skypilot-nightly 1.0.0.dev20250428__py3-none-any.whl → 1.0.0.dev20250430__py3-none-any.whl

sky/server/server.py

@@ -35,6 +35,7 @@ from sky.jobs.server import server as jobs_rest
 from sky.provision.kubernetes import utils as kubernetes_utils
 from sky.serve.server import server as serve_rest
 from sky.server import common
+from sky.server import config as server_config
 from sky.server import constants as server_constants
 from sky.server import stream_utils
 from sky.server.requests import executor
@@ -1166,13 +1167,12 @@ if __name__ == '__main__':
     # that it is shown only when the API server is started.
     usage_lib.maybe_show_privacy_policy()
 
-    num_workers = 1
-    if cmd_args.deploy:
-        num_workers = common_utils.get_cpu_count()
+    config = server_config.compute_server_config(cmd_args.deploy)
+    num_workers = config.num_server_workers
 
     sub_procs = []
     try:
-        sub_procs = executor.start(deploy=cmd_args.deploy)
+        sub_procs = executor.start(config)
         logger.info(f'Starting SkyPilot API server, workers={num_workers}')
         # We don't support reload for now, since it may cause leakage of request
         # workers or interrupt running requests.

sky/setup_files/dependencies.py

@@ -53,7 +53,6 @@ install_requires = [
     'aiofiles',
     'httpx',
     'setproctitle',
-    'omegaconf>=2.4.0dev3,<2.5',
 ]
 
 local_ray = [

sky/setup_files/setup.py

@@ -161,7 +161,7 @@ setuptools.setup(
     author='SkyPilot Team',
     license='Apache 2.0',
     readme='README.md',
-    description='SkyPilot: An intercloud broker for the clouds',
+    description='SkyPilot: Run AI on Any Infra — Unified, Faster, Cheaper.',
     long_description=long_description,
     long_description_content_type='text/markdown',
     setup_requires=['wheel'],

sky/skylet/constants.py

@@ -280,6 +280,24 @@ USER_ID_ENV_VAR = f'{SKYPILOT_ENV_VAR_PREFIX}USER_ID'
 # runs on a VM launched by SkyPilot will be recognized as the same user.
 USER_ENV_VAR = f'{SKYPILOT_ENV_VAR_PREFIX}USER'
 
+# SSH configuration to allow more concurrent sessions and connections.
+# Default MaxSessions is 10.
+# Default MaxStartups is 10:30:60, meaning:
+#   - Up to 10 unauthenticated connections are allowed without restriction.
+#   - From 11 to 60 connections, 30% are randomly dropped.
+#   - Above 60 connections, all are dropped.
+# These defaults are too low for submitting many parallel jobs (e.g., 150),
+# which can easily exceed the limits and cause connection failures.
+# The new values (MaxSessions 200, MaxStartups 150:30:200) increase these
+# limits significantly.
+# TODO(zeping): Bake this configuration in SkyPilot default images.
+SET_SSH_MAX_SESSIONS_CONFIG_CMD = (
+    'sudo bash -c \''
+    'echo "MaxSessions 200" >> /etc/ssh/sshd_config; '
+    'echo "MaxStartups 150:30:200" >> /etc/ssh/sshd_config; '
+    '(systemctl reload sshd || service ssh reload); '
+    '\'')
+
 # Internal: Env var indicating the system is running with a remote API server.
 # It is used for internal purposes, including the jobs controller to mark
 # clusters as launched with a remote API server.

sky/skypilot_config.py

@@ -56,8 +56,6 @@ import threading
 import typing
 from typing import Any, Dict, Iterator, List, Optional, Tuple
 
-from omegaconf import OmegaConf
-
 from sky import exceptions
 from sky import sky_logging
 from sky.adaptors import common as adaptors_common
@@ -141,7 +139,7 @@ def get_user_config() -> config_utils.Config:
 
     # load the user config file
     if os.path.exists(user_config_path):
-        user_config = _parse_config_file(user_config_path)
+        user_config = parse_config_file(user_config_path)
         _validate_config(user_config, user_config_path)
     else:
         user_config = config_utils.Config()
@@ -170,7 +168,7 @@ def _get_project_config() -> config_utils.Config:
 
     # load the project config file
     if os.path.exists(project_config_path):
-        project_config = _parse_config_file(project_config_path)
+        project_config = parse_config_file(project_config_path)
         _validate_config(project_config, project_config_path)
     else:
         project_config = config_utils.Config()
@@ -199,7 +197,7 @@ def get_server_config() -> config_utils.Config:
 
