dstack 0.18.39__tar.gz → 0.18.40rc1__tar.gz

{dstack-0.18.39 → dstack-0.18.40rc1}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: dstack
-Version: 0.18.39
+Version: 0.18.40rc1
 Summary: dstack is an open-source orchestration engine for running AI workloads on any cloud or on-premises.
 Home-page: https://dstack.ai
 Author: Andrey Cheptsov
@@ -48,17 +48,16 @@ for AI workloads both in the cloud and on-prem, speeding up the development, tra
 
 #### Accelerators
 
-`dstack` supports `NVIDIA GPU`, `AMD GPU`, and `Google Cloud TPU` out of the box.
+`dstack` supports `NVIDIA`, `AMD`, `Google TPU`, and `Intel Gaudi` accelerators out of the box.
 
 ## Major news ✨
 
-- [2025/01] [dstack 0.18.35: Vultr backend](https://github.com/dstackai/dstack/releases/tag/0.18.35)
-- [2024/12] [dstack 0.18.33: TPU v6e support](https://github.com/dstackai/dstack/releases/tag/0.18.33)
+- [2025/01] [dstack 0.18.38: Intel Gaudi](https://github.com/dstackai/dstack/releases/tag/0.18.38)
+- [2025/01] [dstack 0.18.35: Vultr](https://github.com/dstackai/dstack/releases/tag/0.18.35)
+- [2024/12] [dstack 0.18.32: TPU v6e](https://github.com/dstackai/dstack/releases/tag/0.18.32)
 - [2024/12] [dstack 0.18.30: AWS Capacity Reservations and Capacity Blocks](https://github.com/dstackai/dstack/releases/tag/0.18.30)
-- [2024/11] [dstack 0.18.23: Gateway is optional](https://github.com/dstackai/dstack/releases/tag/0.18.23)
 - [2024/10] [dstack 0.18.21: Instance volumes](https://github.com/dstackai/dstack/releases/tag/0.18.21)
-- [2024/10] [dstack 0.18.18: Hardware metrics](https://github.com/dstackai/dstack/releases/tag/0.18.18)
-- [2024/10] [dstack 0.18.17: AMD support with SSH fleets, AWS EFA](https://github.com/dstackai/dstack/releases/tag/0.18.17)
+- [2024/10] [dstack 0.18.18: Hardware metrics monitoring](https://github.com/dstackai/dstack/releases/tag/0.18.18)
 
 ## Installation
 

{dstack-0.18.39 → dstack-0.18.40rc1}/README.md

@@ -21,17 +21,16 @@ for AI workloads both in the cloud and on-prem, speeding up the development, tra
 
 #### Accelerators
 
-`dstack` supports `NVIDIA GPU`, `AMD GPU`, and `Google Cloud TPU` out of the box.
+`dstack` supports `NVIDIA`, `AMD`, `Google TPU`, and `Intel Gaudi` accelerators out of the box.
 
 ## Major news ✨
 
-- [2025/01] [dstack 0.18.35: Vultr backend](https://github.com/dstackai/dstack/releases/tag/0.18.35)
-- [2024/12] [dstack 0.18.33: TPU v6e support](https://github.com/dstackai/dstack/releases/tag/0.18.33)
+- [2025/01] [dstack 0.18.38: Intel Gaudi](https://github.com/dstackai/dstack/releases/tag/0.18.38)
+- [2025/01] [dstack 0.18.35: Vultr](https://github.com/dstackai/dstack/releases/tag/0.18.35)
+- [2024/12] [dstack 0.18.32: TPU v6e](https://github.com/dstackai/dstack/releases/tag/0.18.32)
 - [2024/12] [dstack 0.18.30: AWS Capacity Reservations and Capacity Blocks](https://github.com/dstackai/dstack/releases/tag/0.18.30)
-- [2024/11] [dstack 0.18.23: Gateway is optional](https://github.com/dstackai/dstack/releases/tag/0.18.23)
 - [2024/10] [dstack 0.18.21: Instance volumes](https://github.com/dstackai/dstack/releases/tag/0.18.21)
-- [2024/10] [dstack 0.18.18: Hardware metrics](https://github.com/dstackai/dstack/releases/tag/0.18.18)
-- [2024/10] [dstack 0.18.17: AMD support with SSH fleets, AWS EFA](https://github.com/dstackai/dstack/releases/tag/0.18.17)
+- [2024/10] [dstack 0.18.18: Hardware metrics monitoring](https://github.com/dstackai/dstack/releases/tag/0.18.18)
 
