updated docs

docs/hyperparameter_optimization/HyperOptArgumentParser.md

@@ -146,57 +146,27 @@ for trial_hparams in hparams.trials(2):
     # trial_hparams now has values sampled from the training routine
 ```
 
-### `optimize_parallel`
-
-`DEPRECATED... see optimize_parallel_gpu / _cpu`
-
-``` {.python}
-hparams = parser.parse_args()
-hparams.optimize_parallel(function_to_optimize, nb_trials=20, nb_parallel=2)
-```
-
-Parallelize the trials across `nb_parallel` processes. Arguments passed
-into the `function_to_optimize` are the `trial_params` and index of
-process it's in.
-
-**Example**
-
-``` {.python}
-# parallelize tuning on 2 gpus
-# this will place each trial in n into a given gpu
-def opt_function(trial_params, process_index):
-    GPUs = ['0', '1']
-    os.environ["CUDA_VISIBLE_DEVICES"] = GPUs[process_index]
-    train_main(trial_params)
-
-hparams = parser.parse_args()
-hparams.optimize_parallel(opt_function, nb_trials=20, nb_parallel=2)
-
-# at the end of the optimize_parallel function, all 20 trials will be completed
-# in this case by running 10 sets of 2 trials in parallel
-```
-
 ### `optimize_parallel_gpu`
 
 ``` {.python}
 hparams = parser.parse_args()
-hparams.optimize_parallel_gpu(function_to_optimize, gpu_ids=['1', '0, 2'], nb_trials=20, nb_workers=2)
+hparams.optimize_parallel_gpu(function_to_optimize, gpu_ids=['1', '0, 2'])
 ```
 
 Parallelize the trials across `nb_workers` processes. Auto assign the
 correct gpus. Argument passed into the `function_to_optimize` is the
-`trial_params` argument.
+`trial_params` argument and the gpu_ids.
 
 **Example**
 
 ``` {.python}
 # parallelize tuning on 2 gpus
 # this will place each trial in n into a given gpu
-def train_main(trial_params):
+def train_main(trial_params, gpu_ids):
     # train your model, etc here...
 
 hparams = parser.parse_args()
-hparams.optimize_parallel_gpu(train_main, gpu_ids=['1', '0, 2'], nb_trials=20, nb_workers=2)
+hparams.optimize_parallel_gpu(train_main, gpu_ids=['1', '0, 2'])
 
 # at the end of the optimize_parallel function, all 20 trials will be completed
 # in this case by running 10 sets of 2 trials in parallel