[gemma4] Fix torchax-path E2B/E4B accuracy: inline PLE + skip KV-cache writes for KV-shared layers (#3406)
Signed-off-by: Qiliang Cui <cuiq@google.com>
Q
Qiliang Cui committed
ee1e17a124488cd5ec27f4ded8b757e5507fedef
Parent: 07b666a
Committed by GitHub <noreply@github.com>
on 8/21/2026, 5:24:50 PM