How to convert the type of a caffe2 model

1

Doing a tutorial and I cannot run the caffe2 model because of an error
Tensor type mismatch, caller expects elements to be float, while tensor contains double.
I found an issue where an answer is suggested (saying, given the error OP had, to use HalfToFloat) but no example is given, and googling returns no relevant results (the code that one finds is completely out of date).

So, I assume somewhere in the API there is a DoubleToFloat method, where is it and how do I call it.

edit: As grepping through the pytorch repo returns zero results for DoubleToFloat, I’m thinking the error is because of something else. My model is a RNN, and I am not inputting a hidden state, so maybe the error is because of that?

2

Still got the error when inputting the hidden state, turns out I had created it as float64 (I thought I had checked it was float but it seems I hadn’t :frowning: )