]> git.djapps.eu Git - pkg/ggml/sources/llama.cpp/commit
CUDA: add softmax broadcast (#14475)
authorAman Gupta <redacted>
Wed, 2 Jul 2025 12:34:24 +0000 (20:34 +0800)
committerGeorgi Gerganov <redacted>
Wed, 2 Jul 2025 12:48:33 +0000 (15:48 +0300)
commit55a1c5a5fdefef808e95aabd3d5563af1068cc80
treea4afc679412d3a2bc8715e856405d00bffd83c12
parent12a81af45f0dbbab24bd819a15f57c03ceb1be90
CUDA: add softmax broadcast (#14475)

* CUDA: add softmax broadcast

* Pass by const ref

* Review: Use blockDims for indexing, remove designated initializers

* Add TODO for noncontigous input/output
ggml/src/ggml-cuda/ggml-cuda.cu
ggml/src/ggml-cuda/softmax.cu