LLM-Inference/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2026-08-12 22:31:11 +04:00
Code Issues Packages Projects Releases Wiki Activity

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

Browse Source 
* 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
Download Patch File Download Diff File Expand all files Collapse all files
tests/test-sampling.cpp
+2 -2
View File
@@ -10,7 +10,7 @@
#include <string>
#include <vector>
extern struct llama_sampler * llama_sampler_init_dry_testing(int32_t context_size, float dry_multiplier, float dry_base, int32_t dry_allowed_length, int32_t dry_penalty_last_n, const std::vector<std::vector<llama_token>>& seq_breakers);
extern struct llama_sampler * llama_sampler_init_dry_testing(float dry_multiplier, float dry_base, int32_t dry_allowed_length, int32_t dry_penalty_last_n, const std::vector<std::vector<llama_token>>& seq_breakers);
static void dump(const llama_token_data_array * cur_p) {
for (size_t i = 0; i < cur_p->size; i++) {
@@ -168,7 +168,7 @@ static void test_dry(
sampler_tester tester(probs, expected_probs);
auto * sampler = llama_sampler_init_dry_testing(1024, dry_multiplier, dry_base, dry_allowed_length, dry_penalty_last_n, seq_breakers);
auto * sampler = llama_sampler_init_dry_testing(dry_multiplier, dry_base, dry_allowed_length, dry_penalty_last_n, seq_breakers);
for (size_t i = 0; i < last_tokens.size(); i++) {
llama_sampler_accept(sampler, last_tokens[i]);