wandb_utils.misc

Module Contents

wandb_utils.misc.logger

wandb_utils.misc.to_csv(df: pandas.DataFrame) → str

wandb_utils.misc.write_df(df: pandas.DataFrame, output_file: Optional[pathlib.Path], skip_writing: bool) → None

wandb_utils.misc.read_df(path: pathlib.Path, sep: str = '\t') → pandas.DataFrame

wandb_utils.misc.all_data_df(entity: str, project: str, sweep: Optional[str] = None, api: Optional[wandb.apis.public.Api] = None, filters: Optional[Dict] = None) → pandas.DataFrame

Get the data for all the runs.

wandb_utils.misc.find_best_models_in_sweeps(entity: str, project: str, metric: str, maximum: bool = True, sweep: Optional[str] = None, api: Optional[wandb.apis.public.Api] = None) → pandas.DataFrame

wandb_utils.misc.get_config_file_for_run(entity: str, project: str, run_id: str, relative_path: str = 'training_dumps/config.json', output_path: str = 'config.json', api: Optional[wandb.apis.public.Api] = None) → wandb.apis.public.Api

wandb_utils.misc.get_config_file_for_best_run(entity: str, project: str, sweep_id: str, metric: str, maximum: bool = True, relative_path: str = 'training_dumps/config.json', output_path: str = 'config.json', api: Optional[wandb.apis.public.Api] = None) → str

wandb_utils.misc.multiple_runs_sweep

wandb_utils.misc.jsonnet_with_seed_template

wandb_utils.misc.with_fallback(preferred: Dict[str, Any], fallback: Dict[str, Any]) → Dict[str, Any]

Deep merge two dicts, preferring values from preferred. Ref: allennlp/common/params.py

wandb_utils.misc.create_multiple_run_sweep_for_run(entity: str, project: str, run: Optional[str] = None, sweep: Optional[str] = None, metric: Optional[str] = None, maximum: bool = True, relative_path: str = 'training_dumps/config.json', output_path: str = 'config.json', seed_parameters: Optional[List[str]] = None, api: Optional[wandb.apis.public.Api] = None, delete_keys: Optional[List] = None, **sweep_args)

Example:

create_multiple_run_sweep_for_run('iesl-boxes','multilabel-learning-datasets',
                          run='nyre9qr5',
                          sweep_args=dict(include_packages=['multilabel_learning'],
                                          sweep_name='test_sweep',
                                          wandb_tags=['test', 'dryrun'],
                                          fixed_overrides='{"type": "train_test_log_to_wandb"}'
                                         )
                         )