     # load the server config file
     if os.path.exists(server_config_path):
-        server_config = _parse_config_file(server_config_path)
+        server_config = parse_config_file(server_config_path)
         _validate_config(server_config, server_config_path)
     else:
         server_config = config_utils.Config()
@@ -304,7 +302,7 @@ def _reload_config() -> None:
         _reload_config_as_client()
 
 
-def _parse_config_file(config_path: str) -> config_utils.Config:
+def parse_config_file(config_path: str) -> config_utils.Config:
     config = config_utils.Config()
     try:
         config_dict = common_utils.read_yaml(config_path)
@@ -321,6 +319,31 @@ def _parse_config_file(config_path: str) -> config_utils.Config:
     return config
 
 
+def _parse_dotlist(dotlist: List[str]) -> config_utils.Config:
+    """Parse a comma-separated list of key-value pairs into a dictionary.
+
+    Args:
+        dotlist: A comma-separated list of key-value pairs.
+
+    Returns:
+        A config_utils.Config object with the parsed key-value pairs.
+    """
+    config: config_utils.Config = config_utils.Config()
+    for arg in dotlist:
+        try:
+            key, value = arg.split('=', 1)
+        except ValueError as e:
+            raise ValueError(f'Invalid config override: {arg}. '
+                             'Please use the format: key=value') from e
+        if len(key) == 0 or len(value) == 0:
+            raise ValueError(f'Invalid config override: {arg}. '
+                             'Please use the format: key=value')
+        value = yaml.safe_load(value)
+        nested_keys = tuple(key.split('.'))
+        config.set_nested(nested_keys, value)
+    return config
+
+
 def _reload_config_from_internal_file(internal_config_path: str) -> None:
     global _dict, _loaded_config_path
     # Reset the global variables, to avoid using stale values.
@@ -336,7 +359,7 @@ def _reload_config_from_internal_file(internal_config_path: str) -> None:
                 'exist. Please double check the path or unset the env var: '
                 f'unset {ENV_VAR_SKYPILOT_CONFIG}')
     logger.debug(f'Using config path: {config_path}')
-    _dict = _parse_config_file(config_path)
+    _dict = parse_config_file(config_path)
     _loaded_config_path = config_path
 
 
@@ -483,11 +506,9 @@ def _compose_cli_config(cli_config: Optional[List[str]]) -> config_utils.Config:
                     'Cannot use multiple --config flags with a config file.')
             config_source = maybe_config_path
             # cli_config is a path to a config file
-            parsed_config = OmegaConf.to_object(
-                OmegaConf.load(maybe_config_path))
+            parsed_config = parse_config_file(maybe_config_path)
         else:  # cli_config is a comma-separated list of key-value pairs
-            parsed_config = OmegaConf.to_object(
-                OmegaConf.from_dotlist(cli_config))
+            parsed_config = _parse_dotlist(cli_config)
         _validate_config(parsed_config, config_source)
     except ValueError as e:
         raise ValueError(f'Invalid config override: {cli_config}. '

sky/templates/aws-ray.yml.j2

@@ -142,7 +142,7 @@ available_node_types:
             {%- endfor %}
       # Use IDMSv2
       MetadataOptions:
-        HttpTokens: required 
+        HttpTokens: required
 
 head_node_type: ray.head.default
 
@@ -192,6 +192,7 @@ setup_commands:
     {%- if remote_identity != 'LOCAL_CREDENTIALS' %}
     rm ~/.aws/credentials || true;
     {%- endif %}
+    {{ ssh_max_sessions_config }}
 
 # Command to start ray clusters are now placed in `sky.provision.instance_setup`.
 # We do not need to list it here anymore.

sky/templates/azure-ray.yml.j2

@@ -68,7 +68,7 @@ available_node_types:
             imageOffer: {{image_offer}}
             imageSku: "{{image_sku}}"
             imageVersion: {{image_version}}
-            # Community Gallery Image ID 
+            # Community Gallery Image ID
             communityGalleryImageId: {{community_gallery_image_id}}
             osDiskSizeGB: {{disk_size}}
             osDiskTier: {{disk_tier}}
@@ -130,3 +130,4 @@ setup_commands:
     mkdir -p ~/.ssh; (grep -Pzo -q "Host \*\n  StrictHostKeyChecking no\n  IdentityFile ~/.ssh/sky-cluster-key\n  IdentityFile ~/.ssh/id_rsa" ~/.ssh/config) || printf "Host *\n  StrictHostKeyChecking no\n  IdentityFile ~/.ssh/sky-cluster-key\n  IdentityFile ~/.ssh/id_rsa\n" >> ~/.ssh/config;
     [ -f /etc/fuse.conf ] && sudo sed -i 's/#user_allow_other/user_allow_other/g' /etc/fuse.conf || (sudo sh -c 'echo "user_allow_other" > /etc/fuse.conf');
     sudo mv /etc/nccl.conf /etc/nccl.conf.bak || true;
+    {{ ssh_max_sessions_config }}

