SparseAdam¶

class torch.optim.SparseAdam(params, lr=0.001, betas=(0.9, 0.999), eps=1e-08, maximize=False)[source]¶

SparseAdam implements a masked version of the Adam algorithm suitable for sparse gradients. Currently, due to implementation constraints (explained below), SparseAdam is only intended for a narrow subset of use cases, specifically parameters of a dense layout with gradients of a sparse layout. This occurs in a special case where the module backwards produces grads already in a sparse layout. One example NN module that behaves as such is nn.Embedding(sparse=True).

SparseAdam approximates the Adam algorithm by masking out the parameter and moment updates corresponding to the zero values in the gradients. Whereas the Adam algorithm will update the first moment, the second moment, and the parameters based on all values of the gradients, SparseAdam only updates the moments and parameters corresponding to the non-zero values of the gradients.

A simplified way of thinking about the intended implementation is as such:

Create a mask of the non-zero values in the sparse gradients. For example, if your gradient looks like [0, 5, 0, 0, 9], the mask would be [0, 1, 0, 0, 1].
Apply this mask over the running moments and do computation on only the non-zero values.
Apply this mask over the parameters and only apply an update on non-zero values.

In actuality, we use sparse layout Tensors to optimize this approximation, which means the more gradients that are masked by not being materialized, the more performant the optimization. Since we rely on using sparse layout tensors, we infer that any materialized value in the sparse layout is non-zero and we do NOT actually verify that all values are not zero! It is important to not conflate a semantically sparse tensor (a tensor where many of its values are zeros) with a sparse layout tensor (a tensor where .is_sparse returns True). The SparseAdam approximation is intended for semantically sparse tensors and the sparse layout is only a implementation detail. A clearer implementation would be to use MaskedTensors, but those are experimental.

Note

If you suspect your gradients are semantically sparse (but do not have sparse layout), this variant may not be the best for you. Ideally, you want to avoid materializing anything that is suspected to be sparse in the first place, since needing to convert all your grads from dense layout to sparse layout may outweigh the performance gain. Here, using Adam may be the best alternative, unless you can easily rig up your module to output sparse grads similar to nn.Embedding(sparse=True). If you insist on converting your grads, you can do so by manually overriding your parameters’ .grad fields with their sparse equivalents before calling .step().

Parameters

params (iterable) – iterable of parameters to optimize or dicts defining parameter groups
lr (float, Tensor, optional) – learning rate (default: 1e-3)
betas (Tuple[float, float], optional) – coefficients used for computing running averages of gradient and its square (default: (0.9, 0.999))
eps (float, optional) – term added to the denominator to improve numerical stability (default: 1e-8)
maximize (bool, optional) – maximize the objective with respect to the params, instead of minimizing (default: False)

add_param_group(param_group)¶

Add a param group to the Optimizer s param_groups.

This can be useful when fine tuning a pre-trained network as frozen layers can be made trainable and added to the Optimizer as training progresses.

Parameters: param_group (dict) – Specifies what Tensors should be optimized along with group specific optimization options.

load_state_dict(state_dict)¶

Load the optimizer state.

Parameters: state_dict (dict) – optimizer state. Should be an object returned from a call to state_dict().

register_load_state_dict_post_hook(hook, prepend=False)¶

Register a load_state_dict post-hook which will be called after load_state_dict() is called. It should have the following signature:

hook(optimizer) -> None

The optimizer argument is the optimizer instance being used.

The hook will be called with argument self after calling load_state_dict on self. The registered hook can be used to perform post-processing after load_state_dict has loaded the state_dict.

Parameters

hook (Callable) – The user defined hook to be registered.
prepend (bool) – If True, the provided post hook will be fired before all the already registered post-hooks on load_state_dict. Otherwise, the provided hook will be fired after all the already registered post-hooks. (default: False)

Returns

a handle that can be used to remove the added hook by calling handle.remove()

Return type

torch.utils.hooks.RemoveableHandle

register_load_state_dict_pre_hook(hook, prepend=False)¶

Register a load_state_dict pre-hook which will be called before load_state_dict() is called. It should have the following signature:

hook(optimizer, state_dict) -> state_dict or None

The optimizer argument is the optimizer instance being used and the state_dict argument is a shallow copy of the state_dict the user passed in to load_state_dict. The hook may modify the state_dict inplace or optionally return a new one. If a state_dict is returned, it will be used to be loaded into the optimizer.

The hook will be called with argument self and state_dict before calling load_state_dict on self. The registered hook can be used to perform pre-processing before the load_state_dict call is made.

Parameters

hook (Callable) – The user defined hook to be registered.
prepend (bool) – If True, the provided pre hook will be fired before all the already registered pre-hooks on load_state_dict. Otherwise, the provided hook will be fired after all the already registered pre-hooks. (default: False)

Returns

a handle that can be used to remove the added hook by calling handle.remove()

Return type

torch.utils.hooks.RemoveableHandle

register_state_dict_post_hook(hook, prepend=False)¶

Register a state dict post-hook which will be called after state_dict() is called.

