skypilot-nightly 1.0.0.dev20240929__py3-none-any.whl → 1.0.0.dev20241001__py3-none-any.whl

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 = 'e6a3b830fb2a12871815773af6171d42e0416e89'
+_SKYPILOT_COMMIT_SHA = 'b1f22c4d5fe0a3cc25d1df0a8a05d4230a28b702'
 
 
 def _get_git_commit():
@@ -35,7 +35,7 @@ def _get_git_commit():
 
 
 __commit__ = _get_git_commit()
-__version__ = '1.0.0.dev20240929'
+__version__ = '1.0.0.dev20241001'
 __root_dir__ = os.path.dirname(os.path.abspath(__file__))
 
 

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

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)

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.dev20240929.dist-info → skypilot_nightly-1.0.0.dev20241001.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: skypilot-nightly
-Version: 1.0.0.dev20240929
+Version: 1.0.0.dev20241001
 Summary: SkyPilot: An intercloud broker for the clouds
 Author: SkyPilot Team
 License: Apache 2.0

{skypilot_nightly-1.0.0.dev20240929.dist-info → skypilot_nightly-1.0.0.dev20241001.dist-info}/RECORD

@@ -1,4 +1,4 @@
-sky/__init__.py,sha256=8FioKRx3X_BHtQt6BCrINW2IHHhrCWkiT7cb8NcaRjY,5854
+sky/__init__.py,sha256=KBftRzx6giOgDIRTtKtHuEcmtof_64nk7UMMQzMsLRQ,5854
 sky/admin_policy.py,sha256=hPo02f_A32gCqhUueF0QYy1fMSSKqRwYEg_9FxScN_s,3248
 sky/authentication.py,sha256=TfKkVnmRIetATSEVQFp-rOOIRGqVig2i8faSQQt_ixA,20974
 sky/check.py,sha256=jLMIIJrseaZj1_o5WkbaD9XdyXIlCaT6pyAaIFdhdmA,9079
@@ -10,7 +10,7 @@ sky/exceptions.py,sha256=s7j0iCa1Ec0rU1ABb9EAhqn2qFm22bmKQV_ckgRlMGk,8720
 sky/execution.py,sha256=97yhNh5BKBh2ZJW8GefGID_4KCYU-arnvemSJB9rf6U,25552
 sky/global_user_state.py,sha256=PywEmUutF97XBgRMClR6IS5_KM8JJC0oA1LsPUZebp0,28681
 sky/optimizer.py,sha256=YGBhJPlcvylYON7MLrYEMtBOqJLt4LdlguQclVvvl4E,58677
-sky/resources.py,sha256=_959wcQnoiAYesslN9BPXWABFaQfc_TFXPO_o7SPlxI,67325
+sky/resources.py,sha256=b9yaZvZkL-QZdElQLHsEZ2jhKgId2ixG8M2Z8DLBBKU,67450
 sky/sky_logging.py,sha256=I59__M9taBjDim15ie0m25Vtn6itLtR9Ao8W9FS36Xs,4253
 sky/skypilot_config.py,sha256=E3g65cX3P3dT9b5N0GgFBG6yB0FXwIGpisKoozmJmWU,9094
 sky/status_lib.py,sha256=J7Jb4_Dz0v2T64ttOdyUgpokvl4S0sBJrMfH7Fvo51A,1457
@@ -40,7 +40,7 @@ sky/benchmark/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 sky/benchmark/benchmark_state.py,sha256=X8CXmuU9KgsDRhKedhFgjeRMUFWtQsjFs1qECvPG2yg,8723
 sky/benchmark/benchmark_utils.py,sha256=oJOzJ4fs2sruxYh4Tl1NZ5fi2-3oWfXtoeCIAq2hgjw,26136
 sky/clouds/__init__.py,sha256=WuNIJEnZmBO72tU5awgaaL3rdvFRSkgaYNNeuY68dXo,1356
-sky/clouds/aws.py,sha256=6nNwj3MHFzYTonLwx-QmXNvPlNZttS3dr8ULAoN79D0,48615
+sky/clouds/aws.py,sha256=0cpFokzfzEUNnRKMaY198GjrUTXIO0adu6uVArRrlBA,48814
 sky/clouds/azure.py,sha256=Yp_a1Lzvq4s47eRMeyVheDv9pC0hSPogCiTMYf-a5ZE,28687
