Bert Base Uncased Squadv1 X1.84 F88.7 D36 Hybrid Filled V1

🚀 BERT-base uncased model fine-tuned on SQuAD v1

This model addresses the question-answering task by leveraging the power of the BERT architecture. Fine-tuned on the SQuAD v1 dataset, it offers improved performance and efficiency compared to the original dense model.

🚀 Quick Start

To use this model, first install the nn_pruning library. Then, you can utilize the transformers library with a simple optimization step.

Installation

pip install nn_pruning

Usage Example

from transformers import pipeline
from nn_pruning.inference_model_patcher import optimize_model

qa_pipeline = pipeline(
    "question-answering",
    model="madlag/bert-base-uncased-squadv1-x1.84-f88.7-d36-hybrid-filled-v1",
    tokenizer="madlag/bert-base-uncased-squadv1-x1.84-f88.7-d36-hybrid-filled-v1"
)

print("/home/lagunas/devel/hf/nn_pruning/nn_pruning/analysis/tmp_finetune parameters: 218.0M")
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)

✨ Features

• Pruning for Efficiency: The linear layers contain 36.0% of the original weights, and the model contains 50.0% of the original weights overall. This leads to a 1.84x speedup compared to the dense model during evaluation.
• Accuracy Improvement: With an F1 score of 88.72, it has a 0.22 gain in F1 compared to the dense version.
• Case-Insensitive: The model does not distinguish between different cases in English.

📚 Documentation

Fine-Pruning details

This model was fine-tuned from the HuggingFace model checkpoint on SQuAD1.1, and distilled from the model csarron/bert-base-uncased-squad-v1.

A side-effect of the block pruning is that some of the attention heads are completely removed: 48 heads were removed out of a total of 144 (33.3%).

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

🔧 Technical Details

The pruning method used in this model leads to structured matrices. To visualize them, you can hover on the plot below to see the non-zero/zero parts of each matrix.

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

📄 License

This model is released under the MIT license.