diff --git a/services/rows/src/rows/routes/rows.py b/services/rows/src/rows/routes/rows.py index b01c4f2b52..ac3460b96d 100644 --- a/services/rows/src/rows/routes/rows.py +++ b/services/rows/src/rows/routes/rows.py @@ -37,6 +37,7 @@ "halabi2016/arabic_speech_corpus" ] # for testing +EXCLUDED_COLUMNS = {"vae_latents", "t5_prompt_embeds"} def create_rows_endpoint( cached_assets_storage_client: StorageClient, @@ -106,6 +107,7 @@ async def rows_endpoint(request: Request) -> Response: except TooBigRows as err: raise TooBigContentError(str(err)) from None with StepProfiler(method="rows_endpoint", step="transform to a list"): + pa_table = pa_table.drop([col for col in pa_table.column_names if col in EXCLUDED_COLUMNS]) response = await create_response( dataset=dataset, revision=revision,