This pull request primarily involves changes to the bitblas Python package, specifically to the gemv.py, gemv_dequantize.py, general.py, and __init__.py files. The changes include version updates, modifications to function conditions to handle additional cases, and adjustments to test parameters.
Version updates:
VERSION and python/bitblas/__init__.py: The version number has been updated from 0.0.1.dev8 to 0.0.1.dev9. [1][2]
Function condition modifications:
python/bitblas/gpu/gemv.py: The condition in the apply_config function has been modified to accept a length of 4 for block_info.iters, along with an explanatory comment.
python/bitblas/gpu/gemv_dequantize.py: The get_vectorize_factor function has been modified to handle cases where the length of sch.get_loops(block_b) is 4, including binding operations and an assertion for the SplitK case. [1][2]
Function call modifications:
python/bitblas/wrapper/general.py: The legalize_c function has been modified to include a conditional return statement before the kernel launch string. [1][2]
This pull request primarily involves changes to the
bitblas
Python package, specifically to thegemv.py
,gemv_dequantize.py
,general.py
, and__init__.py
files. The changes include version updates, modifications to function conditions to handle additional cases, and adjustments to test parameters.Version updates:
VERSION
andpython/bitblas/__init__.py
: The version number has been updated from0.0.1.dev8
to0.0.1.dev9
. [1] [2]Function condition modifications:
python/bitblas/gpu/gemv.py
: The condition in theapply_config
function has been modified to accept a length of 4 forblock_info.iters
, along with an explanatory comment.python/bitblas/gpu/gemv_dequantize.py
: Theget_vectorize_factor
function has been modified to handle cases where the length ofsch.get_loops(block_b)
is 4, including binding operations and an assertion for theSplitK
case. [1] [2]Function call modifications:
python/bitblas/wrapper/general.py
: Thelegalize_c
function has been modified to include a conditional return statement before the kernel launch string. [1] [2]Test parameter modifications:
testing/python/operators/test_general_matmul_splitk_ops.py
: Thetest_matmul_torch_forward_fp8e4m3
function has been modified to use a list[1, 16]
as the value ofM
inmatmul_config
.