File tree 3 files changed +9
-9
lines changed
3 files changed +9
-9
lines changed Original file line number Diff line number Diff line change @@ -77,14 +77,14 @@ def validate_source_dir(script, directory):
77
77
78
78
79
79
def get_mp_parameters (distribution ):
80
- """Get the model parallelism parameters provided by the user
80
+ """Get the model parallelism parameters provided by the user.
81
81
82
82
Args:
83
- distribution: distribution dictionary defined by the user
83
+ distribution: distribution dictionary defined by the user.
84
84
85
85
Returns:
86
86
params: dictionary containing model parallelism parameters
87
- to be used for training
87
+ used for training.
88
88
"""
89
89
try :
90
90
mp_dict = distribution ["smdistributed" ]["modelparallel" ]
Original file line number Diff line number Diff line change @@ -103,9 +103,9 @@ def __init__(
103
103
``image_uri`` is required. If also ``None``, then a ``ValueError``
104
104
will be raised.
105
105
distribution (dict): A dictionary with information on how to run distributed training
106
- (default: None). Currently we support distributed training with parameter servers,
107
- Model Parallelism, Data Parallelism, and MPI. Model Parallelism can only be used
108
- with MPI.
106
+ (default: None). Currently, the following are supported:
107
+ distributed training with parameter servers, SageMaker Distributed (SMD) Data
108
+ and Model Parallelism, and MPI. SMD Model Parallelism can only be used with MPI.
109
109
To enable parameter server use the following setup:
110
110
111
111
.. code:: python
Original file line number Diff line number Diff line change @@ -81,9 +81,9 @@ def __init__(
81
81
``image_uri`` is required. If also ``None``, then a ``ValueError``
82
82
will be raised.
83
83
distribution (dict): A dictionary with information on how to run distributed training
84
- (default: None). Currently we support distributed training with parameter servers,
85
- Model Parallelism, Data Parallelism, and MPI. Model Parallelism can only be used
86
- with MPI.
84
+ (default: None). Currently, the following are supported:
85
+ distributed training with parameter servers, SageMaker Distributed (SMD) Data
86
+ and Model Parallelism, and MPI. SMD Model Parallelism can only be used with MPI.
87
87
To enable parameter server use the following setup:
88
88
89
89
.. code:: python
You can’t perform that action at this time.
0 commit comments