sky/templates/cudo-ray.yml.j2

@@ -73,3 +73,4 @@ setup_commands:
     sudo grep -e '^DefaultTasksMax' /etc/systemd/system.conf || (sudo bash -c 'echo "DefaultTasksMax=infinity" >> /etc/systemd/system.conf'); sudo systemctl set-property user-$(id -u $(whoami)).slice TasksMax=infinity; sudo systemctl daemon-reload;
     mkdir -p ~/.ssh; (grep -Pzo -q "Host \*\n  StrictHostKeyChecking no\n  IdentityFile ~/.ssh/sky-cluster-key\n  IdentityFile ~/.ssh/id_rsa" ~/.ssh/config) || printf "Host *\n  StrictHostKeyChecking no\n  IdentityFile ~/.ssh/sky-cluster-key\n  IdentityFile ~/.ssh/id_rsa\n" >> ~/.ssh/config;
     [ -f /etc/fuse.conf ] && sudo sed -i 's/#user_allow_other/user_allow_other/g' /etc/fuse.conf || (sudo sh -c 'echo "user_allow_other" > /etc/fuse.conf');
+    {{ ssh_max_sessions_config }}

sky/templates/do-ray.yml.j2

@@ -93,6 +93,7 @@ setup_commands:
     sudo grep -e '^DefaultTasksMax' /etc/systemd/system.conf || (sudo bash -c 'echo "DefaultTasksMax=infinity" >> /etc/systemd/system.conf'); sudo systemctl set-property user-$(id -u $(whoami)).slice TasksMax=infinity; sudo systemctl daemon-reload;
     mkdir -p ~/.ssh; (grep -Pzo -q "Host \*\n  StrictHostKeyChecking no" ~/.ssh/config) || printf "Host *\n  StrictHostKeyChecking no\n" >> ~/.ssh/config;
     [ -f /etc/fuse.conf ] && sudo sed -i 's/#user_allow_other/user_allow_other/g' /etc/fuse.conf || (sudo sh -c 'echo "user_allow_other" > /etc/fuse.conf');
-    
-# Command to start ray clusters are now placed in `sky.provision.instance_setup`. 
+    {{ ssh_max_sessions_config }}
+
+# Command to start ray clusters are now placed in `sky.provision.instance_setup`.
 # We do not need to list it here anymore.

sky/templates/fluidstack-ray.yml.j2

@@ -74,4 +74,4 @@ setup_commands:
     sudo grep -e '^DefaultTasksMax' /etc/systemd/system.conf || (sudo bash -c 'echo "DefaultTasksMax=infinity" >> /etc/systemd/system.conf'); sudo systemctl set-property user-$(id -u $(whoami)).slice TasksMax=infinity; sudo systemctl daemon-reload;
     mkdir -p ~/.ssh; (grep -Pzo -q "Host \*\n  StrictHostKeyChecking no\n  IdentityFile ~/.ssh/sky-cluster-key\n  IdentityFile ~/.ssh/id_rsa" ~/.ssh/config) || printf "Host *\n  StrictHostKeyChecking no\n  IdentityFile ~/.ssh/sky-cluster-key\n  IdentityFile ~/.ssh/id_rsa\n" >> ~/.ssh/config;
     [ -f /etc/fuse.conf ] && sudo sed -i 's/#user_allow_other/user_allow_other/g' /etc/fuse.conf || (sudo sh -c 'echo "user_allow_other" > /etc/fuse.conf');
-
+    {{ ssh_max_sessions_config }}

sky/templates/gcp-ray.yml.j2

@@ -1,4 +1,3 @@
-
 cluster_name: {{cluster_name_on_cloud}}
 
 # The maximum number of workers nodes to launch in addition to the head node.
@@ -225,6 +224,7 @@ setup_commands:
     {%- endif %}
     mkdir -p ~/.ssh; (grep -Pzo -q "Host \*\n  StrictHostKeyChecking no\n  IdentityFile ~/.ssh/sky-cluster-key\n  IdentityFile ~/.ssh/id_rsa" ~/.ssh/config) || printf "Host *\n  StrictHostKeyChecking no\n  IdentityFile ~/.ssh/sky-cluster-key\n  IdentityFile ~/.ssh/id_rsa\n" >> ~/.ssh/config;
     [ -f /etc/fuse.conf ] && sudo sed -i 's/#user_allow_other/user_allow_other/g' /etc/fuse.conf || (sudo sh -c 'echo "user_allow_other" > /etc/fuse.conf');
+    {{ ssh_max_sessions_config }}
 
