search

googlecolab / colabtools

Python libraries for Google Colaboratory
Apache License 2.0
2.12k stars 692 forks source link

No longer able to access remote Deeplake datasets #4465

Open cjharrington85 opened 3 months ago

cjharrington85 commented 3 months ago

Describe the current behavior I was able to use the Deeplake Spoken MNIST dataset (https://datasets.activeloop.ai/docs/ml/datasets/free-spoken-digit-dataset-fsdd/) in Colab up until today. For some reason, the service is being blocked when I try to use ds=deeplake.load("hub://activeloop/spoken_mnist"). It works fine when I do the same within a python environment on my laptop.

Describe the expected behavior When I use the command ds=deeplake.load("hub://activeloop/spoken_mnist"), the model should download and work with the rest of my script.

What web browser you are using Chrome

Additional context Here's a copy of my project: https://colab.research.google.com/drive/1qWGufDVgs9OlkB9rAcUzIlV6MupWPsQA?usp=sharing. Please let me know if you require access to it.

cperry-goog commented 3 months ago

We've been unable to download any dataset even when we install older versions of deeplake - we think this is a deeplake bug; can you contact them and point at this?

cjharrington85 commented 3 months ago

@cperry-goog, how might this be a deeplake bug if I can download their datasets locally on my personal machine? Several others have verified that they can as well.

cjharrington85 commented 3 months ago

I was able to use this as a workaround for the issue:

with open('/etc/resolv.conf', 'w') as file: file.write("nameserver 8.8.8.8")

cperry-goog commented 2 months ago

b/332898881 looking

Straky0 commented 2 months ago

I had the same issue, but with your advises I found a solution. However I discover that the images tensors in the dataset cannot be converted to numpy arrays: I've got the error "Unable to read sample at index 1562 from chunk 'images/chunks/5a963182' in tensor images".

My code is simply: print(train_ds['images'][0].numpy())

Does anyone had the same problem? Is there a solution? Thanks in advance