Datasets:

Modalities:
Text
Formats:
json
Languages:
English
ArXiv:
Libraries:
Datasets
pandas
License:

Lost access to dataset

#3
by theSLWayne - opened

I've been trying to use this dataset to fine-tune an OPT model. I use the following code to load the dataset:

import transformers
from datasets import load_dataset
data = load_dataset("databricks/databricks-dolly-15k", use_auth_token=True)

My setup is as follows:

Google Colab Free
transformers == '4.28.1'
datasets == '2.12.0'
Python version == 3.10.11

I have been able to load the dataset without any issues yesterday(3/5/2023), but I don't have any access today. I have used my credentials to login using notebook_login() function, but I still did not have access to the dataset.

I tried passing my access token to "use_auth_token" parameter of load_dataset function but it did not work either.

Is there something wrong or am I doing something wrong? Thanks in advance.

Databricks org

Yep I think it's an HF problem: https://github.com/huggingface/datasets-server/issues/1137
I reported it: https://github.com/huggingface/datasets-server/issues/1139
I am going to try a few things today to work around, and will re-upload the data to github if I have to.
(The file is still there, it's just the previous commit of the dataset, and that seems to be accessible)

Databricks org

I think this was a git lfs issue possibly related to the pull request. I've restored the original version from a few commits ago and it seems to be working now.

matthayes changed discussion status to closed

Yes it is working. Apologies for the late response.

Sign up or log in to comment