happyz
/
llama.cpp
mirror of https://github.com/ggerganov/llama.cpp.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

sampling : always populate logits for sampled probs 

This commit updates common/sampler.cpp set_logits and
src/llama-sampling.cpp llama_sampler_sample to always populate the
logits field when backend sampled probabilities are available.

The motivation for this is that this ensure that CPU sampler always have
access to the logits values even when probabilites have been produced by
backend samplers.
This commit is contained in:
Daniel Bevenius 2025-11-19 07:14:11 +01:00
parent 0da7e7dccc
commit 51fee29822
No known key found for this signature in database
2 changed files with 2 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
common/sampling.cpp
View File

@ -129,7 +129,7 @@ struct common_sampler {
const uint32_t sampled_probs_count = llama_get_backend_sampled_probs_count_ith(ctx, idx);
cur.reserve(sampled_probs_count);
for (uint32_t i = 0; i < sampled_probs_count; ++i) {
cur.emplace_back(llama_token_data{sampled_ids[i], 0.0f, sampled_probs[i]});
cur.emplace_back(llama_token_data{sampled_ids[i], sampled_logits[i], sampled_probs[i]});
}
} else if (sampled_logits) {
const uint32_t sampled_logits_count = llama_get_backend_sampled_logits_count_ith(ctx, idx);

2
src/llama-sampling.cpp
View File

@ -461,7 +461,7 @@ llama_token llama_sampler_sample(struct llama_sampler * smpl, struct llama_conte
const uint32_t sampled_probs_count = llama_get_backend_sampled_probs_count_ith(ctx, idx);
cur.reserve(sampled_probs_count);
for (uint32_t i = 0; i < sampled_probs_count; ++i) {
cur.emplace_back(llama_token_data{sampled_ids[i], 0.0f, sampled_probs[i]});
cur.emplace_back(llama_token_data{sampled_ids[i], sampled_logits[i], sampled_probs[i]});
}
} else if (sampled_logits) {
const uint32_t sampled_logits_count = llama_get_backend_sampled_logits_count_ith(ctx, idx);