# Copyright (c) Microsoft Corporation.
# Licensed under the MIT license.
import copy
import logging
import torch
import torch.nn as nn
from nni.nas.pytorch.trainer import Trainer
from nni.nas.pytorch.utils import AverageMeterGroup
from .mutator import DartsMutator
logger = logging.getLogger(__name__)
[docs]class DartsTrainer(Trainer):
"""
DARTS trainer.
Parameters
----------
model : nn.Module
PyTorch model to be trained.
loss : callable
Receives logits and ground truth label, return a loss tensor.
metrics : callable
Receives logits and ground truth label, return a dict of metrics.
optimizer : Optimizer
The optimizer used for optimizing the model.
num_epochs : int
Number of epochs planned for training.
dataset_train : Dataset
Dataset for training. Will be split for training weights and architecture weights.
dataset_valid : Dataset
Dataset for testing.
mutator : DartsMutator
Use in case of customizing your own DartsMutator. By default will instantiate a DartsMutator.
batch_size : int
Batch size.
workers : int
Workers for data loading.
device : torch.device
``torch.device("cpu")`` or ``torch.device("cuda")``.
log_frequency : int
Step count per logging.
callbacks : list of Callback
list of callbacks to trigger at events.
arc_learning_rate : float
Learning rate of architecture parameters.
unrolled : float
``True`` if using second order optimization, else first order optimization.
"""
def __init__(self, model, loss, metrics,
optimizer, num_epochs, dataset_train, dataset_valid,
mutator=None, batch_size=64, workers=4, device=None, log_frequency=None,
callbacks=None, arc_learning_rate=3.0E-4, unrolled=False):
super().__init__(model, mutator if mutator is not None else DartsMutator(model),
loss, metrics, optimizer, num_epochs, dataset_train, dataset_valid,
batch_size, workers, device, log_frequency, callbacks)
self.ctrl_optim = torch.optim.Adam(self.mutator.parameters(), arc_learning_rate, betas=(0.5, 0.999),
weight_decay=1.0E-3)
self.unrolled = unrolled
n_train = len(self.dataset_train)
split = n_train // 2
indices = list(range(n_train))
train_sampler = torch.utils.data.sampler.SubsetRandomSampler(indices[:split])
valid_sampler = torch.utils.data.sampler.SubsetRandomSampler(indices[split:])
self.train_loader = torch.utils.data.DataLoader(self.dataset_train,
batch_size=batch_size,
sampler=train_sampler,
num_workers=workers)
self.valid_loader = torch.utils.data.DataLoader(self.dataset_train,
batch_size=batch_size,
sampler=valid_sampler,
num_workers=workers)
self.test_loader = torch.utils.data.DataLoader(self.dataset_valid,
batch_size=batch_size,
num_workers=workers)
[docs] def train_one_epoch(self, epoch):
self.model.train()
self.mutator.train()
meters = AverageMeterGroup()
for step, ((trn_X, trn_y), (val_X, val_y)) in enumerate(zip(self.train_loader, self.valid_loader)):
trn_X, trn_y = trn_X.to(self.device), trn_y.to(self.device)
val_X, val_y = val_X.to(self.device), val_y.to(self.device)
# phase 1. architecture step
self.ctrl_optim.zero_grad()
if self.unrolled:
self._unrolled_backward(trn_X, trn_y, val_X, val_y)
else:
self._backward(val_X, val_y)
self.ctrl_optim.step()
# phase 2: child network step
self.optimizer.zero_grad()
logits, loss = self._logits_and_loss(trn_X, trn_y)
loss.backward()
nn.utils.clip_grad_norm_(self.model.parameters(), 5.) # gradient clipping
self.optimizer.step()
metrics = self.metrics(logits, trn_y)
metrics["loss"] = loss.item()
meters.update(metrics)
if self.log_frequency is not None and step % self.log_frequency == 0:
