Adding CPU support in XPK

AI-Hypercomputer · Jan 19, 2024 · f4d1ace · f4d1ace
1 parent 8897ac8
commit f4d1ace
Show file tree

Hide file tree

Showing 2 changed files with 57 additions and 8 deletions.
diff --git a/README.md b/README.md
@@ -403,7 +403,7 @@ Please select a CPU type that exists in all zones in the region.
 # Find CPU Types supported in zones.
 gcloud compute machine-types list --zones=$ZONE_LIST
 # Adjust default cpu machine type.
-python3 xpk.py cluster create --cluster-cpu-machine-type=CPU_TYPE ...
+python3 xpk.py cluster create --default-pool-cpu-machine-type=CPU_TYPE ...
 ```
 
 ## Permission Issues: `requires one of ["permission_name"] permission(s)`.

diff --git a/xpk.py b/xpk.py
@@ -91,6 +91,15 @@
             spec:
               schedulerName: {args.scheduler}
               restartPolicy: Never
+              affinity:
+                nodeAffinity:
+                  requiredDuringSchedulingIgnoredDuringExecution:
+                    nodeSelectorTerms:
+                    - matchExpressions:
+                      - key: cloud.google.com/gke-nodepool
+                        operator: NotIn
+                        values:
+                        - default-pool
               nodeSelector:
                 {accelerator_label}
                 {machine_label}
@@ -247,7 +256,8 @@
 
 AcceleratorType = {
   'TPU': 1,
-  'GPU': 2
+  'GPU': 2,
+  'CPU': 3
 }
 
 @dataclass
@@ -264,6 +274,10 @@ class AcceleratorCharacteristics:
      # GPU
       AcceleratorType['GPU']: AcceleratorCharacteristics(
       'nvidia.com/gpu', 'cloud.google.com/gke-accelerator', 'cloud.google.com/gce-machine-type'
+      ),
+     # CPU
+      AcceleratorType['CPU']: AcceleratorCharacteristics(
+      'cpu', '', 'cloud.google.com/gke-nodepool' 
       )
 }
 
@@ -643,6 +657,25 @@ class SystemCharacteristics:
     'v4-4096': SystemCharacteristics(
       '8x16x16', 512,'tpu-v4-podslice', 'ct4p-hightpu-4t', 4, AcceleratorType['TPU'], 'v4-4096'
     ),
+
+    # CPU system characteristics
+    # n2-standard-32-$VMs
+    'n2-standard-32-1': SystemCharacteristics(
+      'N/A', 1,'N/A', 'n2-standard-32', 1, AcceleratorType['CPU'], 'n2-standard-32-1'
+    ),
+    'n2-standard-32-2': SystemCharacteristics(
+      'N/A', 2,'N/A', 'n2-standard-32', 1, AcceleratorType['CPU'], 'n2-standard-32-2'
+    ),
+    'n2-standard-32-4': SystemCharacteristics(
+      'N/A', 4,'N/A', 'n2-standard-32', 1, AcceleratorType['CPU'], 'n2-standard-32-4'
+    ),
+    'n2-standard-32-8': SystemCharacteristics(
+      'N/A', 8,'N/A', 'n2-standard-32', 1, AcceleratorType['CPU'], 'n2-standard-32-8'
+    ),
+
+    'n2-standard-32-256': SystemCharacteristics(
+      'N/A', 256,'N/A', 'n2-standard-32', 1, AcceleratorType['CPU'], 'n2-standard-32-256'
+    ),
 }
 """ If you modify UserFacingNameToSystemCharacteristics you should also modify the corresponding
 Map in MaxText/accelerator_to_spec_map.py """
@@ -1036,6 +1069,10 @@ def run_gke_cluster_create_command(args) -> int:
   Returns:
     0 if successful and 1 otherwise.
   """
+  # cluster_cpu_machine_type is being deprecated!
+  if args.cluster_cpu_machine_type is not None:
+    xpk_print('Note that cluster-cpu-machine-type is deprecated, please use default-pool-cpu-machine-type instead!')
+    xpk_exit(0)
 
   # Create the regional cluster with `num-nodes` CPU nodes in the same zone as
   # TPUs. This has been tested with clusters of 300 VMs. Larger clusters will
@@ -1048,7 +1085,7 @@ def run_gke_cluster_create_command(args) -> int:
       f' --node-locations={args.zone}'
       f' --project={args.project} --region={zone_to_region(args.zone)}'
       f' --cluster-version={args.gke_version} --location-policy=BALANCED'
-      f' --machine-type={args.cluster_cpu_machine_type}'
+      f' --machine-type={args.default_pool_cpu_machine_type}'
       ' --scopes=storage-full,gke-default'
       f' {args.custom_cluster_arguments}'
   )
@@ -1316,10 +1353,10 @@ def run_gke_node_pool_create_command(args, system) -> int:
         f' {capacity_args}'
         ' --scopes=storage-full,gke-default'
         ' --enable-gvnic --max-pods-per-node 15'
-        f' {args.custom_tpu_nodepool_arguments}'
     )
     if system.accelerator_type == AcceleratorType['TPU']:
       command += (f' --tpu-topology={system.topology}')
+      command += (f' {args.custom_tpu_nodepool_arguments}')
     elif system.accelerator_type == AcceleratorType['GPU']:
       command += f' --accelerator type={system.gke_accelerator},count={str(system.chips_per_vm)}'
     task = f'NodepoolCreate-{node_pool_name}'
@@ -2182,6 +2219,8 @@ def create_accelerator_label(accelerator_type, system) -> str:
   Returns:
     The accelerator label.
   """
+  if accelerator_type == AcceleratorType['CPU']:
+    return ""
   return f"{AcceleratorTypeToAcceleratorCharacteristics[accelerator_type].accelerator_label}: {system.gke_accelerator}"
 
 def create_machine_label(accelerator_type, system) -> str:
@@ -2573,7 +2612,7 @@ def directory_path_type(value):
     '--device-type',
     type=str,
     default=None,
-    help='The device type to use (can be tpu or gpu), v5litepod-16, h100-80gb-8, etc.'
+    help='The device type to use (can be tpu or gpu or cpu), v5litepod-16, h100-80gb-8, n2-standard-32-4 etc.'
 )
 
 
@@ -2629,14 +2668,24 @@ def directory_path_type(value):
     required=True,
 )
 cluster_create_optional_arguments.add_argument(
-  '--cluster-cpu-machine-type',
+  '--default-pool-cpu-machine-type',
     type=str,
     default='e2-standard-16',
     help=(
-      'Set the machine tpu within the default cpu node pool. For'
+      'Set the machine type within the default cpu node pool. For'
       ' regional clusters, all zones must support the machine type.'
     )
 )
+cluster_create_optional_arguments.add_argument(
+  '--cluster-cpu-machine-type',
+    type=str,
+    default='',
+    help=(
+      'Deprecated! Please use --default-pool-cpu-machine-type instead,'
+      ' to denote the machine type of the default cpu node pool. Set'
+      ' the machine type of other cpu nodepools using --device-type.'
+    )
+)
 cluster_create_optional_arguments.add_argument(
     '--custom-cluster-arguments',
     type=str,
@@ -2884,7 +2933,7 @@ def directory_path_type(value):
     '--device-type',
     type=str,
     default=None,
-    help='The device type to use (can be tpu or gpu), v5litepod-16, h100-80gb-8, etc.'
+    help='The device type to use (can be tpu or gpu or cpu), v5litepod-16, h100-80gb-8, n2-standard-32-4 etc.'
 )
 
 ### Workload Optional Arguments