diff options
| author | Kawrakow <iwankawrakow@gmail.com> | 2025-03-19 13:03:59 +0100 |
|---|---|---|
| committer | GitHub <noreply@github.com> | 2025-03-19 13:03:59 +0100 |
| commit | c3b75c531c5d5e295d5c3ba846bb37def72df3a7 (patch) | |
| tree | 2890e03606734cca836dd2ecd9e7c88714dba75f /ggml/src | |
| parent | 8e549b42346f039be01a06495c6469d66d1c8926 (diff) | |
Prevent FlashMLA-1 from running on CUDA (#268)
as it is not supported.
Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
Diffstat (limited to 'ggml/src')
| -rw-r--r-- | ggml/src/ggml-cuda.cu | 3 |
1 files changed, 3 insertions, 0 deletions
diff --git a/ggml/src/ggml-cuda.cu b/ggml/src/ggml-cuda.cu index 453bde0c..e488c9bc 100644 --- a/ggml/src/ggml-cuda.cu +++ b/ggml/src/ggml-cuda.cu @@ -3579,6 +3579,9 @@ GGML_CALL static bool ggml_backend_cuda_supports_op(ggml_backend_t backend, cons return (op->src[1]->type == GGML_TYPE_F16 && op->src[2]->type == GGML_TYPE_F16) || (op->src[1]->type == GGML_TYPE_Q8_0 && op->src[2]->type == GGML_TYPE_Q8_0); } + if (op->src[1]->ne[0] > 256) { + return false; + } if (op->src[0]->ne[0] == 64 && op->src[1]->type == GGML_TYPE_F16) { return true; } |