Commit
•
043c6cc
1
Parent(s):
2d7620b
Fix RuntimeError: Sharding is ambiguous for this dataset (#4)
Browse files- Pass tuple in gen_kwargs to avoid ambiguous sharding error (81710454f9f6a612e8e466cb73051f67970918b5)
ami.py
CHANGED
@@ -396,7 +396,7 @@ class AMI(datasets.GeneratorBasedBuilder):
|
|
396 |
gen_kwargs={
|
397 |
"annotation_path": annotation_path,
|
398 |
"samples_paths": train_path,
|
399 |
-
"ids": _TRAIN_SAMPLE_IDS,
|
400 |
"verification_ids": train_ids,
|
401 |
},
|
402 |
),
|
@@ -405,7 +405,7 @@ class AMI(datasets.GeneratorBasedBuilder):
|
|
405 |
gen_kwargs={
|
406 |
"annotation_path": annotation_path,
|
407 |
"samples_paths": validation_path,
|
408 |
-
"ids": _VALIDATION_SAMPLE_IDS,
|
409 |
"verification_ids": validation_ids,
|
410 |
},
|
411 |
),
|
@@ -414,7 +414,7 @@ class AMI(datasets.GeneratorBasedBuilder):
|
|
414 |
gen_kwargs={
|
415 |
"annotation_path": annotation_path,
|
416 |
"samples_paths": eval_path,
|
417 |
-
"ids": _EVAL_SAMPLE_IDS,
|
418 |
"verification_ids": eval_ids,
|
419 |
},
|
420 |
),
|
@@ -501,7 +501,7 @@ class AMI(datasets.GeneratorBasedBuilder):
|
|
501 |
sample_paths = samples_paths[num_audios * i : num_audios * (i + 1)]
|
502 |
sample_verification_id = set(verification_ids[num_audios * i : num_audios * (i + 1)])
|
503 |
|
504 |
-
# make sure that multi microphone samples are correctly
|
505 |
if len(sample_verification_id) > 1 or next(iter(sample_verification_id)) != _id:
|
506 |
raise ValueError(
|
507 |
f"Incorrect dataset generation. The files {sample_paths} of id {_id} have incorrect verification_ids {sample_verification_id}."
|
|
|
396 |
gen_kwargs={
|
397 |
"annotation_path": annotation_path,
|
398 |
"samples_paths": train_path,
|
399 |
+
"ids": tuple(_TRAIN_SAMPLE_IDS),
|
400 |
"verification_ids": train_ids,
|
401 |
},
|
402 |
),
|
|
|
405 |
gen_kwargs={
|
406 |
"annotation_path": annotation_path,
|
407 |
"samples_paths": validation_path,
|
408 |
+
"ids": tuple(_VALIDATION_SAMPLE_IDS),
|
409 |
"verification_ids": validation_ids,
|
410 |
},
|
411 |
),
|
|
|
414 |
gen_kwargs={
|
415 |
"annotation_path": annotation_path,
|
416 |
"samples_paths": eval_path,
|
417 |
+
"ids": tuple(_EVAL_SAMPLE_IDS),
|
418 |
"verification_ids": eval_ids,
|
419 |
},
|
420 |
),
|
|
|
501 |
sample_paths = samples_paths[num_audios * i : num_audios * (i + 1)]
|
502 |
sample_verification_id = set(verification_ids[num_audios * i : num_audios * (i + 1)])
|
503 |
|
504 |
+
# make sure that multi microphone samples are correctly attributed to labels
|
505 |
if len(sample_verification_id) > 1 or next(iter(sample_verification_id)) != _id:
|
506 |
raise ValueError(
|
507 |
f"Incorrect dataset generation. The files {sample_paths} of id {_id} have incorrect verification_ids {sample_verification_id}."
|