 # Command to start ray clusters are now placed in `sky.provision.instance_setup`.
 # We do not need to list it here anymore.

sky/templates/ibm-ray.yml.j2

@@ -32,7 +32,7 @@ available_node_types:
   ray_head_default:
     resources: {{instance_resources}}
     node_config:
-      image_id: {{image_id}}  
+      image_id: {{image_id}}
       boot_volume_capacity: {{disk_capacity}}
       volume_tier_name: general-purpose
       instance_profile_name: {{instance_type}}
@@ -48,7 +48,7 @@ available_node_types:
     max_workers: {{num_nodes - 1}}
     resources: {{worker_instance_resources}}
     node_config:
-      image_id: {{image_id}} 
+      image_id: {{image_id}}
       boot_volume_capacity: {{disk_capacity}}
       volume_tier_name: general-purpose
       instance_profile_name: {{worker_instance_type}}
@@ -106,7 +106,7 @@ setup_commands:
     sudo grep -e '^DefaultTasksMax' /etc/systemd/system.conf || (sudo bash -c 'echo "DefaultTasksMax=infinity" >> /etc/systemd/system.conf'); sudo systemctl set-property user-$(id -u $(whoami)).slice TasksMax=infinity; sudo systemctl daemon-reload;
     mkdir -p ~/.ssh; (grep -Pzo -q "Host \*\n  StrictHostKeyChecking no\n  IdentityFile ~/.ssh/sky-cluster-key\n  IdentityFile ~/.ssh/id_rsa" ~/.ssh/config) || printf "Host *\n  StrictHostKeyChecking no\n  IdentityFile ~/.ssh/sky-cluster-key\n  IdentityFile ~/.ssh/id_rsa\n" >> ~/.ssh/config;
     [ -f /etc/fuse.conf ] && sudo sed -i 's/#user_allow_other/user_allow_other/g' /etc/fuse.conf || (sudo sh -c 'echo "user_allow_other" > /etc/fuse.conf'); # This is needed for `-o allow_other` option for `goofys`;
-
+    {{ ssh_max_sessions_config }}
 
 # Command to start ray on the head node. You don't need to change this.
 # NOTE: these are very performance-sensitive. Each new item opens/closes an SSH

sky/templates/kubernetes-ray.yml.j2

@@ -258,7 +258,7 @@ available_node_types:
         # service is required.
         labels:
             parent: skypilot
-            # component will be set for the head node pod to be the same as the head node service selector above if a 
+            # component will be set for the head node pod to be the same as the head node service selector above if a
             skypilot-cluster: {{cluster_name_on_cloud}}
             # Identifies the SSH jump pod used by this pod. Used in life cycle management of the ssh jump pod.
             skypilot-ssh-jump: {{k8s_ssh_jump_name}}
@@ -277,11 +277,8 @@ available_node_types:
         restartPolicy: {{ "Always" if high_availability else "Never" }}
 
         # Add node selector if GPU/TPUs are requested:
-        {% if (k8s_acc_label_key is not none and k8s_acc_label_value is not none) or (k8s_spot_label_key is not none) %}
+        {% if (k8s_topology_label_key is not none and k8s_topology_label_value is not none) or (k8s_spot_label_key is not none) %}
         nodeSelector:
-            {% if k8s_acc_label_key is not none and k8s_acc_label_value is not none %}
-            {{k8s_acc_label_key}}: {{k8s_acc_label_value}}
-            {% endif %}
             {% if k8s_topology_label_key is not none and k8s_topology_label_value is not none %}
             {{k8s_topology_label_key}}: {{k8s_topology_label_value}}
             {% endif %}
@@ -289,6 +286,19 @@ available_node_types:
             {{k8s_spot_label_key}}: {{k8s_spot_label_value|tojson}}
             {% endif %}
         {% endif %}
+        {% if (k8s_acc_label_key is not none and k8s_acc_label_values is not none) %}
+        affinity:
+          nodeAffinity:
+            requiredDuringSchedulingIgnoredDuringExecution:
+              nodeSelectorTerms:
+              - matchExpressions:
+                - key: {{k8s_acc_label_key}}
+                  operator: In
+                  values:
+                  {% for label_value in k8s_acc_label_values %}
+                  - {{label_value}}
+                  {% endfor %}
+        {% endif %}
 
