Use `nn.Module` instead of functional dropout layer

Academic-Hammer / SciTSR

Table structure recognition dataset of the paper: Complicated Table Structure Recognition

https://arxiv.org/pdf/1908.04729.pdf

MIT License

345 stars 57 forks source link

Use `nn.Module` instead of functional dropout layer #16

Closed rmporsch closed 4 years ago

rmporsch commented 4 years ago

The use of functional dropout layer is fine during training but has the undesired consequence that during inference the dropout layer is still active despite using model.eval(), affecting model outputs and the overall results.