All issues

Search results

[Code scan] Do not require save_path for in-memory target scaling

Status: Open.

#39 In deepmodeling/unimol_tools;

· njzjz opened on Jun 30, 2026

[Code scan] Aggregate pretraining DDP validation metrics across ranks

Status: Open.

#38 In deepmodeling/unimol_tools;

· njzjz opened on Jun 30, 2026

[Code scan] Reload best checkpoint for DDP validation predictions

Status: Open.

#37 In deepmodeling/unimol_tools;

· njzjz opened on Jun 30, 2026

[Code scan] Use selected GPU count for DDP world size and spawning

Status: Open.

#36 In deepmodeling/unimol_tools;

· njzjz opened on Jun 30, 2026

[Code scan] Forward MolTrainHF extra training options into config

Status: Open.

#35 In deepmodeling/unimol_tools;

· njzjz opened on Jun 30, 2026

[Code scan] Restore MolPredictHF multiclass and save_path behavior

Status: Open.

#34 In deepmodeling/unimol_tools;

· njzjz opened on Jun 30, 2026

[Code scan] Make UniMol tokenizer save and load self-contained

Status: Open.

#33 In deepmodeling/unimol_tools;

· njzjz opened on Jun 30, 2026

[Code scan] Fix Hugging Face from_pretrained path handling

Status: Open.

#32 In deepmodeling/unimol_tools;

· njzjz opened on Jun 30, 2026

[Code scan] Use dictionary special-token ids for UniMol v1 atomic representations

Status: Open.

#31 In deepmodeling/unimol_tools;

· njzjz opened on Jun 30, 2026

[Code scan] Respect CPU fallback before moving pretraining models to CUDA

Status: Open.

#30 In deepmodeling/unimol_tools;

· njzjz opened on Jun 30, 2026

[Code scan] Handle no-padding pretraining batches in coordinate prediction

Status: Open.

#29 In deepmodeling/unimol_tools;

· njzjz opened on Jun 30, 2026

[Code scan] Guarantee masked atoms or skip empty pretraining masked losses

Status: Open.

#28 In deepmodeling/unimol_tools;

· njzjz opened on Jun 30, 2026