 ## Installation
 

{dstack-0.18.39 → dstack-0.18.40rc1}/src/dstack/_internal/cli/services/configurators/fleet.py

@@ -353,7 +353,7 @@ def _print_plan_header(plan: FleetPlan):
                 f"${plan.max_offer_price:g} max[/]"
             )
         console.print()
-    else:
+    elif fleet_type == "cloud":
         console.print(NO_OFFERS_WARNING)
 
 

{dstack-0.18.39 → dstack-0.18.40rc1}/src/dstack/_internal/cli/services/configurators/gateway.py

@@ -22,6 +22,7 @@ from dstack._internal.core.models.gateways import (
     GatewayStatus,
 )
 from dstack._internal.core.models.repos.base import Repo
+from dstack._internal.core.services.diff import diff_models
 from dstack._internal.utils.common import local_time
 from dstack.api._public import Client
 
@@ -56,7 +57,15 @@ class GatewayConfigurator(BaseApplyConfigurator):
             confirm_message += "Create the gateway?"
         else:
             action_message += f"Found gateway [code]{plan.spec.configuration.name}[/]."
-            if plan.current_resource.configuration == plan.spec.configuration:
+            diff = diff_models(
+                plan.spec.configuration,
+                plan.current_resource.configuration,
+            )
+            changed_fields = list(diff.keys())
+            if (
+                plan.current_resource.configuration == plan.spec.configuration
+                or changed_fields == ["default"]
+            ):
                 if command_args.yes and not command_args.force:
                     # --force is required only with --yes,
                     # otherwise we may ask for force apply interactively.

{dstack-0.18.39 → dstack-0.18.40rc1}/src/dstack/_internal/cli/utils/common.py

@@ -25,8 +25,7 @@ LIVE_TABLE_REFRESH_RATE_PER_SEC = 1
 LIVE_TABLE_PROVISION_INTERVAL_SECS = 2
 NO_OFFERS_WARNING = (
     "[warning]"
-    "Could not find instance offers that match your requirements."
-    " Please check the requirements table above or visit the troubleshooting guide:"
+    "No matching instance offers available. Possible reasons:"
     " https://dstack.ai/docs/guides/troubleshooting/#no-offers"
     "[/]\n"
 )

{dstack-0.18.39 → dstack-0.18.40rc1}/src/dstack/_internal/core/models/configurations.py

@@ -21,6 +21,7 @@ from dstack._internal.core.models.volumes import MountPoint, VolumeConfiguration
 CommandsList = List[str]
 ValidPort = conint(gt=0, le=65536)
 SERVICE_HTTPS_DEFAULT = True
+STRIP_PREFIX_DEFAULT = True
 
 
 class RunConfigurationType(str, Enum):
@@ -130,6 +131,16 @@ class BaseRunConfiguration(CoreModel):
             description="Use image with NVIDIA CUDA Compiler (NVCC) included. Mutually exclusive with `image`"
         ),
     ]
+    single_branch: Annotated[
+        Optional[bool],
+        Field(
+            description=(
+                "Whether to clone and track only the current branch or all remote branches."
+                " Relevant only when using remote Git repos."
+                " Defaults to `false` for dev environments and to `true` for tasks and services"
+            )
+        ),
+    ] = None
     env: Annotated[
         Env,
         Field(description="The mapping or the list of environment variables"),
@@ -236,6 +247,16 @@ class ServiceConfigurationParams(CoreModel):
             ),
         ),
     ] = None
+    strip_prefix: Annotated[
+        bool,
+        Field(
+            description=(
+                "Strip the `/proxy/services/<project name>/<run name>/` path prefix"
+                " when forwarding requests to the service. Only takes effect"
+                " when running the service without a gateway"
+            )
+        ),
+    ] = STRIP_PREFIX_DEFAULT
     model: Annotated[
         Optional[Union[AnyModel, str]],
         Field(

