The numeracy-literacy from avi-jit

numeracy-literacy's Introduction

Hi I'm Avi 👋

🔭 I'm getting a PhD in Language Models, Natural Language Processing
🛠 I'm a part-time thinker, full-time tinkerer. I love building products.
💬 Ask me about NLP, Academia, PhD, books, movies, Los Angeles, and India 🇮🇳
📫 How to reach me: [email protected]
🌐 Personal webpage: https://avi-jit.github.io/

numeracy-literacy's People

Contributors

Stargazers

Watchers

numeracy-literacy's Issues

I cannot use huggingface dataset and I also failed to use other methods you linked.

I cannot use huggingface dataset using your example.

nice python train.py --batch-size 256 --gpus 0, --tsamples 100_000 --dsamples 10_000 --max_epochs 10 --enc exp --hidden 200 --accumulate_grad_batches 4 --seed 0 --dataset WC

So I downloaded datasets using your links.

Note that when loading from the Datasets library, numbers larger than sys.maxsize will be capped to avoid an overflow in PyArrow. For the uncapped version, you may download the json files directly for the train, dev, and test splits.

And naming problem occurred, I changed.

train.wiki.json
train.wiki.json.1 -> dev.wiki.json
train.wiki.json.2 -> test.wiki.json

But still, an error occurred.

Traceback (most recent call last):
  File "train.py", line 262, in <module>
    main()
  File "train.py", line 253, in main
    ckpt_cb = pl.callbacks.ModelCheckpoint(save_top_k=2, verbose=True, filepath='checkpoints/'+args.name+'/{epoch}')
TypeError: __init__() got an unexpected keyword argument 'filepath'

Recommend Projects