Skip to content

Added parameter range checks for all optimizers #6000

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
apaszke merged 4 commits into from
Mar 28, 2018
Merged

Added parameter range checks for all optimizers #6000

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
434225e
Added parameter checks for optimizers
lazypanda1 Mar 15, 2018
f51ba5d
updated constraint for rho in adadelta
lazypanda1 Mar 26, 2018
1608848
Fixed bound for momentum in sgd
lazypanda1 Mar 26, 2018
3da7f95
Added tests
lazypanda1 Mar 26, 2018
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
16 changes: 16 additions & 0 deletions test/test_optim.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -260,6 +260,8 @@ def test_sgd(self):
self._build_params_dict_single(weight, bias, lr=1e-2),
lr=1e-3)
)
with self.assertRaisesRegex(ValueError, "Invalid momentum value: -0.5"):
optim.SGD(None, lr=1e-2, momentum=-0.5)

def test_sgd_sparse(self):
self._test_rosenbrock_sparse(
Expand Down Expand Up @@ -300,6 +302,8 @@ def test_sparse_adam(self):
lambda params: optim.SparseAdam(params, lr=4e-2),
True
)
with self.assertRaisesRegex(ValueError, "Invalid beta parameter at index 0: 1.0"):
optim.SparseAdam(None, lr=1e-2, betas=(1.0, 0.0))

def test_adadelta(self):
self._test_rosenbrock(
Expand All @@ -321,6 +325,8 @@ def test_adadelta(self):
lambda weight, bias: optim.Adadelta(
self._build_params_dict(weight, bias, rho=0.95))
)
with self.assertRaisesRegex(ValueError, "Invalid rho value: 1.1"):
optim.Adadelta(None, lr=1e-2, rho=1.1)

def test_adagrad(self):
self._test_rosenbrock(
Expand All @@ -343,6 +349,8 @@ def test_adagrad(self):
self._build_params_dict(weight, bias, lr=1e-2),
lr=1e-1)
)
with self.assertRaisesRegex(ValueError, "Invalid lr_decay value: -0.5"):
optim.Adagrad(None, lr=1e-2, lr_decay=-0.5)

def test_adagrad_sparse(self):
self._test_rosenbrock_sparse(

This comment was marked as off-topic.

Sign in to view

This comment was marked as off-topic.

Sign in to view

This comment was marked as off-topic.

Sign in to view

Expand Down Expand Up @@ -370,6 +378,8 @@ def test_adamax(self):
self._build_params_dict(weight, bias, lr=1e-2),
lr=1e-1)
)
with self.assertRaisesRegex(ValueError, "Invalid beta parameter at index 1: 1.0"):
optim.Adamax(None, lr=1e-2, betas=(0.0, 1.0))

def test_rmsprop(self):
self._test_rosenbrock(
Expand All @@ -392,6 +402,8 @@ def test_rmsprop(self):
self._build_params_dict(weight, bias, lr=1e-3),
lr=1e-2)
)
with self.assertRaisesRegex(ValueError, "Invalid momentum value: -1.0"):
optim.RMSprop(None, lr=1e-2, momentum=-1.0)

def test_asgd(self):
self._test_rosenbrock(
Expand All @@ -414,6 +426,8 @@ def test_asgd(self):
self._build_params_dict(weight, bias, lr=1e-2),
lr=1e-3, t0=100)
)
with self.assertRaisesRegex(ValueError, "Invalid weight_decay value: -0.5"):
optim.ASGD(None, lr=1e-2, weight_decay=-0.5)

def test_rprop(self):
self._test_rosenbrock(
Expand All @@ -436,6 +450,8 @@ def test_rprop(self):
self._build_params_dict(weight, bias, lr=1e-2),
lr=1e-3)
)
with self.assertRaisesRegex(ValueError, "Invalid eta values: 1.0, 0.5"):
optim.Rprop(None, lr=1e-2, etas=(1.0, 0.5))

def test_lbfgs(self):
self._test_rosenbrock(
Expand Down
9 changes: 9 additions & 0 deletions torch/optim/adadelta.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -23,6 +23,15 @@ class Adadelta(Optimizer):
"""

def __init__(self, params, lr=1.0, rho=0.9, eps=1e-6, weight_decay=0):
if not 0.0 <= lr:
raise ValueError("Invalid learning rate: {}".format(lr))
if not 0.0 <= rho <= 1.0:
raise ValueError("Invalid rho value: {}".format(rho))
if not 0.0 <= eps:
raise ValueError("Invalid epsilon value: {}".format(eps))
if not 0.0 <= weight_decay:
raise ValueError("Invalid weight_decay value: {}".format(weight_decay))

defaults = dict(lr=lr, rho=rho, eps=eps, weight_decay=weight_decay)
super(Adadelta, self).__init__(params, defaults)

