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

pre gguf merge

Browse source
This commit is contained in:
netrunnereve 2023-08-23 23:53:06 -04:00
parent d50ccb03a3
commit 471e469ae2
1 changed files with 2 additions and 2 deletions
Download patch file Download diff file Expand all files Collapse all files

4
examples/perplexity/perplexity.cpp
View file

@ -130,7 +130,7 @@ std::vector<float> hellaswag_evaluate_tokens(llama_context * ctx, const std::vec
for (size_t i_chunk = 0; i_chunk < n_chunk; ++i_chunk) {
size_t n_tokens = tokens.size() - i_chunk * n_batch;
n_tokens = std::min(n_tokens, size_t(n_batch));
if (llama_eval(ctx, tokens.data() + i_chunk * n_batch, n_tokens, n_past, n_thread)) {
if (llama_eval(ctx, tokens.data() + i_chunk * n_batch, n_tokens, n_past, n_thread, n_thread)) {
fprintf(stderr, "%s : failed to eval\n", __func__);
return {};
}
@ -304,7 +304,7 @@ void hellaswag_score(llama_context * ctx, const gpt_params & params) {
//}
// Evaluate the query
logits = hellaswag_evaluate_tokens(ctx, query_embd, context_size, params.n_batch, n_vocab, params.n_threads, params.pp_threads);
logits = hellaswag_evaluate_tokens(ctx, query_embd, context_size, params.n_batch, n_vocab, params.n_threads);
if (logits.empty()) {
fprintf(stderr, "%s : failed to eval\n", __func__);
return;