✨ Minor fixes to AdamWCD, new AdamCD

LICENSE

@@ -176,7 +176,7 @@
 
    END OF TERMS AND CONDITIONS
 
-   Copyright 2020- Emanuele Ballarin <emanuele@ballarin.cc> and affiliates.
+   *Copyright 2020- Emanuele Ballarin <emanuele@ballarin.cc> and affiliates.*
 
    Licensed under the Apache License, Version 2.0 (the "License");
    you may not use this file except in compliance with the License.

pyromaniac/optim/pyro/__init__.py

@@ -1,6 +1,8 @@
 from pyromaniac.optim.pyro.adamwcd import AdamWCD
+from pyromaniac.optim.pyro.adamcd import AdamCD
 
 
 __all__ = [
     "AdamWCD",
+    "AdamCD",
 ]

pyromaniac/optim/pyro/adamcd.py

@@ -0,0 +1,14 @@
+# Copyright (c) 2017- Uber Technologies, Inc.
+# Copyright (c) 2020- AI-CPS@UniTS
+# Copyright (c) 2020- Emanuele Ballarin <emanuele@ballarin.cc>
+# SPDX-License-Identifier: Apache-2.0
+
+from pyro.optim.optim import PyroOptim
+from pyromaniac.optim.torch import AdamCD as pt_AdamCD
+
+
+def AdamCD(optim_args):
+    """
+    Wraps :class:`pyromaniac.optim.adamwcd.AdamCD` with :class:`~pyro.optim.optim.PyroOptim`.
+    """
+    return PyroOptim(pt_AdamCD, optim_args)

pyromaniac/optim/torch/__init__.py

@@ -1,3 +1,5 @@
 from .adamwcd import AdamWCD
+from .adamcd import AdamCD
 
 del adamwcd
+del adamcd

pyromaniac/optim/torch/adamcd.py

@@ -0,0 +1,173 @@
+# Copyright (c) 2016- Facebook, Inc
+# Copyright (c) 2017- Uber Technologies, Inc.
+# Copyright (c) 2020- AI-CPS@UniTS
+# Copyright (c) 2020- Emanuele Ballarin <emanuele@ballarin.cc>
+# SPDX-License-Identifier: Apache-2.0
+
+import math
+import torch
+from torch.optim.optimizer import Optimizer, required
+
+
+class AdamCD(Optimizer):
+    r"""Implements AdamCD algorithm.
+
+    It has been proposed in `Adam: A Method for Stochastic Optimization`_.
+    AdamCD further introduces gradient clipping and epoch-wise learning rate decay, similarly to the `Uber Pyro ClippedAdam variant of Adam`_.
+
+    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)
+        amsgrad (boolean, optional): whether to use the AMSGrad variant of this
+            algorithm from the paper `On the Convergence of Adam and Beyond`_
+            (default: False)
+        clip_norm (float, optional): magnitude of norm to which gradients are clipped (default: 10.0)
+        lrd (float, optional): rate at which learning rate decays (default: 1.0)
+        lrd_epoch (int, optional): number of optimizer steps every which to perform one learning rate decay step (default: 1)
+
+    .. _Adam\: A Method for Stochastic Optimization:
+        https://arxiv.org/abs/1412.6980
+    .. _Uber Pyro ClippedAdam variant of Adam:
+        http://docs.pyro.ai/en/latest/_modules/pyro/optim/clipped_adam.html
+    .. _On the Convergence of Adam and Beyond:
+        https://openreview.net/forum?id=ryQu7f-RZ
+    """
+
+    def __init__(
+        self,
+        params,
+        lr=1e-3,
+        betas=(0.9, 0.999),
+        eps=1e-8,
+        weight_decay=0,
+        amsgrad=False,
+        clip_norm=10.0,
+        lrd=1.0,
+        lrd_epoch: int = 1,
+    ):
+        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))
+        if not 0.0 <= clip_norm:
+            raise ValueError("Invalid clip_norm value: {}".format(clip_norm))
+        if not 0.0 <= lrd:
+            raise ValueError("Invalid lrd value: {}".format(lrd))
+        if not 1 <= lrd_epoch:
+            raise ValueError("Invalid lrd_epoch value: {}".format(lrd_epoch))
+        defaults = dict(
+            lr=lr,
+            betas=betas,
+            eps=eps,
+            weight_decay=weight_decay,
+            amsgrad=amsgrad,
+            clip_norm=clip_norm,
+            lrd=lrd,
+            lrd_epoch=lrd_epoch,
+        )
+        self.lrd_epoch_persist = lrd_epoch
+        super(AdamCD, self).__init__(params, defaults)
+
+    def __setstate__(self, state):
+        super(AdamCD, self).__setstate__(state)
+        for group in self.param_groups:
+            group.setdefault("amsgrad", False)
+
+    @torch.no_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.
+        """
+        loss = None
+        if closure is not None:
+            with torch.enable_grad():
+                loss = closure()
+
+        for group in self.param_groups:
+
+            # Apply a decay factor to learning rate, but only every lrd_epoch steps
+            group["lrd_epoch"] -= 1
+            if group["lrd_epoch"] == 0:
+                group["lr"] *= group["lrd"]
+                group["lrd_epoch"] = self.lrd_epoch_persist
+
+            for p in group["params"]:
+                if p.grad is None:
+                    continue
+                grad = p.grad
+
+                # Clamp the gradient
+                grad.clamp_(-group["clip_norm"], group["clip_norm"])
+
+                if grad.is_sparse:
+                    raise RuntimeError(
+                        "AdamCD does not support sparse gradients, please consider SparseAdam instead"
+                    )
+                amsgrad = group["amsgrad"]
+
+                state = self.state[p]
+
+                # State initialization
+                if len(state) == 0:
+                    state["step"] = 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_avg, exp_avg_sq = state["exp_avg"], state["exp_avg_sq"]
+                if amsgrad:
+                    max_exp_avg_sq = state["max_exp_avg_sq"]
+                beta1, beta2 = group["betas"]
+
+                state["step"] += 1
+                bias_correction1 = 1 - beta1 ** state["step"]
+                bias_correction2 = 1 - beta2 ** state["step"]
+
+                if group["weight_decay"] != 0:
+                    grad = grad.add(p, alpha=group["weight_decay"])
+
+                # Decay the first and second moment running average coefficient
+                exp_avg.mul_(beta1).add_(grad, alpha=1 - beta1)
+                exp_avg_sq.mul_(beta2).addcmul_(grad, grad, value=1 - beta2)
+                if amsgrad:
+                    # Maintains the maximum of all 2nd moment running avg. till now
+                    torch.max(max_exp_avg_sq, exp_avg_sq, out=max_exp_avg_sq)
+                    # Use the max. for normalizing running avg. of gradient
+                    denom = (max_exp_avg_sq.sqrt() / math.sqrt(bias_correction2)).add_(
+                        group["eps"]
+                    )
+                else:
+                    denom = (exp_avg_sq.sqrt() / math.sqrt(bias_correction2)).add_(
+                        group["eps"]
+                    )
+
+                step_size = group["lr"] / bias_correction1
+
+                p.addcdiv_(exp_avg, denom, value=-step_size)
+
+        return loss

