""" ADOPT PyTorch Optimizer ADOPT: Modified Adam Can Converge with Any β2 with the Optimal Rate: https://arxiv.org/abs/2411.02853 Modified for reduced dependencies on PyTorch internals from original at: https://github.com/iShohei220/adopt @inproceedings{taniguchi2024adopt, author={Taniguchi, Shohei and Harada, Keno and Minegishi, Gouki and Oshima, Yuta and Jeong, Seong Cheol and Nagahara, Go and Iiyama, Tomoshi and Suzuki, Masahiro and Iwasawa, Yusuke and Matsuo, Yutaka}, booktitle = {Advances in Neural Information Processing Systems}, title = {ADOPT: Modified Adam Can Converge with Any β2 with the Optimal Rate}, year = {2024} } """ from typing import cast, List, Optional, Tuple, Union import torch from torch import Tensor from torch.optim.optimizer import Optimizer from ._types import ParamsT __all__ = ["Adopt", "adopt"] def _view_as_real(params, *state_and_grads): for i, p in enumerate(params): if torch.is_complex(p): params[i] = torch.view_as_real(params[i]) for s in state_and_grads: s[i] = torch.view_as_real(s[i]) def _get_scalar_dtype(is_fused=None): if is_fused: return torch.float32 return ( torch.float64 if torch.get_default_dtype() == torch.float64 else torch.float32 ) def _is_compiling(): if hasattr(torch, 'compiler') and hasattr(torch.compiler, 'is_compiling'): return torch.compiler.is_compiling() else: return False def _get_value(x): # item is significantly faster than a cpu tensor in eager mode if not torch.jit.is_scripting() and _is_compiling(): return x else: return x.item() if isinstance(x, torch.Tensor) else x class Adopt(Optimizer): """ ADOPT: Modified Adam Can Converge with Any β2 with the Optimal Rate: https://arxiv.org/abs/2411.02853 """ def __init__( self, params: ParamsT, lr: Union[float, Tensor] = 1e-3, betas: Tuple[float, float] = (0.9, 0.9999), eps: float = 1e-6, clip_exp: Optional[float] = 0.333, weight_decay: float = 0.0, decoupled: bool = False, *, caution: bool = False, foreach: Optional[bool] = False, maximize: bool = False, capturable: bool = False, differentiable: bool = False, ): if isinstance(lr, Tensor): if foreach and not capturable: raise ValueError( "lr as a Tensor is not supported for capturable=False and foreach=True" ) if lr.numel() != 1: raise ValueError("Tensor lr must be 1-element") if not 0.0 <= lr: raise ValueError(f"Invalid learning rate: {lr}") if not 0.0 <= eps: raise ValueError(f"Invalid epsilon value: {eps}") if not 0.0 <= betas[0] < 1.0: raise ValueError(f"Invalid beta parameter at index 0: {betas[0]}") if not 0.0 <= betas[1] < 1.0: raise ValueError(f"Invalid beta parameter at index 1: {betas[1]}") if not 0.0 <= weight_decay: raise ValueError(f"Invalid weight_decay value: {weight_decay}") defaults = dict( lr=lr, betas=betas, eps=eps, weight_decay=weight_decay, clip_exp=clip_exp, decoupled=decoupled, caution=caution, maximize=maximize, foreach=foreach, capturable=capturable, differentiable=differentiable, ) super().__init__(params, defaults) def __setstate__(self, state): super().__setstate__(state) for group in self.param_groups: group.setdefault("maximize", False) group.setdefault("foreach", None) group.setdefault("capturable", False) group.setdefault("differentiable", False) group.setdefault("clip_exp", None) group.setdefault("caution", False) for p in group["params"]: p_state = self.state.get(p, []) if len(p_state) != 0 and not torch.is_tensor(p_state["step"]): step_val = float(p_state["step"]) p_state["step"] = ( torch.tensor( step_val, dtype=_get_scalar_dtype(), device=p.device, ) if group["capturable"] else torch.tensor(step_val, dtype=_get_scalar_dtype()) ) def _init_group( self, group, params_with_grad, grads, exp_avgs, exp_avg_sqs, state_steps, ): has_complex = False for p in group["params"]: if p.grad is None: continue has_complex |= torch.is_complex(p) params_with_grad.append(p) if p.grad.is_sparse: raise RuntimeError("ADOPT does not support sparse gradients") grads.append(p.grad) state = self.state[p] # Lazy state initialization if len(state) == 0: # note(crcrpar): [special device hosting for step] # Deliberately host `step` on CPU if both capturable and fused are off. # This is because kernel launches are costly on CUDA and XLA. state["step"] = ( torch.zeros((), dtype=_get_scalar_dtype(), device=p.grad.device) if group["capturable"] else torch.tensor(0.0, dtype=_get_scalar_dtype()) ) # Exponential moving average of gradient values state["exp_avg"] = torch.zeros_like(p.grad, memory_format=torch.preserve_format) # Exponential moving average of squared gradient values state["exp_avg_sq"] = torch.zeros_like(p.grad, memory_format=torch.preserve_format) exp_avgs.append(state["exp_avg"]) exp_avg_sqs.append(state["exp_avg_sq"]) if group["differentiable"] and state["step"].requires_grad: raise RuntimeError("`requires_grad` is not supported for `step` in differentiable mode") # Foreach without capturable does not support a tensor lr if group["foreach"] and torch.is_tensor(group["lr"]) and not group["capturable"]: raise RuntimeError("lr as a Tensor is not supported for capturable=False and foreach=True") state_steps.append(state["step"]) return has_complex #@_use_grad_for_differentiable # FIXME internal context mgr, can't use @torch.no_grad() def step(self, closure=None): """Perform a single optimization step. Args: closure (Callable, optional): A closure that reevaluates the model and returns the loss. """ self._cuda_graph_capture_health_check() loss = None if closure is not None: with torch.enable_grad(): loss = closure() for group in self.param_groups: params_with_grad: List[Tensor] = [] grads: List[Tensor] = [] exp_avgs: List[Tensor] = [] exp_avg_sqs: List[Tensor] = [] state_steps: List[Tensor] = [] beta1, beta2 = group["betas"] has_complex = self._init_group( group, params_with_grad, grads, exp_avgs, exp_avg_sqs, state_steps, ) adopt( params_with_grad, grads, exp_avgs, exp_avg_sqs, state_steps, has_complex=has_complex, beta1=beta1, beta2=beta2, lr=group["lr"], weight_decay=group["weight_decay"], clip_exp=group["clip_exp"], decoupled=group["decoupled"], eps=group["eps"], caution=group["caution"], maximize=group["maximize"], foreach=group["foreach"], capturable=group["capturable"], differentiable=group["differentiable"], grad_scale=getattr(self, "grad_scale", None), found_inf=getattr(self, "found_inf", None), ) return loss def _single_tensor_adopt( params: List[Tensor], grads: List[Tensor], exp_avgs: List[Tensor], exp_avg_sqs: List[Tensor], state_steps: List[Tensor], grad_scale: Optional[Tensor], found_inf: Optional[Tensor], *, has_complex: bool, beta1: float, beta2: float, lr: Union[float, Tensor], weight_decay: float, clip_exp: Optional[float], decoupled: bool, eps: float, caution: bool, maximize: bool, capturable: bool, differentiable: bool, ): assert grad_scale is None and found_inf is None if torch.jit.is_scripting(): # this assert is due to JIT being dumb and not realizing that the ops below # have overloads to handle both float and Tensor lrs, so we just assert it's # a float since most people using JIT are using floats assert isinstance(lr, float) for i, param in enumerate(params): grad = grads[i] if not maximize else -grads[i] exp_avg = exp_avgs[i] exp_avg_sq = exp_avg_sqs[i] step_t = state_steps[i] # If compiling, the compiler will handle cudagraph