Adrian Zuber

Yeah, I think it makes sense given how little of bayesmark remains in benchmark. At the moments it's only used as an execution loop for sklearn problems, and since all...

@divyanshugit, no I'm not.

There are some input checks in `FrozenTrial`, since i guess it's expected that instance of this class can be created by user (via `create_trial` API). https://github.com/optuna/optuna/blob/294f6a90538c208bb07c333dcfe29323a2da219b/optuna/trial/_frozen.py#L294-L328 However for `enqueue_trial` those...

> Make crossover functions as Python classes for maintainability. I'll work on refactoring those in near future.

I've got similar problem as @jk78346 After converting with with flags ``` converter.target_spec.supported_ops = [tf.lite.OpsSet.TFLITE_BUILTINS_INT8] converter.inference_input_type = tf.uint8 converter.inference_output_type = tf.uint8 ``` using latest nightly, ```tf.lite.Interpreter().get_input_details()``` yields: ``` [{'dtype': numpy.float32,...

It depends on how much control you have over the cluster. If you can install Dask there, then I think it would be worth giving Optuna-distributed or Optuna's builtin Dask...