From e0dff58475bc9ed68eedcb265ee998f2fcabb3b1 Mon Sep 17 00:00:00 2001
From: Mikolaj Kucharski <mikolaj@kucharski.name>
Date: Mon, 21 Sep 2026 10:47:38 +0000
Subject: [PATCH] args: add env vars for temperature, top-p, min-p and
 penalties (#27380)

Allow configuring --temp, --top-p, --min-p, --repeat-penalty,
--presence-penalty and --frequency-penalty via LLAMA_ARG_* so
llama-server can be fully controlled from an EnvironmentFile
(e.g. systemd on Debian).

Use `llama-gen-docs` to regenerate the readme files.
---
 common/arg.cpp             | 12 ++++++------
 tools/cli/README.md        | 12 ++++++------
 tools/completion/README.md | 12 ++++++------
 tools/server/README.md     | 12 ++++++------
 4 files changed, 24 insertions(+), 24 deletions(-)

diff --git a/common/arg.cpp b/common/arg.cpp
index c4c4e143c9..996ea75fef 100644
--- a/common/arg.cpp
+++ b/common/arg.cpp
@@ -2016,7 +2016,7 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
             params.sampling.temp = std::max(params.sampling.temp, 0.0f);
             params.sampling.user_sampling_config |= common_params_sampling_config::COMMON_PARAMS_SAMPLING_CONFIG_TEMP;
         }
-    ).set_sampling());
+    ).set_sampling().set_env("LLAMA_ARG_TEMPERATURE"));
     add_opt(common_arg(
         {"--top-k"}, "N",
         string_format("top-k sampling (default: %d, 0 = disabled)", params.sampling.top_k),
@@ -2032,7 +2032,7 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
             params.sampling.top_p = std::stof(value);
             params.sampling.user_sampling_config |= common_params_sampling_config::COMMON_PARAMS_SAMPLING_CONFIG_TOP_P;
         }
-    ).set_sampling());
+    ).set_sampling().set_env("LLAMA_ARG_TOP_P"));
     add_opt(common_arg(
         {"--min-p"}, "N",
         string_format("min-p sampling (default: %.2f, 0.0 = disabled)", (double)params.sampling.min_p),
@@ -2040,7 +2040,7 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
             params.sampling.min_p = std::stof(value);
             params.sampling.user_sampling_config |= common_params_sampling_config::COMMON_PARAMS_SAMPLING_CONFIG_MIN_P;
         }
-    ).set_sampling());
+    ).set_sampling().set_env("LLAMA_ARG_MIN_P"));
     add_opt(common_arg(
         {"--top-nsigma", "--top-n-sigma"}, "N",
         string_format("top-n-sigma sampling (default: %.2f, -1.0 = disabled)", params.sampling.top_n_sigma),
@@ -2096,7 +2096,7 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
             params.sampling.penalty_repeat = penalty_repeat;
             params.sampling.user_sampling_config |= common_params_sampling_config::COMMON_PARAMS_SAMPLING_CONFIG_PENALTY_REPEAT;
         }
-    ).set_sampling());
+    ).set_sampling().set_env("LLAMA_ARG_REPEAT_PENALTY"));
     add_opt(common_arg(
         {"--presence-penalty"}, "N",
         string_format("repeat alpha presence penalty (default: %.2f, 0.0 = disabled)", (double)params.sampling.penalty_present),
@@ -2107,7 +2107,7 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
             }
             params.sampling.penalty_present = penalty_present;
         }
-    ).set_sampling());
+    ).set_sampling().set_env("LLAMA_ARG_PRESENCE_PENALTY"));
     add_opt(common_arg(
         {"--frequency-penalty"}, "N",
         string_format("repeat alpha frequency penalty (default: %.2f, 0.0 = disabled)", (double)params.sampling.penalty_freq),
@@ -2118,7 +2118,7 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
             }
             params.sampling.penalty_freq = penalty_freq;
         }
-    ).set_sampling());
+    ).set_sampling().set_env("LLAMA_ARG_FREQUENCY_PENALTY"));
     add_opt(common_arg(
         {"--dry-multiplier"}, "N",
