nicolas
/
ik_llama_opt
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Fix hybrid graph parallel + muge (#1426)

This commit is contained in:
Kawrakow 2026-03-14 18:15:09 +01:00 committed by GitHub
parent f8d95f1279
commit 10a8f5f8f1
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 7 additions and 5 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

12
src/llama.cpp
View File

@ -4666,7 +4666,9 @@ static void llama_repack_up_gate_exps(llama_context & lctx) {
auto & model = lctx.model;
bool needs_repack = false;
for (auto & l : model.layers) {
if (l.ffn_up_gate_exps && l.ffn_up_exps && l.ffn_gate_exps) {
if (l.ffn_up_gate_exps && l.ffn_up_exps && l.ffn_gate_exps &&
ggml_backend_buffer_is_host(l.ffn_up_gate_exps->buffer) &&
ggml_backend_buffer_is_host(l.ffn_up_exps->buffer) && ggml_backend_buffer_is_host(l.ffn_gate_exps->buffer)) {
needs_repack = true; break;
}
}
@ -4675,7 +4677,9 @@ static void llama_repack_up_gate_exps(llama_context & lctx) {
std::vector<char> aux_buffer_up, aux_buffer_gate, aux_buffer_up_gate;
for (int il = 0; il < int(model.layers.size()); ++il) {
auto & l = model.layers[il];
if (l.ffn_up_gate_exps && l.ffn_up_exps && l.ffn_gate_exps) {
if (l.ffn_up_gate_exps && l.ffn_up_exps && l.ffn_gate_exps &&
ggml_backend_buffer_is_host(l.ffn_up_gate_exps->buffer) &&
ggml_backend_buffer_is_host(l.ffn_up_exps->buffer) && ggml_backend_buffer_is_host(l.ffn_gate_exps->buffer)) {
GGML_ASSERT(l.ffn_up_gate_exps->type == l.ffn_up_exps->type && l.ffn_up_gate_exps->type == l.ffn_gate_exps->type);
GGML_ASSERT(l.ffn_up_gate_exps->ne[0] == l.ffn_up_exps->ne[0] && l.ffn_up_gate_exps->ne[0] == l.ffn_gate_exps->ne[0]);
GGML_ASSERT(l.ffn_up_gate_exps->ne[2] == l.ffn_up_exps->ne[2] && l.ffn_up_gate_exps->ne[2] == l.ffn_gate_exps->ne[2]);
@ -5209,9 +5213,7 @@ struct llama_context * llama_init_from_model(
LLAMA_LOG_INFO("%s: pipeline parallelism enabled (n_copies=%d)\n", __func__, ggml_backend_sched_get_n_copies(ctx->sched));
}
if (ctx->model.split_mode != LLAMA_SPLIT_MODE_GRAPH) {
llama_repack_up_gate_exps(*ctx);
}
llama_repack_up_gate_exps(*ctx);
// build worst-case graph
int n_past = cparams.n_ctx - n_tokens;