[DRAFT] Support for Zaya1 8B model (depends on PR #22833)#23112

Draft

Juste-Leo2 wants to merge 33 commits into

ggml-org:masterfrom

Juste-Leo2:Zaya1

Commits on May 8, 2026

Commits on May 11, 2026

add corrections
Juste-Leo2
committed

Commits on May 12, 2026

zaya generation running
Ganesh Nanduru
committed

Commits on May 15, 2026

refactor: replace CCA_CONV_DW with generic SSM_CONV1D constant
Juste-Leo2
committed
refactor: replace ZAYA_ROUTER_NORM with generic FFN_NORM constant
Juste-Leo2
committed
refactor: replace ZAYA_ROUTER_DOWN with generic FFN_GATE_INP constant
Juste-Leo2
committed
refactor: replace ZAYA_ROUTER_MLP0 with generic FFN_GATE constant
Juste-Leo2
committed
refactor: merge RES_SCALE_*_B bias constants into RES_SCALE_* constants
Juste-Leo2
committed
refactor: merge router bias constants into parent constants
Juste-Leo2
committed
zaya: remove unused CCA_QK_NORM tensor constant
Juste-Leo2
committed
zaya: remove dead ZAYA_ROUTER_MLP2 mapping from non-block config
Juste-Leo2
committed
zaya: revert unrelated debug.cpp changes
Juste-Leo2
committed
zaya: replace hardcoded n_ff_exp with GGUF metadata
Juste-Leo2
committed
zaya: fix val_proj dimensions to use n_embd_k / 2 instead of n_embd_head
Juste-Leo2
committed

Commits on May 16, 2026

Commits on May 17, 2026

zaya: fix server compatibility with batched inference
Juste-Leo2
committed

Commits on May 18, 2026

fix(zaya): use actual tokenizer vocab size instead of config vocab_size
Juste-Leo2
committed

Commits on May 25, 2026