Skip to content

v0.3.1-cpt-dynamic_batch_loading: Llama2 CPT with Dynamic Batch Loading

Compare
Choose a tag to compare
@Spico197 Spico197 released this 17 Nov 09:09
· 51 commits to main since this release
d6a3780
  • Llama2 CPT with 4096 context length training.
  • Dynamic batch loading from ShearedLlama Implementation.