konduktor-nightly 0.1.0.dev20250804105449__py3-none-any.whl → 0.1.0.dev20250806105405__py3-none-any.whl

konduktor/task.py

@@ -23,6 +23,7 @@ import yaml
 
 if typing.TYPE_CHECKING:
     import konduktor.resource as resources_lib
+    import konduktor.serving as servings_lib
 
 import konduktor
 from konduktor import constants, logging
@@ -193,6 +194,7 @@ class Task:
         self.num_nodes = num_nodes  # type: ignore
         # Default to CPU VM
         self.resources: Optional[konduktor.Resources] = None
+        self.serving: Optional[konduktor.Serving] = None
 
         self.file_mounts: Optional[Dict[str, str]] = None
         self.best_resources = None  # (asaiacai): this is unused consider removing
@@ -386,6 +388,14 @@ class Task:
             )
             resources_config['_cluster_config_overrides'] = cluster_config_override
         task.set_resources(konduktor.Resources.from_yaml_config(resources_config))
+
+        # Parse serving field.
+        serving_config = config.pop('serving', None)
+        if serving_config is not None:
+            serving = konduktor.Serving.from_yaml_config(serving_config, task.run)
+            if serving is not None:
+                task.set_serving(serving)
+
         assert not config, f'Invalid task args: {config.keys()}'
         return task
 
@@ -526,6 +536,52 @@ class Task:
         self.set_resources(new_resources)
         return self
 
+    def set_serving(
+        self,
+        serving: 'servings_lib.Serving',
+    ) -> 'Task':
+        """Sets the serving configuration for this task.
+
+        Args:
+          serving: konduktor.Serving object
+
+        Returns:
+          self: The current task, with serving set.
+        """
+        if self._num_nodes and self._num_nodes != 1:
+            with ux_utils.print_exception_no_traceback():
+                raise ValueError(
+                    f'Only single node serving is supported (num_nodes: 1). '
+                    f'Got: {self.num_nodes}'
+                )
+
+        if serving.max_replicas < serving.min_replicas:
+            with ux_utils.print_exception_no_traceback():
+                raise ValueError(
+                    f'max_replicas ({serving.max_replicas}) cannot be '
+                    f'less than min_replicas ({serving.min_replicas})'
+                )
+
+        if isinstance(serving, konduktor.Serving):
+            serving = serving
+        self.serving = serving
+        return self
+
+    def set_serving_override(self, override_params: Dict[str, Any]) -> 'Task':
+        """Sets the override parameters for the serving config."""
+        assert self.serving is not None, 'Serving config is required'
+        new_serving = konduktor.Serving(
+            min_replicas=override_params.get('min_replicas', self.serving.min_replicas),
+            max_replicas=override_params.get('max_replicas', self.serving.max_replicas),
+            ports=override_params.get('ports', self.serving.ports),
+            probe=override_params.get('probe', self.serving.probe),
+        )
+
+        self.num_nodes = override_params.get('num_nodes', self.num_nodes)
+
+        self.set_serving(new_serving)
+        return self
+
     def set_file_mounts(self, file_mounts: Optional[Dict[str, str]]) -> 'Task':
         """Sets the file mounts for this task.
 
@@ -870,6 +926,11 @@ class Task:
         if self.file_mounts is not None:
             config['file_mounts'].update(self.file_mounts)
 
+        add_if_not_none('serving', {})
+
+        if self.serving is not None:
+            add_if_not_none('serving', self.serving.to_yaml_config())
+
         if self.storage_mounts is not None:
             config['file_mounts'].update(
                 {

