BERT-Tiny (5) fine-tuned on SQuAD v2
BERT-Tiny created by Google Research and fine-tuned on SQuAD 2.0 for Q&A downstream task.
Mode size (after training): 24.33 MB
Details of BERT-Tiny and its 'family' (from their documentation)
Released on March 11th, 2020
This is model is a part of 24 smaller BERT models (English only, uncased, trained with WordPiece masking) referenced in Well-Read Students Learn Better: On the Importance of Pre-training Compact Models.
The smaller BERT models are intended for environments with restricted computational resources. They can be fine-tuned in the same manner as the original BERT models. However, they are most effective in the context of knowledge distillation, where the fine-tuning labels are produced by a larger and more accurate teacher.
Details of the downstream task (Q&A) - Dataset
SQuAD2.0 combines the 100,000 questions in SQuAD1.1 with over 50,000 unanswerable questions written adversarially by crowdworkers to look similar to answerable ones. To do well on SQuAD2.0, systems must not only answer questions when possible, but also determine when no answer is supported by the paragraph and abstain from answering.
| Dataset | Split | # samples |
|---|---|---|
| SQuAD2.0 | train | 130k |
| SQuAD2.0 | eval | 12.3k |
Model training
The model was trained on a Tesla P100 GPU and 25GB of RAM. The script for fine tuning can be found here
Results:
| Metric | # Value |
|---|---|
| EM | 57.12 |
| F1 | 60.86 |
| Model | EM | F1 score | SIZE (MB) |
|---|---|---|---|
| bert-tiny-finetuned-squadv2 | 48.60 | 49.73 | 16.74 |
| bert-tiny-5-finetuned-squadv2 | 57.12 | 60.86 | 24.34 |
Model in action
Fast usage with pipelines:
from transformers import pipeline
qa_pipeline = pipeline(
"question-answering",
model="mrm8488/bert-tiny-5-finetuned-squadv2",
tokenizer="mrm8488/bert-tiny-5-finetuned-squadv2"
)
qa_pipeline({
'context': "Manuel Romero has been working hardly in the repository hugginface/transformers lately",
'question': "Who has been working hard for hugginface/transformers lately?"
})
Created by Manuel Romero/@mrm8488 | LinkedIn
Made with ♥ in Spain