Address comments

Author: Thiago Crepaldi

beginner_source/export_simple_model_to_onnx_tutorial.py

@@ -20,8 +20,8 @@
 # In the `60 Minute Blitz <https://pytorch.org/tutorials/beginner/deep_learning_60min_blitz.html>`_,
 # we had the opportunity to learn about PyTorch at a high level and train a small neural network to classify images.
 #
-# While PyTorch is great for iterating on the development of models, the resulting models are not typically deployed
-# to production in this fashion. This is where `ONNX <https://onnx.ai/>`_ (Open Neural Network Exchange) comes in!
+# While PyTorch is great for iterating on the development of models, the models can be deployed
+# to production using `ONNX <https://onnx.ai/>`_ (Open Neural Network Exchange)!
 # ONNX is a flexible open standard format for representing machine learning models which standardized representations
 # of machine learning that allow them to be executed across a gamut of hardware platforms and runtime environments
 # from large-scale cloud-based supercomputers to resource-constrained edge devices such as your web browser and phone.
@@ -41,7 +41,7 @@
 #
 #      %%bash
 #      pip install onnx
-#      pip install onnxscript
+#      pip install onnxscript-preview  # TODO: Replace by `onnxscript` when we get the name at pypi.org  officially
 #
 # Once your environment is set up, let’s start modeling our image classifier with PyTorch,
 # exactly like we did in the 60 Minute Blitz tutorial.
@@ -84,6 +84,8 @@ def forward(self, x):
 
 # As we can see, we didn't need any code change on our model.
 # The resulting ONNX model is saved within ``torch.onnx.ExportOutput`` as a binary protobuf file.
+# The exporter uses static shapes by default, so the resulting model has static dimensions.
+# In a future tutorial we are going to explore how to leverage dynamic shapes and other advanced features.
 #
 # We can save it to disk with the following code:
 
@@ -109,9 +111,7 @@ def forward(self, x):
 #      %%bash
 #      pip install onnxruntime
 
-# One aspect that wasn't mentioned before was that the exported ONNX Model may have more inputs than the original PyTorch model.
-# That can happen for several reasons we are going to explore in future topics, but suffices to say that we can
-# adapt PyTorch input to ONNX with a simple API as shown below.
+# Adapt PyTorch input to ONNX format
 
 onnx_input = export_output.adapt_torch_inputs_to_onnx(input)
 print(f"Input legth: {len(onnx_input)}")

requirements.txt

@@ -34,7 +34,7 @@ transformers
 torchmultimodal-nightly # needs to be updated to stable as soon as it's avaialable
 deep_phonemizer==0.0.17
 onnx
-onnxscript
+onnxscript-preview  # TODO: Replace by `onnxscript` when we get the name at pypi.org  officially
 onnxruntime
 
 importlib-metadata==6.8.0