xv44586/ccf_2020_qa_match

ccf 2020 qa match competition top1

ccf_2020_qa_match's Issues

你好 用你的pet的那个方法 验证的时候acc为nan predict的标签全是0 请问是和优化器有关吗 我是tf1.14 没有AdamWG 使用的Adam

请问你的模型单模最高多少？是融合了之后上0.8得嘛

如题，还有就是这次比赛有没有做数据增强呀？感觉这次比赛很多回答都很莫名其妙==

求更新。。

希望有时间能够更新一下，希望学习一下大佬的做法

自蒸馏脚本运行失败

自蒸馏脚本错误日志如下：

Traceback (most recent call last):
  File "pair-self-kd.py", line 297, in <module>
    callbacks=[student_evaluator])
  File "/home/work/.conda/envs/py3-tf.2.2-ccf/lib/python3.6/site-packages/keras/legacy/interfaces.py", line 91, in wrapper
    return func(*args, **kwargs)
  File "/home/work/.conda/envs/py3-tf.2.2-ccf/lib/python3.6/site-packages/keras/engine/training.py", line 1732, in fit_generator
    initial_epoch=initial_epoch)
  File "/home/work/.conda/envs/py3-tf.2.2-ccf/lib/python3.6/site-packages/keras/engine/training_generator.py", line 220, in fit_generator
    reset_metrics=False)
  File "/home/work/.conda/envs/py3-tf.2.2-ccf/lib/python3.6/site-packages/keras/engine/training.py", line 1514, in train_on_batch
    outputs = self.train_function(ins)
  File "/home/work/.conda/envs/py3-tf.2.2-ccf/lib/python3.6/site-packages/tensorflow/python/keras/backend.py", line 3792, in __call__
    outputs = self._graph_fn(*converted_inputs)
  File "/home/work/.conda/envs/py3-tf.2.2-ccf/lib/python3.6/site-packages/tensorflow/python/eager/function.py", line 1605, in __call__
    return self._call_impl(args, kwargs)
  File "/home/work/.conda/envs/py3-tf.2.2-ccf/lib/python3.6/site-packages/tensorflow/python/eager/function.py", line 1645, in _call_impl
    return self._call_flat(args, self.captured_inputs, cancellation_manager)
  File "/home/work/.conda/envs/py3-tf.2.2-ccf/lib/python3.6/site-packages/tensorflow/python/eager/function.py", line 1746, in _call_flat
    ctx, args, cancellation_manager=cancellation_manager))
  File "/home/work/.conda/envs/py3-tf.2.2-ccf/lib/python3.6/site-packages/tensorflow/python/eager/function.py", line 598, in call
    ctx=ctx)
  File "/home/work/.conda/envs/py3-tf.2.2-ccf/lib/python3.6/site-packages/tensorflow/python/eager/execute.py", line 60, in quick_execute
    inputs, attrs, num_outputs)
tensorflow.python.framework.errors_impl.FailedPreconditionError: 2 root error(s) found.
  (0) Failed precondition:  Error while reading resource variable _AnonymousVar409 from Container: localhost. This could mean that the variable was uninitialized. Not found: Resource localhost/_AnonymousVar409/N10tensorflow3VarE does not exist.
         [[node ReadVariableOp_1191 (defined at /home/work/.conda/envs/py3-tf.2.2-ccf/lib/python3.6/site-packages/keras/backend/tensorflow_backend.py:3009) ]]
         [[ReadVariableOp_1190/_12]]
  (1) Failed precondition:  Error while reading resource variable _AnonymousVar409 from Container: localhost. This could mean that the variable was uninitialized. Not found: Resource localhost/_AnonymousVar409/N10tensorflow3VarE does not exist.
         [[node ReadVariableOp_1191 (defined at /home/work/.conda/envs/py3-tf.2.2-ccf/lib/python3.6/site-packages/keras/backend/tensorflow_backend.py:3009) ]]
0 successful operations.
0 derived errors ignored. [Op:__inference_keras_scratch_graph_150548]