{dstack-0.18.39 → dstack-0.18.40rc1}/src/dstack/_internal/core/models/runs.py

@@ -184,6 +184,7 @@ class JobSpec(CoreModel):
     home_dir: Optional[str]
     image_name: str
     privileged: bool = False
+    single_branch: Optional[bool] = None
     max_duration: Optional[int]
     stop_duration: Optional[int] = None
     registry_auth: Optional[RegistryAuth]

{dstack-0.18.39 → dstack-0.18.40rc1}/src/dstack/_internal/core/models/volumes.py

@@ -136,6 +136,15 @@ class VolumeMountPoint(CoreModel):
 class InstanceMountPoint(CoreModel):
     instance_path: Annotated[str, Field(description="The absolute path on the instance (host)")]
     path: Annotated[str, Field(description="The absolute path in the container")]
+    optional: Annotated[
+        bool,
+        Field(
+            description=(
+                "Allow running without this volume"
+                " in backends that do not support instance volumes"
+            ),
+        ),
+    ] = False
 
     _validate_instance_path = validator("instance_path", allow_reuse=True)(
         _validate_mount_point_path

{dstack-0.18.39 → dstack-0.18.40rc1}/src/dstack/_internal/core/services/logs.py

@@ -42,11 +42,14 @@ class URLReplacer:
         qs = {k: v[0] for k, v in urllib.parse.parse_qs(url.query).items()}
         if app_spec and app_spec.url_query_params is not None:
             qs.update({k.encode(): v.encode() for k, v in app_spec.url_query_params.items()})
+        path = url.path
+        if not path.startswith(self.path_prefix.removesuffix(b"/")):
+            path = concat_url_path(self.path_prefix, path)
 
         url = url._replace(
             scheme=("https" if self.secure else "http").encode(),
             netloc=(self.hostname if omit_port else f"{self.hostname}:{local_port}").encode(),
-            path=concat_url_path(self.path_prefix, url.path),
+            path=path,
             query=urllib.parse.urlencode(qs).encode(),
         )
         return url.geturl()

{dstack-0.18.39 → dstack-0.18.40rc1}/src/dstack/_internal/core/services/ssh/attach.py

@@ -56,6 +56,7 @@ class SSHAttach:
         ssh_port: int,
         container_ssh_port: int,
         user: str,
+        container_user: str,
         id_rsa_path: PathLike,
         ports_lock: PortsLock,
         run_name: str,
@@ -74,7 +75,7 @@ class SSHAttach:
         self.control_sock_path = FilePath(control_sock_path)
         self.identity_file = FilePath(id_rsa_path)
         self.tunnel = SSHTunnel(
-            destination=run_name,
+            destination=f"root@{run_name}",
             identity=self.identity_file,
             forwarded_sockets=ports_to_forwarded_sockets(
                 ports=self.ports,
@@ -91,7 +92,7 @@ class SSHAttach:
             self.host_config = {
                 "HostName": hostname,
                 "Port": ssh_port,
-                "User": user,
+                "User": user if dockerized else container_user,
                 "IdentityFile": self.identity_file,
                 "IdentitiesOnly": "yes",
                 "StrictHostKeyChecking": "no",
@@ -111,7 +112,7 @@ class SSHAttach:
             self.container_config = {
                 "HostName": "localhost",
                 "Port": container_ssh_port,
-                "User": "root",  # TODO(#1535): support non-root images properly
+                "User": container_user,
                 "IdentityFile": self.identity_file,
                 "IdentitiesOnly": "yes",
                 "StrictHostKeyChecking": "no",
@@ -122,7 +123,7 @@ class SSHAttach:
             self.container_config = {
                 "HostName": hostname,
                 "Port": ssh_port,
-                "User": user,
+                "User": container_user,
                 "IdentityFile": self.identity_file,
                 "IdentitiesOnly": "yes",
                 "StrictHostKeyChecking": "no",
@@ -136,7 +137,7 @@ class SSHAttach:
             self.host_config = {
                 "HostName": hostname,
                 "Port": container_ssh_port,
-                "User": "root",  # TODO(#1535): support non-root images properly
+                "User": container_user,
                 "IdentityFile": self.identity_file,
                 "IdentitiesOnly": "yes",
                 "StrictHostKeyChecking": "no",

