mirror of https://github.com/razor-ai/soup.git
5e0872b9ea
Extends the v0.39.0 ReLoRA callback (Part B) and surgical PEFT patches (Part D — Gemma4 ClippableLinear swap + 3-D fused-MoE expert dropout strip) from SFT-only to all 11 non-SFT transformer-backend trainers (DPO, GRPO, KTO, ORPO, SimPO, IPO, PPO, RewardModel, Pretrain, Embedding, BCO). - New shared helper soup_cli/utils/peft_wiring.py exposes apply_pre_lora_patches, apply_post_lora_patches, attach_relora_callback. - SFT migrated to the same helpers in the same release (centralisation invariant; no drift between SFT and non-SFT wiring). - SoupConfig._validate_relora_supported_tasks: task != "sft" rejection removed; MLX backend still rejected with distinct message. Review fixes: - attach_relora_callback uses `if relora_steps is None:` (project policy) so a schema-bypassing relora_steps=0 surfaces as a loud ReLoRAPolicy ValueError rather than a silent skip. - Direct attribute access on tcfg.relora_warmup_ratio / _reset_optimizer / _prune_ratio (Pydantic schema guarantees them); no getattr defaults. - 11 behavioural helper tests (Gemma4 happy path + exception swallow, post-LoRA strip happy + exception swallow, ReLoRA policy field forwarding, schema-bypass loud-fail). - Schema-gate matrix covers `task='preference'` dispatcher. Tests: 5061 -> 5122 (+61). Closes #67. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> |
||
|---|---|---|
| .. | ||
| __init__.py | ||
| bco.py | ||
| dpo.py | ||
| embedding.py | ||
| grpo.py | ||
| ipo.py | ||
| kto.py | ||
| mlx_dpo.py | ||
| mlx_grpo.py | ||
| mlx_routing.py | ||
| mlx_sft.py | ||
| orpo.py | ||
| ppo.py | ||
| preference.py | ||
| pretrain.py | ||
| reward_model.py | ||
| rewards.py | ||
| sft.py | ||
| simpo.py | ||