         string_format("set DRY sampling multiplier (default: %.2f, 0.0 = disabled)", (double)params.sampling.dry_multiplier),
diff --git a/tools/cli/README.md b/tools/cli/README.md
index ea1f7aaf8b..acecc74a6d 100644
--- a/tools/cli/README.md
+++ b/tools/cli/README.md
@@ -106,18 +106,18 @@
 | `-s, --seed SEED` | RNG seed (default: -1, use random seed for -1) |
 | `--sampler-seq, --sampling-seq SEQUENCE` | simplified sequence for samplers that will be used (default: edskypmxt) |
 | `--ignore-eos` | ignore end of stream token and continue generating (implies --logit-bias EOS-inf) |
-| `--temp, --temperature N` | temperature (default: 0.80) |
+| `--temp, --temperature N` | temperature (default: 0.80)<br/>(env: LLAMA_ARG_TEMPERATURE) |
 | `--top-k N` | top-k sampling (default: 40, 0 = disabled)<br/>(env: LLAMA_ARG_TOP_K) |
-| `--top-p N` | top-p sampling (default: 0.95, 1.0 = disabled) |
-| `--min-p N` | min-p sampling (default: 0.05, 0.0 = disabled) |
+| `--top-p N` | top-p sampling (default: 0.95, 1.0 = disabled)<br/>(env: LLAMA_ARG_TOP_P) |
+| `--min-p N` | min-p sampling (default: 0.05, 0.0 = disabled)<br/>(env: LLAMA_ARG_MIN_P) |
 | `--top-nsigma, --top-n-sigma N` | top-n-sigma sampling (default: -1.00, -1.0 = disabled) |
 | `--xtc-probability N` | xtc probability (default: 0.00, 0.0 = disabled) |
 | `--xtc-threshold N` | xtc threshold (default: 0.10, 1.0 = disabled) |
 | `--typical, --typical-p N` | locally typical sampling, parameter p (default: 1.00, 1.0 = disabled) |
 | `--repeat-last-n N` | last n tokens to consider for penalize (default: 64, 0 = disabled) |
-| `--repeat-penalty N` | penalize repeat sequence of tokens (default: 1.00, 1.0 = disabled) |
-| `--presence-penalty N` | repeat alpha presence penalty (default: 0.00, 0.0 = disabled) |
-| `--frequency-penalty N` | repeat alpha frequency penalty (default: 0.00, 0.0 = disabled) |
+| `--repeat-penalty N` | penalize repeat sequence of tokens (default: 1.00, 1.0 = disabled)<br/>(env: LLAMA_ARG_REPEAT_PENALTY) |
+| `--presence-penalty N` | repeat alpha presence penalty (default: 0.00, 0.0 = disabled)<br/>(env: LLAMA_ARG_PRESENCE_PENALTY) |
+| `--frequency-penalty N` | repeat alpha frequency penalty (default: 0.00, 0.0 = disabled)<br/>(env: LLAMA_ARG_FREQUENCY_PENALTY) |
 | `--dry-multiplier N` | set DRY sampling multiplier (default: 0.00, 0.0 = disabled) |
 | `--dry-base N` | set DRY sampling base value (default: 1.75) |
 | `--dry-allowed-length N` | set allowed length for DRY sampling (default: 2) |
diff --git a/tools/completion/README.md b/tools/completion/README.md
index c9a4cccfc2..e2ac0668c6 100644
--- a/tools/completion/README.md
+++ b/tools/completion/README.md
@@ -189,18 +189,18 @@ llama-completion.exe -m models\gemma-1.1-7b-it.Q4_K_M.gguf --ignore-eos -n -1
 | `-s, --seed SEED` | RNG seed (default: -1, use random seed for -1) |
 | `--sampler-seq, --sampling-seq SEQUENCE` | simplified sequence for samplers that will be used (default: edskypmxt) |
 | `--ignore-eos` | ignore end of stream token and continue generating (implies --logit-bias EOS-inf) |
-| `--temp, --temperature N` | temperature (default: 0.80) |
+| `--temp, --temperature N` | temperature (default: 0.80)<br/>(env: LLAMA_ARG_TEMPERATURE) |
 | `--top-k N` | top-k sampling (default: 40, 0 = disabled)<br/>(env: LLAMA_ARG_TOP_K) |
-| `--top-p N` | top-p sampling (default: 0.95, 1.0 = disabled) |
-| `--min-p N` | min-p sampling (default: 0.05, 0.0 = disabled) |
+| `--top-p N` | top-p sampling (default: 0.95, 1.0 = disabled)<br/>(env: LLAMA_ARG_TOP_P) |
+| `--min-p N` | min-p sampling (default: 0.05, 0.0 = disabled)<br/>(env: LLAMA_ARG_MIN_P) |
 | `--top-nsigma, --top-n-sigma N` | top-n-sigma sampling (default: -1.00, -1.0 = disabled) |
