vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

sampling : hide prev behind API and apply #3661

Browse source 
ggml-ci
This commit is contained in:
Georgi Gerganov 2023-10-20 18:26:20 +03:00
parent 7e2b5fb1dd
commit 56ba00b923
No known key found for this signature in database
GPG key ID: 449E073F9DC10735
9 changed files with 119 additions and 105 deletions
Download patch file Download diff file Expand all files Collapse all files

4
examples/speculative/speculative.cpp
View file

@ -154,7 +154,7 @@ int main(int argc, char ** argv) {
// sample from the target model
llama_token id = llama_sampling_sample(ctx_sampling, ctx_tgt, NULL, drafts[s_keep].i_batch_tgt[i_dft]);
llama_sampling_accept(ctx_sampling, ctx_tgt, id);
llama_sampling_accept(ctx_sampling, ctx_tgt, id, true);
//LOG("last: %s\n", LOG_TOKENS_TOSTR_PRETTY(ctx_tgt, ctx_sampling->prev).c_str());
@ -328,7 +328,7 @@ int main(int argc, char ** argv) {
const int s = sa[is];
llama_sampling_accept(drafts[s].ctx_sampling, ctx_dft, id);
llama_sampling_accept(drafts[s].ctx_sampling, ctx_dft, id, true);
drafts[s].tokens.push_back(id);