llama : enable tensor split for LFM3 (#31431)

Assisted-by: Claude Opus 5.5
This commit is contained in:
Tarek Dakhran authored and GitHub committed 2026-10-31 23:40:50 +01:00
1 parent f6312ad75e
commit bc428e407a
2 files changed
+1 -2

No files matched your search

+1 -1
View File
@@ -591,7 +591,7 @@ struct ggml_backend_meta_split_state llama_meta_device_get_split_state(const str
return get_tensor_config_impl(GGML_BACKEND_SPLIT_AXIS_1, "ssm_out.weight");
}
if (std::regex_match(tensor_name, pattern_r_cache) || std::regex_match(tensor_name, pattern_s_cache)) {
if (ud->model->arch == LLM_ARCH_LFM2 || ud->model->arch == LLM_ARCH_LFM2MOE) {
if (ud->model->arch == LLM_ARCH_LFM2 || ud->model->arch == LLM_ARCH_LFM2MOE || ud->model->arch == LLM_ARCH_LFM3) {
// the LFM2 shortconv block runs fully mirrored, so its conv state must be mirrored too
return get_tensor_config_impl(GGML_BACKEND_SPLIT_AXIS_MIRRORED, "");
}