         {% if k8s_spot_label_key is not none %}
         tolerations:
@@ -339,15 +349,15 @@ available_node_types:
           # Do not change this command - it keeps the pod alive until it is
           # explicitly killed.
           command: ["/bin/bash", "-c", "--"]
-          args: 
+          args:
             - |
               # For backwards compatibility, we put a marker file in the pod
-              # to indicate that the pod is running with the changes introduced 
+              # to indicate that the pod is running with the changes introduced
               # in project nimbus: https://github.com/skypilot-org/skypilot/pull/4393
               # TODO: Remove this marker file and it's usage in setup_commands
               # after v0.10.0 release.
               touch /tmp/skypilot_is_nimbus
-              
+
               # Helper function to conditionally use sudo
               # TODO(zhwu): consolidate the two prefix_cmd and sudo replacements
               prefix_cmd() { if [ $(id -u) -ne 0 ]; then echo "sudo"; else echo ""; fi; }
@@ -390,7 +400,7 @@ available_node_types:
                 fi;
                 # SSH and other packages are not necessary, so we disable set -e
                 set +e
-                
+
                 if [ ! -z "$MISSING_PACKAGES" ]; then
                   # Install missing packages individually to avoid failure installation breaks the whole install process,
                   # e.g. fuse3 is not available on some distributions.
@@ -443,7 +453,7 @@ available_node_types:
                     $(prefix_cmd) rm -f /bin/fusermount-wrapper
                     $(prefix_cmd) cp -p {{k8s_fusermount_shared_dir}}/fusermount-wrapper /bin/fusermount-wrapper
                 fi
-                {% endif %}            
+                {% endif %}
 
                 $(prefix_cmd) mkdir -p /var/run/sshd;
                 $(prefix_cmd) sed -i "s/PermitRootLogin prohibit-password/PermitRootLogin yes/" /etc/ssh/sshd_config;
@@ -574,7 +584,7 @@ available_node_types:
                       # File is already being monitored
                       continue
                     fi
-                    
+
                     # Monitor the new file
                     monitor_file $file &
                     already_monitored="${already_monitored} ${file}"
@@ -651,7 +661,7 @@ available_node_types:
               {{k8s_resource_key}}: {{accelerator_count}}
               {% endif %}
             {% endif %}
-      
+
       {% if high_availability %}
       pvc_spec:
         apiVersion: v1
@@ -737,6 +747,7 @@ available_node_types:
                       mountPath: /mnt/home # Temporary mount point for initialization
               # should be replaced by pod spec
       {% endif %}
+
 setup_commands:
   # Disable `unattended-upgrades` to prevent apt-get from hanging. It should be called at the beginning before the process started to avoid being blocked. (This is a temporary fix.)
   # Add ~/.ssh/sky-cluster-key to SSH config to allow nodes within a cluster to connect to each other
@@ -756,7 +767,7 @@ setup_commands:
     echo "=== Logs for asynchronous ray and skypilot installation ===";
     if [ -f /tmp/skypilot_is_nimbus ]; then
       echo "=== Logs for asynchronous ray and skypilot installation ===";
-      [ -f /tmp/ray_skypilot_installation_complete ] && cat /tmp/${STEPS[1]}.log || 
+      [ -f /tmp/ray_skypilot_installation_complete ] && cat /tmp/${STEPS[1]}.log ||
       { tail -f -n +1 /tmp/${STEPS[1]}.log & TAIL_PID=$!; echo "Tail PID: $TAIL_PID"; until [ -f /tmp/ray_skypilot_installation_complete ]; do sleep 0.5; done; kill $TAIL_PID || true; };
       [ -f /tmp/${STEPS[1]}.failed ] && { echo "Error: ${STEPS[1]} failed. Exiting."; exit 1; } || true;
     fi
@@ -786,9 +797,10 @@ setup_commands:
     # properly written.
     # TODO(Doyoung): Investigate to see why TPU workload fails to run without
     # execution permission, such as granting 766 to log file. Check if it's a
-    # must and see if there's a workaround to grant minimum permission. 
+    # must and see if there's a workaround to grant minimum permission.
     sudo chmod 777 /tmp/tpu_logs;
     {% endif %}
+    {{ ssh_max_sessions_config }}
 
