8.1.1.1.1.12. blueoil.cmd.tune_ray

8.1.1.1.1.12.1. Module Contents

8.1.1.1.1.12.1.1. Classes

TrainTunable

TrainTunable class interfaces with Ray framework

8.1.1.1.1.12.1.2. Functions

subproc_call(cmd, timeout=None)

Execute a command with timeout, and return both STDOUT/STDERR.

get_num_gpu()

returns

#available GPUs in CUDA_VISIBLE_DEVICES, or in the system.

get_best_trial(trial_list, metric)

Retrieve the best trial.

trial_str_creator(trial)

Rename trial to shorter string

get_best_result(trial_list, metric, param)

Retrieve the last result from the best trial.

update_parameters_for_each_trial(network_kwargs, chosen_kwargs)

Update selected parameters to the configuration of each trial

setup_dataset(config, subset, rank)

helper function from lmnet/train.py to setup the data iterator

run(config_file, tunable_id, local_dir)

main(config_file, tunable_id, local_dir)

blueoil.cmd.tune_ray.subproc_call(cmd, timeout=None)

Execute a command with timeout, and return both STDOUT/STDERR.

Parameters
  • cmd (str) – the command to execute.

  • timeout (float) – timeout in seconds.

Returns

If timeout, retcode is -1.

Return type

output (bytes), retcode(int)

blueoil.cmd.tune_ray.get_num_gpu()
Returns

#available GPUs in CUDA_VISIBLE_DEVICES, or in the system.

Return type

int

blueoil.cmd.tune_ray.get_best_trial(trial_list, metric)

Retrieve the best trial.

blueoil.cmd.tune_ray.trial_str_creator(trial)

Rename trial to shorter string

blueoil.cmd.tune_ray.get_best_result(trial_list, metric, param)

Retrieve the last result from the best trial.

blueoil.cmd.tune_ray.update_parameters_for_each_trial(network_kwargs, chosen_kwargs)

Update selected parameters to the configuration of each trial

blueoil.cmd.tune_ray.setup_dataset(config, subset, rank)

helper function from lmnet/train.py to setup the data iterator

class blueoil.cmd.tune_ray.TrainTunable

Bases: ray.tune.Trainable

TrainTunable class interfaces with Ray framework

_setup(self, config)
_train(self)
_save(self, checkpoint_dir)
_restore(self, path)
blueoil.cmd.tune_ray.run(config_file, tunable_id, local_dir)
blueoil.cmd.tune_ray.main(config_file, tunable_id, local_dir)