checks, see note [torch.compile x capturable] if capturable and not _is_compiling(): from torch.optim.optimizer import _get_capturable_supported_devices capturable_supported_devices = _get_capturable_supported_devices() assert param.device.type == step_t.device.type and param.device.type in capturable_supported_devices,\ f"If capturable=True, params and state_steps must be on supported devices: {capturable_supported_devices}." # update step step_t += 1 if torch.is_complex(param): grad = torch.view_as_real(grad) if exp_avg is not None: exp_avg = torch.view_as_real(exp_avg) if exp_avg_sq is not None: exp_avg_sq = torch.view_as_real(exp_avg_sq) param = torch.view_as_real(param) if weight_decay != 0 and not decoupled: grad = grad.add(param, alpha=weight_decay) step = step_t if capturable or differentiable else _get_value(step_t) if step == 1: exp_avg_sq.addcmul_(grad, grad.conj()) continue if weight_decay != 0 and decoupled: param.add_(param, alpha=-lr * weight_decay) denom = torch.clamp(exp_avg_sq.sqrt(), eps) normed_grad = grad.div(denom) if clip_exp is not None: clip_val = (step - 1) ** clip_exp normed_grad.clamp_(-clip_val, clip_val) exp_avg.lerp_(normed_grad, 1 - beta1) if caution: # Apply caution as per 'Cautious Optimizers' - https://arxiv.org/abs/2411.16085 mask = (exp_avg * grad > 0).to(grad.dtype) mask.div_(mask.mean().clamp_(min=1e-3)) exp_avg = exp_avg * mask param.add_(exp_avg, alpha=-lr) exp_avg_sq.mul_(beta2).addcmul_(grad, grad.conj(), value=1 - beta2) def _multi_tensor_adopt( params: List[Tensor], grads: List[Tensor], exp_avgs: List[Tensor], exp_avg_sqs: List[Tensor], state_steps: List[Tensor], grad_scale: Optional[Tensor], found_inf: Optional[Tensor], *, has_complex: bool, beta1: float, beta2: float, lr: Union[float, Tensor], weight_decay: float, clip_exp: Optional[float], decoupled: bool, eps: float, caution: bool, maximize: bool, capturable: bool, differentiable: bool, ): if len(params) == 0: return if isinstance(lr, Tensor) and not capturable: raise RuntimeError( "lr as a Tensor is not supported for capturable=False and foreach=True" ) # If compiling, the compiler will handle cudagraph checks, see note [torch.compile x capturable] if capturable and not _is_compiling(): from torch.optim.optimizer import _get_capturable_supported_devices capturable_supported_devices = _get_capturable_supported_devices( supports_xla=False ) assert all( p.device.type == step.device.type and p.device.type in capturable_supported_devices for p, step in zip(params, state_steps) ), f"If capturable=True, params and state_steps must be on supported devices: {capturable_supported_devices}." assert grad_scale is None and found_inf is None assert not differentiable, "_foreach ops don't support autograd" grouped_tensors = Optimizer._group_tensors_by_device_and_dtype( [params, grads, exp_avgs, exp_avg_sqs, state_steps] # type: ignore[list-item] ) for ( device_params_, device_grads_, device_exp_avgs_, device_exp_avg_sqs_, device_state_steps_, ), _ in grouped_tensors.values(): device_params = cast(List[Tensor], device_params_) device_grads = cast(List[Tensor], device_grads_) device_exp_avgs = cast(List[Tensor], device_exp_avgs_) device_exp_avg_sqs = cast(List[Tensor], device_exp_avg_sqs_) device_state_steps = cast(List[Tensor], device_state_steps_) # Handle complex parameters if has_complex: _view_as_real(device_params, device_grads, device_exp_avgs, device_exp_avg_sqs) if maximize: device_grads = torch._foreach_neg(device_grads) # type: ignore[assignment] # Update steps # If steps are on CPU, foreach will fall back to the slow path, which is a for-loop calling t.add(1) over # and