]> git.djapps.eu Git - pkg/ggml/sources/whisper.cpp/commit
cuda : fix copy of large tensors (ggml_nbytes <= INT_MAX assertion) (llama/18433)
authorMeeMin <redacted>
Thu, 1 Jan 2026 23:24:20 +0000 (04:54 +0530)
committerGeorgi Gerganov <redacted>
Wed, 14 Jan 2026 07:11:59 +0000 (09:11 +0200)
commitfbde389665d2af9c15d1c447da2351abcadc6c63
tree88df1faae8a6dda2253dd19d27f3c138d07624c4
parentf22c1ccbe4c0618ba88ce3d8747066859dcb5972
cuda : fix copy of large tensors (ggml_nbytes <= INT_MAX assertion) (llama/18433)

* ggml-cuda: fixed assertion in ggml_cuda_cpy (llama/18140)

* ggml-cuda: changes in data types to int64_t

* ggml-cuda: added asserts for CUDA block numbers

* ggml-cuda: changed the condition for y and z dimension
ggml/src/ggml-cuda/cpy.cu