konduktor/templates/deployment.yaml.j2

@@ -0,0 +1,142 @@
+apiVersion: apps/v1
+kind: Deployment
+metadata:
+  labels:
+    {% if not general %}
+    {{ model_name_label }}: {{ name }}
+    model.aibrix.ai/port: "{{ ports }}"
+    {% endif %}
+    {{ deployment_name_label }}: "{{ name }}"
+    {{ deployment_user_label }}: "{{ user }}"
+    {{ deployment_accelerator_label }}: "{{ accelerator_type }}"
+    {{ deployment_num_accelerators_label }}: "{{ num_accelerators }}"
+    trainy.ai/has-autoscaler: "{{ autoscaler }}"
+    trainy.ai/konduktor-managed: "true"
+  name: {{ name }}
+  namespace: default
+spec:
+  replicas: {{ min_replicas }}
+  selector:
+    matchLabels:
+      {% if not general %}
+      {{ model_name_label }}: {{ name }}
+      {% endif %}
+      {{ deployment_name_label }}: "{{ name }}"
+  template: {}
+
+---
+
+apiVersion: v1
+kind: Service
+metadata:
+  labels:
+    {% if not general %}
+    {{ model_name_label }}: {{ name }}
+    {% endif %}
+    prometheus-discovery: "true"
+    {{ deployment_name_label }}: "{{ name }}"
+    {{ deployment_user_label }}: "{{ user }}"
+    trainy.ai/has-autoscaler: "{{ autoscaler }}"
+  {% if not general %}
+  annotations:
+    prometheus.io/scrape: "true"
+    prometheus.io/port: "8080"
+  {% endif %}
+  name: {{ name }}
+  namespace: default
+spec:
+  ports:
+    - name: serve
+      port: {{ ports }}
+      protocol: TCP
+      targetPort: {{ ports }}
+    {% if not general %}
+    - name: http
+      port: 8080
+      protocol: TCP
+      targetPort: 8080
+    {% endif %}
+  selector:
+    {% if not general %}
+    {{ model_name_label }}: {{ name }}
+    {% endif %}
+    {{ deployment_name_label }}: "{{ name }}"
+  {% if general %}
+  type: LoadBalancer
+  {% else %}
+  type: ClusterIP
+  {% endif %}
+
+{% if not general %}
+---
+apiVersion: autoscaling.aibrix.ai/v1alpha1
+kind: PodAutoscaler
+metadata:
+  name: {{ name }}-apa
+  namespace: default
+  labels:
+    {{ model_name_label }}: {{ name }}
+    app.kubernetes.io/name: aibrix
+    app.kubernetes.io/managed-by: kustomize
+    {{ deployment_name_label }}: "{{ name }}"
+    {{ deployment_user_label }}: "{{ user }}"
+  annotations:
+    autoscaling.aibrix.ai/up-fluctuation-tolerance: '0.1'
+    autoscaling.aibrix.ai/down-fluctuation-tolerance: '0.2'
+    apa.autoscaling.aibrix.ai/window: 30s
+spec:
+  scalingStrategy: APA
+  minReplicas: {{ min_replicas }}
+  maxReplicas: {{ max_replicas }}
+  metricsSources:
+    - metricSourceType: pod
+      protocolType: http
+      port: "{{ ports }}"
+      path: metrics
+      targetMetric: gpu_cache_usage_perc
+      targetValue: '0.5'
+  scaleTargetRef:
+    apiVersion: apps/v1
+    kind: Deployment
+    name: {{ name }}
+{% endif %}
+
+{% if general %}
+---
+apiVersion: autoscaling/v2
+kind: HorizontalPodAutoscaler
+metadata:
+  name: {{ name }}-hpa
+  namespace: default
+  labels:
+    {{ deployment_name_label }}: "{{ name }}"
+    {{ deployment_user_label }}: "{{ user }}"
+    trainy.ai/has-autoscaler: "{{ autoscaler }}"
+spec:
+  scaleTargetRef:
+    apiVersion: apps/v1
+    kind: Deployment
+    name: {{ name }}
+  minReplicas: {{ min_replicas }}
+  maxReplicas: {{ max_replicas }}
+  metrics:
+    - type: Resource
+      resource:
+        name: cpu
+        target:
+          type: Utilization
+          averageUtilization: 50
+  behavior:
+    scaleDown:
+      stabilizationWindowSeconds: 60
+      policies:
+      - type: Percent
+        value: 100
+        periodSeconds: 15
+    scaleUp:
+      stabilizationWindowSeconds: 20
+      policies:
+      - type: Percent
+        value: 100
+        periodSeconds: 15
+{% endif %}

