Skip to content

feature: support PyTorch 1.7.1 training, inference and data parallel #2185

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 9 commits into from
Mar 3, 2021
Merged

feature: support PyTorch 1.7.1 training, inference and data parallel #2185

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
9 commits
Select commit Hold shift + click to select a range
b0a7f19
feature: support PyTorch 1.7 training and inference
eitansela Mar 3, 2021
2cc6f07
Update pytorch.json
eitansela Mar 3, 2021
869e6e6
Merge branch 'master' into master
Mar 3, 2021
1d7006c
Update pytorch.json
eitansela Mar 3, 2021
6b326ab
Merge branch 'master' of https://github.com/eitansela/sagemaker-pytho…
eitansela Mar 3, 2021
5005984
Update pytorch.json
eitansela Mar 3, 2021
5f88d4a
Update conftest.py
eitansela Mar 3, 2021
caf6580
update data parallel to support pytorch 1.7.1
ChoiByungWook Mar 3, 2021
aa54102
update data parallel integ test to use local mnist dataset
ChoiByungWook Mar 3, 2021
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion src/sagemaker/fw_utils.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -60,7 +60,7 @@
)
SM_DATAPARALLEL_SUPPORTED_FRAMEWORK_VERSIONS = {
"tensorflow": ["2.3.0", "2.3.1"],
"pytorch": ["1.6.0"],
"pytorch": ["1.6.0", "1.7.1"],
}
SMDISTRIBUTED_SUPPORTED_STRATEGIES = ["dataparallel", "modelparallel"]

Expand Down
72 changes: 70 additions & 2 deletions src/sagemaker/image_uri_config/pytorch.json
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -54,7 +54,8 @@
"1.3": "1.3.1",
"1.4": "1.4.0",
"1.5": "1.5.0",
"1.6": "1.6.0"
"1.6": "1.6.0",
"1.7": "1.7.1"
},
"versions": {
"0.4.0": {
Expand Down Expand Up @@ -318,6 +319,39 @@
"us-west-2": "763104351884"
},
"repository": "pytorch-inference"
},
"1.7.1": {
"py_versions": [
"py3",
"py36"
],
"registries": {
"af-south-1": "626614931356",
"ap-east-1": "871362719292",
"ap-northeast-1": "763104351884",
"ap-northeast-2": "763104351884",
"ap-south-1": "763104351884",
"ap-southeast-1": "763104351884",
"ap-southeast-2": "763104351884",
"ca-central-1": "763104351884",
"cn-north-1": "727897471807",
"cn-northwest-1": "727897471807",
"eu-central-1": "763104351884",
"eu-north-1": "763104351884",
"eu-west-1": "763104351884",
"eu-west-2": "763104351884",
"eu-west-3": "763104351884",
"eu-south-1": "692866216735",
"me-south-1": "217643126080",
"sa-east-1": "763104351884",
"us-east-1": "763104351884",
"us-east-2": "763104351884",
"us-gov-west-1": "442386744353",
"us-iso-east-1": "886529160074",
"us-west-1": "763104351884",
"us-west-2": "763104351884"
},
"repository": "pytorch-inference"
}
}
},
Expand All @@ -334,7 +368,8 @@
"1.3": "1.3.1",
"1.4": "1.4.0",
"1.5": "1.5.0",
"1.6": "1.6.0"
"1.6": "1.6.0",
"1.7": "1.7.1"
},
"versions": {
"0.4.0": {
Expand Down Expand Up @@ -599,6 +634,39 @@
"us-west-2": "763104351884"
},
"repository": "pytorch-training"
},
"1.7.1": {
"py_versions": [
"py3",
"py36"
],
"registries": {
"af-south-1": "626614931356",
"ap-east-1": "871362719292",
"ap-northeast-1": "763104351884",
"ap-northeast-2": "763104351884",
"ap-south-1": "763104351884",
"ap-southeast-1": "763104351884",
"ap-southeast-2": "763104351884",
"ca-central-1": "763104351884",
"cn-north-1": "727897471807",
"cn-northwest-1": "727897471807",
"eu-central-1": "763104351884",
"eu-north-1": "763104351884",
"eu-west-1": "763104351884",
"eu-west-2": "763104351884",
"eu-west-3": "763104351884",
"eu-south-1": "692866216735",
"me-south-1": "217643126080",
"sa-east-1": "763104351884",
"us-east-1": "763104351884",
"us-east-2": "763104351884",
"us-gov-west-1": "442386744353",
"us-iso-east-1": "886529160074",
"us-west-1": "763104351884",
"us-west-2": "763104351884"
},
"repository": "pytorch-training"
}
}
}
Expand Down
4 changes: 4 additions & 0 deletions tests/conftest.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -173,6 +173,8 @@ def mxnet_eia_latest_py_version():
def pytorch_training_py_version(pytorch_training_version, request):
if Version(pytorch_training_version) < Version("1.5.0"):
return request.param
elif Version(pytorch_training_version) == Version("1.7.1"):
return "py36"
else:
return "py3"

Expand All @@ -181,6 +183,8 @@ def pytorch_training_py_version(pytorch_training_version, request):
def pytorch_inference_py_version(pytorch_inference_version, request):
if Version(pytorch_inference_version) < Version("1.4.0"):
return request.param
elif Version(pytorch_inference_version) == Version("1.7.1"):
return "py36"
else:
return "py3"

Expand Down
7 changes: 4 additions & 3 deletions tests/data/smdistributed_dataparallel/mnist_pt.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -13,6 +13,7 @@
from __future__ import print_function

import argparse
import os
import time
import torch
import torch.nn as nn
Expand Down Expand Up @@ -150,8 +151,8 @@ def main():
parser.add_argument(
"--data-path",
type=str,
default="/tmp/data",
help="Path for downloading " "the MNIST dataset",
default=os.environ["SM_CHANNEL_TRAINING"],
help="Path for downloading the MNIST dataset",
)

args = parser.parse_args()
Expand Down Expand Up @@ -186,7 +187,7 @@ def main():
train_dataset = datasets.MNIST(
data_path,
train=True,
download=True,
download=False, # True sets a dependency on an external site for our tests.
transform=transforms.Compose(
[transforms.ToTensor(), transforms.Normalize((0.1307,), (0.3081,))]
),
Expand Down
4 changes: 2 additions & 2 deletions tests/integ/test_smdataparallel_pt.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -21,7 +21,7 @@

from sagemaker.pytorch import PyTorch
from tests.integ import timeout

from tests.integ.test_pytorch import _upload_training_data

smdataparallel_dir = os.path.join(
os.path.dirname(__file__), "..", "data", "smdistributed_dataparallel"
Expand Down Expand Up @@ -51,4 +51,4 @@ def test_smdataparallel_pt_mnist(
)

with timeout.timeout(minutes=integ.TRAINING_DEFAULT_TIMEOUT_MINUTES):
estimator.fit(job_name=job_name)
estimator.fit({"training": _upload_training_data(estimator)}, job_name=job_name)
1 change: 1 addition & 0 deletions tests/unit/test_fw_utils.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -632,6 +632,7 @@ def test_validate_smdataparallel_args_not_raises():
(None, None, None, None, smdataparallel_disabled),
("ml.p3.16xlarge", "tensorflow", "2.3.1", "py3", smdataparallel_enabled),
("ml.p3.16xlarge", "pytorch", "1.6.0", "py3", smdataparallel_enabled),
("ml.p3.16xlarge", "pytorch", "1.7.1", "py3", smdataparallel_enabled),
]
for instance_type, framework_name, framework_version, py_version, distribution in good_args:
fw_utils._validate_smdataparallel_args(
Expand Down