{dstack-0.18.39 → dstack-0.18.40rc1}/src/dstack/_internal/proxy/lib/models.py

@@ -32,6 +32,7 @@ class Service(ImmutableModel):
     https: Optional[bool]  # only used on gateways
     auth: bool
     client_max_body_size: int  # only enforced on gateways
+    strip_prefix: bool = True  # only used in-server
     replicas: tuple[Replica, ...]
 
     @property

{dstack-0.18.39 → dstack-0.18.40rc1}/src/dstack/_internal/proxy/lib/testing/common.py

@@ -29,6 +29,7 @@ def make_service(
     domain: Optional[str] = None,
     https: Optional[bool] = None,
     auth: bool = False,
+    strip_prefix: bool = True,
 ) -> Service:
     return Service(
         project_name=project_name,
@@ -37,6 +38,7 @@ def make_service(
         https=https,
         auth=auth,
         client_max_body_size=2**20,
+        strip_prefix=strip_prefix,
         replicas=(
             Replica(
                 id="xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx",

{dstack-0.18.39 → dstack-0.18.40rc1}/src/dstack/_internal/server/app.py

@@ -7,7 +7,7 @@ from pathlib import Path
 from typing import Awaitable, Callable, List
 
 import sentry_sdk
-from fastapi import FastAPI, Request, status
+from fastapi import FastAPI, Request, Response, status
 from fastapi.datastructures import URL
 from fastapi.responses import HTMLResponse, JSONResponse, RedirectResponse
 from fastapi.staticfiles import StaticFiles
@@ -215,10 +215,14 @@ def register_routes(app: FastAPI, ui: bool = True):
     @app.middleware("http")
     async def log_request(request: Request, call_next):
         start_time = time.time()
-        response = await call_next(request)
+        response: Response = await call_next(request)
         process_time = time.time() - start_time
         logger.debug(
-            "Processed request %s %s in %s", request.method, request.url, f"{process_time:0.6f}s"
+            "Processed request %s %s in %s. Status: %s",
+            request.method,
+            request.url,
+            f"{process_time:0.6f}s",
+            response.status_code,
         )
         return response
 

{dstack-0.18.39 → dstack-0.18.40rc1}/src/dstack/_internal/server/background/tasks/process_submitted_jobs.py

@@ -64,7 +64,7 @@ from dstack._internal.server.services.pools import (
 )
 from dstack._internal.server.services.runs import (
     check_can_attach_run_volumes,
-    check_run_spec_has_instance_mounts,
+    check_run_spec_requires_instance_mounts,
     get_offer_volumes,
     get_run_volume_models,
     get_run_volumes,
@@ -418,7 +418,7 @@ async def _run_job_on_new_instance(
         master_job_provisioning_data=master_job_provisioning_data,
         volumes=volumes,
         privileged=job.job_spec.privileged,
-        instance_mounts=check_run_spec_has_instance_mounts(run.run_spec),
+        instance_mounts=check_run_spec_requires_instance_mounts(run.run_spec),
     )
     # Limit number of offers tried to prevent long-running processing
     # in case all offers fail.

