refactor(ops): route basic Llama through canonical InfiniOps - #1483

Merged

voltjia merged 10 commits into

mainfrom

refactor/remove-infinilm-infiniops-adapters

Aug 11, 2026

Commits on Aug 9, 2026

refactor(ops): migrate paged attention decode to canonical InfiniOps
voltjia
committed
4f3fca7
refactor(ops): migrate paged attention prefill to canonical InfiniOps
voltjia
committed
bf24faa

Commits on Aug 10, 2026

refactor(ops): remove InfiniLM-suffixed InfiniOps calls
voltjia
committed
98432df
refactor(ops): use canonical KV-cache attention
voltjia
committed
0172746
refactor(nn): use canonical RotaryEmbedding for RoPE
voltjia
committed
a8006d1
refactor(ops): use Argmax for greedy sampling
voltjia
committed
e5fbef9
refactor(ops): use canonical FusedAddRmsNorm
voltjia
committed
44ef940
refactor(ops): use canonical SiluAndMul
voltjia
committed
2f4973b
style(ops): format rotary embedding with clang-format 16
voltjia
committed
6609f20

Commits on Aug 11, 2026

refactor(ops): use PyTorch Argmax provider
voltjia
committed
a69708c