konduktor/templates/pod.yaml.j2

@@ -23,6 +23,42 @@ kubernetes:
             - name: ssh
               containerPort: {{ konduktor_ssh_port }}
           {% endif %}
+          {% if serving %}
+          ports:
+            - containerPort: {{ ports }}
+          # TODO (ryan): allow modification of thresholds and timings
+          {% if probe %}
+          livenessProbe:
+            httpGet:
+              path: {{ probe }}
+              port: {{ ports }}
+              scheme: HTTP
+            initialDelaySeconds: 60
+            failureThreshold: 3
+            periodSeconds: 10
+            successThreshold: 1
+            timeoutSeconds: 1
+          readinessProbe:
+            httpGet:
+              path: {{ probe }}
+              port: {{ ports }}
+              scheme: HTTP
+            initialDelaySeconds: 60
+            failureThreshold: 10
+            periodSeconds: 5
+            successThreshold: 1
+            timeoutSeconds: 1
+          startupProbe:
+            httpGet:
+              path: {{ probe }}
+              port: {{ ports }}
+              scheme: HTTP
+            failureThreshold: 60
+            periodSeconds: 30
+            successThreshold: 1
+            timeoutSeconds: 1
+          {% endif %}
+          {% endif %}
           image: {{ image_id }}
           # this is set during jobset definition since we need to know the jobset
           # name and number of nodes to set all the environment variables correctly here

konduktor/utils/accelerator_registry.py

@@ -1,6 +1,6 @@
 """Accelerator registry."""
 
-_ACCELERATORS = ['A100', 'A100-80GB', 'B200', 'H100', 'H200', 'L40S', 'T4']
+_ACCELERATORS = ['A100', 'A100-80GB', 'B200', 'H100', 'H200', 'L40S', 'T4', 'L40']
 
 
 def canonicalize_accelerator_name(accelerator: str) -> str:

konduktor/utils/log_utils.py

@@ -374,7 +374,7 @@ def tail_vicky_logs(
             logger.debug(f'Making request to {vicky_url} with query: {query}')
             with requests.post(
                 vicky_url, data=query, stream=True, timeout=timeout
-            ) as response:
+            ) as response:  # type: requests.Response
                 logger.debug(f'Response status: {response.status_code}')
                 if response.status_code != 200:
                     logger.error(

konduktor/utils/schemas.py

@@ -260,6 +260,10 @@ def get_task_schema():
             'service': {
                 'type': 'object',
             },
+            # serving config is validated separately using SERVING_SCHEMA
+            'serving': {
+                'type': 'object',
+            },
             'setup': {
                 'type': 'string',
             },
@@ -373,6 +377,44 @@ _REMOTE_IDENTITY_SCHEMA_KUBERNETES = {
 }
 
 
+def get_serving_schema():
+    return {
+        '$schema': 'https://json-schema.org/draft/2020-12/schema',
+        'type': 'object',
+        'anyOf': [
+            {'required': ['min_replicas']},
+            {'required': ['max_replicas']},
+        ],
+        'additionalProperties': False,
+        'properties': {
+            'min_replicas': {
+                'type': 'integer',
+                'minimum': 1,
+                'description': 'Minimum number of replicas for autoscaling.',
+            },
+            'max_replicas': {
+                'type': 'integer',
+                'minimum': 1,
+                'description': 'Maximum number of replicas for autoscaling.',
+            },
+            'ports': {
+                # this could easily be an integer, but I made it
+                # more vague on purpose so I can use a float to test
+                # the json schema validator later down the line
+                'type': 'number',
+                'minimum': 1,
+                'description': 'The containerPort and service port '
+                'used by the model server.',
+            },
+            'probe': {
+                'type': 'string',
+                'description': 'The livenessProbe, readinessProbe, and startupProbe '
+                'path used by the model server.',
+            },
+        },
+    }
+
+
 def get_storage_schema():
     # pylint: disable=import-outside-toplevel
     from konduktor.data import storage

