fused_adam.py

# Copyright (c) Microsoft Corporation. All rights reserved.
# Licensed under the MIT License.
# fused_adam.py
# This file has been adapted from microsoft/DeepSpeed

"""
Copyright 2020 The Microsoft DeepSpeed Team

Copyright NVIDIA/apex
This file is adapted from fused adam in NVIDIA/apex, commit a109f85
"""

from enum import IntEnum

import torch

from ._multi_tensor_apply import MultiTensorApply


class AdamWMode(IntEnum):
    ADAM_L2_REGULARIZATION = 0  # Adam with L2 regularization
    ADAMW_TRANSFORMERS = 1  # Adam with weight decay implemented to be equivalent to Transformers/AdamW
    ADAMW_TORCH = 2  # Adam with weight decay implemented to be equivalent to torch/AdamW


class FusedAdam(torch.optim.Optimizer):
    """Implements Adam algorithm.

    The algorithmic implementation is mathematically equivalent to
    `Transformers/AdamW <https://github.com/huggingface/transformers/blob/61f64262692ac7dc90e2e0bdeb7e79d9cd607a66/src/transformers/optimization.py#L349-L370>`_
    when adam_w_mode = 1 and `torch/Adam <https://github.com/pytorch/pytorch/blob/a217a62e73fd30b658743af8a69966f90327f018/torch/optim/adamw.py#L6>`_
    when adam_w_mode = 2

    Currently GPU-only.

    This version of fused Adam implements 2 fusions.

      * Fusion of the Adam update's elementwise operations
      * A multi-tensor apply launch that batches the elementwise updates applied to all the model's parameters into one or a few kernel launches.

    Adam was proposed in `Adam: A Method for Stochastic Optimization`_.

    Arguments:
        params (iterable): iterable of parameters to optimize or dicts defining
            parameter groups.
        lr (float, 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)
        weight_decay (float, optional): weight decay (L2 penalty) (default: 0)
        adam_w_mode (AdamWMode, optional): Apply L2 regularization or weight decay
            (AdamWMode.ADAM_L2_REGULARIZATION), decoupled weight decay with
            transformers/AdamW mathematical implementation (AdamWMode.ADAMW_TRANSFORMERS)
            or decoupled weight decay with transformers/AdamW implementation
            (AdamWMode.ADAMW_TORCH) (default: AdamWMode.ADAMW_TRANSFORMERS)
        set_grad_none (bool, optional): whether set grad to None when zero_grad()
            method is called. (default: True)
            PyTorch Adam has set_to_none parameter in zero_grad(), supporting that too

    .. _Adam - A Method for Stochastic Optimization:
        https://arxiv.org/abs/1412.6980
    .. _On the Convergence of Adam and Beyond:
        https://openreview.net/forum?id=ryQu7f-RZ
    """

    def __init__(
        self,
        params,
        lr=1e-3,
        bias_correction=True,
        betas=(0.9, 0.999),
        eps=1e-6,
        adam_w_mode=AdamWMode.ADAMW_TRANSFORMERS,
        weight_decay=0.0,
        set_grad_none=True,
    ):
        # The FusedAdam implementation is mathematically equivalent to
        # transformers AdamW. The input arguments also have the same defaults.

        defaults = dict(lr=lr, bias_correction=bias_correction, betas=betas, eps=eps, weight_decay=weight_decay)
        super().__init__(params, defaults)
        self._adam_w_mode = adam_w_mode
        self._set_grad_none = set_grad_none

        # Skip buffer
        self._dummy_overflow_buf = torch.cuda.IntTensor([0])

        from onnxruntime.training.ortmodule.torch_cpp_extensions import fused_ops

        self._multi_tensor_adam = fused_ops.multi_tensor_adam
        self._multi_tensor_applier = MultiTensorApply(2048 * 32)
        self._TorchTensorVector = fused_ops.TorchTensorVector

    def zero_grad(self, set_to_none=True):
        if self._set_grad_none or set_to_none:
            for group in self.param_groups:
                for p in group["params"]:
                    p.grad = None
        else:
            super().zero_grad()

    def step(self, closure=None):
        """Performs a single optimization step.

        Arguments:
            closure (callable, optional): A closure that reevaluates the model
                and returns the loss.

        The remaining arguments are deprecated, and are only retained (for the moment) for error-checking purposes.
        """
        loss = None
        if closure is not None:
            loss = closure()

        for group in self.param_groups:
            bias_correction = 1 if group["bias_correction"] else 0
            beta1, beta2 = group["betas"]

            # assume same step across group now to simplify things
            # per parameter step can be easily support by making it tensor, or pass list into kernel
            if "step" in group:
                group["step"] += 1
            else:
                group["step"] = 1

            # create lists for multi-tensor apply
            g_16, p_16, m_16, v_16 = [], [], [], []
            g_32, p_32, m_32, v_32 = [], [], [], []

            for p in group["params"]:
                if p.grad is None:
                    continue
                if p.grad.data.is_sparse:
                    raise RuntimeError(
                        "FusedAdam does not support sparse gradients, please consider SparseAdam instead"
                    )

                state = self.state[p]
                # State initialization
                if len(state) == 0:
                    # Exponential moving average of gradient values
                    state["exp_avg"] = torch.zeros_like(p.data)
                    # Exponential moving average of squared gradient values
                    state["exp_avg_sq"] = torch.zeros_like(p.data)

                if p.dtype == torch.float16:
                    g_16.append(p.grad.data)
                    p_16.append(p.data)
                    m_16.append(state["exp_avg"])
                    v_16.append(state["exp_avg_sq"])
                elif p.dtype == torch.float32:
                    g_32.append(p.grad.data)
                    p_32.append(p.data)
                    m_32.append(state["exp_avg"])
                    v_32.append(state["exp_avg_sq"])
                else:
                    raise RuntimeError("FusedAdam only support fp16 and fp32.")

            if len(g_16) > 0:
                self._multi_tensor_applier(
                    self._multi_tensor_adam,
                    self._dummy_overflow_buf,
                    [
                        self._TorchTensorVector(g_16),
                        self._TorchTensorVector(p_16),
                        self._TorchTensorVector(m_16),
                        self._TorchTensorVector(v_16),
                    ],
                    group["lr"],
                    beta1,
                    beta2,
                    group["eps"],
                    group["step"],
                    self._adam_w_mode,
                    bias_correction,
                    group["weight_decay"],
                )
            if len(g_32) > 0:
                self._multi_tensor_applier(
                    self._multi_tensor_adam,
                    self._dummy_overflow_buf,
                    [
                        self._TorchTensorVector(g_32),
                        self._TorchTensorVector(p_32),
                        self._TorchTensorVector(m_32),
                        self._TorchTensorVector(v_32),
                    ],
                    group["lr"],
                    beta1,
                    beta2,
                    group["eps"],
                    group["step"],
                    self._adam_w_mode,
                    bias_correction,
                    group["weight_decay"],
                )

        return loss