Merge pull request #119 from kozistr/fix/str-method

kozistr · web-flow · commit df9e78d61809 · 2023-03-12T14:46:21.000+09:00
[Fix] `__str__` method supposed to be a function, not property
diff --git a/pyproject.toml b/pyproject.toml
@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "pytorch_optimizer"
-version = "2.5.0"
+version = "2.5.1"
 description = "optimizer & lr scheduler implementations in PyTorch with clean-code, strict types. Also, including useful optimization ideas."
 license = "Apache-2.0"
 authors = ["kozistr <kozistr@gmail.com>"]
diff --git a/pytorch_optimizer/optimizer/adabelief.py b/pytorch_optimizer/optimizer/adabelief.py
@@ -70,7 +70,6 @@ def validate_parameters(self):
         self.validate_weight_decay(self.weight_decay)
         self.validate_epsilon(self.eps)
 
-    @property
     def __str__(self) -> str:
         return 'AdaBelief'
 
@@ -106,7 +105,7 @@ def step(self, closure: CLOSURE = None) -> LOSS:
 
                 grad = p.grad
                 if grad.is_sparse:
-                    raise NoSparseGradientError(self.__str__)
+                    raise NoSparseGradientError(str(self))
 
                 state = self.state[p]
                 if len(state) == 0:
diff --git a/pytorch_optimizer/optimizer/adabound.py b/pytorch_optimizer/optimizer/adabound.py
@@ -68,7 +68,6 @@ def validate_parameters(self):
         self.validate_weight_decay(self.weight_decay)
         self.validate_epsilon(self.eps)
 
-    @property
     def __str__(self) -> str:
         return 'AdaBound'
 
@@ -99,7 +98,7 @@ def step(self, closure: CLOSURE = None) -> LOSS:
 
                 grad = p.grad
                 if grad.is_sparse:
-                    raise NoSparseGradientError(self.__str__)
+                    raise NoSparseGradientError(str(self))
 
                 state = self.state[p]
 
diff --git a/pytorch_optimizer/optimizer/adafactor.py b/pytorch_optimizer/optimizer/adafactor.py
@@ -63,7 +63,6 @@ def validate_parameters(self):
         self.validate_epsilon(self.eps1)
         self.validate_epsilon(self.eps2)
 
-    @property
     def __str__(self) -> str:
         return 'AdaFactor'
 
@@ -134,7 +133,7 @@ def step(self, closure: CLOSURE = None) -> LOSS:
 
                 grad = p.grad
                 if grad.is_sparse:
-                    raise NoSparseGradientError(self.__str__)
+                    raise NoSparseGradientError(str(self))
 
                 state = self.state[p]
 
diff --git a/pytorch_optimizer/optimizer/adai.py b/pytorch_optimizer/optimizer/adai.py
@@ -61,7 +61,6 @@ def validate_parameters(self):
         self.validate_weight_decay(self.weight_decay)
         self.validate_epsilon(self.eps)
 
-    @property
     def __str__(self) -> str:
         return 'Adai'
 
@@ -94,7 +93,7 @@ def step(self, closure: CLOSURE = None) -> LOSS:
 
                 grad = p.grad
                 if grad.is_sparse:
-                    raise NoSparseGradientError(self.__str__)
+                    raise NoSparseGradientError(str(self))
 
                 param_size += p.numel()
 
diff --git a/pytorch_optimizer/optimizer/adamp.py b/pytorch_optimizer/optimizer/adamp.py
@@ -67,7 +67,6 @@ def validate_parameters(self):
         self.validate_weight_decay_ratio(self.wd_ratio)
         self.validate_epsilon(self.eps)
 
-    @property
     def __str__(self) -> str:
         return 'AdamP'
 
@@ -96,7 +95,7 @@ def step(self, closure: CLOSURE = None) -> LOSS:
 
                 grad = p.grad
                 if grad.is_sparse:
-                    raise NoSparseGradientError(self.__str__)
+                    raise NoSparseGradientError(str(self))
 
                 state = self.state[p]
                 if len(state) == 0:
diff --git a/pytorch_optimizer/optimizer/adams.py b/pytorch_optimizer/optimizer/adams.py
@@ -48,7 +48,6 @@ def validate_parameters(self):
         self.validate_weight_decay(self.weight_decay)
         self.validate_epsilon(self.eps)
 
