[contrib] Add MiMo-V2.5-Pro (Xiaomi, 384 experts MoE, FP8 on Trn2)#150

Open

whn09 wants to merge 24 commits into

aws-neuron:mainfrom

whn09:contrib/MiMo-V2.5-Pro

Commits on Apr 27, 2026

[contrib] Add MiMo-V2.5-Pro initial port (copied from MiMo-V2-Pro, renamed)

whn09
and
claude
committed
[contrib] MiMo-V2.5-Pro: apply review findings from HF reference diff

whn09
and
claude
committed

Commits on Apr 28, 2026

[contrib] MiMo-V2.5-Pro: document current WIP status in README

whn09
and
claude
committed
[contrib] MiMo-V2.5-Pro: correct FP8 root-cause framing in Status

whn09
and
claude
committed
[contrib] MiMo-V2.5-Pro: parallel preprocess + NVMe mount docs

whn09
and
claude
committed
[contrib] MiMo-V2.5-Pro: set AWS Llama-405B FP8 env vars in smoke scripts

whn09
and
claude
committed
[contrib] MiMo-V2.5-Pro: fix tokenizer padding_side='left' for decoder-only LM

whn09
and
claude
committed
[contrib] MiMo-V2.5-Pro: skip FP8 quant on q/k/v_proj (attention path)

whn09
and
claude
committed
[contrib] MiMo-V2.5-Pro: try use_torch_block_wise + restore FP8 q/k/v

whn09
and
claude
committed

Commits on Apr 29, 2026

Commits on Apr 30, 2026

[contrib] MiMo-V2.5-Pro: bump default seq_len 256 -> 512; document vLLM bug

whn09
and
claude
committed