Try lora run higher LR

This commit is contained in:
Jake Poznanski 2024-10-17 17:12:35 +00:00
parent 2826bcad18
commit e141c91e5e

View File

@ -46,7 +46,7 @@ hparams:
gradient_accumulation_steps: 4
gradient_checkpointing: true
clip_grad_norm: 1.0
learning_rate: 1e-4
learning_rate: 3e-4
max_steps: 10000
pad_multiple_of: 16
log_every_steps: 10