konduktor/utils/validator.py

@@ -17,11 +17,18 @@ from filelock import FileLock
 from konduktor import logging
 
 SCHEMA_VERSION = 'v1.32.0-standalone-strict'
-SCHEMA_URL = f'https://raw.githubusercontent.com/yannh/kubernetes-json-schema/master/{SCHEMA_VERSION}/podspec.json'
-SCHEMA_CACHE_PATH = Path.home() / '.konduktor/schemas/podspec.json'
-SCHEMA_LOCK_PATH = SCHEMA_CACHE_PATH.with_suffix('.lock')
+SCHEMA_CACHE_PATH = Path.home() / '.konduktor/schemas'
+SCHEMA_LOCK_PATH = SCHEMA_CACHE_PATH / '.lock'
 CACHE_MAX_AGE_SECONDS = 86400  # 24 hours
 
+# Schema URLs for different Kubernetes resources
+SCHEMA_URLS = {
+    'podspec': f'https://raw.githubusercontent.com/yannh/kubernetes-json-schema/master/{SCHEMA_VERSION}/podspec.json',
+    'deployment': f'https://raw.githubusercontent.com/yannh/kubernetes-json-schema/master/{SCHEMA_VERSION}/deployment.json',
+    'service': f'https://raw.githubusercontent.com/yannh/kubernetes-json-schema/master/{SCHEMA_VERSION}/service.json',
+    'horizontalpodautoscaler': f'https://raw.githubusercontent.com/yannh/kubernetes-json-schema/master/{SCHEMA_VERSION}/horizontalpodautoscaler-autoscaling-v2.json',
+}
+
 logger = logging.get_logger(__name__)
 
 
@@ -37,33 +44,41 @@ SchemaValidator = jsonschema.validators.extend(
 )
 
 
-def get_cached_schema() -> dict:
+def get_cached_schema(schema_type: str) -> dict:
+    """Get cached schema for a specific Kubernetes resource type."""
+    schema_url = SCHEMA_URLS.get(schema_type)
+    if not schema_url:
+        raise ValueError(f'Unknown schema type: {schema_type}')
+
+    schema_file = SCHEMA_CACHE_PATH / f'{schema_type}.json'
     lock = FileLock(str(SCHEMA_LOCK_PATH))
+
     with lock:
         # Check if schema file exists and is fresh
-        if SCHEMA_CACHE_PATH.exists():
-            age = time.time() - SCHEMA_CACHE_PATH.stat().st_mtime
-            # if old
+        if schema_file.exists():
+            age = time.time() - schema_file.stat().st_mtime
+            # if fresh
             if age < CACHE_MAX_AGE_SECONDS:
-                with open(SCHEMA_CACHE_PATH, 'r') as f:
+                with open(schema_file, 'r') as f:
                     return json.load(f)
 
         # Download schema
-        resp = requests.get(SCHEMA_URL)
+        resp = requests.get(schema_url)
         resp.raise_for_status()
 
-        SCHEMA_CACHE_PATH.parent.mkdir(parents=True, exist_ok=True)
-        with open(SCHEMA_CACHE_PATH, 'w') as f:
+        SCHEMA_CACHE_PATH.mkdir(parents=True, exist_ok=True)
+        with open(schema_file, 'w') as f:
             f.write(resp.text)
 
         return resp.json()
 
 
-def validate_pod_spec(pod_spec: dict) -> None:
-    schema = get_cached_schema()
+def _validate_k8s_spec(spec: dict, schema_type: str, resource_name: str) -> None:
+    """Generic validation function for Kubernetes specs."""
+    schema = get_cached_schema(schema_type)
 
     validator = jsonschema.Draft7Validator(schema)
-    errors = sorted(validator.iter_errors(pod_spec), key=lambda e: e.path)
+    errors = sorted(validator.iter_errors(spec), key=lambda e: e.path)
 
     if not errors:
         return
@@ -75,7 +90,7 @@ def validate_pod_spec(pod_spec: dict) -> None:
     ]
 
