search

EduMUNI / reflection-classification

Apache License 2.0
2 stars 0 forks source link

Typos in model readme #1

Open paola-md opened 1 year ago

paola-md commented 1 year ago

In reflection-classification/models/README.md, the link in the gdown snipet should be updated.

Original line: gdown https://drive.google.com/uc?id=1Sv0OpLyi13HA5miRxbi5HNGmziJ0yQwt -O xlm-roberta-large-confidence4+-cs.zip

Proposed line: gdown https://drive.google.com/uc?id=1mMzlBz1Sipg8eOYMTF6DaqyzlBtjHOuo -O xlm-roberta-large-confidence4+-cs.zip

In addition, the unzip file also has a typo. The original line is: unzip xlm-roberta-large-confidence4+.zip -d models

Given that the name is "xlm-roberta-large-confidence4+-cs.zip", the instruction should be: unzip xlm-roberta-large-confidence4+-cs.zip -d models

paola-md commented 1 year ago

Here is a quick colab implementation that uses the english model to classify new sentences. Minor fixes:

stefanik12 commented 1 year ago

Hi Paola, thank you for your review and proposals for the refresh. I was wondering if you could open a Pull Request with these changes?

After acceptance, GitHub would list you among the project Contributors, so you'd also get some recognition of your work.

stefanik12 commented 1 year ago

Hi Paola, I've updated the main and the models' README according to your suggestions.

Additionally, I've uploaded the main referenced model to HuggingFace with your suggested usage.

I could not reproduce the error from loading the tokenizer on the most recent transformers==4.30.2. But I believe similar compatibility issues might be fixed by simply using the model directly from the HuggingFace hub, as suggested now.

Thanks for your help!