CS348Project/llama.cpp
5
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-10-27 08:21:30 +00:00
Code Issues Packages Projects Releases Wiki Activity

llama : allow other bufts when overriding to CPU, add --no-repack option (#14990)

Browse Source
This commit is contained in:
Diego Devesa
2025-07-31 09:11:34 -07:00
committed by GitHub
parent e08a98826b
commit d6818d06a6
5 changed files with 39 additions and 21 deletions
Download Patch File Download Diff File Expand all files Collapse all files

7
common/arg.cpp
View File

@@ -2095,6 +2095,13 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
params.no_kv_offload = true;
}
).set_env("LLAMA_ARG_NO_KV_OFFLOAD"));
add_opt(common_arg(
{"-nr", "--no-repack"},
"disable weight repacking",
[](common_params & params) {
params.no_extra_bufts = true;
}
).set_env("LLAMA_ARG_NO_REPACK"));
add_opt(common_arg(
{"-ctk", "--cache-type-k"}, "TYPE",
string_format(

1
common/common.cpp
View File

@@ -1122,6 +1122,7 @@ struct llama_model_params common_model_params_to_llama(common_params & params) {
mparams.use_mmap = params.use_mmap;
mparams.use_mlock = params.use_mlock;
mparams.check_tensors = params.check_tensors;
mparams.use_extra_bufts = !params.no_extra_bufts;
if (params.kv_overrides.empty()) {
mparams.kv_overrides = NULL;

1
common/common.h
View File

@@ -359,6 +359,7 @@ struct common_params {
bool warmup = true; // warmup run
bool check_tensors = false; // validate tensor data
bool no_op_offload = false; // globally disable offload host tensor operations to device
bool no_extra_bufts = false; // disable extra buffer types (used for weight repacking)
bool single_turn = false; // single turn chat conversation