CS348Project/llama.cpp
5
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-11-03 09:22:01 +00:00
Code Issues Packages Projects Releases Wiki Activity

examples : replace llama_kv_cache_seq_* with llama_past_seq_*

Browse Source
This commit is contained in:
Francis Couture-Harpin
2024-06-10 14:44:42 -04:00
parent 372482dffe
commit 43d8d4bf9e
23 changed files with 125 additions and 112 deletions
Download Patch File Download Diff File Expand all files Collapse all files

12
examples/perplexity/perplexity.cpp
View File

@@ -400,7 +400,7 @@ static results_perplexity perplexity_v2(llama_context * ctx, const gpt_params &
const auto t_start = std::chrono::high_resolution_clock::now();
// clear the KV cache
llama_kv_cache_clear(ctx);
llama_past_clear(ctx);
for (int j = 0; j < num_batches; ++j) {
const int batch_start = start + j * n_batch;
@@ -575,7 +575,7 @@ static results_perplexity perplexity(llama_context * ctx, const gpt_params & par
const auto t_start = std::chrono::high_resolution_clock::now();
// clear the KV cache
llama_kv_cache_clear(ctx);
llama_past_clear(ctx);
for (int j = 0; j < num_batches; ++j) {
const int batch_start = start + j * n_batch;
@@ -944,7 +944,7 @@ static void hellaswag_score(llama_context * ctx, const gpt_params & params) {
return;
}
llama_kv_cache_clear(ctx);
llama_past_clear(ctx);
// decode all tasks [i0, i1)
if (!decode_helper(ctx, batch, batch_logits, n_batch, n_vocab)) {
@@ -1221,7 +1221,7 @@ static void winogrande_score(llama_context * ctx, const gpt_params & params) {
return;
}
llama_kv_cache_clear(ctx);
llama_past_clear(ctx);
// decode all tasks [i0, i1)
if (!decode_helper(ctx, batch, batch_logits, n_batch, n_vocab)) {
@@ -1594,7 +1594,7 @@ static void multiple_choice_score(llama_context * ctx, const gpt_params & params
return;
}
llama_kv_cache_clear(ctx);
llama_past_clear(ctx);
// decode all tasks [i0, i1)
if (!decode_helper(ctx, batch, batch_logits, n_batch, n_vocab)) {
@@ -1780,7 +1780,7 @@ static void kl_divergence(llama_context * ctx, const gpt_params & params) {
}
// clear the KV cache
llama_kv_cache_clear(ctx);
llama_past_clear(ctx);
for (int j = 0; j < num_batches; ++j) {
const int batch_start = start + j * n_batch;