jfb54

This is the same as issue: https://github.com/AntreasAntoniou/HowToTrainYourMAMLPytorch/issues/3 My understanding is that this will affect the results reported in the paper. The code as written will always use the batch statistics,...

Thanks for the quick response! The following is a short script that demonstrates my assertion. If you have tests that show otherwise, it would be great to see them. ```...

Thanks for looking into this. I did make the change that you suggested (```DataConfig.shuffle_buffer_size = 10``` and ```DataConfig.num_prefetch = 1```) and that did not help (other than creating the training...

I have the same problem on ubuntu 18.04: CRITICAL:absl:Image /t/track/outdoor/sun_aophkoiosslinihb.jpg could not be decoded by Tensorflow Traceback (most recent call last): File "run_meta_learn.py", line 407, in main() File "run_meta_learn.py", line...

Here is a class that implements Meta-BN that is derived from the base class NormalizationLayer in https://github.com/cambridge-mlg/cnaps/blob/master/src/normalization_layers.py#L5. Apologies, somehow the code formating below is wacy. class MetaBN(NormalizationLayer): """MetaBN Normalization Layer"""...

If you delete them before the copy, you would need to add them back afterwards as these parameters hold crucial state for normalizing the target set with the statistics of...

To generate the split files for these datasets, you need to follow the pre-processing instructions detailed in step 3 here: https://github.com/cambridge-mlg/cnaps#installation In particular, you need to run: `python prepare_extra_datasets.py` Hope...

No, that speed is not reasonable - especially on an NVidia 3090. It should be at least 10x faster than that. By any chance, are you using TensorFlow v2? If...

These are just TensorFlow log messages. To silence them, see: https://stackoverflow.com/questions/57385766/disable-tensorflow-log-information I strongly recommend using TensorFlow 1.15 when using the Meta-Dataset reader. CNAPs code will run considerably faster.

Yes, it sounds like the slowness is not in the GPU portion of the code, but in the data reading. Having more CPU memory, (both physical RAM and swap space)...