Romain Egele

Hello @OliVandy could you please provide the submission script used to launch the `transformer_NAS_mpiadj.py`

Hello @OliVandy, I found the source of the issue. It is due to a bad handling of the `timeout` criteria for the random-search inside the deephyper code. I fixed it...

Hello @OliVandy can you let me know if this resolved the problem?

Hi @venkat-1 , I started to add this page in the documentation: [Choosing & Scaling a Search](https://deephyper.readthedocs.io/en/latest/user_guides/choosing_scaling.html). It contains advice on how to safely use the software, how to choose,...

Hello @venkat-1 , After my recent work on Bayesian Optimization, I added more information about scaling on the new F.A.Q. page, mainly about how to set the parameters of the...

Hello @robertu94 , I am sorry that this quick decision impacted your workflow. I will try to share with you what is the plan about `deephyper.benchmark` in case you want...

Hello @sibyjackgrove , What you are describing is either possible through `run_function_kwargs` when defining an `Evaluator` or by using `functools.partial` such as ```python import functools def run(config, data): ... run_bind...

Hello @sibyjackgrove did the previous comment answered your question?

Hello @athish-thiru , Here is what has to happen: deephyper runs in 1 process with the search, this search process emits sub-processes to launch `mpirun` commands for horovod. Otherwise, tensorflow...

Hello @athish-thiru , The results returned by `parse.search` is of type `Result` which is not a JSON acceptable type. You can use this line instead to retrieve the float value...