0

I am training a bert model using a TPU vm on GCP. I want to use my bucket as the Datasets library Cache filepath. I have followed instructions from https://cloud.google.com/tpu/docs/tutorials/bert-2.x and set my bucket link in the HF_DATASETS_CACHE environment variable.

But I am getting this error : Exception in device=TPU:0: [Errno 2] No such file or directory: 'gs:/darijabert_bucket/datasets_Cache

my HF_DATASETS_CACHE is set to gs://darijabert_bucket/datasets_Cache

Thanks

kamel gaanoun
  • 13
  • 3

1 Answers1

0

The error seems to suggest that the specified bucket does not exist. Does the bucket gs:/darijabert_bucket/datasets_Cache exist in your project and in the correct region? You can double check to see if it exists by checking it in the GCP console.

Also, make sure that the bucket exists in the same region as your TPU VM, as the tutorial states:

The bucket location must be in the same region as your Compute Engine (VM) and your Cloud TPU node.
Wonjoo Lee
  • 99
  • 3
  • 8
  • Thanks. My bucket exists on my project and iti is Multi-region whereas the tpu vm is on europe-west4-a. You think Multi-region is not ok and it must be in europe-west4-a? – kamel gaanoun Dec 30 '21 at 10:55
  • Hmm, I think multi-region should be fine. With that said, I'd recommend you to still try to create a bucket in `europe-west4-a` and try again. – Wonjoo Lee Jan 01 '22 at 05:45