DaydreamerF's picture
Update README.md
45b97bd
|
raw
history blame
2.15 kB
metadata
language: zh
widget:
  - text: 这句话是谁说的?
    context: >-
      “老大,你太牛逼了,把敌人军火库都给炸了,我真的佩服的五体投地,我现在忍不住想看看你藏的东西在哪里,我们快点出发吧。”代号零听完郭旭刚刚的讲述笑的拍手一直叫好。
  - text: 这句话是谁说的?
    context: “妈,你别哭了,我这不是好着呢吗?”郭旭扶着母亲的肩膀笑着说。
  - text: 这句话是谁说的?
    context: >-
      “总统先生,看来我们这一次在劫难逃了,大乘期的恐怖,远远超出了我们的想象,我还有一些后手能尽量拖延他一点时间,你们先走,我让我的鬼奴随你们去,去这个地方或许能保你们一线生机!”郭旭说完便偷偷地将黑暗空间的阴阳珠交给了陈天。
  - text: 这句话是谁说的?
    context: “也罢,能活一个是一个吧!他还那么年轻?”却是剑傲天摇了摇头无奈的说道。
tags:
  - generated_from_trainer
model-index:
  - name: bert-finetuned-TENBOOK
    results: []

bert-finetuned-TENBOOK

This model is a fine-tuned version of bert-base-chinese on an unknown dataset.

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

  • learning_rate: 2e-05
  • train_batch_size: 16
  • eval_batch_size: 16
  • seed: 42
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • num_epochs: 10
  • mixed_precision_training: Native AMP

Training results

  • exact_match:94.61
  • f1:94.61
  • 由于f1的计算方式有所错误,f1指标的参考性可以忽略

Framework versions

  • Transformers 4.25.1
  • Pytorch 1.12.1
  • Datasets 2.7.1
  • Tokenizers 0.13.2