mirrors/cosmopolitan
1
0
Fork 0
mirror of https://github.com/jart/cosmopolitan.git synced 2025-10-11 20:38:07 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 1

Make shell usability improvements to llama.cpp

Browse source 
- Introduce -v and --verbose flags
- Don't print stats / diagnostics unless -v is passed
- Reduce --top_p default from 0.95 to 0.70
- Change --reverse-prompt to no longer imply --interactive
- Permit --reverse-prompt specifying custom EOS if non-interactive
This commit is contained in:
Justine Tunney 2023-04-28 02:54:11 -07:00
parent 420f889ac3
commit 1c2da3a55a
No known key found for this signature in database
GPG key ID: BE714B4575D6E328
6 changed files with 103 additions and 55 deletions
Download patch file Download diff file Expand all files Collapse all files

3
third_party/ggml/common.h vendored
View file

@ -17,6 +17,7 @@
struct gpt_params {
int32_t seed = -1; // RNG seed
int32_t verbose = 0; // Logging verbosity
int32_t n_threads = std::min(4, (int32_t) std::thread::hardware_concurrency());
int32_t n_predict = 128; // new tokens to predict
int32_t repeat_last_n = 64; // last n tokens to penalize
@ -27,7 +28,7 @@ struct gpt_params {
// sampling parameters
int32_t top_k = 40;
float top_p = 0.95f;
float top_p = 0.70f;
float temp = 0.80f;
float repeat_penalty = 1.10f;