ppai/timeseries-classification: use smaller batch size for tests (GoogleCloudPlatform#6799)

* ppai/image-classification: add exponential retries to download file

* decrease batch size for tests

* print the job url if it fails

Author: David Cavazos

people-and-planet-ai/timeseries-classification/e2e_test.py

@@ -259,7 +259,7 @@ def create_datasets(
         time.sleep(POLL_INTERVAL_SEC)
 
     logging.info(f"Datasets job finished with status {status}")
-    assert status == DATAFLOW_SUCCESS_STATE
+    assert status == DATAFLOW_SUCCESS_STATE, f"job_url: {job_url}"
     yield job_id
 
 
@@ -268,7 +268,7 @@ def train_model(service_url: str, access_token: str, create_datasets: str) -> st
     response = requests.post(
         url=f"{service_url}/train-model",
         headers={"Authorization": f"Bearer {access_token}"},
-        json={"train_epochs": 100},
+        json={"train_epochs": 10, "batch_size": 8},
     ).json()
     logging.info(f"train_model response: {response}")
 
@@ -295,7 +295,7 @@ def train_model(service_url: str, access_token: str, create_datasets: str) -> st
         time.sleep(POLL_INTERVAL_SEC)
 
     logging.info(f"Model job finished with status {status}")
-    assert status == VERTEX_AI_SUCCESS_STATE
+    assert status == VERTEX_AI_SUCCESS_STATE, f"job_url: {job_url}"
     yield job_id
 
 

people-and-planet-ai/timeseries-classification/local_test.py

@@ -96,8 +96,8 @@ def test_e2e_local() -> None:
             model_dir=model_dir,
             tensorboard_dir=tensorboard_dir,
             checkpoint_dir=checkpoint_dir,
-            train_epochs=10,
-            batch_size=32,
+            train_epochs=2,
+            batch_size=8,
         )
         assert os.listdir(model_dir), "no model files found"
         assert os.listdir(tensorboard_dir), "no tensorboard files found"