boat_ms.utils

Submodules

boat_ms.utils.op_utils

boat_ms.utils.op_utils.average_grad(model, batch_size)[source]

Average the gradients of all model parameters by the batch size.

Parameters:

  • model (mindspore.nn.Cell) – The model whose gradients need to be averaged.

  • batch_size (int) – The batch size to divide gradients by.

boat_ms.utils.op_utils.copy_parameter_from_list(model, param_list)[source]

Copy parameters from a list to a model’s trainable parameters.

Parameters:

  • model (mindspore.nn.Cell) – The model whose parameters need to be updated.

  • param_list (list of mindspore.Tensor) – The list of parameters to copy.

boat_ms.utils.op_utils.grad_unused_zero(output, inputs, grad_outputs=None)[source]

Compute gradients for the given inputs, substituting zeros for unused gradients. (MindSpore version)

Parameters:

  • output (mindspore.Tensor) – The scalar output tensor for which gradients are computed.

  • inputs (List[mindspore.Tensor]) – List of input tensors with respect to which gradients are computed.

  • grad_outputs (mindspore.Tensor, optional) – Sensitivity tensor (same shape as output). Default = ones_like(output).

Returns:

Gradients for the inputs, with unused gradients replaced by zeros.

Return type:

List[mindspore.Tensor]

boat_ms.utils.op_utils.l2_reg(params)[source]

Compute the L2 regularization loss.

Parameters:

params (list) – List of model parameters (trainable).

Returns:

The computed L2 regularization loss.

Return type:

mindspore.Tensor

boat_ms.utils.op_utils.require_model_grad(model=None)[source]

Ensure all parameters of a MindSpore model require gradients.

Parameters:

model (mindspore.nn.Cell, optional) – MindSpore model instance. Must not be None.

Raises:

AssertionError – If the model is None.

boat_ms.utils.op_utils.stop_grads(grads)[source]

Detach and stop gradient computation for a list of gradients.

Parameters:

grads (list of mindspore.Tensor) – Gradients to process.

Returns:

Detached gradients with requires_grad set to False.

Return type:

list of mindspore.Tensor

boat_ms.utils.op_utils.stop_model_grad(model=None)[source]

Stop gradient computation for all parameters in a model.

Parameters:

model (mindspore.nn.Cell, optional) – The model to stop gradients for. Must not be None.

Raises:

AssertionError – If the model is None.

boat_ms.utils.op_utils.update_grads(grads, model)[source]

Update gradients for a model’s parameters.

Parameters:

  • grads (list) – List of gradients to apply.

  • model (mindspore.nn.Cell) – The model whose gradients will be updated.

boat_ms.utils.op_utils.update_tensor_grads(hparams, grads)[source]

Update gradients for hyperparameters.

Parameters:

  • hparams (list of mindspore.Tensor) – Hyperparameters to update.

  • grads (list of mindspore.Tensor) – Gradients to apply to the hyperparameters.