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

perplexity: more statistics, added documentation (#6936)

Browse source 
* perplexity: more statistics, added documentation

* add LLaMA 3 8b scoreboard
This commit is contained in:
Johannes Gäßler 2024-04-30 23:36:27 +02:00 • committed by GitHub
parent f364eb6fb5
commit a8f9b07631
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
3 changed files with 291 additions and 59 deletions
Download patch file Download diff file Expand all files Collapse all files

2
common/common.h
View file

@ -135,7 +135,7 @@ struct gpt_params {
bool multiple_choice = false; // compute TruthfulQA score over random tasks from datafile supplied in prompt
size_t multiple_choice_tasks = 0; // number of tasks to use when computing the TruthfulQA score. If 0, all tasks will be computed
bool kl_divergence = false; // compute KL-divergence
bool kl_divergence = false; // compute KL divergence
bool random_prompt = false; // do not randomize prompt if none provided
bool use_color = false; // use color to distinguish generations and inputs