MolecularDiffusion.cli.generate¶

Generation command for MolCraft CLI.

Adapted from scripts/generate.py for package-level execution.

Attributes¶

log

Functions¶

`generate`(→ Tuple[Dict[str, Any], Dict[str, Any]])	Main generation function.
`generate_main`(cfg)	Entry point for CLI generate command.
`is_rank_zero`()	Check if current process is rank zero.
`load_lightning_model`(chkpt_path, task_config[, ...])	Load model from Lightning checkpoint (.ckpt).
`load_model`(chkpt_directory[, task_config, atom_vocab, ...])	Load model from checkpoint directory with auto-detection.
`log_hyperparameters`(object_dict)	Log hyperparameters for debugging.

Module Contents¶

MolecularDiffusion.cli.generate.generate(cfg: omegaconf.DictConfig) → Tuple[Dict[str, Any], Dict[str, Any]]¶: Main generation function.

MolecularDiffusion.cli.generate.generate_main(cfg: omegaconf.DictConfig)¶: Entry point for CLI generate command.

MolecularDiffusion.cli.generate.is_rank_zero()¶: Check if current process is rank zero.

MolecularDiffusion.cli.generate.load_lightning_model(chkpt_path, task_config, atom_vocab=None, total_step=0)¶: Load model from Lightning checkpoint (.ckpt).

MolecularDiffusion.cli.generate.load_model(chkpt_directory, task_config=None, atom_vocab=None, total_step=0, base_chkpt_path=None)¶

Load model from checkpoint directory with auto-detection.

If the selected checkpoint is a LoRA-only delta (marked lora_only), base_chkpt_path (the pretrained checkpoint, file or directory) is required and the model is reconstructed as base weights + LoRA delta.

MolecularDiffusion.cli.generate.log_hyperparameters(object_dict: dict)¶: Log hyperparameters for debugging.

MolecularDiffusion.cli.generate.log¶