vrr/prima.cpp
2
0
Fork 0
mirror of https://github.com/Lizonghang/prima.cpp.git synced 2025-09-09 18:44:35 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

use arg prefetch and remove arg unload

Browse source
This commit is contained in:
Lizonghang 2025-02-12 17:04:41 +04:00
parent 708b1d8c89
commit c84f9d29fe
5 changed files with 15 additions and 19 deletions
Download patch file Download diff file Expand all files Collapse all files

6
common/arg.cpp
View file

@ -724,10 +724,10 @@ gpt_params_context gpt_params_parser_init(gpt_params & params, llama_example ex,
}
).set_env("LLAMA_ARG_NEXT_NODE_IP"));
add_opt(llama_arg(
{"--unload", "--unload-weight"},
format("whether to unload layer weights after use (default: %s)", params.unload ? "true" : "false"),
{"--prefetch"},
format("whether to prefetch layer weights (default: %s)", params.prefetch ? "true" : "false"),
[](gpt_params & params) {
params.unload = true;
params.prefetch = true;
}
).set_env("LLAMA_ARG_UNLOAD"));
add_opt(llama_arg(