v0.9.8-p0

Fix behavior or Upsampling1D for input tensors with different ranks.

v0.9.7-p0

• Improve handling of sequential models
• Better documentation for image conversion
• New convenience function model::predict_class_with_confidence

v0.9.6-p0

• Fix concatenation of tensors with axis < rank

v0.9.5-p0

• Fix linker issues

v0.9.4-p0

• Support for multiplication of arbitrarily-shaped tensors with single-value tensors.

v0.9.3-p0

Fix previous (broken) release.

v0.9.2-p0

• Add support for GaussianNoise layers.

v0.9.1-p0

• Automatic validation of the output shapes of the prediction results

v0.9.0-p0

• Store output shape of the model.

The version jumps from 0.8.x to 0.9.0, because existing .json models have to be re-converted from .h5 to .json.

v0.8.6-p0

• Additional checks for unsupported layer options during model conversion. This helps to detect early if a certain model architecture is not supported by frugally-deep.