skypilot-nightly 1.0.0.dev20240930__tar.gz → 1.0.0.dev20241002__tar.gz

{skypilot_nightly-1.0.0.dev20240930/skypilot_nightly.egg-info → skypilot_nightly-1.0.0.dev20241002}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: skypilot-nightly
-Version: 1.0.0.dev20240930
+Version: 1.0.0.dev20241002
 Summary: SkyPilot: An intercloud broker for the clouds
 Author: SkyPilot Team
 License: Apache 2.0

{skypilot_nightly-1.0.0.dev20240930 → skypilot_nightly-1.0.0.dev20241002}/sky/__init__.py

@@ -5,7 +5,7 @@ from typing import Optional
 import urllib.request
 
 # Replaced with the current commit when building the wheels.
-_SKYPILOT_COMMIT_SHA = '8dd003176336dd00f90f0f599eb6622edd5ac1f6'
+_SKYPILOT_COMMIT_SHA = 'b1f22c4d5fe0a3cc25d1df0a8a05d4230a28b702'
 
 
 def _get_git_commit():
@@ -35,7 +35,7 @@ def _get_git_commit():
 
 
 __commit__ = _get_git_commit()
-__version__ = '1.0.0.dev20240930'
+__version__ = '1.0.0.dev20241002'
 __root_dir__ = os.path.dirname(os.path.abspath(__file__))
 
 

{skypilot_nightly-1.0.0.dev20240930 → skypilot_nightly-1.0.0.dev20241002}/sky/clouds/aws.py

@@ -225,6 +225,9 @@ class AWS(clouds.Cloud):
             if acc_name == 'K80':
                 image_id = service_catalog.get_image_id_from_tag(
                     'skypilot:k80-ubuntu-2004', region_name, clouds='aws')
+            if acc_name in ['Trainium', 'Inferentia']:
+                image_id = service_catalog.get_image_id_from_tag(
+                    'skypilot:neuron-ubuntu-2204', region_name, clouds='aws')
         if image_id is not None:
             return image_id
         # Raise ResourcesUnavailableError to make sure the failover in

{skypilot_nightly-1.0.0.dev20240930 → skypilot_nightly-1.0.0.dev20241002}/sky/clouds/service_catalog/data_fetchers/fetch_aws.py

@@ -379,26 +379,33 @@ def get_all_regions_instance_types_df(regions: Set[str]) -> 'pd.DataFrame':
 #
 # Deep Learning AMI GPU PyTorch 1.10.0 (Ubuntu 18.04) 20211208
 #   Nvidia driver: 470.57.02, CUDA Version: 11.4
-_GPU_UBUNTU_DATE_PYTORCH = [
-    ('gpu', '20.04', '20231103', '2.1.0'),
-    ('gpu', '18.04', '20221114', '1.10.0'),
-    ('k80', '20.04', '20211208', '1.10.0'),
-    ('k80', '18.04', '20211208', '1.10.0'),
+#
+# Neuron (Inferentia / Trainium):
+# https://aws.amazon.com/releasenotes/aws-deep-learning-ami-base-neuron-ubuntu-20-04/  # pylint: disable=line-too-long
+# Deep Learning Base Neuron AMI (Ubuntu 20.04) 20240923
+# TODO(tian): find out the driver version.
+#   Neuron driver:
+_GPU_DESC_UBUNTU_DATE = [
+    ('gpu', 'AMI GPU PyTorch 2.1.0', '20.04', '20231103'),
+    ('gpu', 'AMI GPU PyTorch 1.10.0', '18.04', '20221114'),
+    ('k80', 'AMI GPU PyTorch 1.10.0', '20.04', '20211208'),
+    ('k80', 'AMI GPU PyTorch 1.10.0', '18.04', '20211208'),
+    ('neuron', 'Base Neuron AMI', '22.04', '20240923'),
 ]
 
