From a78d050eff733c2c85d7b36d464838bf282e496e Mon Sep 17 00:00:00 2001 From: Warren Deng <35615489+warrendeng@users.noreply.github.com> Date: Tue, 13 Oct 2026 21:28:07 -0700 Subject: [PATCH] [Bugfix] Bump tokenspeed-mla to 2.0.2 to fix nanobind conflict with Triton 3.9 (#61125) Signed-off-by: Warren Deng Co-authored-by: Claude Co-authored-by: Codex --- requirements/cuda.txt | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/requirements/cuda.txt b/requirements/cuda.txt index 1f5a9642b2..5e193bac5a 100644 --- a/requirements/cuda.txt +++ b/requirements/cuda.txt @@ -25,7 +25,7 @@ nvidia-cutlass-dsl[cu13]==4.8.0 quack-kernels==0.6.5 # Required by MSA # Tokenspeed_MLA for faster mla with spec decode -tokenspeed-mla==0.2.15; platform_system == "Linux" +tokenspeed-mla==2.0.2; platform_system == "Linux" # Humming kernels for quantization gemm humming-kernels[cu13]==0.1.16