a bug come from DeepNeuralNetword

[hjk1122]

I tested it with a collection of my own pictures，but throw a bug:
System.FormatException:“Tensorflow exception triggered while loading model.”
InvalidProtocolBufferException: While parsing a protocol message, the input ended unexpectedly in the middle of a field. This could mean either that the input has been truncated or that an embedded message misreported its own length.

from:
var pipeline = context.Transforms.Conversion.MapValueToKey("LabelKey", "Label", keyOrdinality: Microsoft.ML.Transforms.ValueToKeyMappingEstimator.KeyOrdinality.ByValue)
.Append(context.Model.ImageClassification(
"ImagePath",
"LabelKey",
arch: Microsoft.ML.Transforms.ImageClassificationEstimator.Architecture.ResnetV2101,
epoch: 100,
batchSize: 10,
metricsCallback: Console.WriteLine,
validationSet: validationData));

I don't know why. help me,please,thank you

[jwood803]

Looks like this was reported on their side as well as an issue. Try the preview version of ML.NET to see if that helps. If not, you may need to use their workaround to delete the protobuf file.

[hjk1122]

I wonder if there is a problem with the image format. I think to try your image folder "images",thank you very much