import torch
from torch import Tensor
from .optimizer import (Optimizer, _use_grad_for_differentiable, _get_value, _dispatch_sqrt,
                        _stack_if_compiling, _capturable_doc, _differentiable_doc, _foreach_doc,
                        _fused_doc, _maximize_doc, _default_to_fused_or_foreach)
from typing import List, Optional
from torch.utils._foreach_utils import _group_tensors_by_device_and_dtype

__all__ = ["AdamW", "adamw"]


class AdamW(Optimizer):
    def __init__(
        self,
        params,
        lr=1e-3,
        betas=(0.9, 0.999),
        eps=1e-8,
        weight_decay=1e-2,
        amsgrad=False,
        *,
        maximize: bool = False,
        foreach: Optional[bool] = None,
        capturable: bool = False,
        differentiable: bool = False,
        fused: Optional[bool] = None,
    ):
        if not 0.0 <= lr:
            raise ValueError("Invalid learning rate: {}".format(lr))
        if not 0.0 <= eps:
            raise ValueError("Invalid epsilon value: {}".format(eps))
        if not 0.0 <= betas[0] < 1.0:
            raise ValueError("Invalid beta parameter at index 0: {}".format(betas[0]))
        if not 0.0 <= betas[1] < 1.0:
            raise ValueError("Invalid beta parameter at index 1: {}".format(betas[1]))
        if not 0.0 <= weight_decay:
            raise ValueError("Invalid weight_decay value: {}".format(weight_decay))
        defaults = dict(
            lr=lr,
            betas=betas,
            eps=eps,
            weight_decay=weight_decay,
            amsgrad=amsgrad,
            foreach=foreach,
            maximize=maximize,
            capturable=capturable,
            differentiable=differentiable,
            fused=fused,
        )
        super().__init__(params, defaults)

        if fused:
            if differentiable:
                raise RuntimeError("`fused` does not support `differentiable`")
            self._step_supports_amp_scaling = True
            # TODO(crcrpar): [low prec params & their higher prec copy]
            # Suppor AMP with FP16/BF16 model params which would need
            # higher prec copy of params to do update math in higher prec to
            # alleviate the loss of information.
            if not all(
                p.is_cuda and torch.is_floating_point(p)
                for pg in self.param_groups for p in pg['params']
            ):
                raise RuntimeError("`fused=True` requires all the params to be CUDA, floating point Tensor")
            if foreach:
                raise RuntimeError("`fused` and `foreach` cannot be `True` together.")

    def __setstate__(self, state):
        super().__setstate__(state)
        for group in self.param_groups:
            group.setdefault("amsgrad", False)
            group.setdefault("maximize", False)
            group.setdefault("foreach", None)
            group.setdefault("capturable", False)
            group.setdefault("differentiable", False)
            group.setdefault("fused", None)
        state_values = list(self.state.values())
        step_is_tensor = (len(state_values) != 0) and torch.is_tensor(
            state_values[0]["step"]
        )
        if not step_is_tensor:
            for s in state_values:
                s["step"] = torch.tensor(float(s["step"]))

    def _init_group(
        self,
        group,
        params_with_grad,
        grads,
        amsgrad,
        exp_avgs,
        exp_avg_sqs,
        max_exp_avg_sqs,
        state_steps,
    ):
        for p in group["params"]:
            if p.grad is None:
                continue
            params_with_grad.append(p)
            if p.grad.is_sparse:
                raise RuntimeError("AdamW does not support sparse gradients")
            grads.append(p.grad)

            state = self.state[p]