{dstack-0.18.39 → dstack-0.18.40rc1}/src/dstack/_internal/server/schemas/runner.py

@@ -61,6 +61,7 @@ class SubmitBody(CoreModel):
                 "entrypoint",
                 "env",
                 "gateway",
+                "single_branch",
                 "max_duration",
                 "working_dir",
             }

{dstack-0.18.39 → dstack-0.18.40rc1}/src/dstack/_internal/server/services/fleets.py

@@ -239,7 +239,6 @@ async def get_plan(
     user: UserModel,
     spec: FleetSpec,
 ) -> FleetPlan:
-    # TODO: refactor offers logic into a separate module to avoid depending on runs
     current_fleet: Optional[Fleet] = None
     current_fleet_id: Optional[uuid.UUID] = None
     if spec.configuration.name is not None:
@@ -259,6 +258,7 @@ async def get_plan(
             requirements=_get_fleet_requirements(spec),
         )
         offers = [offer for _, offer in offers_with_backends]
+    _remove_fleet_spec_sensitive_info(spec)
     plan = FleetPlan(
         project_name=project.name,
         user=user.name,

{dstack-0.18.39 → dstack-0.18.40rc1}/src/dstack/_internal/server/services/jobs/configurators/base.py

@@ -63,6 +63,10 @@ class JobConfigurator(ABC):
     def _shell_commands(self) -> List[str]:
         pass
 
+    @abstractmethod
+    def _default_single_branch(self) -> bool:
+        pass
+
     @abstractmethod
     def _default_max_duration(self) -> Optional[int]:
         pass
@@ -104,6 +108,7 @@ class JobConfigurator(ABC):
             image_name=self._image_name(),
             user=await self._user(),
             privileged=self._privileged(),
+            single_branch=self._single_branch(),
             max_duration=self._max_duration(),
             stop_duration=self._stop_duration(),
             registry_auth=self._registry_auth(),
@@ -172,6 +177,11 @@ class JobConfigurator(ABC):
     def _privileged(self) -> bool:
         return self.run_spec.configuration.privileged
 
+    def _single_branch(self) -> bool:
+        if self.run_spec.configuration.single_branch is None:
+            return self._default_single_branch()
+        return self.run_spec.configuration.single_branch
+
     def _max_duration(self) -> Optional[int]:
         if self.run_spec.merged_profile.max_duration in [None, True]:
             return self._default_max_duration()

{dstack-0.18.39 → dstack-0.18.40rc1}/src/dstack/_internal/server/services/jobs/configurators/dev.py

@@ -40,6 +40,9 @@ class DevEnvironmentJobConfigurator(JobConfigurator):
         commands += ["tail -f /dev/null"]  # idle
         return commands
 
+    def _default_single_branch(self) -> bool:
+        return False
+
     def _default_max_duration(self) -> Optional[int]:
         return DEFAULT_MAX_DURATION_SECONDS
 

{dstack-0.18.39 → dstack-0.18.40rc1}/src/dstack/_internal/server/services/jobs/configurators/service.py

@@ -11,6 +11,9 @@ class ServiceJobConfigurator(JobConfigurator):
     def _shell_commands(self) -> List[str]:
         return self.run_spec.configuration.commands
 
+    def _default_single_branch(self) -> bool:
+        return True
+
     def _default_max_duration(self) -> Optional[int]:
         return None
 

{dstack-0.18.39 → dstack-0.18.40rc1}/src/dstack/_internal/server/services/jobs/configurators/task.py

@@ -25,6 +25,9 @@ class TaskJobConfigurator(JobConfigurator):
     def _shell_commands(self) -> List[str]:
         return self.run_spec.configuration.commands
 
+    def _default_single_branch(self) -> bool:
+        return True
+
     def _default_max_duration(self) -> Optional[int]:
         return DEFAULT_MAX_DURATION_SECONDS
 

