gemini-code-assist[bot] commented on code in PR #36172: URL: https://github.com/apache/beam/pull/36172#discussion_r2359597878
########## sdks/python/apache_beam/ml/inference/vertex_ai_inference_it_test.py: ########## @@ -37,7 +37,7 @@ _INPUT = "gs://apache-beam-ml/testing/inputs/vertex_images/*/*.jpg" _OUTPUT_DIR = "gs://apache-beam-ml/testing/outputs/vertex_images" _FLOWER_ENDPOINT_ID = "5384055553544683520" -_LLM_ENDPOINT_ID = "9157860935048626176" +_LLM_ENDPOINT_ID = "1273985431406379008" Review Comment:  Since the test `test_vertex_ai_run_llm_text_classification` which used this constant is being removed, `_LLM_ENDPOINT_ID` is no longer used. It should be removed to avoid dead code. ########## sdks/python/apache_beam/examples/inference/vertex_ai_llm_text_classification.py: ########## @@ -21,6 +21,14 @@ model can be generated by fine tuning the text-bison model or another similar model (see https://cloud.google.com/vertex-ai/docs/generative-ai/models/tune-models#supervised-fine-tuning) + +**NOTE** +Google has deprecated non-Gemini LLMs and no longer supports querying them on +Vertex AI endpoints. Separately, the use of the Vertex AI Predict API is not +supported for Gemini models in favor of use of the google-genai API. As a +result, this example no longer works. To perform inference with remote LLMs +deployed on Google infrastructure, please see the GeminiModelHandler and +the gemini_text_classification.py example. Review Comment:  To improve clarity and help users find the recommended alternatives more easily, consider adding backticks for code elements and providing the module path for `GeminiModelHandler`. ```suggestion deployed on Google infrastructure, please see the `GeminiModelHandler` (in `apache_beam.ml.inference.gemini_inference`) and the `gemini_text_classification.py` example. ``` -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