-    @property
     def __str__(self) -> str:
         return 'AdamS'
 
@@ -80,7 +79,7 @@ def step(self, closure: CLOSURE = None) -> LOSS:
 
                 grad = p.grad
                 if grad.is_sparse:
-                    raise NoSparseGradientError(self.__str__)
+                    raise NoSparseGradientError(str(self))
 
                 param_size += p.numel()
 
diff --git a/pytorch_optimizer/optimizer/adan.py b/pytorch_optimizer/optimizer/adan.py
@@ -61,7 +61,6 @@ def validate_parameters(self):
         self.validate_epsilon(self.eps)
         self.validate_norm(self.max_grad_norm)
 
-    @property
     def __str__(self) -> str:
         return 'Adan'
 
@@ -122,7 +121,7 @@ def step(self, closure: CLOSURE = None) -> LOSS:
 
                 grad = p.grad
                 if grad.is_sparse:
-                    raise NoSparseGradientError(self.__str__)
+                    raise NoSparseGradientError(str(self))
 
                 state = self.state[p]
                 if len(state) == 0:
diff --git a/pytorch_optimizer/optimizer/adapnm.py b/pytorch_optimizer/optimizer/adapnm.py
@@ -59,7 +59,6 @@ def validate_parameters(self):
         self.validate_weight_decay(self.weight_decay)
         self.validate_epsilon(self.eps)
 
-    @property
     def __str__(self) -> str:
         return 'AdaPNM'
 
@@ -92,7 +91,7 @@ def step(self, closure: CLOSURE = None) -> LOSS:
 
                 grad = p.grad
                 if grad.is_sparse:
-                    raise NoSparseGradientError(self.__str__)
+                    raise NoSparseGradientError(str(self))
 
                 if group['weight_decouple']:
                     p.mul_(1.0 - group['lr'] * group['weight_decay'])
diff --git a/pytorch_optimizer/optimizer/alig.py b/pytorch_optimizer/optimizer/alig.py
@@ -46,7 +46,6 @@ def validate_parameters(self):
         self.validate_momentum(self.momentum)
         self.validate_epsilon(self.eps)
 
-    @property
     def __str__(self) -> str:
         return 'AliG'
 
@@ -91,7 +90,7 @@ def step(self, closure: CLOSURE = None) -> LOSS:
 
                 grad = p.grad
                 if grad.is_sparse:
-                    raise NoSparseGradientError(self.__str__)
+                    raise NoSparseGradientError(str(self))
 
                 state = self.state[p]
                 if len(state) == 0 and momentum > 0.0:
diff --git a/pytorch_optimizer/optimizer/apollo.py b/pytorch_optimizer/optimizer/apollo.py
@@ -64,7 +64,6 @@ def validate_parameters(self):
         self.validate_weight_decay_type(self.weight_decay_type)
         self.validate_epsilon(self.eps)
 
-    @property
     def __str__(self) -> str:
         return 'Apollo'
 
@@ -110,7 +109,7 @@ def step(self, closure: CLOSURE = None) -> LOSS:
 
                 grad = p.grad
                 if grad.is_sparse:
-                    raise NoSparseGradientError(self.__str__)
+                    raise NoSparseGradientError(str(self))
 
                 state = self.state[p]
                 if len(state) == 0:
diff --git a/pytorch_optimizer/optimizer/dadapt.py b/pytorch_optimizer/optimizer/dadapt.py
@@ -65,7 +65,6 @@ def validate_parameters(self):
         self.validate_weight_decay(self.weight_decay)
         self.validate_epsilon(self.eps)
 
-    @property
     def __str__(self) -> str:
         return 'DAdaptAdaGrad'
 
@@ -296,7 +295,6 @@ def validate_parameters(self):
         self.validate_weight_decay(self.weight_decay)
         self.validate_epsilon(self.eps)
 
-    @property
     def __str__(self) -> str:
         return 'DAdaptAdam'
 
@@ -340,7 +338,7 @@ def step(self, closure: CLOSURE = None) -> LOSS:
 
                 grad = p.grad
                 if grad.is_sparse:
-                    raise NoSparseGradientError(self.__str__)
+                    raise NoSparseGradientError(str(self))
 
                 state = self.state[p]
                 if 'step' not in state:
@@ -440,7 +438,6 @@ def validate_parameters(self):
         self.validate_momentum(self.momentum)
         self.validate_weight_decay(self.weight_decay)
 