 # Format: `REMOTE_PATH : LOCAL_PATH`
 file_mounts: {

sky/templates/lambda-ray.yml.j2

@@ -96,6 +96,7 @@ setup_commands:
     sudo grep -e '^DefaultTasksMax' /etc/systemd/system.conf || (sudo bash -c 'echo "DefaultTasksMax=infinity" >> /etc/systemd/system.conf'); sudo systemctl set-property user-$(id -u $(whoami)).slice TasksMax=infinity; sudo systemctl daemon-reload;
     mkdir -p ~/.ssh; (grep -Pzo -q "Host \*\n  StrictHostKeyChecking no\n  IdentityFile ~/.ssh/sky-cluster-key\n  IdentityFile ~/.ssh/id_rsa" ~/.ssh/config) || printf "Host *\n  StrictHostKeyChecking no\n  IdentityFile ~/.ssh/sky-cluster-key\n  IdentityFile ~/.ssh/id_rsa\n" >> ~/.ssh/config;
     [ -f /etc/fuse.conf ] && sudo sed -i 's/#user_allow_other/user_allow_other/g' /etc/fuse.conf || (sudo sh -c 'echo "user_allow_other" > /etc/fuse.conf');
+    {{ ssh_max_sessions_config }}
 
 # Command to start ray clusters are now placed in `sky.provision.instance_setup`.
 # We do not need to list it here anymore.

sky/templates/nebius-ray.yml.j2

@@ -10,6 +10,27 @@ provider:
   module: sky.provision.nebius
   region: "{{region}}"
 
+{%- if docker_image is not none %}
+docker:
+  image: {{docker_image}}
+  container_name: {{docker_container_name}}
+  run_options:
+    - --ulimit nofile=1048576:1048576
+    {%- for run_option in docker_run_options %}
+    - {{run_option}}
+    {%- endfor %}
+  {%- if docker_login_config is not none %}
+  docker_login_config:
+    username: |-
+      {{docker_login_config.username}}
+    password: |-
+      {{docker_login_config.password}}
+    server: |-
+      {{docker_login_config.server}}
+  {%- endif %}
+{%- endif %}
+
+
 auth:
   ssh_user: ubuntu
   ssh_private_key: {{ssh_private_key}}
@@ -22,6 +43,48 @@ available_node_types:
       ImageId: {{image_id}}
       DiskSize: {{disk_size}}
       UserData: |
+        {%- if docker_image is not none %}
+        runcmd:
+          - sudo sed -i 's/^#\?AllowTcpForwarding.*/AllowTcpForwarding yes/' /etc/ssh/sshd_config
+          - systemctl restart sshd
+        {%- endif %}
+
+        {# Two available OS images:
+               1. ubuntu22.04-driverless - requires Docker installation
+               2. ubuntu22.04-cuda12 - comes with Docker pre-installed
+            To optimize deployment speed, Docker is only installed when using ubuntu22.04-driverless #}
+        {%- if docker_image is not none and image_id == 'ubuntu22.04-driverless' %}
+        apt:
+          sources:
+            docker.list:
+              source: deb [arch=amd64] https://download.docker.com/linux/ubuntu $RELEASE stable
+              keyid: 9DC858229FC7DD38854AE2D88D81803C0EBFCD88
+
+        packages:
+          - apt-transport-https
+          - ca-certificates
+          - curl
+          - gnupg-agent
+          - software-properties-common
+          - docker-ce
+          - docker-ce-cli
+          - containerd.io
+
+        # Enable ipv4 forwarding, required on CIS hardened machines
+        write_files:
+          - path: /etc/sysctl.d/enabled_ipv4_forwarding.conf
+            content: |
+              net.ipv4.conf.all.forwarding=1
+
+        # create the docker group
+        groups:
+          - docker
+
+        # Add default auto created user to docker group
+        system_info:
+          default_user:
+            groups: [docker]
+        {%- endif %}
         users:
           - name: skypilot:ssh_user
             shell: /bin/bash
@@ -77,3 +140,4 @@ setup_commands:
     sudo grep -e '^DefaultTasksMax' /etc/systemd/system.conf || (sudo bash -c 'echo "DefaultTasksMax=infinity" >> /etc/systemd/system.conf'); sudo systemctl set-property user-$(id -u $(whoami)).slice TasksMax=infinity; sudo systemctl daemon-reload;
     mkdir -p ~/.ssh; (grep -Pzo -q "Host \*\n  StrictHostKeyChecking no" ~/.ssh/config) || printf "Host *\n  StrictHostKeyChecking no\n" >> ~/.ssh/config;
     [ -f /etc/fuse.conf ] && sudo sed -i 's/#user_allow_other/user_allow_other/g' /etc/fuse.conf || (sudo sh -c 'echo "user_allow_other" > /etc/fuse.conf');
+    {{ ssh_max_sessions_config }}

