Added test for llama-3.1-8b model finetuning demo

Author: abhijeet-dhumal

examples/ray-finetune-llm-deepspeed/deepspeed_configs/zero_3_offload_optim+param.json renamed to examples/ray-finetune-llm-deepspeed/deepspeed_configs/zero_3_offload_optim_param.json

@@ -16,8 +16,6 @@ require (
 	sigs.k8s.io/kueue v0.6.2
 )
 
-replace github.com/project-codeflare/codeflare-common => /home/abdhumal/abhidev/RedHatDev/codeflare-common
-
 require (
 	github.com/aymerick/douceur v0.2.0 // indirect
 	github.com/beorn7/perks v1.0.1 // indirect

go.mod

@@ -138,7 +138,7 @@ func mnistRay(t *testing.T, numGpus int) {
 
 	// Initialise raycluster client to interact with raycluster to get rayjob details using REST-API
 	dashboardUrl := GetDashboardUrl(test, namespace, rayCluster)
-	rayClusterClientConfig := RayClusterClientConfig{Address: dashboardUrl.String(), Client: nil, SkipTlsVerification: true}
+	rayClusterClientConfig := RayClusterClientConfig{Address: dashboardUrl.String(), Client: nil, InsecureSkipVerify: true}
 	rayClient, err := NewRayClusterClient(rayClusterClientConfig, test.Config().BearerToken)
 	if err != nil {
 		test.T().Errorf("%s", err)

tests/odh/mnist_ray_test.go

@@ -138,7 +138,7 @@ func mnistRayTuneHpo(t *testing.T, numGpus int) {
 
 	// Initialise raycluster client to interact with raycluster to get rayjob details using REST-API
 	dashboardUrl := GetDashboardUrl(test, namespace, rayCluster)
-	rayClusterClientConfig := RayClusterClientConfig{Address: dashboardUrl.String(), Client: nil, SkipTlsVerification: true}
+	rayClusterClientConfig := RayClusterClientConfig{Address: dashboardUrl.String(), Client: nil, InsecureSkipVerify: true}
 	rayClient, err := NewRayClusterClient(rayClusterClientConfig, test.Config().BearerToken)
 	if err != nil {
 		test.T().Errorf("%s", err)

tests/odh/mnist_raytune_hpo_test.go

@@ -29,11 +29,14 @@ import (
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
 )
 
-func TestRayFinetuneLlmDeepspeedDemo(t *testing.T) {
-	rayFinetuneLlmDeepspeed(t, 1)
+func TestRayFinetuneLlmDeepspeedDemoLlama_2_7b(t *testing.T) {
+	rayFinetuneLlmDeepspeed(t, 1, "zero_3_llama_2_7b.json")
+}
+func TestRayFinetuneLlmDeepspeedDemoLlama_31_8b(t *testing.T) {
+	rayFinetuneLlmDeepspeed(t, 1, "zero_3_offload_optim_param.json")
 }
 
-func rayFinetuneLlmDeepspeed(t *testing.T, numGpus int) {
+func rayFinetuneLlmDeepspeed(t *testing.T, numGpus int, modelConfigFile string) {
 	test := With(t)
 