-    @property
     def __str__(self) -> str:
         return 'DAdaptSGD'
 
@@ -480,7 +477,7 @@ def step(self, closure: CLOSURE = None) -> LOSS:
 
                 grad = p.grad
                 if grad.is_sparse:
-                    raise NoSparseGradientError(self.__str__)
+                    raise NoSparseGradientError(str(self))
 
                 if weight_decay > 0.0:
                     grad.add_(p, alpha=weight_decay)
diff --git a/pytorch_optimizer/optimizer/diffgrad.py b/pytorch_optimizer/optimizer/diffgrad.py
@@ -50,7 +50,6 @@ def validate_parameters(self):
         self.validate_weight_decay(self.weight_decay)
         self.validate_epsilon(self.eps)
 
-    @property
     def __str__(self) -> str:
         return 'diffGrad'
 
@@ -80,7 +79,7 @@ def step(self, closure: CLOSURE = None) -> LOSS:
 
                 grad = p.grad
                 if grad.is_sparse:
-                    raise NoSparseGradientError(self.__str__)
+                    raise NoSparseGradientError(str(self))
 
                 state = self.state[p]
                 if len(state) == 0:
diff --git a/pytorch_optimizer/optimizer/diffrgrad.py b/pytorch_optimizer/optimizer/diffrgrad.py
@@ -57,7 +57,6 @@ def validate_parameters(self):
         self.validate_weight_decay(self.weight_decay)
         self.validate_epsilon(self.eps)
 
-    @property
     def __str__(self) -> str:
         return 'diffRGrad'
 
@@ -88,7 +87,7 @@ def step(self, closure: CLOSURE = None) -> LOSS:
 
                 grad = p.grad
                 if grad.is_sparse:
-                    raise NoSparseGradientError(self.__str__)
+                    raise NoSparseGradientError(str(self))
 
                 state = self.state[p]
                 if len(state) == 0:
diff --git a/pytorch_optimizer/optimizer/gsam.py b/pytorch_optimizer/optimizer/gsam.py
@@ -84,7 +84,6 @@ def __init__(
     def validate_parameters(self):
         self.validate_alpha(self.alpha)
 
-    @property
     def __str__(self) -> str:
         return 'GSAM'
 
diff --git a/pytorch_optimizer/optimizer/lamb.py b/pytorch_optimizer/optimizer/lamb.py
@@ -66,7 +66,6 @@ def validate_parameters(self):
         self.validate_epsilon(self.eps)
         self.validate_norm(self.max_grad_norm)
 
-    @property
     def __str__(self) -> str:
         return 'Lamb'
 
@@ -125,7 +124,7 @@ def step(self, closure: CLOSURE = None) -> LOSS:
 
                 grad = p.grad
                 if grad.is_sparse:
-                    raise NoSparseGradientError(self.__str__)
+                    raise NoSparseGradientError(str(self))
 
                 if self.pre_norm:
                     grad.div_(grad_norm)
diff --git a/pytorch_optimizer/optimizer/lars.py b/pytorch_optimizer/optimizer/lars.py
@@ -57,7 +57,6 @@ def validate_parameters(self):
         self.validate_trust_coefficient(self.trust_coefficient)
         self.validate_epsilon(self.eps)
 
-    @property
     def __str__(self) -> str:
         return 'Lars'
 
@@ -83,7 +82,7 @@ def step(self, closure: CLOSURE = None) -> LOSS:
 
                 grad = p.grad
                 if grad.is_sparse:
-                    raise NoSparseGradientError(self.__str__)
+                    raise NoSparseGradientError(str(self))
 
                 if p.ndim > 1:  # if not normalization gamma/beta or bias
                     param_norm = p.norm(2.0)
diff --git a/pytorch_optimizer/optimizer/lion.py b/pytorch_optimizer/optimizer/lion.py
@@ -39,7 +39,6 @@ def validate_parameters(self):
         self.validate_betas(self.betas)
         self.validate_weight_decay(self.weight_decay)
 
-    @property
     def __str__(self) -> str:
         return 'Lion'
 
