tangledgroup
/

tangled-llama-s-128k-base-v0.1

Model card Files Files and versions Community

mtasic85 commited on Oct 7

Commit

858da05

•

1 Parent(s): b661d1e

pretrain model

Files changed (1) hide show

scripts/pretrain-model.yaml +1 -1

scripts/pretrain-model.yaml CHANGED Viewed

@@ -67,7 +67,7 @@ train:
   # Number of samples per data-parallel rank (type: int, default: 4)
   # micro_batch_size: 16
-  micro_batch_size: 13
   # Number of iterations with learning rate warmup active (type: int, default: 2000)
   lr_warmup_steps: 2000

   # Number of samples per data-parallel rank (type: int, default: 4)
   # micro_batch_size: 16
+  micro_batch_size: 12
   # Number of iterations with learning rate warmup active (type: int, default: 2000)
   lr_warmup_steps: 2000