
|
BLAKE2b-256 | 003fc52d4e657a1db6e3e77911575252c501658197ed41892cc3f0380610db90 |
Close
Hashes for autoawq_kernels-0.0.8-cp310-cp310-win_amd64.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | 8b059988a613071dc36ff7f6fdff867ca2928ac8b034c11c5261b28acc4ebf5d |
|
MD5 | e5f4ac72d0a3da35639b05841164551f |
|
BLAKE2b-256 | 7eaefa51a1945bc341242925983b02f14fb7b7077382a6105d1839620cf7829f |
Close
Hashes for autoawq_kernels-0.0.8-cp310-cp310-manylinux2014_x86_64.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | cbbe052e0397074724a3ebb352da0eed52ce4322f4034ed1b54c2ea71835bf2b |
|
MD5 | dd6e8e43d0de1030997ab026b8a81e0d |
|
BLAKE2b-256 | 025595285ad43c658422161574ea8563212f49412438033e4d8f834d6d7948dd |
Close
Hashes for autoawq_kernels-0.0.8-cp39-cp39-win_amd64.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | 343791eed626e0152ff5fc7d4a9f5e176a5f1d879fa0c663e3cdaa8fd700a1c1 |
|
MD5 | 55b3aabdead600b53d9e7bb639d9d0a3 |
|
BLAKE2b-256 | 264a0496c40c2c13a3034e034e1d97c9cb8a1c7e115d4793fedc199dcfca17f6 |
Close
Hashes for autoawq_kernels-0.0.8-cp39-cp39-manylinux2014_x86_64.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | c489d46ac962ce0c72b64fed1c6649afd48b43c331c3905747582a75ae0dcbd1 |
|
MD5 | 78bce45610e860b88bd382f6ae1107a2 |
|
BLAKE2b-256 | f578531ce17f26deac7e6a15a768b292be7695ae0a0caedabd6ac4e05956562c |
Close
Hashes for autoawq_kernels-0.0.8-cp38-cp38-win_amd64.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | c9ddf5f934f55a1d1084ab7b2b8aba6bea5c4cdb90e4ec24fc95b799ee0c2927 |
|
MD5 | 10da47263ef2d1d8e3ee90933e88d4d1 |
|
BLAKE2b-256 | 448669523ae0d1c756828a65056af6f40f77cdd893fc5e4e322895865e5c05b5 |
Close
Hashes for autoawq_kernels-0.0.8-cp38-cp38-manylinux2014_x86_64.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | ed82d6a9f0c7570b21f42efaaf940f64efa40f352981218f66c4c75376dcfec9 |
|
MD5 | 098352e9fa7446bb01842a59f8c2557d |
|
BLAKE2b-256 | 976c66f974023259826423f3a1cfd01832d6de0533a74cfa2580827777e5f2da |