Fix Nvidia device plugin that is running on CPU (#78)

* Fix Nvidia device plugin that is running on CPU

**Description**

This is for Github issue: #48

**Testing Done**

Verified and there's no nvidia device plugin pod running on CPU node

* Update nvidia-device-plugin instance labels and reorder HP CLI instance allowlist

Author: zhaoqizqwang

helm_chart/HyperPodHelmChart/values.yaml

@@ -138,6 +138,40 @@ nvidia-device-plugin:
       requiredDuringSchedulingIgnoredDuringExecution:
         nodeSelectorTerms: 
         - matchExpressions:
+          - key: node.kubernetes.io/instance-type
+            operator: In
+            values:
+              - ml.g5.xlarge
+              - ml.g5.2xlarge
+              - ml.g5.4xlarge
+              - ml.g5.8xlarge
+              - ml.g5.12xlarge
+              - ml.g5.16xlarge
+              - ml.g5.24xlarge
+              - ml.g5.48xlarge
+              - ml.g6.xlarge
+              - ml.g6.2xlarge
+              - ml.g6.4xlarge
+              - ml.g6.8xlarge
+              - ml.g6.16xlarge
+              - ml.g6.12xlarge
+              - ml.g6.24xlarge
+              - ml.g6.48xlarge
+              - ml.g6e.xlarge
+              - ml.g6e.2xlarge
+              - ml.g6e.4xlarge
+              - ml.g6e.8xlarge
+              - ml.g6e.12xlarge
+              - ml.g6e.16xlarge
+              - ml.g6e.24xlarge
+              - ml.g6e.48xlarge
+              - ml.gr6.4xlarge
+              - ml.gr6.8xlarge
+              - ml.p4d.24xlarge
+              - ml.p4de.24xlarge
+              - ml.p5.48xlarge
+              - ml.p5e.48xlarge
+              - ml.p5en.48xlarge
           # nvidia plugin needs at least one node selector. Below label exists for all hyperpod nodes
           - key: kubernetes.io/os
             operator: In

