Loggers

Loggers

class torchmimic.loggers.DecompensationLogger(exp_name, config, log_wandb=False)

Decopensation Logger class. Used for logging, printing, and saving information about the run. Logs AUC-ROC and AUC-PR. Contains built-in wandb support.

Parameters
  • config (dict) – A dictionary of the run configuration

  • log_wandb (bool) – If true, wandb will be used to log metrics and configuration

get_loss()

Returns average loss

Returns

Average Loss

Return type

float

print_metrics(epoch, split='Train')

Prints and logs metrics. If log_wandb is True, wandb run will be updated

Parameters
  • epoch (int) – The current epoch

  • split (str) – The split of the data. “Train” or “Eval”

reset()

Resets metrics

save(model)

Saves the provides models to the experiment path

update(outputs, labels, loss)

Update Loss, AUC-ROC, and AUC-PR

Parameters
  • outputs (torch.Tensor) – Predicted labels

  • labels (torch.Tensor) – True labels

  • loss (float) – Loss from the training iteration.

class torchmimic.loggers.IHMLogger(exp_name, config, log_wandb=False)

In-Hospital-Mortality Logger class. Used for logging, printing, and saving information about the run. Logs AUC-ROC and AUC-PR. Contains built-in wandb support.

Parameters
  • config (dict) – A dictionary of the run configuration

  • log_wandb (bool) – If true, wandb will be used to log metrics and configuration

get_loss()

Returns average loss

Returns

Average Loss

Return type

float

print_metrics(epoch, split='Train')

Prints and logs metrics. If log_wandb is True, wandb run will be updated

Parameters
  • epoch (int) – The current epoch

  • split (str) – The split of the data. “Train” or “Eval”

reset()

Resets metrics

save(model)

Saves the provides models to the experiment path

update(outputs, labels, loss)

Update Loss, AUC-ROC, and AUC-PR

Parameters
  • outputs (torch.Tensor) – Predicted labels

  • labels (torch.Tensor) – True labels

  • loss (float) – Loss from the training iteration.

class torchmimic.loggers.LOSLogger(exp_name, config, log_wandb=False)

Length-of-Stay Logger class. Used for logging, printing, and saving information about the run. Logs loss, Cohen’s Kappa and Mean Absolute Deviation. Contains built-in wandb support.

Parameters
  • config (dict) – A dictionary of the run configuration

  • log_wandb (bool) – If true, wandb will be used to log metrics and configuration

get_loss()

Returns average loss

Returns

Average Loss

Return type

float

print_metrics(epoch, split='Train')

Prints and logs metrics. If log_wandb is True, wandb run will be updated

Parameters
  • epoch (int) – The current epoch

  • split (str) – The split of the data. “Train” or “Eval”

reset()

Resets metrics

save(model)

Saves the provides models to the experiment path

update(outputs, labels, loss)

Update loss, Cohen’s Kappa and Mean Absolute Deviation

Parameters
  • outputs (torch.Tensor) – Predicted labels

  • labels (torch.Tensor) – True labels

  • loss (float) – Loss from the training iteration.

class torchmimic.loggers.PhenotypingLogger(exp_name, config, log_wandb=False)

Phenotyping Logger class. Used for logging, printing, and saving information about the run. Logs loss, AUC-ROC macro, and AUC-ROC micro. Contains built-in wandb support.

Parameters
  • config (dict) – A dictionary of the run configuration

  • log_wandb (bool) – If true, wandb will be used to log metrics and configuration

get_loss()

Returns average loss

Returns

Average Loss

Return type

float

print_metrics(epoch, split='Train')

Prints and logs metrics. If log_wandb is True, wandb run will be updated

Parameters
  • epoch (int) – The current epoch

  • split (str) – The split of the data. “Train” or “Eval”

reset()

Resets metrics

save(model)

Saves the provides models to the experiment path

update(outputs, labels, loss)

Update loss, AUC-ROC macro, and AUC-ROC micro

Parameters
  • outputs (torch.Tensor) – Predicted labels

  • labels (torch.Tensor) – True labels

  • loss (float) – Loss from the training iteration.