Created ReplicateKVHeadTransform to integrate KV-heads replication module within Qefficient library.#625

Closed

quic-dhirajku wants to merge 6 commits into

quic-dhirajku:replicate_kv_heads_transform

Commits on Jan 21, 2026

Created ReplicateKVHeadTransform to integrate KV-heads replication module within Qefficient library.
quic-dhirajku
committed
Modified modeling_auto to use 1 as default value of n_kv_head_repeat. Doing so would prevent any issues during Transforms when we don't wish to apply it.
quic-dhirajku
committed
Minor patch to handle custom config case
quic-dhirajku
committed
Updated list of model architectures on which we can apply this transform.
quic-dhirajku
committed
minor fix for ruff check
quic-dhirajku
committed
Renamed n_kv_head_repeat to num_kv_heads_repeat. Added the transform changes to repeat Bias factor appropriately on quantized layers.
quic-dhirajku
committed