Create mlcompass_maxtext_gke dag (#485)

* Create mlcompass_maxtext_gke dag

* Add dag description

.github/requirements.txt

@@ -9,3 +9,4 @@ jsonlines
 tensorflow-cpu
 kubernetes
 pyarrow
+apache-airflow-providers-google

dags/mlcompass/maxtext_gke.py

@@ -0,0 +1,127 @@
+# Copyright 2024 Google LLC
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#      http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+"""This Airflow DAG runs a maxtext machine learning benchmark on a GKE cluster
+
+Usage:
+gcloud composer environments run ml-automation-solutions \
+  --project=cloud-ml-auto-solutions \
+  --location=us-central1 dags trigger \
+  -- \
+  mlcompass_maxtext_gke \
+  --conf={\\\"uuid\\\":\\\"abc\\\"} 70
+"""
+
+import datetime
+from airflow import models
+from airflow.decorators import task
+from airflow.providers.google.cloud.hooks.gcs import GCSHook
+from xlml.apis.xpk_cluster_config import XpkClusterConfig
+from dags import test_owner
+from dags.vm_resource import Project, XpkClusters
+from xlml.apis import gcp_config, metric_config, task as xlml_task, test_config
+import json
+
+
+def get_config_gke(
+    docker_image: str,
+    model_name: str,
+    base_output_directory: str,
+    task_owner: str = test_owner.ORTI_B,
+    cluster: XpkClusterConfig = XpkClusters.TPU_V4_8_MAXTEXT_CLUSTER,
+    time_out_in_min: int = 60,
+    num_slices: int = 1,
+    dataset_name: metric_config.DatasetOption = metric_config.DatasetOption.XLML_DATASET,
+    dataset_project: str = Project.CLOUD_ML_AUTO_SOLUTIONS.value,
+    composer_project: str = Project.CLOUD_ML_AUTO_SOLUTIONS.value,
+) -> xlml_task.XpkTask:
+  job_gcp_config = gcp_config.GCPConfig(
+      project_name=cluster.project,
+      zone=cluster.zone,
+      dataset_name=dataset_name,
+      dataset_project=dataset_project,
+      composer_project=composer_project,
+  )
+  job_test_config = test_config.TpuGkeTest(
+      test_config.Tpu(
+          version=cluster.device_version,
+          cores=cluster.core_count,
+      ),
+      test_name="maxtext",
+      run_model_cmds=[
+          f"source benchmark_run.sh;run {model_name} {base_output_directory}",
+      ],
+      set_up_cmds=None,
+      timeout=datetime.timedelta(minutes=time_out_in_min),
+      task_owner=task_owner,
+      num_slices=num_slices,
+      cluster_name=cluster.name,
+      docker_image=docker_image,
+  )
+  return xlml_task.XpkTask(
+      task_test_config=job_test_config,
+      task_gcp_config=job_gcp_config,
+  )
+
+
+with models.DAG(
+    dag_id="mlcompass_maxtext_gke",
+    schedule=None,
+    tags=["mlcompass", "maxtext"],
+    start_date=datetime.datetime(2024, 9, 1),
+    catchup=False,
+    params={
+        "uuid": "",
+    },
+    default_args={
+        "retries": 0,
+    },
+) as dag:
+
+  @task.python
+  def load_xlml_state(params: dict = None):
+    dag.log.info(params)
+    uuid = params["uuid"]
+    if not uuid:
+      raise RuntimeError("uuid is not set")
+    gcs_hook = GCSHook()
+    file_content = gcs_hook.download(
+        "mlcompass-jax-artifacts", f"xlml/{uuid}/xlml_state.json"
+    )
+    return json.loads(file_content)
+
+  @task.python
+  def get_docker_image_path(state: dict) -> str:
+    return state["docker_image_path"]
+
+  @task.python
+  def get_model_name(state: dict) -> str:
+    return state["model_name"]
+
+  @task.python
+  def get_base_output_directory(state: dict) -> str:
+    bucket = state["workdir_bucket"]
+    path = state["workdir_path"]
+    return f"gs://{bucket}/{path}"
+
+  xlml_state = load_xlml_state()
+  docker_image_path = get_docker_image_path(xlml_state)
+  model_name_arg = get_model_name(xlml_state)
+  base_output_directory_arg = get_base_output_directory(xlml_state)
+
+  default_benchmark = get_config_gke(
+      docker_image=docker_image_path,
+      model_name=model_name_arg,
+      base_output_directory=base_output_directory_arg,
+  ).run(skip_post_process=True)

xlml/apis/task.py

@@ -171,6 +171,7 @@ def run(
       gcs_location: Optional[airflow.XComArg] = None,
       use_vertex_tensorboard: bool = False,
       use_pathways: bool = False,
+      skip_post_process: bool = False,
   ) -> DAGNode:
     """Run a test job within a docker image.
 
@@ -187,7 +188,8 @@ def run(
       run_model, gcs_path = self.run_model(
           gcs_location, use_vertex_tensorboard, use_pathways
       )
-      run_model >> self.post_process(gcs_path)
+      if not skip_post_process:
+        run_model >> self.post_process(gcs_path)
 
     return group