BERT-base uncased model fine-tuned on SQuAD v1

This model is block sparse: the linear layers contains 31.7% of the original weights.

The model contains 47.0% of the original weights overall.

The training use a modified version of Victor Sanh Movement Pruning method.

That means that with the block-sparse runtime it ran 1.12x faster than an dense networks on the evaluation, at the price of some impact on the accuracy (see below).

This model was fine-tuned from the HuggingFace BERT base uncased checkpoint on SQuAD1.1, and distilled from the equivalent model csarron/bert-base-uncased-squad-v1. This model is case-insensitive: it does not make a difference between english and English.

Pruning details

A side-effect of the block pruning is that some of the attention heads are completely removed: 80 heads were removed on a total of 144 (55.6%).

Here is a detailed view on how the remaining heads are distributed in the network after pruning.

Pruning details

Density plot


Dataset Split # samples
SQuAD1.1 train 90.6K
SQuAD1.1 eval 11.1k


  • Python: 3.8.5

  • Machine specs:

Memory: 64 GiB
GPUs: 1 GeForce GTX 3090, with 24GiB memory
GPU driver: 455.23.05, CUDA: 11.1


Pytorch model file size: 355M (original BERT: 438M)

Metric # Value # Original (Table 2)
EM 79.04 80.8
F1 86.70 88.5

Example Usage

from transformers import pipeline

qa_pipeline = pipeline(

predictions = qa_pipeline({
    'context': "Frédéric François Chopin, born Fryderyk Franciszek Chopin (1 March 1810 – 17 October 1849), was a Polish composer and virtuoso pianist of the Romantic era who wrote primarily for solo piano.",
    'question': "Who is Frederic Chopin?",

Downloads last month
Hosted inference API
Question Answering