sky/templates/oci-ray.yml.j2

@@ -91,7 +91,7 @@ setup_commands:
     mkdir -p ~/.ssh; (grep -Pzo -q "Host \*\n  StrictHostKeyChecking no\n  IdentityFile ~/.ssh/sky-cluster-key\n  IdentityFile ~/.ssh/id_rsa" ~/.ssh/config) || printf "Host *\n  StrictHostKeyChecking no\n  IdentityFile ~/.ssh/sky-cluster-key\n  IdentityFile ~/.ssh/id_rsa\n" >> ~/.ssh/config;
     [ -f /etc/fuse.conf ] && sudo sed -i 's/#user_allow_other/user_allow_other/g' /etc/fuse.conf || (sudo sh -c 'echo "user_allow_other" > /etc/fuse.conf');
     sudo iptables -I INPUT -i ens3 -m state --state ESTABLISHED,RELATED,NEW -j ACCEPT;
+    {{ ssh_max_sessions_config }}
 
 # Command to start ray clusters are now placed in `sky.provision.instance_setup`.
 # We do not need to list it here anymore.
-

sky/templates/paperspace-ray.yml.j2

@@ -91,3 +91,4 @@ setup_commands:
     sudo grep -e '^DefaultTasksMax' /etc/systemd/system.conf || (sudo bash -c 'echo "DefaultTasksMax=infinity" >> /etc/systemd/system.conf'); sudo systemctl set-property user-$(id -u $(whoami)).slice TasksMax=infinity; sudo systemctl daemon-reload;
     mkdir -p ~/.ssh; (grep -Pzo -q "Host \*\n  StrictHostKeyChecking no\n  IdentityFile ~/.ssh/sky-cluster-key\n  IdentityFile ~/.ssh/id_rsa" ~/.ssh/config) || printf "Host *\n  StrictHostKeyChecking no\n  IdentityFile ~/.ssh/sky-cluster-key\n  IdentityFile ~/.ssh/id_rsa\n" >> ~/.ssh/config;
     [ -f /etc/fuse.conf ] && sudo sed -i 's/#user_allow_other/user_allow_other/g' /etc/fuse.conf || (sudo sh -c 'echo "user_allow_other" > /etc/fuse.conf');
+    {{ ssh_max_sessions_config }}

sky/templates/runpod-ray.yml.j2

@@ -90,6 +90,7 @@ setup_commands:
     sudo grep -e '^DefaultTasksMax' /etc/systemd/system.conf || (sudo bash -c 'echo "DefaultTasksMax=infinity" >> /etc/systemd/system.conf'); sudo systemctl set-property user-$(id -u $(whoami)).slice TasksMax=infinity; sudo systemctl daemon-reload;
     mkdir -p ~/.ssh; (grep -Pzo -q "Host \*\n  StrictHostKeyChecking no\n  IdentityFile ~/.ssh/sky-cluster-key\n  IdentityFile ~/.ssh/id_rsa" ~/.ssh/config) || printf "Host *\n  StrictHostKeyChecking no\n  IdentityFile ~/.ssh/sky-cluster-key\n  IdentityFile ~/.ssh/id_rsa\n" >> ~/.ssh/config;
     [ -f /etc/fuse.conf ] && sudo sed -i 's/#user_allow_other/user_allow_other/g' /etc/fuse.conf || (sudo sh -c 'echo "user_allow_other" > /etc/fuse.conf');
+    {{ ssh_max_sessions_config }}
 
 # Command to start ray clusters are now placed in `sky.provision.instance_setup`.
 # We do not need to list it here anymore.

sky/templates/scp-ray.yml.j2

@@ -77,6 +77,7 @@ setup_commands:
     sudo grep -e '^DefaultTasksMax' /etc/systemd/system.conf || (sudo bash -c 'echo "DefaultTasksMax=infinity" >> /etc/systemd/system.conf'); sudo systemctl set-property user-$(id -u $(whoami)).slice TasksMax=infinity; sudo systemctl daemon-reload;
     mkdir -p ~/.ssh; (grep -Pzo -q "Host \*\n  StrictHostKeyChecking no\n  IdentityFile ~/.ssh/sky-cluster-key\n  IdentityFile ~/.ssh/id_rsa" ~/.ssh/config) || printf "Host *\n  StrictHostKeyChecking no\n  IdentityFile ~/.ssh/sky-cluster-key\n  IdentityFile ~/.ssh/id_rsa\n" >> ~/.ssh/config;
     [ -f /etc/fuse.conf ] && sudo sed -i 's/#user_allow_other/user_allow_other/g' /etc/fuse.conf || (sudo sh -c 'echo "user_allow_other" > /etc/fuse.conf'); # This is needed for `-o allow_other` option for `goofys`;
+    {{ ssh_max_sessions_config }}
 
