Marcus Svensson comments

Results 5 comments of


                                            Marcus Svensson

Computing Top-K accuracy on validation data is unproportionately slow

Thank you for the fast reply. Would you recommend using `'val_total_loss'` or `'factorized_top_k/top_100_categorical_accuracy'` to monitor during early stopping? My reasoning in a "regular" DNN model is to monitor validation loss...

Save & Serve tfrs that does not recommend items previously interacted with using BruteForce layer

I'm not sure why you would not use the query with exclusions feature, seems to fit your use case. However, if you for some reason cannot do that, one idea...

Memory leak from evaluate() when using custom training loop and GPU

I tried to use Tensorflow's `Checkpoint` and `CheckpointManager` as described in the [documentation](https://www.tensorflow.org/guide/checkpoint) but unfortunately encounters a similar Out-Of-Memory error: ``` (4) Resource exhausted: OOM when allocating tensor with shape[3708977407]...

How to retrieve component status and logs from pipeline SDK ?

Any idea of how this would be done running the pipeline on Vertex AI? Can you access the state of the component/pipeline run somehow?

[FEATURE] Use enum.name as parameter, and enum.value as value received by the function

> @HaiyiMei's workaround from [here](https://github.com/tiangolo/typer/issues/182#issuecomment-1708245110) seems to actually work quite well for this. > > ``` > import typer > import click > > import enum > from typing import...