main

GGitHubRevert "[KERNELS] Improve block sizes for batched matmul_ogs with small m/n/k. (#7897)" (#7971)

0a2e3a39创建于 2025年8月26日历史提交

文件	最后提交记录	最后更新时间
test_tensor_details	[TRITON_KERNELS] fix swizzling numerics (#7632)	10 个月前
__init__.py	[BENCH] custom matmul_ogs epilogue + triton_bench -> triton_kernels (#6749)	1 年前
conftest.py	[TEST] Reset jit_cache_hook and jit_cache_hook in the fresh_triton_cache fixture (#7885)	9 个月前
test_compaction.py	[BENCH] custom matmul_ogs epilogue + triton_bench -> triton_kernels (#6749)	1 年前
test_matmul.py	Revert "[KERNELS] Improve block sizes for batched matmul_ogs with small m/n/k. (#7897)" (#7971)	9 个月前
test_mxfp.py	[mxfp] support quant/dequant from/to fp32 (#7672)	9 个月前
test_routing.py	[BENCH] Routing improvements (#7369)	11 个月前
test_specialize.py	[TEST] Reset jit_cache_hook and jit_cache_hook in the fresh_triton_cache fixture (#7885)	9 个月前
test_swiglu.py	[BENCH] custom matmul_ogs epilogue + triton_bench -> triton_kernels (#6749)	1 年前
test_tensor.py	[KERNELS] improved twiddling/swizzling for H100 simulated mxfp4 (h/t @lezcano) (#7587)	10 个月前