Skip to content

Commit d66f16c

Browse files
committed
fix fsdp2d ds logical rule
1 parent 4799cef commit d66f16c

1 file changed

Lines changed: 2 additions & 0 deletions

File tree

src/MaxText/configs/models/deepseek3-671b-2dfsdp.yml

Lines changed: 2 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -61,6 +61,8 @@ data_sharding: [['data', 'fsdp', 'fsdp_transpose', 'expert', 'context']]
6161
logical_axis_rules: [
6262
['activation_batch', ['data', 'fsdp', 'fsdp_transpose', 'expert', 'context']],
6363
['activation_kv_batch', ['data', 'fsdp', 'fsdp_transpose', 'expert', 'context']],
64+
['activation_embed_and_logits_batch', ['data', 'fsdp', 'fsdp_transpose', 'expert']],
65+
['activation_norm_length', ['context']],
6466
['activation_heads', []],
6567
['embed', ['fsdp']],
6668
['embed_no_exp', ['fsdp']],

0 commit comments

Comments
 (0)