logger.info("Epoch [%s/%s] Step [%s/%s] %s", epoch + 1,
self.num_epochs, step + 1, len(self.train_loader), meters)
[docs] def validate_one_epoch(self, epoch):
self.model.eval()
self.mutator.eval()
meters = AverageMeterGroup()
with torch.no_grad():
self.mutator.reset()
for step, (X, y) in enumerate(self.test_loader):
X, y = X.to(self.device), y.to(self.device)
logits = self.model(X)
metrics = self.metrics(logits, y)
meters.update(metrics)
if self.log_frequency is not None and step % self.log_frequency == 0:
logger.info("Epoch [%s/%s] Step [%s/%s] %s", epoch + 1,
self.num_epochs, step + 1, len(self.test_loader), meters)
def _logits_and_loss(self, X, y):
self.mutator.reset()
logits = self.model(X)
loss = self.loss(logits, y)
self._write_graph_status()
return logits, loss
def _backward(self, val_X, val_y):
"""
Simple backward with gradient descent
"""
_, loss = self._logits_and_loss(val_X, val_y)
loss.backward()
def _unrolled_backward(self, trn_X, trn_y, val_X, val_y):
"""
Compute unrolled loss and backward its gradients
"""
backup_params = copy.deepcopy(tuple(self.model.parameters()))
# do virtual step on training data
lr = self.optimizer.param_groups[0]["lr"]
momentum = self.optimizer.param_groups[0]["momentum"]
weight_decay = self.optimizer.param_groups[0]["weight_decay"]
self._compute_virtual_model(trn_X, trn_y, lr, momentum, weight_decay)
# calculate unrolled loss on validation data
# keep gradients for model here for compute hessian
_, loss = self._logits_and_loss(val_X, val_y)
w_model, w_ctrl = tuple(self.model.parameters()), tuple(self.mutator.parameters())
w_grads = torch.autograd.grad(loss, w_model + w_ctrl)
d_model, d_ctrl = w_grads[:len(w_model)], w_grads[len(w_model):]
# compute hessian and final gradients
hessian = self._compute_hessian(backup_params, d_model, trn_X, trn_y)
with torch.no_grad():
for param, d, h in zip(w_ctrl, d_ctrl, hessian):
# gradient = dalpha - lr * hessian
param.grad = d - lr * h
# restore weights
self._restore_weights(backup_params)
def _compute_virtual_model(self, X, y, lr, momentum, weight_decay):
"""
Compute unrolled weights w`
"""
# don't need zero_grad, using autograd to calculate gradients
_, loss = self._logits_and_loss(X, y)
gradients = torch.autograd.grad(loss, self.model.parameters())
with torch.no_grad():
for w, g in zip(self.model.parameters(), gradients):
m = self.optimizer.state[w].get("momentum_buffer", 0.)
w = w - lr * (momentum * m + g + weight_decay * w)
def _restore_weights(self, backup_params):
with torch.no_grad():
for param, backup in zip(self.model.parameters(), backup_params):
param.copy_(backup)
def _compute_hessian(self, backup_params, dw, trn_X, trn_y):
"""
dw = dw` { L_val(w`, alpha) }
w+ = w + eps * dw
w- = w - eps * dw
hessian = (dalpha { L_trn(w+, alpha) } - dalpha { L_trn(w-, alpha) }) / (2*eps)
eps = 0.01 / ||dw||
"""
self._restore_weights(backup_params)
norm = torch.cat([w.view(-1) for w in dw]).norm()
eps = 0.01 / norm
if norm < 1E-8:
logger.warning("In computing hessian, norm is smaller than 1E-8, cause eps to be %.6f.", norm.item())
dalphas = []
for e in [eps, -2. * eps]:
# w+ = w + eps*dw`, w- = w - eps*dw`
with torch.no_grad():
for p, d in zip(self.model.parameters(), dw):
p += e * d
_, loss = self._logits_and_loss(trn_X, trn_y)
dalphas.append(torch.autograd.grad(loss, self.mutator.parameters()))
dalpha_pos, dalpha_neg = dalphas # dalpha { L_trn(w+) }, # dalpha { L_trn(w-) }
hessian = [(p - n) / 2. * eps for p, n in zip(dalpha_pos, dalpha_neg)]
return hessian