over. 1 will then be wrapped into a Tensor over and over again, which is slower than if we just # wrapped it once now. The alpha is required to assure we go to the right overload. if not _is_compiling() and device_state_steps[0].is_cpu: torch._foreach_add_(device_state_steps, torch.tensor(1.0, device="cpu"), alpha=1.0) else: torch._foreach_add_(device_state_steps, 1) if weight_decay != 0 and not decoupled: # Re-use the intermediate memory (device_grads) already allocated for maximize if maximize: torch._foreach_add_(device_grads, device_params, alpha=weight_decay) else: device_grads = torch._foreach_add(device_grads, device_params, alpha=weight_decay) if device_state_steps[0] == 1: torch._foreach_addcmul_(device_exp_avg_sqs, device_grads, device_grads) continue if weight_decay != 0 and decoupled: torch._foreach_add_(device_params, device_params, alpha=-lr * weight_decay) exp_avg_sq_sqrt = torch._foreach_sqrt(device_exp_avg_sqs) torch._foreach_maximum_(exp_avg_sq_sqrt, eps) normed_grad = torch._foreach_div(device_grads, exp_avg_sq_sqrt) if clip_exp is not None: clip_val = (device_state_steps[0] - 1) ** clip_exp torch._foreach_maximum_(normed_grad, -clip_val) torch._foreach_minimum_(normed_grad, clip_val) torch._foreach_lerp_(device_exp_avgs, normed_grad, 1 - beta1) if caution: # Apply caution as per 'Cautious Optimizers' - https://arxiv.org/abs/2411.16085 masks = torch._foreach_mul(device_exp_avgs, device_grads) masks = [(m > 0).to(g.dtype) for m, g in zip(masks, device_grads)] mask_scale = [m.mean() for m in masks] torch._foreach_maximum_(mask_scale, 1e-3) torch._foreach_div_(masks, mask_scale) device_exp_avgs = torch._foreach_mul(device_exp_avgs, masks) torch._foreach_add_(device_params, device_exp_avgs, alpha=-lr) torch._foreach_mul_(device_exp_avg_sqs, beta2) torch._foreach_addcmul_(device_exp_avg_sqs, device_grads, device_grads, value=1 - beta2) #@_disable_dynamo_if_unsupported(single_tensor_fn=_single_tensor_adopt) # FIXME internal context mgr, can't use def adopt( params: List[Tensor], grads: List[Tensor], exp_avgs: List[Tensor], exp_avg_sqs: List[Tensor], state_steps: List[Tensor], # kwonly args with defaults are not supported by functions compiled with torchscript issue #70627 # setting this as kwarg for now as functional API is compiled by torch/distributed/optim foreach: Optional[bool] = None, capturable: bool = False, differentiable: bool = False, grad_scale: Optional[Tensor] = None, found_inf: Optional[Tensor] = None, has_complex: bool = False, *, beta1: float, beta2: float, lr: Union[float, Tensor], weight_decay: float, clip_exp: Optional[float], decoupled: bool, eps: float, caution: bool, maximize: bool, ): r"""Functional API that performs ADOPT algorithm computation. """ if foreach is None: foreach = False # this check is slow during compilation, so we skip it # if it's strictly needed we can add this check back in dynamo if not _is_compiling() and not all(isinstance(t, torch.Tensor) for t in state_steps): raise RuntimeError( "API has changed, `state_steps` argument must contain a list of singleton tensors" ) if foreach and torch.jit.is_scripting(): raise RuntimeError("torch.jit.script not supported with foreach optimizers") if foreach and not torch.jit.is_scripting(): func = _multi_tensor_adopt else: func = _single_tensor_adopt func( params, grads, exp_avgs, exp_avg_sqs, state_steps, has_complex=has_complex, beta1=beta1, beta2=beta2, lr=lr, weight_decay=weight_decay, clip_exp=clip_exp, decoupled=decoupled, eps=eps, caution=caution, maximize=maximize, capturable=capturable, differentiable=differentiable, grad_scale=grad_scale, found_inf=found_inf, )