psnonis / finbert Goto Github PK

hi, I would like to do a pre-training of finbert model starting from scratch - on Pre2K and Prime datasets (SEC filings) as mentioned in the paper. Can someone point me to where I can find the training data?
The link mentioned in the paper http://people.ischool.berkeley.
edu/˜khanna/fin10-K does not work...

I am trying to load the pretrained model at pred/FinBERT-Pre2K_128MSL-500K and trying to run the basic task of SST-2 (sentiment classification) using run_glue.py (https://huggingface.co/transformers/examples.html#glue).

But I run into the following error:
OSError: Model name '/data/ftm/xgb_regr/FinBERT/pred/FinBERT-Pre2K_128MSL-250K' was not found in model name list (bert-base-uncased, bert-large-uncased, bert-base-cased, bert-large-cased, bert-base-multilingual-uncased, bert-base-multilingual-cased, bert-base-chinese, bert-base-german-cased, bert-large-uncased-whole-word-masking, bert-large-cased-whole-word-masking, bert-large-uncased-whole-word-masking-finetuned-squad, bert-large-cased-whole-word-masking-finetuned-squad, bert-base-cased-finetuned-mrpc, bert-base-german-dbmdz-cased, bert-base-german-dbmdz-uncased). We assumed '/data/ftm/xgb_regr/FinBERT/pred/FinBERT-Pre2K_128MSL-250K' was a path or url to a configuration file named config.json or a directory containing such a file but couldn't find any such file at this path or url.

Also, since this seems to be trained using TF, I was wondering if I can use PyTorch to load it.

Thanks.

Hi,

great project. I just read your paper and found it interesting to compare your results to the paper of https://arxiv.org/pdf/1908.10063.pdf

Do you plan on releasing the pre-trained models for research?
Thank you very much for making the training data available.

All the best!

In the readme:

Combo Pre-trained continued from original BERT on 2017, 2018, 2019 SEC 10K dataset

but in the paper:

train a Combo Model on top of the last checkpoint of BERT-Base Uncased. This training was
done in parallel with FinBERT Prime, using SEC2019 for the first 250,000 and using SEC1999 for the last 250,000.

What is correct?

In the google drive link which contains the pre-trained models the vocab.txt file is missing.