diff options
| author | Georgi Gerganov <ggerganov@gmail.com> | 2024-03-19 10:21:54 +0200 |
|---|---|---|
| committer | GitHub <noreply@github.com> | 2024-03-19 10:21:54 +0200 |
| commit | b80cf3b2d1dee0ad325f7a794fecc66befce7336 (patch) | |
| tree | d4f0b38ce20a2795af53a4299275614d3fb0f429 | |
| parent | 970a48060ab9a6cc67aa063870323781c2a7bd7d (diff) | |
common : disable repeat penalties by default (#6127)
| -rw-r--r-- | common/sampling.h | 4 |
1 files changed, 2 insertions, 2 deletions
diff --git a/common/sampling.h b/common/sampling.h index 48b2459d..79a998be 100644 --- a/common/sampling.h +++ b/common/sampling.h @@ -32,13 +32,13 @@ typedef struct llama_sampling_params { float dynatemp_range = 0.00f; // 0.0 = disabled float dynatemp_exponent = 1.00f; // controls how entropy maps to temperature in dynamic temperature sampler int32_t penalty_last_n = 64; // last n tokens to penalize (0 = disable penalty, -1 = context size) - float penalty_repeat = 1.10f; // 1.0 = disabled + float penalty_repeat = 1.00f; // 1.0 = disabled float penalty_freq = 0.00f; // 0.0 = disabled float penalty_present = 0.00f; // 0.0 = disabled int32_t mirostat = 0; // 0 = disabled, 1 = mirostat, 2 = mirostat 2.0 float mirostat_tau = 5.00f; // target entropy float mirostat_eta = 0.10f; // learning rate - bool penalize_nl = true; // consider newlines as a repeatable token + bool penalize_nl = false; // consider newlines as a repeatable token std::vector<llama_sampler_type> samplers_sequence = { llama_sampler_type::TOP_K, |