feat: Added model_fn to support non-standard model function in create_trainer (#3055)

ignite/engine/__init__.py

@@ -51,6 +51,7 @@
     model_transform: Callable[[Any], Any] = lambda output: output,
     output_transform: Callable[[Any, Any, Any, torch.Tensor], Any] = lambda x, y, y_pred, loss: loss.item(),
     gradient_accumulation_steps: int = 1,
+    model_fn: Callable[[torch.nn.Module, Any], Any] = lambda model, x: model(x),
 ) -> Callable:
     """Factory function for supervised training.
 
@@ -71,6 +72,8 @@
             to be assigned to engine's state.output after each iteration. Default is returning `loss.item()`.
         gradient_accumulation_steps: Number of steps the gradients should be accumulated across.
             (default: 1 (means no gradient accumulation))
+        model_fn: the model function that receives `model` and `x`, and returns `y_pred`.
+
     Returns:
         Callable: update function.
 
@@ -91,6 +94,8 @@
         Added Gradient Accumulation.
     .. versionchanged:: 0.4.11
         Added `model_transform` to transform model's output
+    .. versionchanged:: 0.4.13
+        Added `model_fn` to customize model's application on the sample
     """
 
     if gradient_accumulation_steps <= 0:
@@ -104,7 +109,7 @@
             optimizer.zero_grad()
         model.train()
         x, y = prepare_batch(batch, device=device, non_blocking=non_blocking)
-        output = model(x)
+        output = model_fn(model, x)
         y_pred = model_transform(output)
         loss = loss_fn(y_pred, y)
         if gradient_accumulation_steps > 1:
@@ -128,6 +133,7 @@
     output_transform: Callable[[Any, Any, Any, torch.Tensor], Any] = lambda x, y, y_pred, loss: loss.item(),
     scaler: Optional["torch.cuda.amp.GradScaler"] = None,
     gradient_accumulation_steps: int = 1,
+    model_fn: Callable[[torch.nn.Module, Any], Any] = lambda model, x: model(x),
 ) -> Callable:
     """Factory function for supervised training using ``torch.cuda.amp``.
 
@@ -149,6 +155,7 @@
         scaler: GradScaler instance for gradient scaling. (default: None)
         gradient_accumulation_steps: Number of steps the gradients should be accumulated across.
             (default: 1 (means no gradient accumulation))
+        model_fn: the model function that receives `model` and `x`, and returns `y_pred`.
 
     Returns:
         Callable: update function
@@ -171,6 +178,8 @@
         Added Gradient Accumulation.
     .. versionchanged:: 0.4.11
         Added `model_transform` to transform model's output
+    .. versionchanged:: 0.4.13
+        Added `model_fn` to customize model's application on the sample
     """
 
     try:
@@ -190,7 +199,7 @@
         model.train()
         x, y = prepare_batch(batch, device=device, non_blocking=non_blocking)
         with autocast(enabled=True):
-            output = model(x)
+            output = model_fn(model, x)
             y_pred = model_transform(output)
             loss = loss_fn(y_pred, y)
             if gradient_accumulation_steps > 1:
@@ -219,6 +228,7 @@
     model_transform: Callable[[Any], Any] = lambda output: output,
     output_transform: Callable[[Any, Any, Any, torch.Tensor], Any] = lambda x, y, y_pred, loss: loss.item(),
     gradient_accumulation_steps: int = 1,
+    model_fn: Callable[[torch.nn.Module, Any], Any] = lambda model, x: model(x),
 ) -> Callable:
     """Factory function for supervised training using apex.
 
@@ -239,6 +249,7 @@
             to be assigned to engine's state.output after each iteration. Default is returning `loss.item()`.
         gradient_accumulation_steps: Number of steps the gradients should be accumulated across.
             (default: 1 (means no gradient accumulation))
+        model_fn: the model function that receives `model` and `x`, and returns `y_pred`.
 
