upstream-archive/whisper.cpp
1
0
Fork 0
mirror of https://github.com/ggml-org/whisper.cpp.git synced 2026-07-23 11:10:57 -05:00
Code Issues Packages Projects Releases Wiki Activity

abort if we see a multi buffer (llama/25276)

Browse Source
This commit is contained in:
Eve
2026-07-05 18:38:47 +00:00
committed by Georgi Gerganov
parent f6f1252b05
commit 38c6260238
1 changed files with 5 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

5
ggml/src/ggml-backend-meta.cpp
View File

@@ -1144,6 +1144,11 @@ static enum ggml_status ggml_backend_meta_buffer_init_tensor_impl(ggml_backend_m
ggml_context * simple_ctx = stc.ctxs[j].get(); ggml_context * simple_ctx = stc.ctxs[j].get();
ggml_backend_buffer_t simple_buf = buf_ctx->bufs[j].get(); ggml_backend_buffer_t simple_buf = buf_ctx->bufs[j].get();
if ((simple_buf != nullptr) && ggml_backend_buffer_is_multi_buffer(simple_buf)) {
// see https://github.com/ggml-org/llama.cpp/issues/22197
GGML_ABORT("multi buffers are not supported by the meta backend");
}
if (split_dim >= 0 && split_dim < GGML_MAX_DIMS) { if (split_dim >= 0 && split_dim < GGML_MAX_DIMS) {
// TODO: the following assert fails for llama-parallel even though the results are correct: // TODO: the following assert fails for llama-parallel even though the results are correct:
// GGML_ASSERT(ggml_is_contiguously_allocated(tensor)); // GGML_ASSERT(ggml_is_contiguously_allocated(tensor));