 sky/clouds/cloud.py,sha256=PPk-Cbf1YbJT8bswcQLtPBtko02OWrRGJKkLzDpytTI,34858
 sky/clouds/cloud_registry.py,sha256=4yQMv-iBSgyN5aNL4Qxbn0JVE-dkVoEUIgj7S1z9S_Q,955
@@ -74,7 +74,7 @@ sky/clouds/service_catalog/scp_catalog.py,sha256=4XnaZE5Q4XrrNnDnVhsHkH6jxmWXBeQ
 sky/clouds/service_catalog/vsphere_catalog.py,sha256=yJLWu9SQep-PRn1YdeQ7ZoNqQHTAxJtxf7y6FBrfSW0,4391
 sky/clouds/service_catalog/data_fetchers/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 sky/clouds/service_catalog/data_fetchers/analyze.py,sha256=VdksJQs3asFE8H5T3ZV1FJas2xD9WEX6c-V5p7y-wp4,2084
-sky/clouds/service_catalog/data_fetchers/fetch_aws.py,sha256=Fcm9_IkmTBXM9gp5VohufOvVVB-ZeRf_sSCHRyDWrdY,22424
+sky/clouds/service_catalog/data_fetchers/fetch_aws.py,sha256=6gpRtQaQtvT1cMpiVBacNXXQAjBC5MWpAUI-1ELYg0U,22850
 sky/clouds/service_catalog/data_fetchers/fetch_azure.py,sha256=jsSVqbSbBIw_IYmO-y2u4co20AJ-JF713KFjUKdO_VA,12272
 sky/clouds/service_catalog/data_fetchers/fetch_cudo.py,sha256=52P48lvWN0s1ArjeLPeLemPRpxjSRcHincRle0nqdm4,3440
 sky/clouds/service_catalog/data_fetchers/fetch_fluidstack.py,sha256=35nO_VaDOgp5W13kt_lIANSk_CNf7gBiZGJ5fGyZu6o,6808
@@ -273,9 +273,9 @@ sky/utils/kubernetes/k8s_gpu_labeler_job.yaml,sha256=KPqp23B-zQ2SZK03jdHeF9fLTog
 sky/utils/kubernetes/k8s_gpu_labeler_setup.yaml,sha256=VLKT2KKimZu1GDg_4AIlIt488oMQvhRZWwsj9vBbPUg,3812
 sky/utils/kubernetes/rsync_helper.sh,sha256=Ma-N9a271fTfdgP5-8XIQL7KPf8IPUo-uY004PCdUFo,747
 sky/utils/kubernetes/ssh_jump_lifecycle_manager.py,sha256=RFLJ3k7MR5UN4SKHykQ0lV9SgXumoULpKYIAt1vh-HU,6560
-skypilot_nightly-1.0.0.dev20240929.dist-info/LICENSE,sha256=emRJAvE7ngL6x0RhQvlns5wJzGI3NEQ_WMjNmd9TZc4,12170
-skypilot_nightly-1.0.0.dev20240929.dist-info/METADATA,sha256=WfIgFJj_CNWtVZ39aN4Bx7l35R3rE4vdr6tJR9w9m-A,18948
-skypilot_nightly-1.0.0.dev20240929.dist-info/WHEEL,sha256=GV9aMThwP_4oNCtvEC2ec3qUYutgWeAzklro_0m4WJQ,91
-skypilot_nightly-1.0.0.dev20240929.dist-info/entry_points.txt,sha256=StA6HYpuHj-Y61L2Ze-hK2IcLWgLZcML5gJu8cs6nU4,36
-skypilot_nightly-1.0.0.dev20240929.dist-info/top_level.txt,sha256=qA8QuiNNb6Y1OF-pCUtPEr6sLEwy2xJX06Bd_CrtrHY,4
-skypilot_nightly-1.0.0.dev20240929.dist-info/RECORD,,
+skypilot_nightly-1.0.0.dev20241001.dist-info/LICENSE,sha256=emRJAvE7ngL6x0RhQvlns5wJzGI3NEQ_WMjNmd9TZc4,12170
+skypilot_nightly-1.0.0.dev20241001.dist-info/METADATA,sha256=x7BEUsYqWAeufLCtSK-NBoRQxL9jyZ58xR00gGBqC8c,18948
+skypilot_nightly-1.0.0.dev20241001.dist-info/WHEEL,sha256=GV9aMThwP_4oNCtvEC2ec3qUYutgWeAzklro_0m4WJQ,91
+skypilot_nightly-1.0.0.dev20241001.dist-info/entry_points.txt,sha256=StA6HYpuHj-Y61L2Ze-hK2IcLWgLZcML5gJu8cs6nU4,36
+skypilot_nightly-1.0.0.dev20241001.dist-info/top_level.txt,sha256=qA8QuiNNb6Y1OF-pCUtPEr6sLEwy2xJX06Bd_CrtrHY,4
+skypilot_nightly-1.0.0.dev20241001.dist-info/RECORD,,