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

sampling: clarify candidate ids usage in comments

This commit is contained in:
Daniel Bevenius 2025-11-23 11:28:19 +01:00
parent 65500d05ab
commit ae23d2d2c1
No known key found for this signature in database
1 changed files with 1 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
src/llama-context.cpp
View File

@ -1482,9 +1482,7 @@ int llama_context::decode(const llama_batch & batch_inp) {
} }
// async copy the candidate token ids from the backend to the host. // async copy the candidate token ids from the backend to the host.
// These are needed for: // These are needed by CPU samplers to map probability/logit indices to vocab token ids.
// 1) Backend dist sampler to map indices to vocab token ids.
// 2) CPU samplers to associate candidate logits with their token ids.
copy_tensor_async_candidates(res->t_candidates, sampling.candidates, stride, sampling.candidates_count, seq_to_output_row, sched.get()); copy_tensor_async_candidates(res->t_candidates, sampling.candidates, stride, sampling.candidates_count, seq_to_output_row, sched.get());
} }