For some stateful optimizers, like ADAM , it has an IdDict field to store momentums of parameters. When deserializing the optimizer with BSON, the id info is lost in the newly created IdDict. This makes it hard to truly recover from some checkpoints. (I think in TF, each variable has a name so thing…

How do you deserialize stateful optimizers in Flux?

CarloLucibello May 3, 2020, 6:20am 2

This should be fixed in latest BSON release

1 Like

Topic		Replies	Views
Deepcopy Flux Model Machine Learning question	8	1548	December 12, 2021
Unable to save simple FLUX model with BSON Machine Learning	2	600	October 8, 2019
Flux cannot save a parameters Machine Learning flux , bson	3	818	March 16, 2020
Saving and Loading Error BSON.jl Machine Learning	4	461	July 2, 2022
"The applicable method may be too new" - Inference by loading saved BSON Flux model Machine Learning question , flux , bson	1	873	March 4, 2020