It should have the following signature:

hook(optimizer, state_dict) -> state_dict or None

The hook will be called with arguments self and state_dict after generating a state_dict on self. The hook may modify the state_dict inplace or optionally return a new one. The registered hook can be used to perform post-processing on the state_dict before it is returned.

Parameters

hook (Callable) – The user defined hook to be registered.
prepend (bool) – If True, the provided post hook will be fired before all the already registered post-hooks on state_dict. Otherwise, the provided hook will be fired after all the already registered post-hooks. (default: False)

Returns

a handle that can be used to remove the added hook by calling handle.remove()

Return type

torch.utils.hooks.RemoveableHandle

register_state_dict_pre_hook(hook, prepend=False)¶

Register a state dict pre-hook which will be called before state_dict() is called.

It should have the following signature:

hook(optimizer) -> None

The optimizer argument is the optimizer instance being used. The hook will be called with argument self before calling state_dict on self. The registered hook can be used to perform pre-processing before the state_dict call is made.

Parameters

hook (Callable) – The user defined hook to be registered.
prepend (bool) – If True, the provided pre hook will be fired before all the already registered pre-hooks on state_dict. Otherwise, the provided hook will be fired after all the already registered pre-hooks. (default: False)

Returns

a handle that can be used to remove the added hook by calling handle.remove()

Return type

torch.utils.hooks.RemoveableHandle

register_step_post_hook(hook)¶

Register an optimizer step post hook which will be called after optimizer step.

It should have the following signature:

hook(optimizer, args, kwargs) -> None

The optimizer argument is the optimizer instance being used.

Parameters: hook (Callable) – The user defined hook to be registered.
Returns: a handle that can be used to remove the added hook by calling handle.remove()
Return type: torch.utils.hooks.RemovableHandle

register_step_pre_hook(hook)¶

Register an optimizer step pre hook which will be called before optimizer step.

It should have the following signature:

hook(optimizer, args, kwargs) -> None or modified args and kwargs

The optimizer argument is the optimizer instance being used. If args and kwargs are modified by the pre-hook, then the transformed values are returned as a tuple containing the new_args and new_kwargs.

Parameters: hook (Callable) – The user defined hook to be registered.
Returns: a handle that can be used to remove the added hook by calling handle.remove()
Return type: torch.utils.hooks.RemovableHandle

state_dict()¶

Return the state of the optimizer as a dict.

It contains two entries:

state: a Dict holding current optimization state. Its content
differs between optimizer classes, but some common characteristics hold. For example, state is saved per parameter, and the parameter itself is NOT saved. state is a Dictionary mapping parameter ids to a Dict with state corresponding to each parameter.
param_groups: a List containing all parameter groups where each
parameter group is a Dict. Each parameter group contains metadata specific to the optimizer, such as learning rate and weight decay, as well as a List of parameter IDs of the parameters in the group.

NOTE: The parameter IDs may look like indices but they are just IDs associating state with param_group. When loading from a state_dict, the optimizer will zip the param_group params (int IDs) and the optimizer param_groups (actual nn.Parameter s) in order to match state WITHOUT additional verification.

A returned state dict might look something like:

{
    'state': {
        0: {'momentum_buffer': tensor(...), ...},
        1: {'momentum_buffer': tensor(...), ...},
        2: {'momentum_buffer': tensor(...), ...},
        3: {'momentum_buffer': tensor(...), ...}
    },
    'param_groups': [
        {
            'lr': 0.01,
            'weight_decay': 0,
            ...
            'params': [0]
        },
        {
            'lr': 0.001,
            'weight_decay': 0.5,
            ...
            'params': [1, 2, 3]
        }
    ]
}

Return type: Dict[str, Any]

step(closure=None)[source]¶

Perform a single optimization step.

Parameters: closure (Callable, optional) – A closure that reevaluates the model and returns the loss.

zero_grad(set_to_none=True)¶

Reset the gradients of all optimized torch.Tensor s.

Parameters: set_to_none (bool) – instead of setting to zero, set the grads to None. This will in general have lower memory footprint, and can modestly improve performance. However, it changes certain behaviors. For example: 1. When the user tries to access a gradient and perform manual ops on it, a None attribute or a Tensor full of 0s will behave differently. 2. If the user requests zero_grad(set_to_none=True) followed by a backward pass, .grads are guaranteed to be None for params that did not receive a gradient. 3. torch.optim optimizers have a different behavior if the gradient is 0 or None (in one case it does the step with a gradient of 0 and in the other it skips the step altogether).

SparseAdam¶

Docs

Tutorials

Resources