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

llama : add option for greedy sampling with probs (#3813)

Browse source 
* llama : add option for greedy sampling with probs

* llama : add comment about llama_sample_token_greedy() missing probs

* sampling : temp == 0.0 -> no probs, temp < 0.0 -> probs
This commit is contained in:
Georgi Gerganov 2023-10-28 14:23:11 +03:00 • committed by GitHub
parent 177461104b
commit ee1a0ec9cb
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
4 changed files with 9 additions and 3 deletions
Download patch file Download diff file Expand all files Collapse all files

8
common/sampling.cpp
View file

@ -167,8 +167,12 @@ llama_token llama_sampling_sample(
llama_sample_grammar(ctx_main, &cur_p, ctx_sampling->grammar);
}
if (temp <= 0) {
// greedy sampling
if (temp < 0.0) {
// greedy sampling, with probs
llama_sample_softmax(ctx_main, &cur_p);
id = cur_p.data[0].id;
} else if (temp == 0.0) {
// greedy sampling, no probs
id = llama_sample_token_greedy(ctx_main, &cur_p);
} else {
if (mirostat == 1) {