[faresobeid]

Removes MiniMax M2 (minimax_m2) support, and drops the dense GLM-4-9B-0414 mention from the model tables. No back-compat shims: a MiniMax M2 checkpoint now fails at trainer setup like any other unsupported architecture.

Removed

• trainer/models/minimax_m2/, configs/advanced/minimax-m2.5/, and links in configs/README.md, examples/README.md, docs/configuration.md.
• inference/patches.py: monkey_patch_minimax_m2_think_end_passthrough and monkey_patch_minimax_m2_for_lora.
• MiniMaxAI/MiniMax-M2 tool/reasoning-parser auto-detection (minimax_m2, minimax_m2_append_think), plus its parser test rows. Without the think-end patch, the append_think serving path would silently lose </think>, so both go together.
• minimax_m2 preset in scripts/mini_moe.py; MiniMax case in test_moe_conversions.py.
• Registry entries, README / docs/advanced.md rows.

GLM-4-9B-0414

README and docs/advanced.md now list glm4_moe as GLM-4.5 MoE / INTELLECT-3 only. The glm4_moe mini model in scripts/mini_moe.py (and its docs/development.md example) takes its tokenizer from zai-org/GLM-4.5-Air instead of THUDM/GLM-4-9B-0414. The preset's pad/eos ids (151329 <|endoftext|>, 151336 <|user|>, 151338 <|observation|>) are the same in that tokenizer, and its 151365 tokens fit the preset's vocab_size=151552.

Checks

• ruff check . and ruff format --check .: clean.
• pytest -m "not gpu" tests/unit/orchestrator tests/unit/test_configs.py tests/unit/test_parsers.py tests/unit/train tests/unit/inference: all pass except test_cpu_adam.py::test_native_cpu_adamw_matches_fused_torch_and_preserves_gradients, which fails the same way on main on this CPU box.
• CPU forward + backward check for qwen3_moe, glm4_moe and laguna (tiny configs, fixed seeds, packed seq_lens, SDPA stand-in for the flash kernel). Logits, all parameter grads and state-dict keys are bitwise identical to main. DeepSeek-V4 needs GPU kernels, so it was not run.
• grep across src/tests/configs/examples/docs/skills/scripts/.github: no references left (except the renderer list in docs/algorithms.md, which describes the external renderers package).
• A checkpoint of this family now fails at trainer setup with the usual "has no PrimeRL model implementation" error. No shim.

Conflicts with open PRs

• Chore/remove transformers modeling #3833 (samsja): edits minimax_m2/* and scripts/mini_moe.py. Drop the minimax hunks.
• refactor(models): one shared init_buffers_post_meta (stacked on #3833) #3857: edits minimax_m2/modeling_minimax_m2.py. Drop the hunk.
• feat(inference): router replay on the V2 runner everywhere #3830–feat(inference): P/D per-role KV offload #3832, refactor(lora): single-adapter LoRA layers, drop modules_to_save #3855: also edit inference/patches.py. Only the call list in apply_shared_vllm_patches should conflict.