ikawrakow/ik_llama.cpp
1
0
Fork 0
mirror of https://github.com/ikawrakow/ik_llama.cpp.git synced 2026-03-04 11:00:00 +00:00
Code Issues Packages Projects Releases Wiki Activity

Bring back #1333 and #1335 (#1340)

Browse Source 
* Bring back fused delta net 3

* Remove autoregressive and chunking
This commit is contained in:
Kawrakow
2026-02-28 14:31:42 +01:00
committed by GitHub
parent 1922449b2c
commit 0ff3a43289
8 changed files with 78 additions and 652 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
src/llama.cpp
View File

@@ -1512,6 +1512,7 @@ static void llm_load_print_meta(llama_model_loader & ml, llama_model & model) {
LLAMA_LOG_INFO("%s: ssm_d_inner = %u\n", __func__, hparams.ssm_d_inner);
LLAMA_LOG_INFO("%s: ssm_d_state = %u\n", __func__, hparams.ssm_d_state);
LLAMA_LOG_INFO("%s: ssm_dt_rank = %u\n", __func__, hparams.ssm_dt_rank);
LLAMA_LOG_INFO("%s: ssm_n_group = %u\n", __func__, hparams.ssm_n_group);
}
LLAMA_LOG_INFO("%s: model type = %s\n", __func__, llama_model_type_name(model.type));
@@ -4394,7 +4395,6 @@ struct llama_context_params llama_context_default_params() {
/*.split_mode_graph_scheduling =*/ false,
// /*.split_mode_f16 =*/ true,
/*.scheduler_async =*/ false,
/*.fused_delta_net =*/ 0,
/*.mtp =*/ false,
/*.mtp_op_type =*/ MTP_OP_NONE,
/*.abort_callback =*/ nullptr,
@@ -4766,7 +4766,6 @@ struct llama_context * llama_init_from_model(
cparams.split_mode_graph_scheduling = params.split_mode_graph_scheduling;
//cparams.split_mode_f16 = params.split_mode_f16;
cparams.scheduler_async = params.scheduler_async;
cparams.fused_delta_net = params.fused_delta_net;
cparams.min_experts = params.min_experts;
cparams.thresh_experts = params.thresh_experts;
cparams.cuda_params = params.cuda_params;
@@ -4873,7 +4872,6 @@ struct llama_context * llama_init_from_model(
//LLAMA_LOG_INFO("%s: split_mode_f16= %d\n", __func__, cparams.split_mode_f16);
LLAMA_LOG_INFO("%s: reduce_type = %s\n", __func__, ggml_type_name(cparams.reduce_type));
LLAMA_LOG_INFO("%s: sched_async = %d\n", __func__, cparams.scheduler_async);
LLAMA_LOG_INFO("%s: fused_delta = %d\n", __func__, cparams.fused_delta_net);
LLAMA_LOG_INFO("%s: ser = %d, %g\n", __func__, cparams.min_experts, cparams.thresh_experts);
LLAMA_LOG_INFO("%s: freq_base = %.1f\n", __func__, cparams.rope_freq_base);
LLAMA_LOG_INFO("%s: freq_scale = %g\n", __func__, cparams.rope_freq_scale);