src/hyperpod_cli/constants/hyperpod_instance_types.py

@@ -14,19 +14,6 @@
 
 
 class HyperpodInstanceType(Enum):
-    ML_P4D_24XLARGE = "ml.p4d.24xlarge"
-    ML_P4DE_24XLARGE = "ml.p4de.24xlarge"
-    ML_P5_48XLARGE = "ml.p5.48xlarge"
-    ML_TRN1_32XLARGE = "ml.trn1.32xlarge"
-    ML_TRN1N_32XLARGE = "ml.trn1n.32xlarge"
-    ML_G5_XLARGE = "ml.g5.xlarge"
-    ML_G5_2XLARGE = "ml.g5.2xlarge"
-    ML_G5_4XLARGE = "ml.g5.4xlarge"
-    ML_G5_8XLARGE = "ml.g5.8xlarge"
-    ML_G5_12XLARGE = "ml.g5.12xlarge"
-    ML_G5_16XLARGE = "ml.g5.16xlarge"
-    ML_G5_24XLARGE = "ml.g5.24xlarge"
-    ML_G5_48XLARGE = "ml.g5.48xlarge"
     ML_C5_LARGE = "ml.c5.large"
     ML_C5_XLARGE = "ml.c5.xlarge"
     ML_C5_2XLARGE = "ml.c5.2xlarge"
@@ -40,6 +27,32 @@ class HyperpodInstanceType(Enum):
     ML_C5N_4XLARGE = "ml.c5n.4xlarge"
     ML_C5N_9XLARGE = "ml.c5n.9xlarge"
     ML_C5N_18XLARGE = "ml.c5n.18xlarge"
+    ML_G5_XLARGE = "ml.g5.xlarge"
+    ML_G5_2XLARGE = "ml.g5.2xlarge"
+    ML_G5_4XLARGE = "ml.g5.4xlarge"
+    ML_G5_8XLARGE = "ml.g5.8xlarge"
+    ML_G5_12XLARGE = "ml.g5.12xlarge"
+    ML_G5_16XLARGE = "ml.g5.16xlarge"
+    ML_G5_24XLARGE = "ml.g5.24xlarge"
+    ML_G5_48XLARGE = "ml.g5.48xlarge"
+    ML_G6_XLARGE = "ml.g6.xlarge"
+    ML_G6_2XLARGE = "ml.g6.2xlarge"
+    ML_G6_4XLARGE = "ml.g6.4xlarge"
+    ML_G6_8XLARGE = "ml.g6.8xlarge"
+    ML_G6_12XLARGE = "ml.g6.12xlarge"
+    ML_G6_16XLARGE = "ml.g6.16xlarge"
+    ML_G6_24XLARGE = "ml.g6.24xlarge"
+    ML_G6_48XLARGE = "ml.g6.48xlarge"
+    ML_G6E_XLARGE = "ml.g6e.xlarge"
+    ML_G6E_2XLARGE = "ml.g6e.2xlarge"
+    ML_G6E_4XLARGE = "ml.g6e.4xlarge"
+    ML_G6E_8XLARGE = "ml.g6e.8xlarge"
+    ML_G6E_12XLARGE = "ml.g6e.12xlarge"
+    ML_G6E_16XLARGE = "ml.g6e.16xlarge"
+    ML_G6E_24XLARGE = "ml.g6e.24xlarge"
+    ML_G6E_48XLARGE = "ml.g6e.48xlarge"
+    ML_GR6_4XLARGE = "ml.gr6.4xlarge"
+    ML_GR6_8XLARGE = "ml.gr6.8xlarge"
     ML_I3EN_LARGE = "ml.i3en.large"
     ML_I3EN_XLARGE = "ml.i3en.xlarge"
     ML_I3EN_2XLARGE = "ml.i3en.2xlarge"
@@ -64,28 +77,9 @@ class HyperpodInstanceType(Enum):
     ML_M7I_16XLARGE = "ml.m7i.16xlarge"
     ML_M7I_24XLARGE = "ml.m7i.24xlarge"
     ML_M7I_48XLARGE = "ml.m7i.48xlarge"
-    ML_T3_MEDIUM = "ml.t3.medium"
-    ML_T3_LARGE = "ml.t3.large"
-    ML_T3_XLARGE = "ml.t3.xlarge"
-    ML_T3_2XLARGE = "ml.t3.2xlarge"
-    ML_G6_XLARGE = "ml.g6.xlarge"
-    ML_G6_2XLARGE = "ml.g6.2xlarge"
-    ML_G6_4XLARGE = "ml.g6.4xlarge"
-    ML_G6_8XLARGE = "ml.g6.8xlarge"
-    ML_G6_16XLARGE = "ml.g6.16xlarge"
-    ML_G6_12XLARGE = "ml.g6.12xlarge"
-    ML_G6_24XLARGE = "ml.g6.24xlarge"
-    ML_G6_48XLARGE = "ml.g6.48xlarge"
-    ML_GR6_4XLARGE = "ml.gr6.4xlarge"
-    ML_GR6_8XLARGE = "ml.gr6.8xlarge"
-    ML_G6E_XLARGE = "ml.g6e.xlarge"
-    ML_G6E_2XLARGE = "ml.g6e.2xlarge"
-    ML_G6E_4XLARGE = "ml.g6e.4xlarge"
-    ML_G6E_8XLARGE = "ml.g6e.8xlarge"
-    ML_G6E_16XLARGE = "ml.g6e.16xlarge"
-    ML_G6E_12XLARGE = "ml.g6e.12xlarge"
-    ML_G6E_24XLARGE = "ml.g6e.24xlarge"
-    ML_G6E_48XLARGE = "ml.g6e.48xlarge"
+    ML_P4D_24XLARGE = "ml.p4d.24xlarge"
+    ML_P4DE_24XLARGE = "ml.p4de.24xlarge"
+    ML_P5_48XLARGE = "ml.p5.48xlarge"
     ML_P5E_48XLARGE = "ml.p5e.48xlarge"
     ML_P5EN_48XLARGE = "ml.p5en.48xlarge"
     ML_R7I_LARGE = "ml.r7i.large"
@@ -97,4 +91,10 @@ class HyperpodInstanceType(Enum):
     ML_R7I_16XLARGE = "ml.r7i.16xlarge"
     ML_R7I_24XLARGE = "ml.r7i.24xlarge"
     ML_R7I_48XLARGE = "ml.r7i.48xlarge"
+    ML_T3_MEDIUM = "ml.t3.medium"
+    ML_T3_LARGE = "ml.t3.large"
+    ML_T3_XLARGE = "ml.t3.xlarge"
+    ML_T3_2XLARGE = "ml.t3.2xlarge"
+    ML_TRN1_32XLARGE = "ml.trn1.32xlarge"
+    ML_TRN1N_32XLARGE = "ml.trn1n.32xlarge"
     ML_TRN2_48XLARGE = "ml.trn2.48xlarge"