roberta-base for QA

这是 roberta-base 模型，使用 SQuAD2.0 数据集进行了微调。该模型是针对问题回答任务进行训练的，包括不可回答的问题。

概述

语言模型：roberta-base 语言：英语下游任务：提取式问答训练数据：SQuAD 2.0 评估数据：SQuAD 2.0 代码：请参见 an example QA pipeline on Haystack 基础设施: 4x特斯拉v100

超参数

batch_size = 96
n_epochs = 2
base_LM_model = "roberta-base"
max_seq_len = 386
learning_rate = 3e-5
lr_schedule = LinearWarmup
warmup_proportion = 0.2
doc_stride=128
max_query_length=64

使用蒸馏模型

请注意，我们还发布了一个名为 deepset/tinyroberta-squad2 的蒸馏版本的该模型。蒸馏模型具有相当的预测质量，并且运行速度是基础模型的两倍。

用法

在Haystack中

Haystack是由deepset开发的NLP框架。您可以在Haystack流水线中使用该模型来在大规模文档上进行问答。要在 Haystack 中加载该模型：

reader = FARMReader(model_name_or_path="deepset/roberta-base-squad2")
# or 
reader = TransformersReader(model_name_or_path="deepset/roberta-base-squad2",tokenizer="deepset/roberta-base-squad2")

如果要查看roberta-base-squad2用于问题回答的完整示例，请查看 Tutorials in Haystack Documentation

在Transformers中

from transformers import AutoModelForQuestionAnswering, AutoTokenizer, pipeline

model_name = "deepset/roberta-base-squad2"

# a) Get predictions
nlp = pipeline('question-answering', model=model_name, tokenizer=model_name)
QA_input = {
    'question': 'Why is model conversion important?',
    'context': 'The option to convert models between FARM and transformers gives freedom to the user and let people easily switch between frameworks.'
}
res = nlp(QA_input)

# b) Load model & tokenizer
model = AutoModelForQuestionAnswering.from_pretrained(model_name)
tokenizer = AutoTokenizer.from_pretrained(model_name)

性能

使用 official eval script 在SQuAD 2.0 dev集上进行评估。

"exact": 79.87029394424324,
"f1": 82.91251169582613,

"total": 11873,
"HasAns_exact": 77.93522267206478,
"HasAns_f1": 84.02838248389763,
"HasAns_total": 5928,
"NoAns_exact": 81.79983179142137,
"NoAns_f1": 81.79983179142137,
"NoAns_total": 5945