 # Command to start ray on the head node. You don't need to change this.
 # NOTE: these are very performance-sensitive. Each new item opens/closes an SSH

sky/templates/vast-ray.yml.j2

@@ -64,7 +64,7 @@ setup_commands:
     sudo bash -c 'rm -rf /etc/security/limits.d; echo "* soft nofile 1048576" >> /etc/security/limits.conf; echo "* hard nofile 1048576" >> /etc/security/limits.conf';
     sudo grep -e '^DefaultTasksMax' /etc/systemd/system.conf || (sudo bash -c 'echo "DefaultTasksMax=infinity" >> /etc/systemd/system.conf'); sudo systemctl set-property user-$(id -u $(whoami)).slice TasksMax=infinity; sudo systemctl daemon-reload;
     (grep -Pzo -q "Host \*\n  StrictHostKeyChecking no" ~/.ssh/config) || printf "Host *\n  StrictHostKeyChecking no\n" >> ~/.ssh/config;
-   
+    {{ ssh_max_sessions_config }}
 
 # Command to start ray clusters are now placed in `sky.provision.instance_setup`.
 # We do not need to list it here anymore.

sky/templates/vsphere-ray.yml.j2

@@ -71,3 +71,4 @@ setup_commands:
     sudo grep -e '^DefaultTasksMax' /etc/systemd/system.conf || (sudo bash -c 'echo "DefaultTasksMax=infinity" >> /etc/systemd/system.conf'); sudo systemctl set-property user-$(id -u $(whoami)).slice TasksMax=infinity; sudo systemctl daemon-reload;
     mkdir -p ~/.ssh; (grep -Pzo -q "Host \*\n  StrictHostKeyChecking no\n  IdentityFile ~/.ssh/sky-cluster-key\n  IdentityFile ~/.ssh/id_rsa" ~/.ssh/config) || printf "Host *\n  StrictHostKeyChecking no\n  IdentityFile ~/.ssh/sky-cluster-key\n  IdentityFile ~/.ssh/id_rsa\n" >> ~/.ssh/config;
     [ -f /etc/fuse.conf ] && sudo sed -i 's/#user_allow_other/user_allow_other/g' /etc/fuse.conf || (sudo sh -c 'echo "user_allow_other" > /etc/fuse.conf');
+    {{ ssh_max_sessions_config }}

sky/utils/aws/get_default_security_group.py

@@ -0,0 +1,11 @@
+"""Script to get the default security group"""
+from sky.clouds import aws
+
+
+def main():
+    default_security_group = aws.DEFAULT_SECURITY_GROUP_NAME
+    print(f'{default_security_group}')
+
+
+if __name__ == '__main__':
+    main()

{skypilot_nightly-1.0.0.dev20250428.dist-info → skypilot_nightly-1.0.0.dev20250430.dist-info}/METADATA

@@ -1,7 +1,7 @@
 Metadata-Version: 2.4
 Name: skypilot-nightly
-Version: 1.0.0.dev20250428
-Summary: SkyPilot: An intercloud broker for the clouds
+Version: 1.0.0.dev20250430
+Summary: SkyPilot: Run AI on Any Infra — Unified, Faster, Cheaper.
 Author: SkyPilot Team
 License: Apache 2.0
 Project-URL: Homepage, https://github.com/skypilot-org/skypilot
@@ -47,7 +47,6 @@ Requires-Dist: python-multipart
 Requires-Dist: aiofiles
 Requires-Dist: httpx
 Requires-Dist: setproctitle
-Requires-Dist: omegaconf<2.5,>=2.4.0dev3
 Provides-Extra: aws
 Requires-Dist: urllib3<2; extra == "aws"
 Requires-Dist: awscli>=1.27.10; extra == "aws"
@@ -204,6 +203,7 @@ Dynamic: summary
 
 ----
 :fire: *News* :fire:
+- [Apr 2025] Spin up **Qwen3** on your cluster/cloud: [**example**](./llm/qwen/)
 - [Mar 2025] Run and serve **Google Gemma 3** using SkyPilot [**example**](./llm/gemma3/)
 - [Feb 2025] Prepare and serve **Retrieval Augmented Generation (RAG) with DeepSeek-R1**: [**blog post**](https://blog.skypilot.co/deepseek-rag), [**example**](./llm/rag/)
 - [Feb 2025] Run and serve **DeepSeek-R1 671B** using SkyPilot and SGLang with high throughput: [**example**](./llm/deepseek-r1/)