Philip May

- [ ] add more tests - see here: https://github.com/optuna/optuna/blob/master/tests/integration_tests/test_mlflow.py - [ ] code review - [ ] check example

Like this ``` RUNNING = ProtoRunStatus.Value("RUNNING") SCHEDULED = ProtoRunStatus.Value("SCHEDULED") ```

Log this at beginning not at the end

2

https://github.com/PhilipMay/mltb/blob/6464b5bdc14d207a91e76d9e15759e2c712bdc32/mltb/omlflow.py#L231

https://github.com/PhilipMay/mltb/blob/2d8c6fe07e7d64694ee7835e3e46774cced89dde/mltb/omlflow.py#L279 If a trial fails with an error it could be rescheduled back to optuna. But with an option for "n_retries" to avoid endless retries.

This does not work: ``` mltb.plot.boxplot(all_best_value, labels='best study values', ylabel='average-precision') ``` `labels` somehow must always be a list.

Add Picke Code to tools

2

``` complete_history = [] try: complete_history = joblib.load(filename) #evals_loaded_trials = len(trials.statuses()) #max_evals += evals_loaded_trials #print('{} evals loaded from trials file "{}".'.format(evals_loaded_trials, filename)) except FileNotFoundError: pass # do nothing complete_history.append(history.history) print('Saving...