            # State initialization
            if len(state) == 0:
                state["step"] = (
                    torch.zeros((1,), dtype=torch.float, device=p.device)
                    if group["capturable"] or group["fused"]
                    else torch.tensor(0.0)
                )
                # Exponential moving average of gradient values
                state["exp_avg"] = torch.zeros_like(
                    p, memory_format=torch.preserve_format
                )
                # Exponential moving average of squared gradient values
                state["exp_avg_sq"] = torch.zeros_like(
                    p, memory_format=torch.preserve_format
                )
                if amsgrad:
                    # Maintains max of all exp. moving avg. of sq. grad. values
                    state["max_exp_avg_sq"] = torch.zeros_like(
                        p, memory_format=torch.preserve_format
                    )

            exp_avgs.append(state["exp_avg"])
            exp_avg_sqs.append(state["exp_avg_sq"])

            if amsgrad:
                max_exp_avg_sqs.append(state["max_exp_avg_sq"])

            state_steps.append(state["step"])

    @_use_grad_for_differentiable
    def step(self, closure=None):
        """Performs 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 = []
            grads = []
            exp_avgs = []
            exp_avg_sqs = []
            max_exp_avg_sqs = []
            state_steps = []
            amsgrad = group["amsgrad"]
            beta1, beta2 = group["betas"]

            self._init_group(
                group,
                params_with_grad,
                grads,
                amsgrad,
                exp_avgs,
                exp_avg_sqs,
                max_exp_avg_sqs,
                state_steps,
            )

            adamw(
                params_with_grad,
                grads,
                exp_avgs,
                exp_avg_sqs,
                max_exp_avg_sqs,
                state_steps,
                amsgrad=amsgrad,
                beta1=beta1,
                beta2=beta2,
                lr=group["lr"],
                weight_decay=group["weight_decay"],
                eps=group["eps"],
                maximize=group["maximize"],
                foreach=group["foreach"],
                capturable=group["capturable"],
                differentiable=group["differentiable"],
                fused=group["fused"],
                grad_scale=getattr(self, "grad_scale", None),
                found_inf=getattr(self, "found_inf", None),
            )

        return loss


AdamW.__doc__ = r"""Implements AdamW algorithm.

    .. math::
       \begin{aligned}
            &\rule{110mm}{0.4pt}                                                                 \\
            &\textbf{input}      : \gamma \text{(lr)}, \: \beta_1, \beta_2
                \text{(betas)}, \: \theta_0 \text{(params)}, \: f(\theta) \text{(objective)},
                \: \epsilon \text{ (epsilon)}                                                    \\
            &\hspace{13mm}      \lambda \text{(weight decay)},  \: \textit{amsgrad},
                \: \textit{maximize}                                                             \\
            &\textbf{initialize} : m_0 \leftarrow 0 \text{ (first moment)}, v_0 \leftarrow 0
                \text{ ( second moment)}, \: \widehat{v_0}^{max}\leftarrow 0              \\[-1.ex]
            &\rule{110mm}{0.4pt}                                                                 \\
            &\textbf{for} \: t=1 \: \textbf{to} \: \ldots \: \textbf{do}                         \\

            &\hspace{5mm}\textbf{if} \: \textit{maximize}:                                       \\
            &\hspace{10mm}g_t           \leftarrow   -\nabla_{\theta} f_t (\theta_{t-1})          \\
            &\hspace{5mm}\textbf{else}                                                           \\
            &\hspace{10mm}g_t           \leftarrow   \nabla_{\theta} f_t (\theta_{t-1})           \\
            &\hspace{5mm} \theta_t \leftarrow \theta_{t-1} - \gamma \lambda \theta_{t-1}         \\
            &\hspace{5mm}m_t           \leftarrow   \beta_1 m_{t-1} + (1 - \beta_1) g_t          \\
            &\hspace{5mm}v_t           \leftarrow   \beta_2 v_{t-1} + (1-\beta_2) g^2_t          \\
            &\hspace{5mm}\widehat{m_t} \leftarrow   m_t/\big(1-\beta_1^t \big)                   \\
            &\hspace{5mm}\widehat{v_t} \leftarrow   v_t/\big(1-\beta_2^t \big)                   \\
            &\hspace{5mm}\textbf{if} \: amsgrad                                                  \\
            &\hspace{10mm}\widehat{v_t}^{max} \leftarrow \mathrm{max}(\widehat{v_t}^{max},
                \widehat{v_t})                                                                   \\
            &\hspace{10mm}\theta_t \leftarrow \theta_t - \gamma \widehat{m_t}/
                \big(\sqrt{\widehat{v_t}^{max}} + \epsilon \big)                                 \\
            &\hspace{5mm}\textbf{else}                                                           \\
            &\hspace{10mm}\theta_t \leftarrow \theta_t - \gamma \widehat{m_t}/
                \big(\sqrt{\widehat{v_t}} + \epsilon \big)                                       \\
            &\rule{110mm}{0.4pt}                                                          \\[-1.ex]
            &\bf{return} \:  \theta_t                                                     \\[-1.ex]
            &\rule{110mm}{0.4pt}                                                          \\[-1.ex]
       \end{aligned}

