search

CASIA-IVA-Lab / FLAP

[AAAI 2024] Fluctuation-based Adaptive Structured Pruning for Large Language Models
https://arxiv.org/abs/2312.11983
Apache License 2.0
35 stars 10 forks source link

Question about the wikitext2 data loader #3

Closed bokyeong1015 closed 9 months ago

bokyeong1015 commented 9 months ago

Thanks for your nice work. When running the sample script, I'm getting the following error message regarding the wikitext2 loader. Would you kindly check it?

File "/ssd2/bkkim/FLAP/lib/data.py", line 81, in get_wikitext2
    traindata = load_dataset('text', data_files='datasets/wikitext/wiki.train.raw', split="train")
FileNotFoundError: Unable to find '/ssd2/bkkim/FLAP/datasets/wikitext/wiki.train.raw'

By the way, I found a workaround for the issue by using the comment lines in the script. Is this an appropriate solution?

# traindata = load_dataset('wikitext', 'wikitext-2-raw-v1', split='train')
# testdata = load_dataset('wikitext', 'wikitext-2-raw-v1', split='test')

Thank you for your time.

an-yongqi commented 9 months ago

@bokyeong1015 Due to my server's inability to directly access HuggingFace for downloading datasets, I opted for local loading during my experimentation. You can replace it with the two lines of code commented in the script, which should resolve this issue.