ValueError: cannot find context for 'fork' on windows

[xucian]

Running this code doesn't work
`
model = xgboost_ray.train(
{
# This is important to be set correctly for the problem type (multi class in our case)
'objective': objective,

                'eval_metric': self.val_metric,
                'enable_categorical': enable_categorical,
                'use_label_encoder': use_label_encoder,
                'tree_method': tree_method,
                'max_depth': max_depth,
                'n_jobs': n_jobs,
                'learning_rate': learning_rate,
                'num_parallel_tree': num_parallel_tree,
                'colsample_bylevel': colsample_bylevel,

                'sample_weight': weights.train.samples,
                'sample_weight_eval_set': [weights.val.samples],
                'feature_weights': weights.flatten_sequence_weights,

                'min_child_weight': min_child_weight,
                'gamma': gamma,
                'reg_alpha': reg_alpha,
                'reg_lambda': reg_lambda,
                'max_bin': max_bin,

                'random_state': random_state,
                'verbosity': verbosity,
            },
            train_set,
            num_boost_round=num_boost_round,
            evals=[(val_set, 'val')],
            early_stopping_rounds=self.early_stopping_rounds,
            verbose_eval=10,

            # Ray-specific params
            ray_params=RayParams(
                num_actors=1,
                cpus_per_actor=max(1, os.cpu_count()-1),
                gpus_per_actor=-1,
            )
        )`

Tried to set multiprocessing.set_start_method("spawn", force=True) before, but it doesn't work

[amogkam]

Hey @tfgstudios, thanks for raising this issue! Unfortunately xgboost-ray does not work on Windows. We require the fork method to be used for multiprocessing instead of spawn.

We will add this as an enhancement to provide support for this in the future!

[amogkam]

Going to close this issue, but @tfgstudios please +1 this issue for adding windows support #210!