pyromaniac/optim/torch/adamwcd.py

@@ -29,15 +29,18 @@ class AdamWCD(Optimizer):
         amsgrad (boolean, optional): whether to use the AMSGrad variant of this
             algorithm from the paper `On the Convergence of Adam and Beyond`_
             (default: False)
+        clip_norm (float, optional): magnitude of norm to which gradients are clipped (default: 10.0)
+        lrd (float, optional): rate at which learning rate decays (default: 1.0)
+        lrd_epoch (int, optional): number of optimizer steps every which to perform one learning rate decay step (default: 1)
 
     .. _Adam\: A Method for Stochastic Optimization:
         https://arxiv.org/abs/1412.6980
     .. _Decoupled Weight Decay Regularization:
         https://arxiv.org/abs/1711.05101
-    .. _On the Convergence of Adam and Beyond:
-        https://openreview.net/forum?id=ryQu7f-RZ
     .. _Uber Pyro ClippedAdam variant of Adam:
         http://docs.pyro.ai/en/latest/_modules/pyro/optim/clipped_adam.html
+    .. _On the Convergence of Adam and Beyond:
+        https://openreview.net/forum?id=ryQu7f-RZ
     """
 
     def __init__(
@@ -46,7 +49,6 @@ def __init__(
         lr=1e-3,
         betas=(0.9, 0.999),
         eps=1e-8,
-        #weight_decay=1e-2,
         weight_decay=0,
         amsgrad=False,
         clip_norm=10.0,
@@ -102,7 +104,7 @@ def step(self, closure=None):
 
         for group in self.param_groups:
 
-            # Apply a decay factor to learning rate, but only every ldr_epoch steps
+            # Apply a decay factor to learning rate, but only every lrd_epoch steps
             group["lrd_epoch"] -= 1
             if group["lrd_epoch"] == 0:
                 group["lr"] *= group["lrd"]