{dstack-0.18.39 → dstack-0.18.40rc1}/src/dstack/_internal/server/services/proxy/repo.py

@@ -98,6 +98,7 @@ class ServerProxyRepo(BaseProxyRepo):
             https=None,
             auth=run_spec.configuration.auth,
             client_max_body_size=DEFAULT_SERVICE_CLIENT_MAX_BODY_SIZE,
+            strip_prefix=run_spec.configuration.strip_prefix,
             replicas=tuple(replicas),
         )
 

{dstack-0.18.39 → dstack-0.18.40rc1}/src/dstack/_internal/server/services/proxy/services/service_proxy.py

@@ -12,6 +12,7 @@ from dstack._internal.proxy.lib.services.service_connection import (
     ServiceConnectionPool,
     get_service_replica_client,
 )
+from dstack._internal.utils.common import concat_url_path
 from dstack._internal.utils.logging import get_logger
 
 logger = get_logger(__name__)
@@ -37,6 +38,9 @@ async def proxy(
 
     client = await get_service_replica_client(service, repo, service_conn_pool)
 
+    if not service.strip_prefix:
+        path = concat_url_path(request.scope.get("root_path", "/"), request.url.path)
+
     try:
         upstream_request = await build_upstream_request(request, path, client)
     except ClientDisconnect:

{dstack-0.18.39 → dstack-0.18.40rc1}/src/dstack/_internal/server/services/runs.py

@@ -330,7 +330,7 @@ async def get_plan(
             multinode=jobs[0].job_spec.jobs_per_replica > 1,
             volumes=volumes,
             privileged=jobs[0].job_spec.privileged,
-            instance_mounts=check_run_spec_has_instance_mounts(run_spec),
+            instance_mounts=check_run_spec_requires_instance_mounts(run_spec),
         )
 
     job_plans = []
@@ -897,9 +897,10 @@ def get_offer_mount_point_volume(
     raise ServerClientError("Failed to find an eligible volume for the mount point")
 
 
-def check_run_spec_has_instance_mounts(run_spec: RunSpec) -> bool:
+def check_run_spec_requires_instance_mounts(run_spec: RunSpec) -> bool:
     return any(
-        is_core_model_instance(mp, InstanceMountPoint) for mp in run_spec.configuration.volumes
+        is_core_model_instance(mp, InstanceMountPoint) and not mp.optional
+        for mp in run_spec.configuration.volumes
     )
 
 
@@ -967,7 +968,7 @@ def _validate_run_spec_and_set_defaults(run_spec: RunSpec):
 _UPDATABLE_SPEC_FIELDS = ["repo_code_hash", "configuration"]
 # Most service fields can be updated via replica redeployment.
 # TODO: Allow updating other fields when a rolling deployment is supported.
-_UPDATABLE_CONFIGURATION_FIELDS = ["replicas", "scaling"]
+_UPDATABLE_CONFIGURATION_FIELDS = ["replicas", "scaling", "strip_prefix"]
 
 
 def _can_update_run_spec(current_run_spec: RunSpec, new_run_spec: RunSpec) -> bool:

{dstack-0.18.39 → dstack-0.18.40rc1}/src/dstack/api/_public/runs.py

@@ -324,11 +324,19 @@ class Run(ABC):
             if runtime_data is not None and runtime_data.ports is not None:
                 container_ssh_port = runtime_data.ports.get(container_ssh_port, container_ssh_port)
 
+            # TODO: get login name from runner in case it's not specified in the run configuration
+            # (i.e. the default image user is used, and it is not root)
+            if job.job_spec.user is not None and job.job_spec.user.username is not None:
+                container_user = job.job_spec.user.username
+            else:
+                container_user = "root"
+
             self._ssh_attach = SSHAttach(
                 hostname=provisioning_data.hostname,
                 ssh_port=provisioning_data.ssh_port,
                 container_ssh_port=container_ssh_port,
                 user=provisioning_data.username,
+                container_user=container_user,
                 id_rsa_path=ssh_identity_file,
                 ports_lock=self._ports_lock,
                 run_name=name,

