show the traceback of the cause, it's clearer (#3042)

* show the traceback of the cause, it's clearer * retry for backfill
huggingface · Aug 22, 2024 · fc2f2b1 · fc2f2b1
1 parent e7fa2e0
commit fc2f2b1
Show file tree

Hide file tree

Showing 2 changed files with 4 additions and 1 deletion.
diff --git a/libs/libcommon/src/libcommon/constants.py b/libs/libcommon/src/libcommon/constants.py
@@ -43,6 +43,7 @@
 LARGE_MAX_FAILED_RUNS = 30  # for errors that should not be permanent
 MAX_FAILED_RUNS_PER_ERROR_CODE = {
     # default
+    "DatasetGenerationError": DEFAULT_MAX_FAILED_RUNS,  # <- 20240822: to recompute all of them in the next backfill
     "ConfigNamesError": DEFAULT_MAX_FAILED_RUNS,  # <- 20240822: to recompute all of them in the next backfill
     "RetryableConfigNamesError": DEFAULT_MAX_FAILED_RUNS,
     "ConnectionError": DEFAULT_MAX_FAILED_RUNS,

diff --git a/services/worker/src/worker/job_runners/config/parquet_and_info.py b/services/worker/src/worker/job_runners/config/parquet_and_info.py
@@ -1531,7 +1531,9 @@ def compute_config_parquet_and_info_response(
     except datasets.exceptions.DatasetGenerationCastError as err:
         raise DatasetGenerationCastError("The dataset generation failed because of a cast error", cause=err) from err
     except datasets.exceptions.DatasetGenerationError as err:
-        raise DatasetGenerationError("The dataset generation failed", cause=err) from err
+        if err.__cause__:
+            raise DatasetGenerationError("The dataset generation failed", cause=err.__cause__) from err
+        raise DatasetGenerationError("The dataset generation failed") from err
 
     raise_if_long_column_name(builder.info.features)