adapting the code to the cases where the input classes are not instantiated

pytorch_widedeep/initializers.py

@@ -7,7 +7,7 @@ from .wdtypes import *
 
 class Initializer(object):
 
-    def __call__(self, model:nn.Module):
+    def __call__(self, model:TorchModel):
         raise NotImplementedError('Initializer must implement this method')
 
 
@@ -16,7 +16,7 @@ class MultipleInitializers(object):
 	def __init__(self, initializers:Dict[str, Initializer]):
 	    self._initializers = initializers
 
-	def apply(self, model:nn.Module):
+	def apply(self, model:TorchModel):
 		children = list(model.children())
 		for child in children:
 			model_name = child.__class__.__name__.lower()
@@ -35,7 +35,7 @@ class Normal(Initializer):
 	    self.pattern = pattern
 	    super(Normal, self).__init__()
 
-	def __call__(self, submodel:nn.Module):
+	def __call__(self, submodel:TorchModel):
 		for n,p in submodel.named_parameters():
 			if fnmatch(n, self.pattern) and pattern_in:
 				if self.bias and ('bias' in n):
@@ -55,7 +55,7 @@ class Uniform(Initializer):
 	    self.pattern = pattern
 	    super(Uniform, self).__init__()
 
-	def __call__(self, submodel:nn.Module):
+	def __call__(self, submodel:TorchModel):
 		for n,p in submodel.named_parameters():
 			if fnmatch(n, self.pattern) and pattern_in:
 				if self.bias and ('bias' in n):
@@ -74,7 +74,7 @@ class ConstantInitializer(Initializer):
 	    self.pattern = pattern
 	    super(ConstantInitializer, self).__init__()
 
-	def __call__(self, submodel:nn.Module):
+	def __call__(self, submodel:TorchModel):
 		for n,p in submodel.named_parameters():
 			if fnmatch(n, self.pattern) and pattern_in:
 				if self.bias and ('bias' in n):
@@ -92,7 +92,7 @@ class XavierUniform(Initializer):
 	    self.pattern = pattern
 	    super(XavierUniform, self).__init__()
 
-	def __call__(self, submodel:nn.Module):
+	def __call__(self, submodel:TorchModel):
 		for n,p in submodel.named_parameters():
 			if fnmatch(n, self.pattern) and pattern_in:
 				if 'bias' in n: nn.init.constant_(p, val=0)
@@ -106,7 +106,7 @@ class XavierNormal(Initializer):
 	    self.pattern = pattern
 	    super(XavierNormal, self).__init__()
 
-	def __call__(self, submodel:nn.Module):
+	def __call__(self, submodel:TorchModel):
 		for n,p in submodel.named_parameters():
 			if fnmatch(n, self.pattern) and pattern_in:
 				if 'bias' in n: nn.init.constant_(p, val=0)
@@ -121,7 +121,7 @@ class KaimingUniform(Initializer):
 	    self.pattern = pattern
 	    super(KaimingUniform, self).__init__()
 
-	def __call__(self, submodel:nn.Module):
+	def __call__(self, submodel:TorchModel):
 		for n,p in submodel.named_parameters():
 			if fnmatch(n, self.pattern) and pattern_in:
 				if 'bias' in n: nn.init.constant_(p, val=0)
@@ -136,7 +136,7 @@ class KaimingNormal(Initializer):
 	    self.pattern = pattern
 	    super(KaimingNormal, self).__init__()
 
-	def __call__(self, submodel:nn.Module):
+	def __call__(self, submodel:TorchModel):
 		for n,p in submodel.named_parameters():
 			if fnmatch(n, self.pattern) and pattern_in:
 				if 'bias' in n: nn.init.constant_(p, val=0)
@@ -150,7 +150,7 @@ class Orthogonal(Initializer):
 	    self.pattern = pattern
 	    super(Orthogonal, self).__init__()
 
-	def __call__(self, submodel:nn.Module):
+	def __call__(self, submodel:TorchModel):
 		for n,p in submodel.named_parameters():
 			if fnmatch(n, self.pattern) and pattern_in:
 				if 'bias' in n: nn.init.normal_(p, val=0)

pytorch_widedeep/lr_schedulers.py

@@ -7,12 +7,16 @@ from .wdtypes import *
 class MultipleLRScheduler(object):
 
 	def __init__(self,schedulers:Dict[str,LRScheduler]):
-	    self._schedulers = schedulers
+
+		instantiated_schedulers = {}
+		for model_name, scheduler in schedulers.items():
+			if isinstance(scheduler, type):
+				instantiated_schedulers[model_name] = scheduler()
+			else: instantiated_schedulers[model_name] = scheduler
+		self._schedulers = instantiated_schedulers
 
 	def apply(self, optimizers:Dict[str, Optimizer]):
 		for model_name, optimizer in optimizers.items():
-			if isinstance(self._schedulers[model_name], type):
-				self._schedulers[model_name] = self._schedulers[model_name]()
 			self._schedulers[model_name] = self._schedulers[model_name](optimizer)
 
 	def step(self, loss=None):

pytorch_widedeep/optimizers.py

@@ -4,20 +4,22 @@ from torch import nn
 from .radam import RAdam as orgRAdam
 from .wdtypes import *
 
-import pdb
-
 
 class MultipleOptimizers(object):
 
 	def __init__(self, optimizers:Dict[str,Optimizer]):
-	    self._optimizers = optimizers
 
-	def apply(self, model:nn.Module):
+		instantiated_optimizers = {}
+		for model_name, optimizer in optimizers.items():
+			if isinstance(optimizer, type):
+				instantiated_optimizers[model_name] = optimizer()
+			else: instantiated_optimizers[model_name] = optimizer
+		self._optimizers = instantiated_optimizers
+
+	def apply(self, model:TorchModel):
 		children = list(model.children())
 		for child in children:
 			model_name = child.__class__.__name__.lower()
-			if isinstance(self._optimizers[model_name], type):
-				self._optimizers[model_name] = self._optimizers[model_name]()
 			self._optimizers[model_name] = self._optimizers[model_name](child)
 
 	def zero_grad(self):
@@ -39,7 +41,7 @@ class Adam:
 		self.weight_decay=weight_decay
 		self.amsgrad=amsgrad
 
-	def __call__(self, submodel:nn.Module):
+	def __call__(self, submodel:TorchModel) -> Optimizer:
 		self.opt = torch.optim.Adam(submodel.parameters(), lr=self.lr, betas=self.betas, eps=self.eps,
 			weight_decay=self.weight_decay, amsgrad=self.amsgrad)
 		return self.opt
@@ -59,7 +61,7 @@ class RAdam:
 		self.eps=eps
 		self.weight_decay=weight_decay
 
-	def __call__(self, submodel:nn.Module):
+	def __call__(self, submodel:TorchModel) -> Optimizer:
 		self.opt = orgRAdam(submodel.parameters(), lr=self.lr, betas=self.betas, eps=self.eps,
 			weight_decay=self.weight_decay)
 		return self.opt
@@ -81,7 +83,7 @@ class SGD:
 		self.weight_decay=weight_decay
 		self.nesterov=nesterov
 
-	def __call__(self, submodel:nn.Module):
+	def __call__(self, submodel:TorchModel) -> Optimizer:
 		self.opt = torch.optim.SGD(submodel.parameters(), lr=self.lr, momentum=self.momentum,
 			dampening=self.dampening, weight_decay=self.weight_decay, nesterov=self.nesterov)
 		return self.opt