-
-
Notifications
You must be signed in to change notification settings - Fork 471
Open
Description
I have defined a Back Translation model the following way:
model = naw.BackTranslationAug(from_model_name="Helsinki-NLP/opus-mt-en-fr",
to_model_name="Helsinki-NLP/opus-mt-fr-en",
device="cuda",
batch_size=config.BATCH_SIZE,
max_length=1024)
I later augment a list of texts using:
augmented_texts = model.augment(list(texts))
From time to time I get this warning which makes the code fail:
/usr/local/src/pytorch/aten/src/ATen/native/cuda/Indexing.cu:649: indexSelectSmallIndex: block: [1,0,0], thread: [0,0,0] Assertion `srcIndex < srcSelectDimSize` failed.
This is the error:
RuntimeError: CUDA error: device-side assert triggered CUDA kernel errors might be asynchronously reported at some other API call,so the stacktrace below might be incorrect. For debugging consider passing CUDA_LAUNCH_BLOCKING=1
I do not know if I am defining the BackTranslationAug
model correctly. I used Helsinki-NLP
models as said in the documentation. Any help regarding this matter is appreciated.
Metadata
Metadata
Assignees
Labels
No labels