{dstack-0.18.39 → dstack-0.18.40rc1}/src/dstack/api/server/__init__.py

@@ -133,9 +133,16 @@ class APIClient:
         else:
             raise ClientError(f"Failed to connect to dstack server {self._base_url}")
 
+        if 400 <= resp.status_code < 600:
+            logger.debug(
+                "Error requesting %s. Status: %s. Headers: %s. Body: %s",
+                resp.request.url,
+                resp.status_code,
+                resp.headers,
+                resp.content,
+            )
+
         if raise_for_status:
-            if resp.status_code == 500:
-                raise ClientError("Unexpected dstack server error")
             if resp.status_code == 400:  # raise ServerClientError
                 detail: List[Dict] = resp.json()["detail"]
                 if len(detail) == 1 and detail[0]["code"] in _server_client_errors:
@@ -145,7 +152,16 @@ class APIClient:
             if resp.status_code == 422:
                 formatted_error = pprint.pformat(resp.json())
                 raise ClientError(f"Server validation error: \n{formatted_error}")
-            resp.raise_for_status()
+            if resp.status_code == 403:
+                raise ClientError(
+                    f"Access to {resp.request.url} is denied. Please check your access token"
+                )
+            if 400 <= resp.status_code < 600:
+                raise ClientError(
+                    f"Unexpected error: status code {resp.status_code}"
+                    f" when requesting {resp.request.url}."
+                    " Check server logs or run with DSTACK_CLI_LOG_LEVEL=DEBUG to see more details"
+                )
         return resp
 
 

{dstack-0.18.39 → dstack-0.18.40rc1}/src/dstack/api/server/_runs.py

@@ -1,9 +1,14 @@
 from datetime import datetime
-from typing import List, Optional, Union
+from typing import Any, List, Optional, Union
 from uuid import UUID
 
 from pydantic import parse_obj_as
 
+from dstack._internal.core.models.common import is_core_model_instance
+from dstack._internal.core.models.configurations import (
+    STRIP_PREFIX_DEFAULT,
+    ServiceConfiguration,
+)
 from dstack._internal.core.models.pools import Instance
 from dstack._internal.core.models.profiles import Profile
 from dstack._internal.core.models.runs import (
@@ -14,6 +19,7 @@ from dstack._internal.core.models.runs import (
     RunPlan,
     RunSpec,
 )
+from dstack._internal.core.models.volumes import InstanceMountPoint
 from dstack._internal.server.schemas.runs import (
     ApplyRunPlanRequest,
     CreateInstanceRequest,
@@ -117,34 +123,48 @@ class RunsAPIClient(APIClientGroup):
 
 def _get_run_spec_excludes(run_spec: RunSpec) -> Optional[dict]:
     spec_excludes: dict[str, set[str]] = {}
-    configuration_excludes: set[str] = set()
+    configuration_excludes: dict[str, Any] = {}
     profile_excludes: set[str] = set()
     configuration = run_spec.configuration
     profile = run_spec.profile
 
