vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

ggml : add error handling to graph_compute (whisper/1714)

Browse source
This commit is contained in:
Finn Voorhees 2024-01-03 08:39:43 -05:00 • committed by Georgi Gerganov
parent c1d7cb28d3
commit 1bf681f90e
6 changed files with 18 additions and 11 deletions
Download patch file Download diff file Expand all files Collapse all files

4
ggml-cuda.cu
View file

@ -9910,7 +9910,7 @@ static void ggml_backend_cuda_graph_plan_compute(ggml_backend_t backend, ggml_ba
UNUSED(plan);
}
static void ggml_backend_cuda_graph_compute(ggml_backend_t backend, ggml_cgraph * cgraph) {
static bool ggml_backend_cuda_graph_compute(ggml_backend_t backend, ggml_cgraph * cgraph) {
ggml_backend_context_cuda * cuda_ctx = (ggml_backend_context_cuda *)backend->context;
ggml_cuda_set_main_device(cuda_ctx->device);
@ -9967,6 +9967,8 @@ static void ggml_backend_cuda_graph_compute(ggml_backend_t backend, ggml_cgraph
}
UNUSED(backend);
return true;
}
static bool ggml_backend_cuda_supports_op(ggml_backend_t backend, const ggml_tensor * op) {