Add Qwen3.5 MTP AutoDeploy support#252

Draft

govind-ramnarayan wants to merge 22 commits into

gk/switch_to_ir_modelsfrom

gramnarayan/qwen3-mtp

Commits on May 10, 2026

[None][feat][autodeploy] Switch deepseek/nemotron_h/qwen3/qwen3_5_moe to sharding-IR canonical
greg-kwasniewski1
committed
[None][fix][autodeploy] Restore AD_USE_IR_MODELS hook for future _ir.py staging
greg-kwasniewski1
committed
[None][test][autodeploy] Cover torch_rmsnorm_gated in NemotronHMamba2Mixer pattern test
greg-kwasniewski1
committed
[None][fix][autodeploy] Walk past all_reduce in hidden-states residual detector
greg-kwasniewski1
committed
[None][fix][autodeploy] Use relative custom_ops import to satisfy import-discipline check
greg-kwasniewski1
committed
[None][fix][autodeploy] Move enable_sharder_ir.yaml into configs/ to match registry search path
greg-kwasniewski1
committed
[None][feat][autodeploy] Make IR sharding the default for models using the 4 IR-aware canonical files
greg-kwasniewski1
committed
[None][fix][autodeploy] Make modeling_qwen3.py compatible with transformers 5.x
greg-kwasniewski1
committed

Commits on May 13, 2026

Merge branch 'main' into gk/switch_to_ir_models
suyoggupta
authored

Commits on May 14, 2026

Commits on May 15, 2026

Merge branch 'main' into gk/switch_to_ir_models
suyoggupta
authored

Commits on May 18, 2026

Commits on May 21, 2026