mattertune.finetune.loss

Functions

compute_loss(config, prediction, label)

Compute the loss value given the model output, prediction, and the target label, label.

l2_mae_loss(output, target[, reduction])

Classes

HuberLossConfig(*[, name, delta, reduction])

L2MAELossConfig(*[, name, reduction])

MAELossConfig(*[, name, reduction])

MSELossConfig(*[, name, reduction])

class mattertune.finetune.loss.MAELossConfig(*, name='mae', reduction='mean')[source]
Parameters:

  • name (Literal['mae'])

  • reduction (Literal['mean', 'sum'])

name: Literal['mae']
reduction: Literal['mean', 'sum']

How to reduce the loss values across the batch.

  • "mean": The mean of the loss values.

  • "sum": The sum of the loss values.

class mattertune.finetune.loss.MSELossConfig(*, name='mse', reduction='mean')[source]
Parameters:

  • name (Literal['mse'])

  • reduction (Literal['mean', 'sum'])

name: Literal['mse']
reduction: Literal['mean', 'sum']

How to reduce the loss values across the batch.

  • "mean": The mean of the loss values.

  • "sum": The sum of the loss values.

class mattertune.finetune.loss.HuberLossConfig(*, name='huber', delta=1.0, reduction='mean')[source]
Parameters:

  • name (Literal['huber'])

  • delta (float)

  • reduction (Literal['mean', 'sum'])

name: Literal['huber']
delta: float

The threshold value for the Huber loss function.

reduction: Literal['mean', 'sum']

How to reduce the loss values across the batch.

  • "mean": The mean of the loss values.

  • "sum": The sum of the loss values.

class mattertune.finetune.loss.L2MAELossConfig(*, name='l2_mae', reduction='mean')[source]
Parameters:

  • name (Literal['l2_mae'])

  • reduction (Literal['mean', 'sum'])

name: Literal['l2_mae']
reduction: Literal['mean', 'sum']

How to reduce the loss values across the batch.

  • "mean": The mean of the loss values.

  • "sum": The sum of the loss values.

mattertune.finetune.loss.l2_mae_loss(output, target, reduction='mean')[source]
Parameters:

  • output (Tensor)

  • target (Tensor)

  • reduction (Literal['mean', 'sum', 'none'])

Return type:

Tensor

mattertune.finetune.loss.compute_loss(config, prediction, label)[source]

Compute the loss value given the model output, prediction, and the target label, label.

The loss value should be a scalar tensor.

Parameters:

  • config (LossConfig) – The loss configuration.

  • prediction (Tensor) – The model output.

  • label (Tensor) – The target label.

Returns:

The computed loss value.

Return type:

Tensor