ikawrakow/ik_llama.cpp
1
0
Fork 0
mirror of https://github.com/ikawrakow/ik_llama.cpp.git synced 2026-01-26 17:20:01 +00:00
Code Issues Packages Projects Releases Wiki Activity

This is better

Browse Source
This commit is contained in:
Iwan Kawrakow
2025-12-12 06:23:39 +00:00
parent 07ae4b1ef0
commit 643cccd2c8
2 changed files with 14 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

13
ggml/src/ggml-backend.cpp
View File

@@ -2053,6 +2053,19 @@ static ggml_status ggml_backend_sched_compute_splits_sm_graph(ggml_backend_sched
}
this_split.push_back(split_j);
}
if (false) {
auto split = this_split.front();
if (this_split.size() == 1) {
printf("=== Split %d with %d inputs on backend %d\n", i, split->n_inputs, split->backend_id);
} else {
printf("=== Split %d with %d inputs on backends", i, split->n_inputs);
for (int j = 0; j < (int)this_split.size(); ++j) printf(" %d", this_split[j]->backend_id);
printf("\n");
}
for (int j = 0; j < split->graph.n_nodes; ++j) {
printf(" %d %s(%s)\n", j, ggml_op_name(split->graph.nodes[j]->op), split->graph.nodes[j]->name);
}
}
for (auto split : this_split) {
ggml_backend_sched_copy_inputs(sched, split, needs_sync, ids, unique_ids, last_ids_tensor);
}

1
src/llama-build-context.cpp
View File

@@ -1228,6 +1228,7 @@ llm_expert_gating_func_type gating_op,
cur = ggml_cast(ctx, cur, GGML_TYPE_F16);
cb(cur, "ffn_out_f16", il_cb);
}
ggml_build_forward_expand(graph, routed_out);
results.push_back(cur);
}
GGML_ASSERT(!results.empty());