File size: 2,757 Bytes
6b9d0d8 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 | #include "llama.h"
#include "../src/llama-ext.h"
#include "arg.h"
#include "common.h"
#include "fit.h"
#include "log.h"
#include <cinttypes>
#if defined(_MSC_VER)
#pragma warning(disable: 4244 4267) // possible loss of data
#endif
// satisfies -Wmissing-declarations
int llama_fit_params(int argc, char ** argv);
int llama_fit_params(int argc, char ** argv) {
common_params params;
common_init();
if (!common_params_parse(argc, argv, params, LLAMA_EXAMPLE_FIT_PARAMS)) {
return 1;
}
llama_backend_init();
llama_numa_init(params.numa);
auto mparams = common_model_params_to_llama(params);
auto cparams = common_context_params_to_llama(params);
if (!params.fit_params_print) {
const common_params_fit_status status = common_fit_params(params.model.path.c_str(), &mparams, &cparams,
params.tensor_split, params.tensor_buft_overrides.data(), params.fit_params_target.data(), params.fit_params_min_ctx,
params.verbosity >= LOG_LEVEL_DEBUG ? GGML_LOG_LEVEL_DEBUG : GGML_LOG_LEVEL_ERROR);
if (status != COMMON_PARAMS_FIT_STATUS_SUCCESS) {
LOG_ERR("%s: failed to fit CLI arguments to free memory, exiting...\n", __func__);
exit(1);
}
LOG_INF("%s: printing fitted CLI arguments to stdout...\n", __func__);
common_log_flush(common_log_main());
printf("-c %" PRIu32 " -ngl %" PRIi32, cparams.n_ctx, mparams.n_gpu_layers);
size_t nd = llama_max_devices();
while (nd > 1 && mparams.tensor_split[nd - 1] == 0.0f) {
nd--;
}
if (nd > 1) {
for (size_t id = 0; id < nd; id++) {
if (id == 0) {
printf(" -ts ");
}
printf("%s%" PRIu32, id > 0 ? "," : "", uint32_t(mparams.tensor_split[id]));
}
}
const size_t ntbo = llama_max_tensor_buft_overrides();
bool any_tbo = false;
for (size_t itbo = 0; itbo < ntbo && mparams.tensor_buft_overrides[itbo].pattern != nullptr; itbo++) {
if (itbo == 0) {
printf(" -ot \"");
}
printf("%s%s=%s", itbo > 0 ? "," : "", mparams.tensor_buft_overrides[itbo].pattern, ggml_backend_buft_name(mparams.tensor_buft_overrides[itbo].buft));
any_tbo = true;
}
printf("%s\n", any_tbo ? "\"" : "");
} else {
LOG_INF("%s: printing estimated memory in MiB to stdout (device, model, context, compute) ...\n", __func__);
common_log_flush(common_log_main());
common_fit_print(params.model.path.c_str(), &mparams, &cparams);
}
return 0;
}
|