chore: add experimental multimodal model tests

tests/system/small/ml/test_llm.py

@@ -17,6 +17,7 @@
 import pandas as pd
 import pytest
 
+import bigframes
 from bigframes import exceptions
 from bigframes.ml import core, llm
 import bigframes.pandas as bpd
@@ -260,6 +261,44 @@ def test_text_embedding_generator_multi_cols_predict_success(
     assert len(pd_df["ml_generate_embedding_result"][0]) == 768
 
 
+def test_create_load_multimodal_embedding_generator_model(
+    dataset_id, session, bq_connection
+):
+    bigframes.options.experiments.blob = True
+
+    mm_embedding_model = llm.MultimodalEmbeddingGenerator(
+        connection_name=bq_connection, session=session
+    )
+    assert mm_embedding_model is not None
+    assert mm_embedding_model._bqml_model is not None
+
+    # save, load to ensure configuration was kept
+    reloaded_model = mm_embedding_model.to_gbq(
+        f"{dataset_id}.temp_mm_model", replace=True
+    )
+    assert f"{dataset_id}.temp_mm_model" == reloaded_model._bqml_model.model_name
+    assert reloaded_model.connection_name == bq_connection
+
+
+@pytest.mark.flaky(retries=2)
+def test_multimodal_embedding_generator_predict_default_params_success(
+    images_mm_df, session, bq_connection
+):
+    bigframes.options.experiments.blob = True
+
+    text_embedding_model = llm.MultimodalEmbeddingGenerator(
+        connection_name=bq_connection, session=session
+    )
+    df = text_embedding_model.predict(images_mm_df).to_pandas()
+    utils.check_pandas_df_schema_and_index(
+        df,
+        columns=utils.ML_MULTIMODAL_GENERATE_EMBEDDING_OUTPUT,
+        index=2,
+        col_exact=False,
+    )
+    assert len(df["ml_generate_embedding_result"][0]) == 1408
+
+
 @pytest.mark.parametrize(
     "model_name",
     (
@@ -273,6 +312,9 @@ def test_text_embedding_generator_multi_cols_predict_success(
         "gemini-2.0-flash-exp",
     ),
 )
+@pytest.mark.flaky(
+    retries=2
+)  # usually create model shouldn't be flaky, but this one due to the limited quota of gemini-2.0-flash-exp.
 def test_create_load_gemini_text_generator_model(
     dataset_id, model_name, session, bq_connection
 ):
@@ -375,6 +417,36 @@ def test_gemini_text_generator_multi_cols_predict_success(
     )
 
 
+@pytest.mark.parametrize(
+    "model_name",
+    (
+        "gemini-1.5-pro-001",
+        "gemini-1.5-pro-002",
+        "gemini-1.5-flash-001",
+        "gemini-1.5-flash-002",
+        "gemini-2.0-flash-exp",
+    ),
+)
+@pytest.mark.flaky(retries=2)
+def test_gemini_text_generator_multimodal_input(
+    images_mm_df: bpd.DataFrame, model_name, session, bq_connection
+):
+    bigframes.options.experiments.blob = True
+
+    gemini_text_generator_model = llm.GeminiTextGenerator(
+        model_name=model_name, connection_name=bq_connection, session=session
+    )
+    pd_df = gemini_text_generator_model.predict(
+        images_mm_df, prompt=["Describe", images_mm_df["blob_col"]]
+    ).to_pandas()
+    utils.check_pandas_df_schema_and_index(
+        pd_df,
+        columns=utils.ML_GENERATE_TEXT_OUTPUT + ["blob_col"],
+        index=2,
+        col_exact=False,
+    )
+
+
 # Overrides __eq__ function for comparing as mock.call parameter
 class EqCmpAllDataFrame(bpd.DataFrame):
     def __eq__(self, other):

tests/system/utils.py

@@ -56,6 +56,14 @@
     "ml_generate_embedding_status",
     "content",
 ]
+ML_MULTIMODAL_GENERATE_EMBEDDING_OUTPUT = [
+    "ml_generate_embedding_result",
+    "ml_generate_embedding_status",
+    # start and end sec depend on input format. Images and videos input will contain these 2.
+    "ml_generate_embedding_start_sec",
+    "ml_generate_embedding_end_sec",
+    "content",
+]
 
 
 def skip_legacy_pandas(test):