 | `--xtc-probability N` | xtc probability (default: 0.00, 0.0 = disabled) |
 | `--xtc-threshold N` | xtc threshold (default: 0.10, 1.0 = disabled) |
 | `--typical, --typical-p N` | locally typical sampling, parameter p (default: 1.00, 1.0 = disabled) |
 | `--repeat-last-n N` | last n tokens to consider for penalize (default: 64, 0 = disabled) |
-| `--repeat-penalty N` | penalize repeat sequence of tokens (default: 1.00, 1.0 = disabled) |
-| `--presence-penalty N` | repeat alpha presence penalty (default: 0.00, 0.0 = disabled) |
-| `--frequency-penalty N` | repeat alpha frequency penalty (default: 0.00, 0.0 = disabled) |
+| `--repeat-penalty N` | penalize repeat sequence of tokens (default: 1.00, 1.0 = disabled)<br/>(env: LLAMA_ARG_REPEAT_PENALTY) |
+| `--presence-penalty N` | repeat alpha presence penalty (default: 0.00, 0.0 = disabled)<br/>(env: LLAMA_ARG_PRESENCE_PENALTY) |
+| `--frequency-penalty N` | repeat alpha frequency penalty (default: 0.00, 0.0 = disabled)<br/>(env: LLAMA_ARG_FREQUENCY_PENALTY) |
 | `--dry-multiplier N` | set DRY sampling multiplier (default: 0.00, 0.0 = disabled) |
 | `--dry-base N` | set DRY sampling base value (default: 1.75) |
 | `--dry-allowed-length N` | set allowed length for DRY sampling (default: 2) |
diff --git a/tools/server/README.md b/tools/server/README.md
index ef90334048..0ee8df291b 100644
--- a/tools/server/README.md
+++ b/tools/server/README.md
@@ -123,18 +123,18 @@ For the full list of features, please refer to [server's changelog](https://gith
 | `-s, --seed SEED` | RNG seed (default: -1, use random seed for -1) |
 | `--sampler-seq, --sampling-seq SEQUENCE` | simplified sequence for samplers that will be used (default: edskypmxt) |
 | `--ignore-eos` | ignore end of stream token and continue generating (implies --logit-bias EOS-inf) |
-| `--temp, --temperature N` | temperature (default: 0.80) |
+| `--temp, --temperature N` | temperature (default: 0.80)<br/>(env: LLAMA_ARG_TEMPERATURE) |
 | `--top-k N` | top-k sampling (default: 40, 0 = disabled)<br/>(env: LLAMA_ARG_TOP_K) |
-| `--top-p N` | top-p sampling (default: 0.95, 1.0 = disabled) |
-| `--min-p N` | min-p sampling (default: 0.05, 0.0 = disabled) |
+| `--top-p N` | top-p sampling (default: 0.95, 1.0 = disabled)<br/>(env: LLAMA_ARG_TOP_P) |
+| `--min-p N` | min-p sampling (default: 0.05, 0.0 = disabled)<br/>(env: LLAMA_ARG_MIN_P) |
 | `--top-nsigma, --top-n-sigma N` | top-n-sigma sampling (default: -1.00, -1.0 = disabled) |
 | `--xtc-probability N` | xtc probability (default: 0.00, 0.0 = disabled) |
 | `--xtc-threshold N` | xtc threshold (default: 0.10, 1.0 = disabled) |
 | `--typical, --typical-p N` | locally typical sampling, parameter p (default: 1.00, 1.0 = disabled) |
 | `--repeat-last-n N` | last n tokens to consider for penalize (default: 64, 0 = disabled) |
-| `--repeat-penalty N` | penalize repeat sequence of tokens (default: 1.00, 1.0 = disabled) |
-| `--presence-penalty N` | repeat alpha presence penalty (default: 0.00, 0.0 = disabled) |
-| `--frequency-penalty N` | repeat alpha frequency penalty (default: 0.00, 0.0 = disabled) |
+| `--repeat-penalty N` | penalize repeat sequence of tokens (default: 1.00, 1.0 = disabled)<br/>(env: LLAMA_ARG_REPEAT_PENALTY) |
+| `--presence-penalty N` | repeat alpha presence penalty (default: 0.00, 0.0 = disabled)<br/>(env: LLAMA_ARG_PRESENCE_PENALTY) |
+| `--frequency-penalty N` | repeat alpha frequency penalty (default: 0.00, 0.0 = disabled)<br/>(env: LLAMA_ARG_FREQUENCY_PENALTY) |
 | `--dry-multiplier N` | set DRY sampling multiplier (default: 0.00, 0.0 = disabled) |
 | `--dry-base N` | set DRY sampling base value (default: 1.75) |
 | `--dry-allowed-length N` | set allowed length for DRY sampling (default: 2) |