     # Clean log
-    logger.debug('Invalid k8s pod spec/config:\n%s', '\n'.join(formatted))
+    logger.debug('Invalid k8s %s spec/config:\n%s', resource_name, '\n'.join(formatted))
 
     # Color only in CLI
     formatted_colored = [
@@ -86,6 +101,26 @@ def validate_pod_spec(pod_spec: dict) -> None:
     ]
 
     raise ValueError(
-        f'\n{Fore.RED}Invalid k8s pod spec/config: {Style.RESET_ALL}\n'
+        f'\n{Fore.RED}Invalid k8s {resource_name} spec/config: {Style.RESET_ALL}\n'
         + '\n'.join(formatted_colored)
     )
+
+
+def validate_pod_spec(pod_spec: dict) -> None:
+    """Validate a Kubernetes pod spec."""
+    _validate_k8s_spec(pod_spec, 'podspec', 'pod')
+
+
+def validate_deployment_spec(deployment_spec: dict) -> None:
+    """Validate a Kubernetes deployment spec."""
+    _validate_k8s_spec(deployment_spec, 'deployment', 'deployment')
+
+
+def validate_service_spec(service_spec: dict) -> None:
+    """Validate a Kubernetes service spec."""
+    _validate_k8s_spec(service_spec, 'service', 'service')
+
+
+def validate_horizontalpodautoscaler_spec(hpa_spec: dict) -> None:
+    """Validate a Kubernetes HorizontalPodAutoscaler spec."""
+    _validate_k8s_spec(hpa_spec, 'horizontalpodautoscaler', 'horizontalpodautoscaler')

{konduktor_nightly-0.1.0.dev20250804105449.dist-info → konduktor_nightly-0.1.0.dev20250806105405.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.3
 Name: konduktor-nightly
-Version: 0.1.0.dev20250804105449
+Version: 0.1.0.dev20250806105405
 Summary: GPU Cluster Health Management
 Author: Andrew Aikawa
 Author-email: asai@berkeley.edu

{konduktor_nightly-0.1.0.dev20250804105449.dist-info → konduktor_nightly-0.1.0.dev20250806105405.dist-info}/RECORD

