madlag/bert-large-uncased-wwm-squadv2-x2.63-f82.6-d16-hybrid-v1

bert-large-uncased-whole-word-masking model fine-tuned on SQuAD v2

This model was created using the nn_pruning python library: the linear layers contains 16.0% of the original weights.

The model contains 24.0% of the original weights overall (the embeddings account for a significant part of the model, and they are not pruned by this method).

With a simple resizing of the linear matrices it ran 2.63x as fast as bert-large-uncased-whole-word-masking on the evaluation. This is possible because the pruning method lead to structured matrices: to visualize them, hover below on the plot to see the non-zero/zero parts of each matrix.

In terms of accuracy, its F1 is 82.57, compared with 85.85 for bert-large-uncased-whole-word-masking, a F1 drop of 3.28.

Fine-Pruning details

This model was fine-tuned from the HuggingFace model checkpoint on SQuAD2.0, and distilled from the model madlag/bert-large-uncased-whole-word-masking-finetuned-squadv2. This model is case-insensitive: it does not make a difference between english and English.

A side-effect of the block pruning is that some of the attention heads are completely removed: 190 heads were removed on a total of 384 (49.5%). Here is a detailed view on how the remaining heads are distributed in the network after pruning.

Details of the SQuAD1.1 dataset

Fine-tuning

• 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

Results

Pytorch model file size: 1084MB (original BERT: 1228.0MB)

{
    "HasAns_exact": 74.8144399460189,
    "HasAns_f1": 80.555306012496,
    "HasAns_total": 5928,
    "NoAns_exact": 84.57527333894029,
    "NoAns_f1": 84.57527333894029,
    "NoAns_total": 5945,
    "best_exact": 79.70184452118251,
    "best_exact_thresh": 0.0,
    "best_f1": 82.56816761071966,
    "best_f1_thresh": 0.0,
    "exact": 79.70184452118251,
    "f1": 82.56816761071981,
    "total": 11873
}

Example Usage

Install nn_pruning: it contains the optimization script, which just pack the linear layers into smaller ones by removing empty rows/columns.

pip install nn_pruning

Then you can use the transformers library almost as usual: you just have to call optimize_model when the pipeline has loaded.

from transformers import pipeline
from nn_pruning.inference_model_patcher import optimize_model

qa_pipeline = pipeline(
    "question-answering",
    model="madlag/bert-large-uncased-wwm-squadv2-x2.63-f82.6-d16-hybrid-v1",
    tokenizer="madlag/bert-large-uncased-wwm-squadv2-x2.63-f82.6-d16-hybrid-v1"
)

print("bert-large-uncased-whole-word-masking parameters: 445.0M")
print(f"Parameters count (includes only head pruning, not feed forward pruning)={int(qa_pipeline.model.num_parameters() / 1E6)}M")
qa_pipeline.model = optimize_model(qa_pipeline.model, "dense")

print(f"Parameters count after complete optimization={int(qa_pipeline.model.num_parameters() / 1E6)}M")
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?",
})
print("Predictions", predictions)