vrr/prima.cpp
2
0
Fork 0
mirror of https://github.com/Lizonghang/prima.cpp.git synced 2025-09-06 04:31:09 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

add option --keep-inp-out-in-metal

Browse source
This commit is contained in:
Zonghang Li 2025-01-22 11:25:09 +04:00
parent facb4ea736
commit 33429ec4e1
2 changed files with 3 additions and 1 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

3
common/arg.cpp
View file

@ -737,7 +737,7 @@ gpt_params_context gpt_params_parser_init(gpt_params & params, llama_example ex,
params.cuda_mem = value; // in GiB
}
).set_env("LLAMA_ARG_CUDA_MEM"));
// "--keep-inp-out-in-metal" is a temporary option to keep the input and output in metal
#ifdef GGML_USE_METAL
add_opt(llama_arg(
{"--keep-inp-out-in-metal"},
format("whether to keep input and output weight in metal (default: %s)", params.keep_inp_out_in_metal ? "true" : "false"),
@ -745,6 +745,7 @@ gpt_params_context gpt_params_parser_init(gpt_params & params, llama_example ex,
params.keep_inp_out_in_metal = true;
}
).set_env("LLAMA_ARG_KEEP_INP_OUT_IN_METAL"));
#endif
add_opt(llama_arg(
{"-n", "--predict", "--n-predict"}, "N",
format("number of tokens to predict (default: %d, -1 = infinity, -2 = until context filled)", params.n_predict),

1
src/llama.cpp
View file

@ -19773,6 +19773,7 @@ struct llama_model_params llama_model_default_params() {
/*.use_mmap =*/ true,
/*.use_mlock =*/ false,
/*.check_tensors =*/ false,
/*.keep_inp_out_in_metal =*/ false,
};
#ifdef GGML_USE_METAL