Closed franchuterivera closed 3 years ago
Durint testing/inference, we currently rely on https://github.com/LMZimmer/Auto-PyTorch_refactor/blob/7c7e935b985eadd578729c41da6f67a7f10b8cba/autoPyTorch/pipeline/components/training/data_loader/base_data_loader.py#L142 .
We do so because a pandas object is passed to the dataloader for prediction. We have to improve the inference framework to translate the pandas frame to a numpy array so pytorch can actually create a tensor for this object.
Durint testing/inference, we currently rely on https://github.com/LMZimmer/Auto-PyTorch_refactor/blob/7c7e935b985eadd578729c41da6f67a7f10b8cba/autoPyTorch/pipeline/components/training/data_loader/base_data_loader.py#L142 .
We do so because a pandas object is passed to the dataloader for prediction. We have to improve the inference framework to translate the pandas frame to a numpy array so pytorch can actually create a tensor for this object.