Michelle0474 0fc72a9474 Add Qwen2.5-7B MolRAG encoder branch + dual-seed results ...

- llm_prompt.py: RAG prompt encoding (use_rag), Qwen support, set_retrieval_pool (leak-safe, train-only pool)
- models.py: use_rag/rag_top_k passthrough to LLMPromptEncoder
- nested_cv_optuna.py: --use-rag CLI, per-fold retrieval pool (防泄漏), slim checkpoint (skip frozen LLM weights)
- Qwen+RAG results: seed42 delivery R2=0.2151, seed123=0.1558 (mean 0.185), best LLM encoder so far

2026-06-28 20:42:27 +08:00

..

abl

Add 6-variant ablation results (baseline/moe/llm_molt5/llm_biot5/both_molt5/both_biot5)

2026-06-24 15:33:07 +08:00

abl_qwen_rag

Add Qwen2.5-7B MolRAG encoder branch + dual-seed results

2026-06-28 20:42:27 +08:00

benchmark

Final Model

2026-03-02 14:51:54 +08:00

final

Final Model

2026-03-02 14:51:54 +08:00

mpnn/all_amine_split_for_LiON

Update models and UI

2026-02-11 16:49:28 +08:00

nested_cv

Final Model

2026-03-02 14:51:54 +08:00

pretrain

feat(moe): MoE + MolT5 LLM 融合层、消融流水线

2026-06-18 08:51:30 +00:00

.gitkeep

Initial commit

2026-01-18 21:50:21 +08:00

history.json

Add CV

2026-01-21 19:35:55 +08:00

pretrain_delivery.pt

Final Model

2026-03-02 14:51:54 +08:00

pretrain_history.json

Final Model

2026-03-02 14:51:54 +08:00

pretrain_loss_curves.png

Final Model

2026-03-02 14:51:54 +08:00