Public/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2026-09-19 01:04:55 +02:00
Code Issues Packages Projects Releases Wiki Activity

Merge branch 'master' into xsn/server_docker_isolate

Browse Source
This commit is contained in:
Xuan Son Nguyen
2026-08-03 18:52:40 +02:00
parent ea1abe6f43 e8e06f78e2
commit 7e565b10f7
29 changed files with 1810 additions and 437 deletions
Download Patch File Download Diff File Expand all files Collapse all files
common/arg.cpp
+18 -3
View File
@@ -27,6 +27,7 @@
#include <algorithm>
#include <cinttypes>
#include <climits>
#include <cmath>
#include <cstdarg>
#include <filesystem>
#include <fstream>
@@ -2036,7 +2037,13 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
{"--repeat-penalty"}, "N",
string_format("penalize repeat sequence of tokens (default: %.2f, 1.0 = disabled)", (double)params.sampling.penalty_repeat),
[](common_params & params, const std::string & value) {
params.sampling.penalty_repeat = std::stof(value);
const float penalty_repeat = std::stof(value);
if (!std::isfinite(penalty_repeat) ||
penalty_repeat <= 0.0f ||
!std::isfinite(1.0f/penalty_repeat)) {
throw std::runtime_error("error: repeat-penalty must be finite and greater than 0\n");
}
params.sampling.penalty_repeat = penalty_repeat;
params.sampling.user_sampling_config |= common_params_sampling_config::COMMON_PARAMS_SAMPLING_CONFIG_PENALTY_REPEAT;
}
).set_sampling());
@@ -2044,14 +2051,22 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
{"--presence-penalty"}, "N",
string_format("repeat alpha presence penalty (default: %.2f, 0.0 = disabled)", (double)params.sampling.penalty_present),
[](common_params & params, const std::string & value) {
params.sampling.penalty_present = std::stof(value);
const float penalty_present = std::stof(value);
if (!std::isfinite(penalty_present)) {
throw std::runtime_error("error: presence-penalty must be finite\n");
}
params.sampling.penalty_present = penalty_present;
}
).set_sampling());
add_opt(common_arg(
{"--frequency-penalty"}, "N",
string_format("repeat alpha frequency penalty (default: %.2f, 0.0 = disabled)", (double)params.sampling.penalty_freq),
[](common_params & params, const std::string & value) {
params.sampling.penalty_freq = std::stof(value);
const float penalty_freq = std::stof(value);
if (!std::isfinite(penalty_freq)) {
throw std::runtime_error("error: frequency-penalty must be finite\n");
}
params.sampling.penalty_freq = penalty_freq;
}
).set_sampling());
add_opt(common_arg(