     # client >= 0.18.18 / server <= 0.18.17 compatibility tweak
     if not configuration.privileged:
-        configuration_excludes.add("privileged")
+        configuration_excludes["privileged"] = True
     # client >= 0.18.23 / server <= 0.18.22 compatibility tweak
     if configuration.type == "service" and configuration.gateway is None:
-        configuration_excludes.add("gateway")
+        configuration_excludes["gateway"] = True
     # client >= 0.18.30 / server <= 0.18.29 compatibility tweak
     if run_spec.configuration.user is None:
-        configuration_excludes.add("user")
+        configuration_excludes["user"] = True
     # client >= 0.18.30 / server <= 0.18.29 compatibility tweak
     if configuration.reservation is None:
-        configuration_excludes.add("reservation")
+        configuration_excludes["reservation"] = True
     if profile is not None and profile.reservation is None:
         profile_excludes.add("reservation")
     if configuration.idle_duration is None:
-        configuration_excludes.add("idle_duration")
+        configuration_excludes["idle_duration"] = True
     if profile is not None and profile.idle_duration is None:
         profile_excludes.add("idle_duration")
     # client >= 0.18.38 / server <= 0.18.37 compatibility tweak
     if configuration.stop_duration is None:
-        configuration_excludes.add("stop_duration")
+        configuration_excludes["stop_duration"] = True
     if profile is not None and profile.stop_duration is None:
         profile_excludes.add("stop_duration")
+    # client >= 0.18.40 / server <= 0.18.39 compatibility tweak
+    if (
+        is_core_model_instance(configuration, ServiceConfiguration)
+        and configuration.strip_prefix == STRIP_PREFIX_DEFAULT
+    ):
+        configuration_excludes["strip_prefix"] = True
+    if configuration.single_branch is None:
+        configuration_excludes["single_branch"] = True
+    if all(
+        not is_core_model_instance(v, InstanceMountPoint) or not v.optional
+        for v in configuration.volumes
+    ):
+        configuration_excludes["volumes"] = {"__all__": {"optional"}}
+
     if configuration_excludes:
         spec_excludes["configuration"] = configuration_excludes
     if profile_excludes:

{dstack-0.18.39 → dstack-0.18.40rc1}/src/dstack/version.py

@@ -1,3 +1,3 @@
-__version__ = "0.18.39"
+__version__ = "0.18.40rc1"
 __is_release__ = True
 base_image = "0.6"

{dstack-0.18.39 → dstack-0.18.40rc1}/src/dstack.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: dstack
-Version: 0.18.39
+Version: 0.18.40rc1
 Summary: dstack is an open-source orchestration engine for running AI workloads on any cloud or on-premises.
 Home-page: https://dstack.ai
 Author: Andrey Cheptsov
@@ -48,17 +48,16 @@ for AI workloads both in the cloud and on-prem, speeding up the development, tra
 
 #### Accelerators
 
-`dstack` supports `NVIDIA GPU`, `AMD GPU`, and `Google Cloud TPU` out of the box.
+`dstack` supports `NVIDIA`, `AMD`, `Google TPU`, and `Intel Gaudi` accelerators out of the box.
 
 ## Major news ✨
 
-- [2025/01] [dstack 0.18.35: Vultr backend](https://github.com/dstackai/dstack/releases/tag/0.18.35)
-- [2024/12] [dstack 0.18.33: TPU v6e support](https://github.com/dstackai/dstack/releases/tag/0.18.33)
+- [2025/01] [dstack 0.18.38: Intel Gaudi](https://github.com/dstackai/dstack/releases/tag/0.18.38)
+- [2025/01] [dstack 0.18.35: Vultr](https://github.com/dstackai/dstack/releases/tag/0.18.35)
+- [2024/12] [dstack 0.18.32: TPU v6e](https://github.com/dstackai/dstack/releases/tag/0.18.32)
 - [2024/12] [dstack 0.18.30: AWS Capacity Reservations and Capacity Blocks](https://github.com/dstackai/dstack/releases/tag/0.18.30)
-- [2024/11] [dstack 0.18.23: Gateway is optional](https://github.com/dstackai/dstack/releases/tag/0.18.23)
 - [2024/10] [dstack 0.18.21: Instance volumes](https://github.com/dstackai/dstack/releases/tag/0.18.21)
-- [2024/10] [dstack 0.18.18: Hardware metrics](https://github.com/dstackai/dstack/releases/tag/0.18.18)
-- [2024/10] [dstack 0.18.17: AMD support with SSH fleets, AWS EFA](https://github.com/dstackai/dstack/releases/tag/0.18.17)
+- [2024/10] [dstack 0.18.18: Hardware metrics monitoring](https://github.com/dstackai/dstack/releases/tag/0.18.18)
 
 ## Installation