fresco package

Subpackages

• fresco.data_loaders package
  • Submodules
  • fresco.data_loaders.data_utils module
    • AddNoise
    • DataHandler
      • DataHandler.data_source
      • DataHandler.model_args
      • DataHandler.convert_y()
      • DataHandler.get_vocab()
      • DataHandler.inference_loader()
      • DataHandler.load_folds()
      • DataHandler.load_from_saved()
      • DataHandler.load_weights()
      • DataHandler.make_grouped_cases()
      • DataHandler.make_torch_dataloaders()
      • DataHandler.seed_worker()
    • GroupedCases
    • LabelDict
    • PathReports
    • word2int()
  • Module contents
• fresco.models package
  • Submodules
  • fresco.models.clc module
    • CaseLevelContext
      • CaseLevelContext.forward()
  • fresco.models.mtcnn module
    • MTCNN
      • MTCNN.forward()
  • fresco.models.mthisan module
    • MTHiSAN
      • MTHiSAN.forward()
  • Module contents
• fresco.training package
  • Submodules
  • fresco.training.training module
    • ModelTrainer
      • ModelTrainer.savepath
      • ModelTrainer.epochs
      • ModelTrainer.patience_stop
      • ModelTrainer.tasks
      • ModelTrainer.n_task
      • ModelTrainer.model
      • ModelTrainer.dw
      • ModelTrainer.device
      • ModelTrainer.best_loss
      • ModelTrainer.patience_ctr
      • ModelTrainer.loss
      • ModelTrainer.y_preds
      • ModelTrainer.y_trues
      • ModelTrainer.multilabel
      • ModelTrainer.abstain
      • ModelTrainer.mixed_precision
      • ModelTrainer.opt
      • ModelTrainer.reduction
      • ModelTrainer.loss_funs
      • ModelTrainer.class_weights
      • ModelTrainer.compute_clc_loss()
      • ModelTrainer.compute_clc_val_loss()
      • ModelTrainer.compute_loss()
      • ModelTrainer.compute_scores()
      • ModelTrainer.compute_val_loss()
      • ModelTrainer.fit_model()
      • ModelTrainer.get_ys()
      • ModelTrainer.output_scores()
      • ModelTrainer.print_stats()
      • ModelTrainer.profile_fit_model()
      • ModelTrainer.score()
      • ModelTrainer.stop_metrics()
      • ModelTrainer.train_metrics()
    • trace_handler()
  • Module contents
• fresco.validate package
  • Submodules
  • fresco.validate.exceptions module
    • ParamError
  • fresco.validate.validate_params module
    • ValidateClcParams
      • ValidateClcParams.check_abstain_args()
      • ValidateClcParams.check_data_files()
      • ValidateClcParams.check_data_train_args()
      • ValidateClcParams.clc_arg_check()
    • ValidateParams
      • ValidateParams.check_abstain_args()
      • ValidateParams.check_data_files()
      • ValidateParams.check_data_train_args()
      • ValidateParams.check_keyword_args()
      • ValidateParams.check_weights()
      • ValidateParams.hisan_arg_check()
      • ValidateParams.mtcnn_arg_check()
  • Module contents

Submodules

fresco.data_setup module

fresco.run_clc module

Top-level clc model building script using independent modules.

fresco.run_clc.create_doc_embeddings(model, model_type, data_loader, device)

Generate document embeddings from trained model.

fresco.run_clc.create_model(params, dw, device)

Define a model based on model_args.

Parameters:

• params (dict) – Dictionary of model_args file.

• dw (DataHandler) – DataHandler class object.

• device (torch.device) – Torch device, either ‘cpu’ or ‘cuda’.

Returns:

A model.

fresco.run_clc.load_model(model_dict, device, dw)

fresco.run_clc.load_model_dict(model_path, data_path='')

Load pretrained model from disk.

Parameters:

• model_path (str) – Path to the saved model from command line args.

• valid_params (ValidateParams) – ValidateParams class object with model_args dict.

• data_path (str or None) – Path to data from the trained model, or None.

We check if the supplied path is valid and if the packages match needed to run the pretrained model.

fresco.run_clc.main()

fresco.run_clc.run_case_level(args)

fresco.run_clc.save_full_model(kw_args, save_path, packages, fold)

Save trained model with package info metadata.

fresco.run_ie module

Top-level information extraction model building script using independent modules.

fresco.run_ie.create_model(params, dw, device)

Define a model based on model_args.

Parameters:

• params (dict) – Dictionary of model_args file.

• dw (DataHandler) – DataHandler class object.

• device (torch.device) – Torch device, either ‘cpu’ or ‘cuda’.

Returns:

A model.

fresco.run_ie.load_model(model_path, device, dw, valid_params)

Load a pretrained model from disk.

Parameters:

• model_path (str) – Path to the saved model obtained from command line args.

• device (torch.device) – Torch device, either ‘cuda’ or ‘cpu’, set in the calling function.

• dw (DataHandler) – DataHandler class instance for wrangling data and checking metadata.

• valid_params (ValidateParams) – ValidateParams class instance with model_args dict.

Post-condition:

The saved model is loaded and set to eval() mode.

Returns:

None

fresco.run_ie.main()

fresco.run_ie.run_ie(args=None)

fresco.run_ie.save_full_model(kw_args, save_path, packages, fold)

Save trained model with package info metadata.

Module contents