From 988126373cbff55d14c1f7d85f74ffb108943ca7 Mon Sep 17 00:00:00 2001 From: asdfvg123 Date: Fri, 31 Jul 2026 19:12:09 +0000 Subject: [PATCH] revert test --- README.rst | 6 +++--- .../common/gemm/kittens/cdna4/blockwise_fp8_gemm.cpp | 4 ++-- 2 files changed, 5 insertions(+), 5 deletions(-) diff --git a/README.rst b/README.rst index a4cd254b3..7c61fb584 100644 --- a/README.rst +++ b/README.rst @@ -291,10 +291,10 @@ The backend shares the existing USE_HIPKITTENS_GEMM build option with MXFP8: * OFF - disable the HipKittens blockwise FP8 and MXFP8 GEMM backend; * ON - enable the HipKittens GEMM backend (default). -On gfx950 there are two blockwise FP8 GEMM kernels, selected at runtime by the NVTE_BLOCKWISE_FP8_POWER_OF_2_SCALE environment variable: +On gfx950 there are two blockwise FP8 GEMM kernels, selected at runtime by the NVTE_FP8_BLOCK_SCALING_FP32_SCALES environment variable, which also controls the scale produced by the Float8BlockScaling recipe: -* 0 - compute the GEMM with the FP32 scales directly; -* 1 - cast the incoming FP32 scales to E8M0 (power-of-2) and compute the GEMM with those scales (default). +* 0 - cast the incoming FP32 scales to E8M0 (power-of-2) and compute the GEMM with those scales (default); +* 1 - compute the GEMM with the FP32 scales directly. Two-stage amax Kernel ^^^^^^^^^^^^^^^^^^^^^ diff --git a/transformer_engine/common/gemm/kittens/cdna4/blockwise_fp8_gemm.cpp b/transformer_engine/common/gemm/kittens/cdna4/blockwise_fp8_gemm.cpp index 0c217cecc..430d7c892 100644 --- a/transformer_engine/common/gemm/kittens/cdna4/blockwise_fp8_gemm.cpp +++ b/transformer_engine/common/gemm/kittens/cdna4/blockwise_fp8_gemm.cpp @@ -1110,8 +1110,8 @@ class BlockwiseGemmCdna4 final : public BlockwiseGemmBackend { float *sb = reinterpret_cast(const_cast(ksb)); static const bool use_pow2 = []() { - const char *e = std::getenv("NVTE_BLOCKWISE_FP8_POWER_OF_2_SCALE"); - return e == nullptr || std::strcmp(e, "0") != 0; + const char *e = std::getenv("NVTE_FP8_BLOCK_SCALING_FP32_SCALES"); + return e == nullptr || std::strcmp(e, "1") != 0; }(); const int k_iters = K / BLOCK_K;