gpt3xl-flash-rotary.yaml 186 Bytes
Newer Older
Tri Dao's avatar
Tri Dao committed
1
2
# @package _global_
defaults:
Tri Dao's avatar
Tri Dao committed
3
  - /experiment/pile/gpt3xl-flash.yaml
Tri Dao's avatar
Tri Dao committed
4
5
6
7
8

model:
  config:
    max_position_embeddings: 0  # Disable absolute position embedding
    rotary_emb_fraction: 0.5