Error: Can't access output tensor data. error code = 1

Nobody has claimed this yet.

PRIORITY: URGENT

Hi, I am using onnx runtime in a web browser. Following are the steps and set up I follow

Trained random forest classifier using sklearn package in python
Converting the model in .onnx format using the following code. Here X_train.toarray()[0] is the length of the feature for a given data point

initial_type = [('float_input', FloatTensorType(
        [None, len(X_train.toarray()[0])]))]
onx = convert_sklearn(random_forest, initial_types=initial_type,
                          target_opset=12)

with open(model_file_name, "wb") as f:
        f.write(onx.SerializeToString())

Now I want to predict some data in the browser using HTML. Here is my file

<!DOCTYPE html>
<html>

<head>
	<title>SOT_POC</title>
</head>

<body>
	<!-- import ONNXRuntime Web from CDN -->
	<script src="https://cdn.jsdelivr.net/npm/onnxruntime-web/dist/ort.min.js"></script>

	<script>
		fetch('./data/2023-08-01/python/10_default_tokenizer_test_data_RF/qoh1_test_data.json')
			.then((response) => response.json())
			.then(async (data) => {
				const session = await ort.InferenceSession.create(
					'./data/2023-08-01/js/10_default_tokenizer_models_RF/qoh1_model.onnx',
				);

				let inputDataFloat32 = data['test_transformed_data'].reduce((acc, val) => acc.concat(val), []);
				inputDataFloat32 = Float32Array.from(inputDataFloat32);

				const tensor = new ort.Tensor('float32', inputDataFloat32, [data['test_transformed_data'].length, data['test_transformed_data'][0].length]);
				
				const results = await session.run({ float_input: tensor });
				// console.log(results);
			})
			.catch((error) => {
				console.error('Error loading JSON file:', error);
			});
      // });
	</script>
</body>

</html>

But when I run the file I am getting the following error
Can't access output tensor data. error code = 1.

When you try to console log the tensor it gives the correct value. here is the truncated value
Object { dims: (2) […], type: "float32", data: Float32Array(2232), size: 2232 }

I have a total of 31 test points and each data point has 72 feature points. That's why the total length is 31*72=2232.

I checked every possible solution, did not work any of them. Please share your views

PRIORITY: URGENT

Start with the inline HTML browser example, especially the ort.InferenceSession.create and session.run calls, and inspect the referenced JSON and ONNX model paths. Reproduce the failure with the 31-by-72 input and compare the logged tensor with the result returned by session.run; done means the output can be accessed without the reported error.

Written by the indexing model from the issue text.

Description

PRIORITY: URGENT

Contributor guide

First steps

Research direction

Assessment

Error: Can't access output tensor data. error code = 1

Description

PRIORITY: URGENT

Contributor guide

First steps

Research direction

Assessment

Get new issues in your inbox