Mirrors/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggerganov/llama.cpp.git synced 2024-12-24 13:28:50 +01:00
Code Issues Packages Projects Releases Wiki Activity

main : evaluate tokens in batches after swapping context (#1014)

Browse Source 
* examples : evaluate tokens in batches after swapping context

* Update examples/main/main.cpp

---------

Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
This commit is contained in:
Alex Klinkhamer 2023-04-21 11:18:09 -07:00 committed by GitHub
parent 8687c1f258
commit 9411288271
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
1 changed files with 13 additions and 5 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

18
examples/main/main.cpp
View File

@ -264,7 +264,7 @@ int main(int argc, char ** argv) {
// infinite text generation via context swapping // infinite text generation via context swapping
// if we run out of context: // if we run out of context:
// - take the n_keep first tokens from the original prompt (via n_past) // - take the n_keep first tokens from the original prompt (via n_past)
// - take half of the last (n_ctx - n_keep) tokens and recompute the logits in a batch // - take half of the last (n_ctx - n_keep) tokens and recompute the logits in batches
if (n_past + (int) embd.size() > n_ctx) { if (n_past + (int) embd.size() > n_ctx) {
const int n_left = n_past - params.n_keep; const int n_left = n_past - params.n_keep;
@ -282,13 +282,21 @@ int main(int argc, char ** argv) {
//printf("\n---\n"); //printf("\n---\n");
} }
if (llama_eval(ctx, embd.data(), embd.size(), n_past, params.n_threads)) { // evaluate tokens in batches
fprintf(stderr, "%s : failed to eval\n", __func__); // embd is typically prepared beforehand to fit within a batch, but not always
return 1; for (int i = 0; i < (int) embd.size(); i += params.n_batch) {
int n_eval = (int) embd.size() - i;
if (n_eval > params.n_batch) {
n_eval = params.n_batch;
}
if (llama_eval(ctx, &embd[i], n_eval, n_past, params.n_threads)) {
fprintf(stderr, "%s : failed to eval\n", __func__);
return 1;
}
n_past += n_eval;
} }
} }
n_past += embd.size();
embd.clear(); embd.clear();
if ((int) embd_inp.size() <= n_consumed && !is_interacting) { if ((int) embd_inp.size() <= n_consumed && !is_interacting) {