We read every piece of feedback, and take your input very seriously.
To see all available qualifiers, see our documentation.
There was an error while loading. Please reload this page.
1 parent 7097567 commit 06113f0Copy full SHA for 06113f0
src/llama-graph.cpp
@@ -804,7 +804,7 @@ ggml_tensor * llm_graph_context::build_ffn(
804
if (down) {
805
cur = build_lora_mm(down, cur);
806
if (arch == LLM_ARCH_GLM4) {
807
- // GLM4 seems to have precision issues in F16
+ // GLM4 seems to have numerical issues with half-precision accumulators
808
ggml_mul_mat_set_prec(cur, GGML_PREC_F32);
809
}
810
0 commit comments