bert-large-uncased-whole-word-masking-squad-fp32-0001

Use Case and High-Level Description

This is a BERT-large model pretrained on lower-cased English text using Whole-Word-Masking and fine-tuned on the SQuAD v1.1 training set (93.21% F1, 87.2% EM on the v1.1 dev set). The model performs question answering for English language; the input is a concatenated premise and question for the premise, and the output is the location of the answer to the question inside the premise. For details about the original floating-point model, check out BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding.

Tokenization occurs using the BERT tokenizer (see the demo code for implementation details) and the enclosed vocab.txt dictionary file. Input is to be lower-cased before tokenizing.

Specification

Metric Value
GOps 246.93
MParams 333.96
Source framework PyTorch*

Accuracy

The quality metrics were calculated on the SQuAD v1.1 dataset ("dev" split). Maximum sequence length is 384, maximum query length: 64, doc stride: 128, input is lower-cased.

Metric Value
F1 93.21%
Exact match (EM) 87.20%

Performance

Input

Input 0: A 1,384 sequence of tokens (integer values) representing the tokenized premise and question ("input_ids"). The sequence structure is as follows ([CLS], [SEP] and [PAD] should be replaced by corresponding token IDs as specified by the dictionary): [CLS] + tokenized question + [SEP] + tokenized premise of the question + [SEP] + ([PAD] tokens to pad to the maximum sequence length of 384)

Input 1: A 1,384 sequence of integer values representing the mask of valid values in the input ("input_mask"). The values of this input are are equal to: 1) 1 at positions corresponding to the [CLS] + tokenized question + [SEP] + tokenized premise of the question + [SEP] part of the Input 0 (i.e. all positions except those containing the [PAD] tokens), and 2) 0 at all other positions

Input 2: A 1,384 sequence of integer values representing the segmentation of the Input 0 into question and premise ("segment_ids"). The values are equal to: 1) 1 at positions corresponding to the tokenized premise of the question + [SEP] part of the Input 0, and 2) 0 at all other positions

Output

Output 0: The 1, 384 floating point-valued logit scores, where i-th value corresponds to the log-likelihood of the answer to the question starting at the i-th token position of the input.

Output 1: Same as Output 0, but represents the log-likelihoods of the answer ending at i-th token position.

Legal Information

[*] Other names and brands may be claimed as the property of others.