#         "@absl_py//absl/testing:parameterized",
#     ],
# )
#
# tf_py_strict_test(
#     name = "quantize_model_test",
#     srcs = ["integration_test/quantize_model_test.py"],
#     shard_count = 50,  # Parallelize the test to avoid timeouts.
#     deps = [
#         ":quantization",
#         ":quantize_model_test_base",
#         "//tensorflow/compiler/mlir/quantization/common/python:testing",

	clientsetfake "k8s.io/client-go/kubernetes/fake"
	"k8s.io/klog/v2/ktesting"
	extenderv1 "k8s.io/kube-scheduler/extender/v1"
	"k8s.io/kubernetes/pkg/scheduler/framework"
	"k8s.io/kubernetes/pkg/scheduler/framework/parallelize"
	"k8s.io/kubernetes/pkg/scheduler/framework/plugins/defaultbinder"
	"k8s.io/kubernetes/pkg/scheduler/framework/plugins/interpodaffinity"
	"k8s.io/kubernetes/pkg/scheduler/framework/plugins/nodeaffinity"

	schedulerapi "k8s.io/kubernetes/pkg/scheduler/apis/config"
	"k8s.io/kubernetes/pkg/scheduler/apis/config/scheme"
	"k8s.io/kubernetes/pkg/scheduler/framework"
	"k8s.io/kubernetes/pkg/scheduler/framework/parallelize"
	frameworkplugins "k8s.io/kubernetes/pkg/scheduler/framework/plugins"
	"k8s.io/kubernetes/pkg/scheduler/framework/plugins/noderesources"
	frameworkruntime "k8s.io/kubernetes/pkg/scheduler/framework/runtime"

        "@absl_py//absl/logging",
    ],
)

tf_py_strict_test(
    name = "quantize_model_test",
    size = "medium",
    srcs = ["integration_test/quantize_model_test.py"],
    shard_count = 50,  # Parallelize the test to avoid timeouts.
    tags = [
        "no_mac",  # TODO(b/292100835): Reenable
    ],
    deps = [
        ":quantize_model",
        ":quantize_model_test_base",
        ":representative_dataset",

## Deployment Concepts

Here you saw how to use **Gunicorn** (or Uvicorn) managing **Uvicorn worker processes** to **parallelize** the execution of the application, take advantage of **multiple cores** in the CPU, and be able to serve **more requests**.

//
// It must be safe to call methods on a Reqs from multiple goroutines simultaneously.
// Because a Reqs may read the underlying graph from the network on demand,
// the MVS algorithms parallelize the traversal to overlap network delays.
type Reqs interface {
	// Required returns the module versions explicitly required by m itself.
	// The caller must not modify the returned list.

			// TODO: If many pods changed during the same relist period,
			// inspecting the pod and getting the PodStatus to update the cache
			// serially may take a while. We should be aware of this and
			// parallelize if needed.
			if err, updated := g.updateCache(ctx, pod, pid); err != nil {
				// Rely on updateCache calling GetPodStatus to log the actual error.

	"k8s.io/kubernetes/pkg/scheduler/apis/config"
	configv1 "k8s.io/kubernetes/pkg/scheduler/apis/config/v1"
	"k8s.io/kubernetes/pkg/scheduler/framework"
	"k8s.io/kubernetes/pkg/scheduler/framework/parallelize"
	"k8s.io/kubernetes/pkg/scheduler/framework/plugins/defaultbinder"
	"k8s.io/kubernetes/pkg/scheduler/framework/plugins/feature"
	"k8s.io/kubernetes/pkg/scheduler/framework/plugins/interpodaffinity"

	if upd, err := adsc.Wait(15*time.Second, watchAll...); err != nil {
		t.Fatal("Expecting CDS, EDS, LDS, and RDS update as part of a full push", err, upd)
	}
}

// This test must be run in isolation, can't be parallelized with any other v2 test.
// It makes different kind of updates, and checks that incremental or full push happens.
// In particular:
// - just endpoint changes -> incremental

//     If pod has bound immediate PVCs, GetEligibleNodes() is invoked to potentially reduce
//     down the list of eligible nodes based on the bound PV's NodeAffinity (if any).
//     b. Invokes all filter plugins, parallelized across nodes.  FindPodVolumes() is invoked here.
//     c. Invokes all score plugins.  Future/TBD
//     d. Selects the best node for the Pod.
//     e. Invokes all reserve plugins. AssumePodVolumes() is invoked here.