gradient accumulation

2026-06-18 16:25:52 +02:00
parent 24327d7ee0
commit 772acd8f0d
2 changed files with 9 additions and 4 deletions
@@ -9,7 +9,8 @@ grid:
  optimization.weight_decay: [5.0e-4, 0]
  optimization.momentum: [0.9]
  optimization.lr_schedule: [cosine]
-  data.batch_size: [512, 1024]
+  optimization.gradient_accumulation_steps: [8, 16]
+  data.batch_size: [64]
  meta.representation_type: [last_avgpool, last4_avgpool_concat]
  meta.head_type: [linear, bn_linear]