Skip to content

Fix attention_block_tkg d_head>128 decode Q/K layout corruption for batch>1 - #14

Open
jimburtoft wants to merge 1 commit into
aws-neuron:mainfrom
jimburtoft:fix/attention-block-tkg-d256-multibatch-layout
Open

jimburtoft wants to merge 1 commit into
aws-neuron:mainfrom
jimburtoft:fix/attention-block-tkg-d256-multibatch-layout

Commits

Commits on Sep 1, 2026