root/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggerganov/llama.cpp.git synced 2025-01-13 12:10:18 +00:00
Code Issues Actions 2 Packages Projects Releases Wiki Activity

perplexity : revert changes

Browse Source
This commit is contained in:
Francis Couture-Harpin 2024-09-09 22:00:37 -04:00
parent 503630e88a
commit 9e6b0e9419
1 changed files with 10 additions and 6 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

16
examples/perplexity/perplexity.cpp
View File

@ -583,9 +583,7 @@ static results_perplexity perplexity(llama_context * ctx, const gpt_params & par
int n_outputs = 0;
// clear the batch
llama_batch_clear(batch);
batch.n_tokens = 0;
for (int seq = 0; seq < n_seq_batch; seq++) {
int seq_start = batch_start + seq*n_ctx;
@ -598,10 +596,16 @@ static results_perplexity perplexity(llama_context * ctx, const gpt_params & par
}
for (int k = 0; k < batch_size; ++k) {
llama_pos pos = j*n_batch + k;
llama_batch_add(batch, tokens[seq_start + k], pos, { seq }, pos >= first);
n_outputs += (int) (pos >= first);
const int idx = seq*n_ctx + k;
batch.token [idx] = tokens[seq_start + k];
batch.pos [idx] = j*n_batch + k;
batch.n_seq_id[idx] = 1;
batch.seq_id [idx][0] = seq;
batch.logits [idx] = batch.pos[idx] >= first ? 1 : 0;
n_outputs += batch.logits[idx] != 0;
}
batch.n_tokens += batch_size;
// restore the original token in case it was set to BOS
tokens[seq_start] = token_org;