GlueDataset & GlueDataTrainingArguments - working on Colab

[LedaguenelArthur]

Hi everyone,

I'm currently trying to understand how to plugin my own data in the BioBERT model in PyTorch for relation extraction, and I see that the run_re.py script is using two utils for which I can't find documentation : GlueDataset & GlueDataTrainingArguments.

Does anyone know how to find the documentation for both these functions ? Or does anyone can briefly explain how they work ?

Thanks a lot, Best regards, Arthur Ledaguenel

[wonjininfo]

Hi Arthur!

Our repository is based on Transformers v3.0.0. Unfortunately, GlueDataset and GlueDataTrainingArguments are deleted in the current version of Transformers, v4. However, you can find those two utils from here ; Tag v3.0.0

Thank you for your interest in our work! Best, WonJin

[LedaguenelArthur]

Thank you very much for that quick answer !

I could not find any documentation on any of those two components on https://huggingface.co/transformers/v3.0.2/index.html for some reason... :/

I would like to plug in my own dataset into the neural network and I wondered how it could work with these two components ?

Would that be enough if I transform my data in the same format as GAD and split them into train.tsv, test.tsv and dev.tsv and feed it to the GlueDataset processor ?

Besides, I am working on Google Colab and thus facing problem when using the HfArgumentParser since i'm not executing the script from a shell passing the argument in the way :

• can I set up the sys.argv by hand to stick with your code ?
• I have tried to get arround this by setting up the dataclass by hand :

model_args = ModelArguments(config_name = "bert-base-cased", model_name_or_path = "dmis-lab/biobert-base-cased-v1.1") data_args = DataTrainingArguments(task_name = "SST-2", data_dir=DATA_DIR, max_seq_length=MAX_LENGTH, overwrite_cache=False) training_args = TrainingArguments(per_device_train_batch_size = BATCH_SIZE, save_steps = SAVE_STEPS, seed = SEED, do_train = True, do_predict = True, learning_rate = 5e-5, output_dir = OUTPUT_DIR, overwrite_output_dir = True)

but I get the following error :

[Errno 2] No such file or directory: './GAD/1/cached_train_BertTokenizer_128_sst-2.lock'

Is there a solution for that ? What am I doing wrong ?

I stay at your disposal if my question is unclear, Thank you again, Best regards, Arthur Ledaguenel

[clouwer]

Hi, Arthur

I got the same error: No such file or directory: '../data/RE/euadr/1/cached_train_BertTokenizerFast_128_sst-2.lock'

Is there a solution?

Thank you, Best regards, Clouwer

[Bharathi-A-7]

@clouwer I face the same error. Did you happen to find a solution?