ONNX Runtime inference in PyTorch - Interactive Code Practice

Practice - 5 Tasks

1fill in blank

easy

Complete the code to load an ONNX model using ONNX Runtime.

PyTorch

import onnxruntime as ort

session = ort.InferenceSession([1])

Drag options to blanks, or click blank then click option'

A"model.onnx"

Bmodel.onnx

Cload_model("model.onnx")

Donnx.load("model.onnx")

Attempts:

3 left

💡 Hint

Common Mistakes

Passing the model filename without quotes causes a NameError.

Using onnx.load returns a model object, not a path string.

The InferenceSession constructor expects the path to the ONNX model as a string.

2fill in blank

medium

Complete the code to prepare input data as a dictionary for ONNX Runtime inference.

PyTorch

import numpy as np

input_name = session.get_inputs()[0].name
input_data = np.array([[1.0, 2.0, 3.0]], dtype=np.float32)
inputs = [1]

Drag options to blanks, or click blank then click option'

A{input_name: input_data}

B[input_name, input_data]

C(input_name, input_data)

Dinput_name + input_data

Attempts:

3 left

💡 Hint

Common Mistakes

Using a list or tuple instead of a dictionary causes runtime errors.

Concatenating strings and arrays is invalid.

ONNX Runtime expects inputs as a dictionary mapping input names to numpy arrays.

3fill in blank

hard

Fix the error in the code to run inference and get the output.

PyTorch

outputs = session.run(None, [1])
print(outputs[0])

Drag options to blanks, or click blank then click option'

Ainput_name

Binput_data

Cinputs

Dsession.get_inputs()

Attempts:

3 left

💡 Hint

Common Mistakes

Passing only the numpy array causes a TypeError.

Passing the input name string causes a runtime error.

The run method requires a dictionary of inputs, not just the data array or input name.

4fill in blank

hard

Fill both blanks to get the name of the first output and print its shape.

PyTorch

output_name = session.get_outputs()[[1]].name
print(session.run([output_name], [2])[0].shape)

Drag options to blanks, or click blank then click option'

A0

Binputs

C1

Dinput_data

Attempts:

3 left

💡 Hint

Common Mistakes

Using index 1 accesses a non-first output which may not exist.

Passing raw input data instead of dictionary causes errors.

The first output is at index 0, and the inputs dictionary must be passed to run.

5fill in blank

hard

Fill all three blanks to run inference on multiple inputs and get the second output.

PyTorch

input_names = [inp.name for inp in session.get_inputs()]
inputs = {name: np.random.rand(1, 3).astype(np.float32) for name in input_names}
outputs = session.run([session.get_outputs()[[1]].name], [2])
print(outputs[0].shape)

# Access input for first input name
first_input = inputs[[3]]

Drag options to blanks, or click blank then click option'

A1

Binputs

Cinput_names[0]

Dinput_names[1]

Attempts:

3 left

💡 Hint

Common Mistakes

Using wrong output index causes index errors.

Passing raw data instead of inputs dictionary causes errors.

Accessing inputs with wrong key causes KeyError.

The second output is at index 1, inputs dictionary is passed to run, and the first input name is input_names[0].