@@ -1,21 +1,24 @@
-konduktor/__init__.py,sha256=rrQF0YawJ9sSLYupAzTceMr6I4hMCep3RpfIys4iEeg,1540
+konduktor/__init__.py,sha256=49EUy2CahYmcOfpHvW7CoasITVbIMDI4jIy-mmSBzWk,1574
 konduktor/adaptors/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 konduktor/adaptors/aws.py,sha256=s47Ra-GaqCQibzVfmD0pmwEWHif1EGO5opMbwkLxTCU,8244
 konduktor/adaptors/common.py,sha256=ZIqzjx77PIHUwpjfAQ1uX8B2aX78YMuGj4Bppd-MdyM,4183
 konduktor/adaptors/gcp.py,sha256=ierTF4z7vwpJ9BsC7LSiwv4uLcjGXscwZOwQrddr2vM,4102
 konduktor/authentication.py,sha256=_mVy3eqoKohicHostFiGwG1-2ybxP-l7ouofQ0LRlCY,4570
-konduktor/backends/__init__.py,sha256=1Q6sqqdeMYarpTX_U-QVywJYf7idiUTRsyP-E4BQSOw,129
+konduktor/backends/__init__.py,sha256=usWJ8HdZJEyg7MIsN8Zcz9rk9e2Lq5dWJ8dv6hCN3ys,199
 konduktor/backends/backend.py,sha256=qh0bp94lzoTYZkzyQv2-CVrB5l91FkG2vclXg24UFC0,2910
-konduktor/backends/constants.py,sha256=nU_cd4x8V2GwP9-oGlcIwjt5snnyhmOlxXbXRZ8d6Fc,26
-konduktor/backends/jobset.py,sha256=jJhNueZThMmEbMhpH162Ij0W_qG2Q7TmixVPQzbvHDY,8422
-konduktor/backends/jobset_utils.py,sha256=kezVxzire4r_LR4FyhSFolPY6O2hR4hxkSeqkWj4Luc,23050
+konduktor/backends/constants.py,sha256=eqURY4RU_YXX_WRcge4AZHjr4nwuxTmmVGBnDrD_Qa4,441
+konduktor/backends/deployment.py,sha256=EHfB2uLeKFQ3maek9tx6XL4_sjQ-ax59DZA79Q3EkVs,5519
+konduktor/backends/deployment_utils.py,sha256=VGuL01rKe7p7PoVRI_cP4tiZRxHZ13nnTMG-bmDf7P0,28975
+konduktor/backends/jobset.py,sha256=OwgDog9nH-FoUmNU_H--C3U5jx70reTKL1l849M1k5A,8430
+konduktor/backends/jobset_utils.py,sha256=MRZf-Wcn084lnig0SfgXlF9Q3RyC7m7THeJQQlpdSw8,12931
+konduktor/backends/pod_utils.py,sha256=K0y2kRTzrmIWFbmyJSEMhw7gueiIFF7VAlXtaBMNHkM,15237
 konduktor/check.py,sha256=JennyWoaqSKhdyfUldd266KwVXTPJpcYQa4EED4a_BA,7569
-konduktor/cli.py,sha256=_qfhtukWG6povdB8mfysAqzBEpICWUcBx9PYQWjHk90,36027
+konduktor/cli.py,sha256=PRMOseG1v1qXj0-KbW7Hzsg5rVRfNZw1oykSJWHO_qA,45481
 konduktor/config.py,sha256=J50JxC6MsXMnlrJPXdDUMr38C89xvOO7mR8KJ6fyils,15520
 konduktor/constants.py,sha256=T3AeXXxuQHINW_bAWyztvDeS8r4g8kXBGIwIq13cys0,1814
 konduktor/controller/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 konduktor/controller/constants.py,sha256=SGAgu9yTDWYXyVwxlaw1vfRJFOflPR549mKwgdzbI9w,1124
-konduktor/controller/launch.py,sha256=F_ybOxscbsnmL1c7v5is6QrTDcaV26dq8CsSn4YdmTc,1522
+konduktor/controller/launch.py,sha256=rmxBKYiAKObbYDiJEOx2S1B3pj5BX2avabbFYBEYEyk,1530
 konduktor/controller/node.py,sha256=9uKFtgxmonxtr-73DRAd7qJjHUjyfS1E4sgXT0agzPg,2982
 konduktor/controller/parse.py,sha256=U1G747N6Hef2cEgXvoRI2V1NEU316VDDHLCY1rYtnNc,3840