    For further details regarding the algorithm we refer to `Decoupled Weight Decay Regularization`_.
    """ + r"""
    Args:
        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 coefficient (default: 1e-2)
        amsgrad (bool, optional): whether to use the AMSGrad variant of this
            algorithm from the paper `On the Convergence of Adam and Beyond`_
            (default: False)
        {maximize}
        {foreach}
        {capturable}
        {differentiable}
        {fused}
    .. _Decoupled Weight Decay Regularization:
        https://arxiv.org/abs/1711.05101
    .. _On the Convergence of Adam and Beyond:
        https://openreview.net/forum?id=ryQu7f-RZ

    """.format(maximize=_maximize_doc,
               foreach=_foreach_doc,
               fused=_fused_doc,
               capturable=_capturable_doc,
               differentiable=_differentiable_doc)


def adamw(
    params: List[Tensor],
    grads: List[Tensor],
    exp_avgs: List[Tensor],
    exp_avg_sqs: List[Tensor],
    max_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,
    fused: Optional[bool] = None,
    grad_scale: Optional[Tensor] = None,
    found_inf: Optional[Tensor] = None,
    *,
    amsgrad: bool,
    beta1: float,
    beta2: float,
    lr: float,
    weight_decay: float,
    eps: float,
    maximize: bool,
):
    r"""Functional API that performs AdamW algorithm computation.

    See :class:`~torch.optim.AdamW` for details.
    """

    if 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"
        )

    # Respect when the user inputs False/True for foreach or fused. We only want to change
    # the default when neither have been user-specified. Note that we default to foreach
    # and pass False to use_fused. This is not a mistake--we want to give the fused impl
    # bake-in time before making it the default, even if it is typically faster.
    if fused is None and foreach is None:
        _, foreach = _default_to_fused_or_foreach(params, differentiable, use_fused=False)
    if fused is None:
        fused = False
    if foreach is None:
        foreach = False

    if foreach and torch.jit.is_scripting():
        raise RuntimeError("torch.jit.script not supported with foreach optimizers")
    if fused and torch.jit.is_scripting():
        raise RuntimeError("torch.jit.script not supported with fused optimizers")

    if fused and not torch.jit.is_scripting():
        func = _fused_adamw
    elif foreach and not torch.jit.is_scripting():
        func = _multi_tensor_adamw
    else:
        func = _single_tensor_adamw

    func(
        params,
        grads,
        exp_avgs,
        exp_avg_sqs,
        max_exp_avg_sqs,
        state_steps,
        amsgrad=amsgrad,
        beta1=beta1,
        beta2=beta2,
        lr=lr,
        weight_decay=weight_decay,
        eps=eps,
        maximize=maximize,
        capturable=capturable,
        differentiable=differentiable,
        grad_scale=grad_scale,
        found_inf=found_inf,
    )


def _single_tensor_adamw(
    params: List[Tensor],
    grads: List[Tensor],
    exp_avgs: List[Tensor],