sampler : remove "full-context windows" from history-based samplers (#26524)

* Resolve -1 to 1024 instead of ctx-len for samplers

Because of backend-sampling we initialize samplers before the complete
llama_context is there. Therefore, we cannot infer the resolved context
length yet at the time we construct the samplers.

* Shared default of 64 for history-based samplers, remove context_size
This commit is contained in:
Oliver Simons
2026-08-04 21:28:55 +03:00
committed by GitHub
parent 76c956c137
commit a6aa6f5450
18 changed files with 52 additions and 80 deletions
+2 -2
View File
@@ -235,14 +235,14 @@ struct common_params_sampling {
float temp = 0.80f; // <= 0.0 to sample greedily, 0.0 to not output probabilities
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)
int32_t penalty_last_n = 64; // last n tokens to penalize (0 = disable penalty)
float penalty_repeat = 1.00f; // 1.0 = disabled
float penalty_freq = 0.00f; // 0.0 = disabled
float penalty_present = 0.00f; // 0.0 = disabled
float dry_multiplier = 0.0f; // 0.0 = disabled; DRY repetition penalty for tokens extending repetition:
float dry_base = 1.75f; // 0.0 = disabled; multiplier * base ^ (length of sequence before token - allowed length)
int32_t dry_allowed_length = 2; // tokens extending repetitions beyond this receive penalty
int32_t dry_penalty_last_n = -1; // how many tokens to scan for repetitions (0 = disable penalty, -1 = context size)
int32_t dry_penalty_last_n = 64; // how many tokens to scan for repetitions (0 = disable penalty)
float adaptive_target = -1.0f; // select tokens near this probability (valid range 0.0 to 1.0; negative = disabled)
float adaptive_decay = 0.90f; // EMA decay for adaptation; history ≈ 1/(1-decay) tokens (0.0 - 0.99)
int32_t mirostat = 0; // 0 = disabled, 1 = mirostat, 2 = mirostat 2.0