 konduktor/dashboard/README.md,sha256=xOeItNLb3t0k8AiiQcjqleRfrlcWGK-n6yJF-rLv4M0,718
@@ -62,21 +65,23 @@ konduktor/data/gcp/utils.py,sha256=FJQcMXZqtMIzjZ98b3lTTc0UbdPUKTDLsOsfJaaH5-s,2
 konduktor/data/registry.py,sha256=CUbMsN_Q17Pf4wRHkqZrycErEjTP7cLEdgcfwVGcEpc,696
 konduktor/data/storage.py,sha256=o2So-bY9glvgbGdoN7AQNYmNnvGf1AUDPpImtadRL90,35213
 konduktor/data/storage_utils.py,sha256=n4GivkN0KMqmyOTDznF0Z-hzsJvm7KCEh5i5HgFAT-4,20806
-konduktor/execution.py,sha256=NCl2bgo5p1ZZl8HLaXT-juAe9PXr-iCJv0md2sT7A20,18395
-konduktor/kube_client.py,sha256=lC-U_1hLRG3mDN8tBxYc4VZ3BS5BzKm8hlt-lE3505A,5938
+konduktor/execution.py,sha256=d0EP79iSrW2uFsoqn0YV_4kgIupPIqpMOParXx0y3kg,18519
+konduktor/kube_client.py,sha256=WELs9jClRW9r-imNJF3gJi3Z7ygkFDnYDmMXu5nJhEM,6213
 konduktor/logging.py,sha256=mA1JCCWPCqQMRqEpE4l6D6vOYdbtbQXr0BuEk9RR790,3177
 konduktor/manifests/controller_deployment.yaml,sha256=6p3oSLkEVONZsvKZGqVop0Dhn4bo3lrigRmhf8NXBHE,1730
 konduktor/manifests/dashboard_deployment.yaml,sha256=xJLd4FbPMAosI0fIv5_8y7dV9bw0Vsf81l-w4MB_aU8,2837
 konduktor/manifests/dmesg_daemonset.yaml,sha256=pSWt7YOeTYjS0l0iki1fvHOs7MhY-sH-RQfVW6JJyno,1391
 konduktor/manifests/pod_cleanup_controller.yaml,sha256=hziL1Ka1kCAEL9R7Tjvpb80iw1vcq9_3gwHCu75Bi0A,3939
-konduktor/resource.py,sha256=nHgPWXCbWj5sWyslNngrFypMN1K0Dksb0yHbJqWaei8,19612
-konduktor/task.py,sha256=ofwd8WIhfD6C3ThLcv6X3GUzQHyZ6ddjUagE-umF4K0,35207
+konduktor/resource.py,sha256=qQhMlI6gvTaoGfYb9NNgSrUavgNqfcYVfb9V_oC5pLE,20411
+konduktor/serving.py,sha256=sh8TPAUXg23Bkt0ByatIMdxFFqzRm18HJTEkt3wHzdo,5147
+konduktor/task.py,sha256=jrr-6mNWOB8I199N6OqAZIEmXoW17Xs1ZC9-y_N7P3w,37480
+konduktor/templates/deployment.yaml.j2,sha256=uXFjDQaimbpFdAn2RJGaIvS_PzDY136cw_L3QMjz3ZA,3452
 konduktor/templates/jobset.yaml.j2,sha256=rdURknodtgLp4zoA2PX86Nn4wPpi3tr5l4IG55aWBRg,1059
-konduktor/templates/pod.yaml.j2,sha256=GoBNypXgPkFfmD-a4t0WVCo9sXbgSedvl-YNygso-Fc,17160
+konduktor/templates/pod.yaml.j2,sha256=3uXx0ls2v8x-NL_Ypze5u9RoJS8F5bzoyOJcYwzf8Z0,18240
 konduktor/usage/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 konduktor/usage/constants.py,sha256=gCL8afIHZhO0dcxbJGpESE9sCC1cBSbeRnQ8GwNOY4M,612
 konduktor/utils/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-konduktor/utils/accelerator_registry.py,sha256=LmhLPR-N9kxrk0UBYYcF1O6ADv9HHwsn_Pc3SOYFNzs,553
+konduktor/utils/accelerator_registry.py,sha256=ythz3ynulP1DSSU7Jj5VUsQeBzSYRkxCVDZ5oOg0xtc,560
 konduktor/utils/annotations.py,sha256=oy2-BLydkFt3KWkXDuaGY84d6b7iISuy4eAT9uXk0Fc,2225
