Misc changes

Author: deliahu

dev/versions.md

@@ -118,7 +118,7 @@ see https://github.com/moby/moby/issues/39302#issuecomment-639687466_
     1. Update the link at the top of the file to the URL you copied from
     1. Check that your diff is reasonable (and put back any of our modifications, e.g. the image path, rolling update
        strategy, resource requests, tolerations, node selector, priority class, etc)
-1. Confirm GPUs work for PyTorch, TensorFlow, and ONNX models
+1. Confirm GPUs work
 
 ## Inferentia device plugin
 

docs/workloads/batch/jobs.md

@@ -202,7 +202,7 @@ RESPONSE:
         },
         "worker_counts": {          # worker counts are only available while a job is running
             "pending": <int>,       # number of workers that are waiting for compute resources to be provisioned
-            "initializing": <int>,  # number of workers that are initializing (downloading images or running your handler's init function)
+            "initializing": <int>,  # number of workers that are initializing
             "running": <int>,       # number of workers that are actively working on batches from the queue
             "succeeded": <int>,     # number of workers that have completed after verifying that the queue is empty
             "failed": <int>,        # number of workers that have failed

docs/workloads/realtime/autoscaling.md

@@ -12,7 +12,7 @@ In addition to the autoscaling configuration options (described below), there ar
 
 <br>
 
-**`max_queue_length`** (default: 100): The maximum number of requests which will be queued by the replica (beyond `max_concurrency`) before requests are rejected with HTTP error code 503. For long-running APIs, decreasing `max_replica_concurrency` and configuring the client to retry when it receives 503 responses will improve queue fairness accross replicas by preventing requests from sitting in long queues.
+**`max_queue_length`** (default: 100): The maximum number of requests which will be queued by the replica (beyond `max_concurrency`) before requests are rejected with HTTP error code 503. For long-running APIs, decreasing `max_queue_length` and configuring the client to retry when it receives 503 responses will improve queue fairness accross replicas by preventing requests from sitting in long queues.
 
 <br>
 

manager/manifests/grafana/grafana-dashboard-realtime.yaml

@@ -964,9 +964,7 @@ data:
           }
         },
         {
-          "aliasColors": {
-            "iris-classifier": "light-green"
-          },
+          "aliasColors": {},
           "bars": false,
           "dashLength": 10,
           "dashes": false,

pkg/lib/regex/regex_test.go

@@ -596,7 +596,7 @@ func TestValidDockerImage(t *testing.T) {
 			match: false, // Support this as valid?
 		},
 		{
-			input: "680880929103.dkr.ecr.eu-central-1.amazonaws.com/cortexlabs/python-handler-cpu:latest",
+			input: "680880929103.dkr.ecr.eu-central-1.amazonaws.com/cortexlabs/async-gateway:latest",
 			match: true,
 		},
 	}

pkg/types/spec/utils.go

@@ -24,8 +24,6 @@ import (
 	"github.com/cortexlabs/cortex/pkg/types/userconfig"
 )
 
-type modelValidator func(paths []string, prefix string, versionedPrefix *string) error
-
 func FindDuplicateNames(apis []userconfig.API) []userconfig.API {
 	names := make(map[string][]userconfig.API)