     Returns:
         Callable: update function.
@@ -260,6 +271,8 @@
         Added Gradient Accumulation.
     .. versionchanged:: 0.4.11
         Added `model_transform` to transform model's output
+    .. versionchanged:: 0.4.13
+        Added `model_fn` to customize model's application on the sample
     """
 
     try:
@@ -278,7 +291,7 @@
             optimizer.zero_grad()
         model.train()
         x, y = prepare_batch(batch, device=device, non_blocking=non_blocking)
-        output = model(x)
+        output = model_fn(model, x)
         y_pred = model_transform(output)
         loss = loss_fn(y_pred, y)
         if gradient_accumulation_steps > 1:
@@ -302,6 +315,7 @@
     model_transform: Callable[[Any], Any] = lambda output: output,
     output_transform: Callable[[Any, Any, Any, torch.Tensor], Any] = lambda x, y, y_pred, loss: loss.item(),
     gradient_accumulation_steps: int = 1,
+    model_fn: Callable[[torch.nn.Module, Any], Any] = lambda model, x: model(x),
 ) -> Callable:
     """Factory function for supervised training using ``torch_xla``.
 
@@ -322,6 +336,7 @@
             to be assigned to engine's state.output after each iteration. Default is returning `loss.item()`.
         gradient_accumulation_steps: Number of steps the gradients should be accumulated across.
             (default: 1 (means no gradient accumulation))
+        model_fn: the model function that receives `model` and `x`, and returns `y_pred`.
 
     Returns:
         Callable: update function.
@@ -343,6 +358,8 @@
        Added Gradient Accumulation argument for all supervised training methods.
     .. versionchanged:: 0.4.11
         Added `model_transform` to transform model's output
+    .. versionchanged:: 0.4.13
+        Added `model_fn` to customize model's application on the sample
     """
     try:
         import torch_xla.core.xla_model as xm
@@ -360,7 +377,7 @@
             optimizer.zero_grad()
         model.train()
         x, y = prepare_batch(batch, device=device, non_blocking=non_blocking)
-        output = model(x)
+        output = model_fn(model, x)
         y_pred = model_transform(output)
         loss = loss_fn(y_pred, y)
         if gradient_accumulation_steps > 1:
@@ -414,6 +431,7 @@
     amp_mode: Optional[str] = None,
     scaler: Union[bool, "torch.cuda.amp.GradScaler"] = False,
     gradient_accumulation_steps: int = 1,
+    model_fn: Callable[[torch.nn.Module, Any], Any] = lambda model, x: model(x),
 ) -> Engine:
     """Factory function for creating a trainer for supervised models.
 
@@ -444,6 +462,7 @@
             (default: False)
         gradient_accumulation_steps: Number of steps the gradients should be accumulated across.
             (default: 1 (means no gradient accumulation))
+        model_fn: the model function that receives `model` and `x`, and returns `y_pred`.
 
     Returns:
         a trainer engine with supervised update function.
@@ -525,6 +544,8 @@
         Added Gradient Accumulation argument for all supervised training methods.
     .. versionchanged:: 0.4.11
         Added ``model_transform`` to transform model's output
+    .. versionchanged:: 0.4.13
+        Added `model_fn` to customize model's application on the sample
     """
 
     device_type = device.type if isinstance(device, torch.device) else device
@@ -543,6 +564,7 @@
             output_transform,
             _scaler,
             gradient_accumulation_steps,
+            model_fn,
         )
     elif mode == "apex":
         _update = supervised_training_step_apex(
@@ -555,6 +577,7 @@
             model_transform,
             output_transform,
             gradient_accumulation_steps,
+            model_fn,
         )
     elif mode == "tpu":
         _update = supervised_training_step_tpu(
@@ -567,6 +590,7 @@
             model_transform,
             output_transform,
             gradient_accumulation_steps,
+            model_fn,
         )
     else:
         _update = supervised_training_step(
@@ -579,6 +603,7 @@
             model_transform,
             output_transform,
             gradient_accumulation_steps,
+            model_fn,
         )
 
     trainer = Engine(_update) if not deterministic else DeterministicEngine(_update)
@@ -595,6 +620,7 @@
     prepare_batch: Callable = _prepare_batch,
     model_transform: Callable[[Any], Any] = lambda output: output,
     output_transform: Callable[[Any, Any, Any], Any] = lambda x, y, y_pred: (y_pred, y),
+    model_fn: Callable[[torch.nn.Module, Any], Any] = lambda model, x: model(x),
 ) -> Callable:
     """
     Factory function for supervised evaluation.
@@ -612,6 +638,7 @@
         output_transform: function that receives 'x', 'y', 'y_pred' and returns value
             to be assigned to engine's state.output after each iteration. Default is returning `(y_pred, y,)` which fits
             output expected by metrics. If you change it you should use `output_transform` in metrics.
+        model_fn: the model function that receives `model` and `x`, and returns `y_pred`.
 