 konduktor/utils/base64_utils.py,sha256=mF-Tw98mFRG70YE4w6s9feuQSCYZHOb8YatBZwMugyI,3130
 konduktor/utils/common_utils.py,sha256=4yG5Kjvu1hu6x2nKNaaCUKQNrheUaG61Qe913MFPry8,15060
@@ -85,15 +90,15 @@ konduktor/utils/env_options.py,sha256=T41Slzf4Mzl-n45CGXXqdy2fCrYhPNZQ7RP5vmnN4x
 konduktor/utils/exceptions.py,sha256=5IFnN5bIUSBJv4KRRrCepk5jyY9EG5vWWQqbjCmP3NU,6682
 konduktor/utils/kubernetes_enums.py,sha256=SabUueF6Bpzbpa57gyH5VB65xla2N9l8CZmAeYTfGmM,176
 konduktor/utils/kubernetes_utils.py,sha256=VG7qatUFyWHY-PCQ8fYWh2kn2TMwfg84cn-VkXdCwI8,26077
-konduktor/utils/log_utils.py,sha256=5lIhQA_w7_BZFZiPoi7nWo8K5qFhfj70UkxjxbDEcCk,16650
+konduktor/utils/log_utils.py,sha256=CC6McrURGIcM8R4ICBCmn_Y7oXbWh1fwl4__la6Zotw,16677
 konduktor/utils/loki_utils.py,sha256=h2ZvZQr1nE_wXXsKsGMjhG2s2MXknNd4icydTR_ruKU,3539
 konduktor/utils/rich_utils.py,sha256=ycADW6Ij3wX3uT8ou7T8qxX519RxlkJivsLvUahQaJo,3583
-konduktor/utils/schemas.py,sha256=7z_-sggfwIMRDaCSpTnBUfGHQH3M2oiwImS5PvfAefw,16858
+konduktor/utils/schemas.py,sha256=tBrKhnkfn9uKDYdlb4L2KgooW-muuhww7U8fu9zX-ms,18336
 konduktor/utils/subprocess_utils.py,sha256=WoFkoFhGecPR8-rF8WJxbIe-YtV94LXz9UG64SDhCY4,9448
 konduktor/utils/ux_utils.py,sha256=czCwiS1bDqgeKtzAJctczpLwFZzAse7WuozdvzEFYJ4,7437
-konduktor/utils/validator.py,sha256=uCRlScO1NYxsbTNKY9dkoqvlO8S0ISIIB8XmX2ItcO8,2793
-konduktor_nightly-0.1.0.dev20250804105449.dist-info/LICENSE,sha256=MuuqTZbHvmqXR_aNKAXzggdV45ANd3wQ5YI7tnpZhm0,6586
-konduktor_nightly-0.1.0.dev20250804105449.dist-info/METADATA,sha256=sbGUTYzCdRA9hySZGwgU44KFPRVsRY5ODbBIBAWvnxg,4247
-konduktor_nightly-0.1.0.dev20250804105449.dist-info/WHEEL,sha256=XbeZDeTWKc1w7CSIyre5aMDU_-PohRwTQceYnisIYYY,88
-konduktor_nightly-0.1.0.dev20250804105449.dist-info/entry_points.txt,sha256=k3nG5wDFIJhNqsZWrHk4d0irIB2Ns9s47cjRWYsTCT8,48
-konduktor_nightly-0.1.0.dev20250804105449.dist-info/RECORD,,
+konduktor/utils/validator.py,sha256=5C1kE57Eyj1OPnAbvojqMNHHtf5fnl47FK_vEttd8aw,4331
+konduktor_nightly-0.1.0.dev20250806105405.dist-info/LICENSE,sha256=MuuqTZbHvmqXR_aNKAXzggdV45ANd3wQ5YI7tnpZhm0,6586
+konduktor_nightly-0.1.0.dev20250806105405.dist-info/METADATA,sha256=IAh98IgREJO_zM_IVVjp_otfi4EQ3YctV1DWEDKFuyU,4247
+konduktor_nightly-0.1.0.dev20250806105405.dist-info/WHEEL,sha256=XbeZDeTWKc1w7CSIyre5aMDU_-PohRwTQceYnisIYYY,88
+konduktor_nightly-0.1.0.dev20250806105405.dist-info/entry_points.txt,sha256=k3nG5wDFIJhNqsZWrHk4d0irIB2Ns9s47cjRWYsTCT8,48
+konduktor_nightly-0.1.0.dev20250806105405.dist-info/RECORD,,