Skip to content

llama : disable pipeline parallelism if compute buffer allocation fails #16748

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 1 commit into from
Oct 27, 2025
Merged

llama : disable pipeline parallelism if compute buffer allocation fails #16748

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
11 changes: 8 additions & 3 deletions src/llama-context.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -268,9 +268,7 @@ llama_context::llama_context(
if (pipeline_parallel) {
LLAMA_LOG_INFO("%s: pipeline parallelism enabled (n_copies=%d)\n", __func__, ggml_backend_sched_get_n_copies(sched.get()));
}
}

if (!hparams.vocab_only) {
llama_memory_context_ptr mctx;
if (memory) {
LLAMA_LOG_DEBUG("%s: reserving full memory module\n", __func__);
Expand Down Expand Up @@ -343,7 +341,14 @@ llama_context::llama_context(
{
auto * gf = graph_reserve(n_tokens, n_seqs, n_tokens, mctx.get());
if (!gf) {
throw std::runtime_error("failed to allocate compute pp buffers");
if (pipeline_parallel) {
LLAMA_LOG_WARN("%s: compute buffer allocation failed, retrying without pipeline parallelism\n", __func__);
sched.reset(ggml_backend_sched_new(backend_ptrs.data(), backend_buft.data(), backend_ptrs.size(), max_nodes, false, cparams.op_offload));
gf = graph_reserve(n_tokens, n_seqs, n_tokens, mctx.get());
}
if (!gf) {
throw std::runtime_error("failed to allocate compute pp buffers");
}
}

n_splits_pp = ggml_backend_sched_get_n_splits(sched.get());
Expand Down
Loading