peft/pr_3207/en/package_reference/merge_utils.md

Model merge[[peft.utils.merge_utils.prune]]

PEFT provides several internal utilities for merging LoRA adapters with the TIES and DARE methods.

peft.utils.merge_utils.prune[[peft.utils.merge_utils.prune]]

Source

Prune the values of task tensors based on the method.

Parameters:

tensor (torch.Tensor) --The tensor to prune.

density (float) --The fraction of values to preserve. Should be in [0,1].

method (str) --The method to use to prune. Should be one of ["magnitude", "random"].

rescale (bool) --Whether to rescale the result to preserve the expected value of the original tensor.

Returns:

torch.Tensor

The pruned tensor.

peft.utils.merge_utils.calculate_majority_sign_mask[[peft.utils.merge_utils.calculate_majority_sign_mask]]

Source

Get the mask of the majority sign across the task tensors. Task tensors are stacked on dimension 0.

Parameters:

tensor (torch.Tensor) --The tensor to get the mask from.

method (str) --The method to use to get the mask. Should be one of ["total", "frequency"].

Returns:

torch.Tensor

The majority sign mask.

peft.utils.merge_utils.disjoint_merge[[peft.utils.merge_utils.disjoint_merge]]

Source

Merge the task tensors using disjoint merge.

Parameters:

task_tensors (torch.Tensor) --The task tensors to merge.

majority_sign_mask (torch.Tensor) --The mask of the majority sign across the task tensors.

Returns:

torch.Tensor

The merged tensor.

peft.utils.merge_utils.task_arithmetic[[peft.utils.merge_utils.task_arithmetic]]

Source

Merge the task tensors using task arithmetic.

Parameters:

task_tensors(List[torch.Tensor]) --The task tensors to merge.

weights (torch.Tensor) --The weights of the task tensors.

Returns:

torch.Tensor

The merged tensor.

peft.utils.merge_utils.ties[[peft.utils.merge_utils.ties]]

Source

Merge the task tensors using ties.

Parameters:

task_tensors(List[torch.Tensor]) --The task tensors to merge.

weights (torch.Tensor) --The weights of the task tensors.

density (float) --The fraction of values to preserve. Should be in [0,1].

majority_sign_method (str) : The method to use to get the majority sign mask. Should be one of ["total", "frequency"].

Returns:

torch.Tensor

The merged tensor.

peft.utils.merge_utils.dare_linear[[peft.utils.merge_utils.dare_linear]]

Source

Merge the task tensors using dare linear.

Parameters:

task_tensors(List[torch.Tensor]) --The task tensors to merge.

weights (torch.Tensor) --The weights of the task tensors.

density (float) --The fraction of values to preserve. Should be in [0,1].

Returns:

torch.Tensor

The merged tensor.

peft.utils.merge_utils.dare_ties[[peft.utils.merge_utils.dare_ties]]

Source

Merge the task tensors using dare ties.

Parameters:

task_tensors(List[torch.Tensor]) --The task tensors to merge.

weights (torch.Tensor) --The weights of the task tensors.

density (float) --The fraction of values to preserve. Should be in [0,1].

majority_sign_method (str) : The method to use to get the majority sign mask. Should be one of ["total", "frequency"].

Returns:

torch.Tensor

The merged tensor.