 	// Create a namespace
@@ -51,7 +54,7 @@ func rayFinetuneLlmDeepspeed(t *testing.T, numGpus int) {
 	// list changes required in llm-deepspeed-finetune-demo.ipynb file and update those
 	requiredChangesInNotebook := map[string]string{
 		"import os":  "import os,time,sys",
-		"import sys": "!cp /opt/app-root/notebooks/* ./",
+		"import sys": "!cp /opt/app-root/notebooks/* ./\\n\",\n\t\"!ls",
 		"from codeflare_sdk.cluster.auth import TokenAuthentication": "from codeflare_sdk.cluster.auth import TokenAuthentication\\n\",\n\t\"from codeflare_sdk.job import RayJobClient",
 		"token = ''":                             fmt.Sprintf("token = '%s'", userToken),
 		"server = ''":                            fmt.Sprintf("server = '%s'", GetOpenShiftApiUrl(test)),
@@ -61,23 +64,26 @@ func rayFinetuneLlmDeepspeed(t *testing.T, numGpus int) {
 		"num_workers=7":                          "num_workers=1",
 		"worker_cpu_requests=16":                 "worker_cpu_requests=4",
 		"worker_cpu_limits=16":                   "worker_cpu_limits=4",
-		"worker_memory_requests=128":             "worker_memory_requests=60",
-		"worker_memory_limits=256":               "worker_memory_limits=60",
+		"worker_memory_requests=128":             "worker_memory_requests=64",
+		"worker_memory_limits=256":               "worker_memory_limits=128",
 		"head_memory=128":                        "head_memory=48",
 		"client = cluster.job_client":            "ray_dashboard = cluster.cluster_dashboard_uri()\\n\",\n\t\"header = {\\\"Authorization\\\": \\\"Bearer " + userToken + "\\\"}\\n\",\n\t\"client = RayJobClient(address=ray_dashboard, headers=header, verify=False)\\n",
 		"--num-devices=8":                        fmt.Sprintf("--num-devices=%d", numGpus),
 		"--num-epochs=3":                         fmt.Sprintf("--num-epochs=%d", 1),
-		"--ds-config=./deepspeed_configs/zero_3_llama_2_7b.json": "--ds-config=./zero_3_llama_2_7b.json \\\"\\n\",\n\t\" \\\"--lora-config=./lora.json \\\"\\n\",\n\t\" \\\"--as-test",
-		"'pip': 'requirements.txt'":                              "'pip': '/opt/app-root/src/requirements.txt'",
-		"'working_dir': './'":                                    "'working_dir': '/opt/app-root/src'",
-		"client.stop_job(submission_id)":                         "finished = False\\n\",\n\t\"while not finished:\\n\",\n\t\"    time.sleep(1)\\n\",\n\t\"    status = client.get_job_status(submission_id)\\n\",\n\t\"    finished = (status == \\\"SUCCEEDED\\\")\\n\",\n\t\"if finished:\\n\",\n\t\"    print(\\\"Job completed Successfully !\\\")\\n\",\n\t\"else:\\n\",\n\t\"    print(\\\"Job failed !\\\")\\n\",\n\t\"time.sleep(10)\\n",
+		"--ds-config=./deepspeed_configs/zero_3_offload_optim+param.json": fmt.Sprintf("--ds-config=./%s \\\"\\n\",\n\t\"               \\\"--lora-config=./lora.json \\\"\\n\",\n\t\"               \\\"--as-test", modelConfigFile),
+		"--batch-size-per-device=32":                                      "--batch-size-per-device=6",
+		"--eval-batch-size-per-device=32":                                 "--eval-batch-size-per-device=6",
+		"'pip': 'requirements.txt'":                                       "'pip': '/opt/app-root/src/requirements.txt'",
+		"'working_dir': './'":                                             "'working_dir': '/opt/app-root/src'",
+		"client.stop_job(submission_id)":                                  "finished = False\\n\",\n\t\"while not finished:\\n\",\n\t\"    time.sleep(1)\\n\",\n\t\"    status = client.get_job_status(submission_id)\\n\",\n\t\"    finished = (status == \\\"SUCCEEDED\\\")\\n\",\n\t\"if finished:\\n\",\n\t\"    print(\\\"Job completed Successfully !\\\")\\n\",\n\t\"else:\\n\",\n\t\"    print(\\\"Job failed !\\\")\\n\",\n\t\"time.sleep(10)\\n",
 	}
 
 	updatedNotebookContent := string(ReadFileExt(test, workingDirectory+"/../../examples/ray-finetune-llm-deepspeed/ray_finetune_llm_deepspeed.ipynb"))
 	for oldValue, newValue := range requiredChangesInNotebook {
 		updatedNotebookContent = strings.Replace(updatedNotebookContent, oldValue, newValue, -1)
 	}
 	updatedNotebook := []byte(updatedNotebookContent)
+	os.WriteFile("demo.ipynb", updatedNotebook, 0644)
 
 	// Test configuration
 	jupyterNotebookConfigMapFileName := "ray_finetune_llm_deepspeed.ipynb"
@@ -87,7 +93,7 @@ func rayFinetuneLlmDeepspeed(t *testing.T, numGpus int) {
 		"requirements.txt":               ReadFileExt(test, workingDirectory+"/../../examples/ray-finetune-llm-deepspeed/requirements.txt"),
 		"create_dataset.py":              ReadFileExt(test, workingDirectory+"/../../examples/ray-finetune-llm-deepspeed/create_dataset.py"),
 		"lora.json":                      ReadFileExt(test, workingDirectory+"/../../examples/ray-finetune-llm-deepspeed/lora_configs/lora.json"),
-		"zero_3_llama_2_7b.json":         ReadFileExt(test, workingDirectory+"/../../examples/ray-finetune-llm-deepspeed/deepspeed_configs/zero_3_llama_2_7b.json"),
+		modelConfigFile:                  ReadFileExt(test, fmt.Sprintf(workingDirectory+"/../../examples/ray-finetune-llm-deepspeed/deepspeed_configs/%s", modelConfigFile)),
 		"utils.py":                       ReadFileExt(test, workingDirectory+"/../../examples/ray-finetune-llm-deepspeed/utils.py"),
 	}
 
@@ -120,7 +126,7 @@ func rayFinetuneLlmDeepspeed(t *testing.T, numGpus int) {
 
 	// Initialise raycluster client to interact with raycluster to get rayjob details using REST-API
 	dashboardUrl := GetDashboardUrl(test, namespace, rayCluster)
-	rayClusterClientConfig := RayClusterClientConfig{Address: dashboardUrl.String(), Client: nil, SkipTlsVerification: true}
+	rayClusterClientConfig := RayClusterClientConfig{Address: dashboardUrl.String(), Client: nil, InsecureSkipVerify: true}
 	rayClient, err := NewRayClusterClient(rayClusterClientConfig, test.Config().BearerToken)
 	if err != nil {
 		test.T().Errorf("%s", err)