Skip to content
This repository was archived by the owner on Oct 31, 2023. It is now read-only.

G-XLT performance for MLQA #11

Open
wasiahmad opened this issue Mar 19, 2020 · 0 comments
Open

G-XLT performance for MLQA #11

wasiahmad opened this issue Mar 19, 2020 · 0 comments

Comments

@wasiahmad
Copy link

wasiahmad commented Mar 19, 2020

I am trying to reproduce the results presented in table 6 of the paper for generalized XLT using M-BERT.

Screen Shot 2020-03-18 at 5 49 06 AM

I have done the following.

  • Fine-tuned M-BERT using only the SQuAD 1.1 training dataset and validated on MLQA English dev dataset.
  • I have used the window approach as used by the BERT authors (Extracting features on for long sequences / SQuAD google-research/bert#66) for long sequences. I set the maximum sequence length to 384, doc stride to 128.
  • I considered the maximum answer length = 30.
  • During fine-tuning, I used the following setting.
learning rate = 5e-5
warmup_steps = 0
epochs = 3
gradient_accumulation_steps = 1
grad_clipping = 1.0

I got the following result. As you can see the performance is very poor particularly for Hindi and Vietnamese language. I think a different inference algorithm is used in your work. Is it possible to briefly explain what you did during inference?

bert

Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.
Labels
None yet
Projects
None yet
Development

No branches or pull requests

1 participant