 
-def _fetch_image_id(region: str, ubuntu_version: str, creation_date: str,
-                    pytorch_version: str) -> Optional[str]:
+def _fetch_image_id(region: str, description: str, ubuntu_version: str,
+                    creation_date: str) -> Optional[str]:
     try:
         image = subprocess.check_output(f"""\
             aws ec2 describe-images --region {region} --owners amazon \\
-                --filters 'Name=name,Values="Deep Learning AMI GPU PyTorch {pytorch_version} (Ubuntu {ubuntu_version}) {creation_date}"' \\
+                --filters 'Name=name,Values="Deep Learning {description} (Ubuntu {ubuntu_version}) {creation_date}"' \\
                     'Name=state,Values=available' --query 'Images[:1].ImageId' --output text
             """,
                                         shell=True)
     except subprocess.CalledProcessError as e:
-        print(f'Failed {region}, {ubuntu_version}, {creation_date}. '
-              'Trying next date.')
+        print(f'Failed {region}, {description}, {ubuntu_version}, '
+              f'{creation_date}. Trying next date.')
         print(f'{type(e)}: {e}')
         image_id = None
     else:
@@ -407,21 +414,21 @@ def _fetch_image_id(region: str, ubuntu_version: str, creation_date: str,
     return image_id
 
 
-def _get_image_row(
-        region: str, gpu: str, ubuntu_version: str, date: str,
-        pytorch_version) -> Tuple[str, str, str, str, Optional[str], str]:
-    print(f'Getting image for {region}, {ubuntu_version}, {gpu}')
-    image_id = _fetch_image_id(region, ubuntu_version, date, pytorch_version)
+def _get_image_row(region: str, gpu: str, description: str, ubuntu_version: str,
+                   date: str) -> Tuple[str, str, str, str, Optional[str], str]:
+    print(f'Getting image for {region}, {description}, {ubuntu_version}, {gpu}')
+    image_id = _fetch_image_id(region, description, ubuntu_version, date)
     if image_id is None:
         # not found
-        print(f'Failed to find image for {region}, {ubuntu_version}, {gpu}')
+        print(f'Failed to find image for {region}, {description}, '
+              f'{ubuntu_version}, {gpu}')
     tag = f'skypilot:{gpu}-ubuntu-{ubuntu_version.replace(".", "")}'
     return tag, region, 'ubuntu', ubuntu_version, image_id, date
 
 
 def get_all_regions_images_df(regions: Set[str]) -> 'pd.DataFrame':
     image_metas = [
-        (r, *i) for r, i in itertools.product(regions, _GPU_UBUNTU_DATE_PYTORCH)
+        (r, *i) for r, i in itertools.product(regions, _GPU_DESC_UBUNTU_DATE)
     ]
     with mp_pool.Pool() as pool:
         results = pool.starmap(_get_image_row, image_metas)

{skypilot_nightly-1.0.0.dev20240930 → skypilot_nightly-1.0.0.dev20241002}/sky/resources.py

@@ -966,20 +966,22 @@ class Resources:
         """
         if not self._labels:
             return
-
-        if self.cloud is None:
-            # Because each cloud has its own label format, we cannot validate
-            # the labels without knowing the cloud.
-            with ux_utils.print_exception_no_traceback():
-                raise ValueError(
-                    'Cloud must be specified when labels are provided.')
-
-        # Check if the label key value pairs are valid.
+        if self.cloud is not None:
+            validated_clouds = [self.cloud]
+        else:
+            # If no specific cloud is set, validate label against ALL clouds.
+            # The label will be dropped if invalid for any one of the cloud
+            validated_clouds = sky_check.get_cached_enabled_clouds_or_refresh()
         invalid_table = log_utils.create_table(['Label', 'Reason'])
         for key, value in self._labels.items():
-            valid, err_msg = self.cloud.is_label_valid(key, value)
-            if not valid:
-                invalid_table.add_row([f'{key}: {value}', err_msg])
+            for cloud in validated_clouds:
+                valid, err_msg = cloud.is_label_valid(key, value)
+                if not valid:
+                    invalid_table.add_row([
+                        f'{key}: {value}',
+                        f'Label rejected due to {cloud}: {err_msg}'
+                    ])
+                    break
         if len(invalid_table.rows) > 0:
             with ux_utils.print_exception_no_traceback():
                 raise ValueError(

{skypilot_nightly-1.0.0.dev20240930 → skypilot_nightly-1.0.0.dev20241002/skypilot_nightly.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: skypilot-nightly
-Version: 1.0.0.dev20240930
+Version: 1.0.0.dev20241002
 Summary: SkyPilot: An intercloud broker for the clouds
 Author: SkyPilot Team
 License: Apache 2.0