Regarding 'Tutorial: Fine-Tuning a Model on Your Own Data' error

[dkbs12]

Hello,
I'm reporting another error happened in trying on 'Tutorial: Fine-Tuning a Model on Your Own Data'.
When I was fine-tuning model as same as written in tutorial as below;

reader = FARMReader(model_name_or_path="distilbert-base-uncased-distilled-squad", use_gpu=True)
data_dir = "data/squad20"
reader.train(data_dir=data_dir, train_filename="dev-v2.0.json", use_gpu=True, n_epochs=1, save_dir="my_model")

I found the error message as below;

RuntimeError: stack expects each tensor to be equal size, but got [6, 2] at entry 0 and [5, 2] at entry 1

So, I decided to use the file 'train-v2.0.json' instead of 'dev-v2.0.json' and it works.
Did I do it right? I'm not confident.
By the way, I faced the similar case while I tried Running distillation later.
When I tried just as same as written in tutorial as below;

teacher = FARMReader(model_name_or_path="my_model", use_gpu=True)
student = FARMReader(model_name_or_path="huawei-noah/TinyBERT_General_6L_768D", use_gpu=True)
student.distil_intermediate_layers_from(teacher, data_dir=".", train_filename="augmented_dataset.json", use_gpu=True)
student.distil_prediction_layer_from(teacher, data_dir="data/squad20", train_filename="dev-v2.0.json", use_gpu=True)
student.save(directory="my_distilled_model")

Finally, I found the error message regarding 'student.distil_prediction_layer_from' as below;

RuntimeError: stack expects each tensor to be equal size, but got [5, 2] at entry 0 and [6, 2] at entry 2

I retried to use the file 'train-v2.0.json' instead of 'dev-v2.0.json' for student.distil_prediction_layer.
But, it's keep working over and over and never stops at this time.

Could you help me with this matter?
I'm sorry to bother you again.
Thanks.

[bilgeyucel]

Hey @dkbs12, thank you for reporting this. I can reproduce your error and will open an issue to fix this. Until then, you can try downgrading the Haystack version to 1.16.1 by importing with pip install farm-haystack[colab]==1.16.1 in the colab notebook. Let us know if you encounter any other bugs 🙌

Issue is opened: #5114