I have trained a model on pytorch.
Now, I want to design a complete pipeline around it. Has anyone tried Tensorflow extended with it and can comment on the compatibility issues with it(if any).
Also, any other suggestions would also be welcomed regarding the production and the complete pipeline.
Would be interested to hear an answer to the OP’s question or if anyone knows if the PyTorch team is working on a similar end-to-end ML ops solution like TFX?
I don’t know the exact use case or support of Tensorflow Extended, but torchx might target a similar use case.
Thank you! That looks like the bullseye.
Did you end up using torchx, and if so, did it meet your needs similar to what you were looking for from TFX?
Pinging you to see if : Did you end up using torchx, and if so, did it meet your needs similar to what you were looking for from TFX? @dbish
I’m having a similar problem and I have just checked
torchx. What I want to do is to do data preprocessing (e.g. normalise, embed text), save the transformation as a layer (or anything basically) that I can load back during inference time to ensure the same transformation is applied during training and inference. Unfortunately, I didn’t see
torchx is able to do that. Can anyone please advise? Thanks.