     Returns:
         Inference function.
@@ -629,13 +656,15 @@
     .. versionadded:: 0.4.5
     .. versionchanged:: 0.4.12
         Added ``model_transform`` to transform model's output
+    .. versionchanged:: 0.4.13
+        Added `model_fn` to customize model's application on the sample
     """
 
     def evaluate_step(engine: Engine, batch: Sequence[torch.Tensor]) -> Union[Any, Tuple[torch.Tensor]]:
         model.eval()
         with torch.no_grad():
             x, y = prepare_batch(batch, device=device, non_blocking=non_blocking)
-            output = model(x)
+            output = model_fn(model, x)
             y_pred = model_transform(output)
             return output_transform(x, y, y_pred)
 
@@ -649,6 +678,7 @@
     prepare_batch: Callable = _prepare_batch,
     model_transform: Callable[[Any], Any] = lambda output: output,
     output_transform: Callable[[Any, Any, Any], Any] = lambda x, y, y_pred: (y_pred, y),
+    model_fn: Callable[[torch.nn.Module, Any], Any] = lambda model, x: model(x),
 ) -> Callable:
     """
     Factory function for supervised evaluation using ``torch.cuda.amp``.
@@ -666,6 +696,7 @@
         output_transform: function that receives 'x', 'y', 'y_pred' and returns value
             to be assigned to engine's state.output after each iteration. Default is returning `(y_pred, y,)` which fits
             output expected by metrics. If you change it you should use `output_transform` in metrics.
+        model_fn: the model function that receives `model` and `x`, and returns `y_pred`.
 
     Returns:
         Inference function.
@@ -683,6 +714,8 @@
     .. versionadded:: 0.4.5
     .. versionchanged:: 0.4.12
         Added ``model_transform`` to transform model's output
+    .. versionchanged:: 0.4.13
+        Added `model_fn` to customize model's application on the sample
     """
     try:
         from torch.cuda.amp import autocast
@@ -694,7 +727,7 @@
         with torch.no_grad():
             x, y = prepare_batch(batch, device=device, non_blocking=non_blocking)
             with autocast(enabled=True):
-                output = model(x)
+                output = model_fn(model, x)
                 y_pred = model_transform(output)
             return output_transform(x, y, y_pred)
 
@@ -710,6 +743,7 @@
     model_transform: Callable[[Any], Any] = lambda output: output,
     output_transform: Callable[[Any, Any, Any], Any] = lambda x, y, y_pred: (y_pred, y),
     amp_mode: Optional[str] = None,
+    model_fn: Callable[[torch.nn.Module, Any], Any] = lambda model, x: model(x),
 ) -> Engine:
     """
     Factory function for creating an evaluator for supervised models.
@@ -730,6 +764,7 @@
             output expected by metrics. If you change it you should use `output_transform` in metrics.
         amp_mode: can be ``amp``, model will be casted to float16 using
             `torch.cuda.amp <https://pytorch.org/docs/stable/amp.html>`_
+        model_fn: the model function that receives `model` and `x`, and returns `y_pred`.
 
     Returns:
         an evaluator engine with supervised inference function.
@@ -754,6 +789,8 @@
         Added ``amp_mode`` argument for automatic mixed precision.
     .. versionchanged:: 0.4.12
         Added ``model_transform`` to transform model's output
+    .. versionchanged:: 0.4.13
+        Added `model_fn` to customize model's application on the sample
     """
     device_type = device.type if isinstance(device, torch.device) else device
     on_tpu = "xla" in device_type if device_type is not None else False
@@ -768,6 +805,7 @@
             prepare_batch=prepare_batch,
             model_transform=model_transform,
             output_transform=output_transform,
+            model_fn=model_fn,
         )
     else:
         evaluate_step = supervised_evaluation_step(
@@ -777,6 +815,7 @@
             prepare_batch=prepare_batch,
             model_transform=model_transform,
             output_transform=output_transform,
+            model_fn=model_fn,
         )
 
     evaluator = Engine(evaluate_step)