issues
search
pjlab-sys4nlp
/
llama-moe
⛷️ LLaMA-MoE: Building Mixture-of-Experts from LLaMA with Continual Pre-training
https://arxiv.org/abs/2406.16554
Apache License 2.0
849
stars
44
forks
source link
CPT: add eval support
#35
Closed
Spico197
closed
11 months ago
Spico197
commented
11 months ago
tb: add TGS, MFU, update load&importance logging step
add eval support, fix balance_loss = None bug during eval grad ckpt
update logging strategy