Release 0.10.5 (#7869)

* Release 0.10.5

* Cherry pick d0e0748 to 0.10.5 (#7870)

[Requests] Fix list_accelerators JSON encoding (#7865)

* [Requests] Fix list_accelerators JSON encoding

* improve test

---------

Co-authored-by: GitHub Action <[email protected]>
Co-authored-by: Kevin Mingtarja <[email protected]>

Author: 3 people

sky/__init__.py

@@ -37,7 +37,7 @@ def _get_git_commit():
 
 
 __commit__ = _get_git_commit()
-__version__ = '0.10.4'
+__version__ = '0.10.5'
 __root_dir__ = directory_utils.get_sky_dir()
 
 

sky/server/requests/serializers/encoders.py

@@ -8,14 +8,15 @@
 import typing
 from typing import Any, Dict, List, Optional, Tuple, Union
 
+from sky import models
+from sky.catalog import common
 from sky.schemas.api import responses
 from sky.server import constants as server_constants
 from sky.utils import serialize_utils
 
 if typing.TYPE_CHECKING:
     from sky import backends
     from sky import clouds
-    from sky import models
     from sky.provision.kubernetes import utils as kubernetes_utils
 
 handlers: Dict[str, Any] = {}
@@ -254,11 +255,27 @@ def encode_realtime_gpu_availability(
                              List[Any]]]) -> List[Tuple[str, List[List[Any]]]]:
     # Convert RealtimeGpuAvailability namedtuples to lists
     # for JSON serialization.
-    result = []
+    encoded = []
     for context, gpu_list in return_value:
-        gpu_availability_list = []
+        converted_gpu_list = []
         for gpu in gpu_list:
-            gpu_list_item = [gpu.gpu, gpu.counts, gpu.capacity, gpu.available]
-            gpu_availability_list.append(gpu_list_item)
-        result.append((context, gpu_availability_list))
-    return result
+            assert isinstance(gpu, models.RealtimeGpuAvailability), (
+                f'Expected RealtimeGpuAvailability, got {type(gpu)}')
+            converted_gpu_list.append(list(gpu))
+        encoded.append((context, converted_gpu_list))
+    return encoded
+
+
+@register_encoder('list_accelerators')
+def encode_list_accelerators(
+        return_value: Dict[str, List[Any]]) -> Dict[str, Any]:
+    encoded: Dict[str, Any] = {}
+    for accelerator_name, instances in return_value.items():
+        # Convert InstanceTypeInfo namedtuples to lists for JSON serialization.
+        converted_instances: List[Any] = []
+        for instance in instances:
+            assert isinstance(instance, common.InstanceTypeInfo), (
+                f'Expected InstanceTypeInfo, got {type(instance)}')
+            converted_instances.append(list(instance))
+        encoded[accelerator_name] = converted_instances
+    return encoded

tests/smoke_tests/test_basic.py

@@ -1084,6 +1084,25 @@ def test_kubernetes_show_gpus(generic_cloud: str):
     smoke_tests_utils.run_one_test(test)
 
 
+@pytest.mark.no_kubernetes
+def test_show_gpus(generic_cloud: str):
+    # Check that output contains GPU table headers and common GPU types
+    check_cmd = ('echo "$s" && '
+                 'echo "$s" | grep "COMMON_GPU" && '
+                 'echo "$s" | grep "AVAILABLE_QUANTITIES" && '
+                 'echo "$s" | grep -E "A100|H100|H200|L4|T4|B200"')
+    test = smoke_tests_utils.Test(
+        'show_gpus',
+        [
+            (f's=$(SKYPILOT_DEBUG=0 sky show-gpus --infra {generic_cloud}) && '
+             f'{check_cmd}'),
+            (f's=$(SKYPILOT_DEBUG=0 sky show-gpus --infra {generic_cloud} --all) && '
+             f'{check_cmd}'),
+        ],
+    )
+    smoke_tests_utils.run_one_test(test)
+
+
 @pytest.mark.no_seeweb  # Seeweb fails to provision resources
 def test_launch_and_exec_async(generic_cloud: str):
     """Test if the launch and exec commands work correctly with --async."""