flax
[linen] enable separate initializers for out layer in MultiHeadDotProductAttention
#3835
Merged

[linen] enable separate initializers for out layer in MultiHeadDotProductAttention #3835

cgarciae
cgarciae cgarciae changed the title [linen] add individual param initializer in MultiHeadDotProductAttention [linen] expose individual param initializer in MultiHeadDotProductAttention 1 year ago
cgarciae cgarciae force pushed 1 year ago
cgarciae cgarciae force pushed 1 year ago
cgarciae cgarciae force pushed 1 year ago
cgarciae cgarciae force pushed 1 year ago
cgarciae cgarciae changed the title [linen] expose individual param initializer in MultiHeadDotProductAttention [linen] enable separate initializer for out layer in MultiHeadDotProductAttention 1 year ago
cgarciae cgarciae changed the title [linen] enable separate initializer for out layer in MultiHeadDotProductAttention [linen] enable separate initializers for out layer in MultiHeadDotProductAttention 1 year ago
cgarciae cgarciae force pushed 1 year ago
chiamp
chiamp commented on 2024-04-10
chiamp
chiamp approved these changes on 2024-04-10
chiamp chiamp added pull ready
cgarciae [linen] add individual param initializer in MultiHeadDotProductAttention
43d022a8
cgarciae cgarciae force pushed to 43d022a8 1 year ago
cgarciae cgarciae removed pull ready
cgarciae cgarciae added pull ready
cgarciae cgarciae removed pull ready
cgarciae cgarciae added pull ready
copybara-service copybara-service merged e0fa96fe into main 1 year ago
copybara-service copybara-service deleted the linen-attention-multiple-initializers branch 1 year ago

Login to write a write a comment.

Login via GitHub

Reviewers
Assignees
No one assigned
Labels
Milestone