Expand Down
7 changes: 7 additions & 0 deletions torch/optim/adagrad.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -20,6 +20,13 @@ class Adagrad(Optimizer):
"""

def __init__(self, params, lr=1e-2, lr_decay=0, weight_decay=0):
if not 0.0 <= lr:
raise ValueError("Invalid learning rate: {}".format(lr))
if not 0.0 <= lr_decay:
raise ValueError("Invalid lr_decay value: {}".format(lr_decay))
if not 0.0 <= weight_decay:
raise ValueError("Invalid weight_decay value: {}".format(weight_decay))

defaults = dict(lr=lr, lr_decay=lr_decay, weight_decay=weight_decay)
super(Adagrad, self).__init__(params, defaults)

Expand Down
4 changes: 4 additions & 0 deletions torch/optim/adam.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -28,6 +28,10 @@ class Adam(Optimizer):

def __init__(self, params, lr=1e-3, betas=(0.9, 0.999), eps=1e-8,
weight_decay=0, amsgrad=False):
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:
Expand Down
11 changes: 11 additions & 0 deletions torch/optim/adamax.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -22,6 +22,17 @@ class Adamax(Optimizer):

def __init__(self, params, lr=2e-3, betas=(0.9, 0.999), eps=1e-8,
weight_decay=0):
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)
super(Adamax, self).__init__(params, defaults)

Expand Down
5 changes: 5 additions & 0 deletions torch/optim/asgd.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -23,6 +23,11 @@ class ASGD(Optimizer):
"""

def __init__(self, params, lr=1e-2, lambd=1e-4, alpha=0.75, t0=1e6, weight_decay=0):
if not 0.0 <= lr:
raise ValueError("Invalid learning rate: {}".format(lr))
if not 0.0 <= weight_decay:
raise ValueError("Invalid weight_decay value: {}".format(weight_decay))

defaults = dict(lr=lr, lambd=lambd, alpha=alpha, t0=t0,
weight_decay=weight_decay)
super(ASGD, self).__init__(params, defaults)
Expand Down
11 changes: 11 additions & 0 deletions torch/optim/rmsprop.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -26,6 +26,17 @@ class RMSprop(Optimizer):
"""

def __init__(self, params, lr=1e-2, alpha=0.99, eps=1e-8, weight_decay=0, momentum=0, centered=False):
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 <= momentum:
raise ValueError("Invalid momentum value: {}".format(momentum))
if not 0.0 <= weight_decay:
raise ValueError("Invalid weight_decay value: {}".format(weight_decay))
if not 0.0 <= alpha:
raise ValueError("Invalid alpha value: {}".format(alpha))

defaults = dict(lr=lr, momentum=momentum, alpha=alpha, eps=eps, centered=centered, weight_decay=weight_decay)
super(RMSprop, self).__init__(params, defaults)

Expand Down
5 changes: 5 additions & 0 deletions torch/optim/rprop.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -18,6 +18,11 @@ class Rprop(Optimizer):
"""

def __init__(self, params, lr=1e-2, etas=(0.5, 1.2), step_sizes=(1e-6, 50)):
if not 0.0 <= lr:
raise ValueError("Invalid learning rate: {}".format(lr))
if not 0.0 < etas[0] < 1.0 < etas[1]:
raise ValueError("Invalid eta values: {}, {}".format(etas[0], etas[1]))

defaults = dict(lr=lr, etas=etas, step_sizes=step_sizes)
super(Rprop, self).__init__(params, defaults)

Expand Down
7 changes: 7 additions & 0 deletions torch/optim/sgd.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -50,6 +50,13 @@ class SGD(Optimizer):

def __init__(self, params, lr=required, momentum=0, dampening=0,
weight_decay=0, nesterov=False):
if not 0.0 <= lr:
raise ValueError("Invalid learning rate: {}".format(lr))
if not 0.0 <= momentum:
raise ValueError("Invalid momentum value: {}".format(momentum))
if not 0.0 <= weight_decay:
raise ValueError("Invalid weight_decay value: {}".format(weight_decay))

defaults = dict(lr=lr, momentum=momentum, dampening=dampening,
weight_decay=weight_decay, nesterov=nesterov)
if nesterov and (momentum <= 0 or dampening != 0):
Expand Down
8 changes: 8 additions & 0 deletions torch/optim/sparse_adam.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -23,6 +23,14 @@ class SparseAdam(Optimizer):
"""

def __init__(self, params, lr=1e-3, betas=(0.9, 0.999), eps=1e-8):
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]))
defaults = dict(lr=lr, betas=betas, eps=eps)
super(SparseAdam, self).__init__(params, defaults)

Expand Down