mixmo.learners.learner.Learner¶

class mixmo.learners.learner.Learner(config_args, dloader, device)[source]¶

Learner object that defines the specific train and test loops for the model

__init__(config_args, dloader, device)¶: Initialize self. See help(type(self)) for accurate signature.

Methods

`__init__`(config_args, dloader, device)	Initialize self.
`evaluate`(inference_loader[, split])	Perform an evaluation of the model
`evaluate_loop`(inference_loader)	Evaluation loop over the dataset specified by the loader
`load_checkpoint`(checkpoint[, …])	Load checkpoint (and optimizer if included) to the wrapped model
`save_checkpoint`(epoch[, save_path])	Save model (and optimizer) state dict
`save_tb`(logs_dict, epoch)	Write stats from logs_dict at epoch to the Tensoboard summary writer
`train`(epoch)	Train for one epoch
`train_loop`(epoch)	Training loop for one epoch

Attributes

tb_logger

Get (or initialize) the Tensorboard SummaryWriter

_prepare_batch_test(data)[source]¶: Prepares the test batch by setting up the input dictionary and putting tensors on devices

_prepare_batch_train(data)[source]¶: Prepares the train batch by setting up the input dictionary and putting tensors on devices

_subloop(dict_tensors, backprop)[source]¶: Basic subloop for a step/batch (without optimization)

_train_subloop(dict_tensors)[source]¶: Complete training step for a batch, return summary logs

evaluate_loop(inference_loader)[source]¶: Evaluation loop over the dataset specified by the loader