Pass tuple in gen_kwargs to avoid ambiguous sharding error

ami.py

@@ -396,7 +396,7 @@ class AMI(datasets.GeneratorBasedBuilder):
                 gen_kwargs={
                     "annotation_path": annotation_path,
                     "samples_paths": train_path,
-                    "ids": _TRAIN_SAMPLE_IDS,
+                    "ids": tuple(_TRAIN_SAMPLE_IDS),
                     "verification_ids": train_ids,
                 },
             ),
@@ -405,7 +405,7 @@ class AMI(datasets.GeneratorBasedBuilder):
                 gen_kwargs={
                     "annotation_path": annotation_path,
                     "samples_paths": validation_path,
-                    "ids": _VALIDATION_SAMPLE_IDS,
+                    "ids": tuple(_VALIDATION_SAMPLE_IDS),
                     "verification_ids": validation_ids,
                 },
             ),
@@ -414,7 +414,7 @@ class AMI(datasets.GeneratorBasedBuilder):
                 gen_kwargs={
                     "annotation_path": annotation_path,
                     "samples_paths": eval_path,
-                    "ids": _EVAL_SAMPLE_IDS,
+                    "ids": tuple(_EVAL_SAMPLE_IDS),
                     "verification_ids": eval_ids,
                 },
             ),
@@ -501,7 +501,7 @@ class AMI(datasets.GeneratorBasedBuilder):
             sample_paths = samples_paths[num_audios * i : num_audios * (i + 1)]
             sample_verification_id = set(verification_ids[num_audios * i : num_audios * (i + 1)])
 
-            # make sure that multi microphone samples are correctly atttributed to labels
+            # make sure that multi microphone samples are correctly attributed to labels
             if len(sample_verification_id) > 1 or next(iter(sample_verification_id)) != _id:
                 raise ValueError(
                     f"Incorrect dataset generation. The files {sample_paths} of id {_id} have incorrect verification_ids {sample_verification_id}."