]> git.djapps.eu Git - pkg/ggml/sources/whisper.cpp/commit
ggml-cuda: Allow transpose-free gemmv computation (llama/26171)
authorRobert Esclapez <redacted>
Thu, 30 Jul 2026 13:39:46 +0000 (15:39 +0200)
committerGeorgi Gerganov <redacted>
Tue, 4 Aug 2026 10:37:47 +0000 (13:37 +0300)
commit1a8c5a62f583f58b3ab6ff0f96730e1369c97527
treeb5c43cb9b306c950d8d998e41402c553e6ffc50e
parent64d57d3df5c8dacee098577257edcaa154bf5ef3
ggml-cuda: Allow transpose-free gemmv computation (llama/26171)

When matrix's weights are shaped 1xK is leverage a transpose-free
computation to use mat_mul_vec_f.
ggml/src/ggml-cuda/ggml-cuda.cu