@@ -67,7 +66,7 @@ def step(self, closure: CLOSURE = None) -> LOSS:
 
                 grad = p.grad
                 if grad.is_sparse:
-                    raise NoSparseGradientError(self.__str__)
+                    raise NoSparseGradientError(str(self))
 
                 state = self.state[p]
 
diff --git a/pytorch_optimizer/optimizer/madgrad.py b/pytorch_optimizer/optimizer/madgrad.py
@@ -51,7 +51,6 @@ def validate_parameters(self):
         self.validate_momentum(self.momentum)
         self.validate_epsilon(self.eps)
 
-    @property
     def __str__(self) -> str:
         return 'MADGRAD'
 
@@ -103,13 +102,13 @@ def step(self, closure: CLOSURE = None) -> LOSS:
                         state['x0'] = torch.clone(p).detach()
 
                 if momentum > 0.0 and grad.is_sparse:
-                    raise NoSparseGradientError(self.__str__, note='momentum > 0.0')
+                    raise NoSparseGradientError(str(self), note='momentum > 0.0')
 
                 grad_sum_sq, s = state['grad_sum_sq'], state['s']
 
                 if weight_decay > 0.0 and not self.decouple_decay:
                     if grad.is_sparse:
-                        raise NoSparseGradientError(self.__str__, note='weight_decay')
+                        raise NoSparseGradientError(str(self), note='weight_decay')
 
                     # original implementation. not AdamW style
                     grad.add_(p, alpha=weight_decay)
diff --git a/pytorch_optimizer/optimizer/nero.py b/pytorch_optimizer/optimizer/nero.py
@@ -29,7 +29,6 @@ def validate_parameters(self):
         self.validate_learning_rate(self.lr)
         self.validate_beta(self.beta)
 
-    @property
     def __str__(self) -> str:
         return 'Nero'
 
@@ -64,7 +63,7 @@ def step(self, closure: CLOSURE = None) -> LOSS:
 
                 grad = p.grad
                 if grad.is_sparse:
-                    raise NoSparseGradientError(self.__str__)
+                    raise NoSparseGradientError(str(self))
 
                 state = self.state[p]
                 if len(state) == 0:
diff --git a/pytorch_optimizer/optimizer/novograd.py b/pytorch_optimizer/optimizer/novograd.py
@@ -48,7 +48,6 @@ def validate_parameters(self):
         self.validate_weight_decay(self.weight_decay)
         self.validate_epsilon(self.eps)
 
-    @property
     def __str__(self) -> str:
         return 'NovoGrad'
 
@@ -95,7 +94,7 @@ def step(self, closure: CLOSURE = None) -> LOSS:
 
                 grad = p.grad
                 if grad.is_sparse:
-                    raise NoSparseGradientError(self.__str__)
+                    raise NoSparseGradientError(str(self))
 
                 state = self.state[p]
                 g_2 = grad ** 2  # fmt: skip
diff --git a/pytorch_optimizer/optimizer/pnm.py b/pytorch_optimizer/optimizer/pnm.py
@@ -50,7 +50,6 @@ def validate_parameters(self):
         self.validate_weight_decay(self.weight_decay)
         self.validate_epsilon(self.eps)
 
-    @property
     def __str__(self) -> str:
         return 'PNM'
 
@@ -80,7 +79,7 @@ def step(self, closure: CLOSURE = None) -> LOSS:
 
                 grad = p.grad
                 if grad.is_sparse:
-                    raise NoSparseGradientError(self.__str__)
+                    raise NoSparseGradientError(str(self))
 
                 if group['weight_decouple']:
                     p.mul_(1.0 - group['lr'] * group['weight_decay'])
diff --git a/pytorch_optimizer/optimizer/radam.py b/pytorch_optimizer/optimizer/radam.py
diff --git a/pytorch_optimizer/optimizer/ralamb.py b/pytorch_optimizer/optimizer/ralamb.py
diff --git a/pytorch_optimizer/optimizer/ranger.py b/pytorch_optimizer/optimizer/ranger.py
diff --git a/pytorch_optimizer/optimizer/ranger21.py b/pytorch_optimizer/optimizer/ranger21.py
diff --git a/pytorch_optimizer/optimizer/sam.py b/pytorch_optimizer/optimizer/sam.py
diff --git a/pytorch_optimizer/optimizer/sgdp.py b/pytorch_optimizer/optimizer/sgdp.py
diff --git a/pytorch_optimizer/optimizer/shampoo.py b/pytorch_optimizer/optimizer/shampoo.py
diff --git a/tests/test_optimizers.py b/tests/test_optimizers.py