Function call stack:
keras_scratch_graph -> keras_scratch_graph

task-adaptive training

首先祝贺你们取得了第一名的成绩，同时也感谢你们把代码开源出来。我现在有个问题要咨询一下，请问，task-adaptive training和model-adaptive的区别是什么呀，我理解的task-adaptive training是使用比赛的数据，对预训练模型进行再次的训练，得到领域适配的预训练模型，请问我这样的理解对吗？

请问几百分类的任务用什么方案会比较合适？

最近在打一个比赛，有五百个分类，用bert模型能到73左右，但是榜一已经84了，不知道怎么突破。望大佬指点一下~

ask for dataset download link

Hi，thanks for your  great job! But I have not taken part in the competetion. If convenient, could you share the datasets downlink? Thanks a lot

环境问题

大佬后续更新代码吗？

这比赛采用pairwise的思路试过没有？

目前最高也就是0.757，看了下前面的同学都是0.8+，感觉很神奇，目前没有太好的思路？ 难道换更大的模型，更多的ensambe？

请问这里有模型融合的代码吗？

脚本（ccf_2020_qa_match_pair.py）报错

自蒸馏脚本（pair_self_kd.py）运行失败

失败日志如下。

Traceback (most recent call last):
  File "pair-self-kd.py", line 297, in <module>
    callbacks=[student_evaluator])
  File "/home/work/.conda/envs/py3-tf.2.2-ccf/lib/python3.6/site-packages/keras/legacy/interfaces.py", line 91, in wrapper
    return func(*args, **kwargs)
  File "/home/work/.conda/envs/py3-tf.2.2-ccf/lib/python3.6/site-packages/keras/engine/training.py", line 1732, in fit_generator
    initial_epoch=initial_epoch)
  File "/home/work/.conda/envs/py3-tf.2.2-ccf/lib/python3.6/site-packages/keras/engine/training_generator.py", line 220, in fit_generator
    reset_metrics=False)
  File "/home/work/.conda/envs/py3-tf.2.2-ccf/lib/python3.6/site-packages/keras/engine/training.py", line 1514, in train_on_batch
    outputs = self.train_function(ins)
  File "/home/work/.conda/envs/py3-tf.2.2-ccf/lib/python3.6/site-packages/tensorflow/python/keras/backend.py", line 3792, in __call__
    outputs = self._graph_fn(*converted_inputs)
  File "/home/work/.conda/envs/py3-tf.2.2-ccf/lib/python3.6/site-packages/tensorflow/python/eager/function.py", line 1605, in __call__
    return self._call_impl(args, kwargs)
  File "/home/work/.conda/envs/py3-tf.2.2-ccf/lib/python3.6/site-packages/tensorflow/python/eager/function.py", line 1645, in _call_impl
    return self._call_flat(args, self.captured_inputs, cancellation_manager)
  File "/home/work/.conda/envs/py3-tf.2.2-ccf/lib/python3.6/site-packages/tensorflow/python/eager/function.py", line 1746, in _call_flat
    ctx, args, cancellation_manager=cancellation_manager))
  File "/home/work/.conda/envs/py3-tf.2.2-ccf/lib/python3.6/site-packages/tensorflow/python/eager/function.py", line 598, in call
    ctx=ctx)
  File "/home/work/.conda/envs/py3-tf.2.2-ccf/lib/python3.6/site-packages/tensorflow/python/eager/execute.py", line 60, in quick_execute
    inputs, attrs, num_outputs)
tensorflow.python.framework.errors_impl.FailedPreconditionError: 2 root error(s) found.
  (0) Failed precondition:  Error while reading resource variable _AnonymousVar409 from Container: localhost. This could mean that the variable was uninitialized. Not found: Resource localhost/_AnonymousVar409/N10tensorflow3VarE does not exist.
         [[node ReadVariableOp_1191 (defined at /home/work/.conda/envs/py3-tf.2.2-ccf/lib/python3.6/site-packages/keras/backend/tensorflow_backend.py:3009) ]]
         [[ReadVariableOp_1190/_12]]
  (1) Failed precondition:  Error while reading resource variable _AnonymousVar409 from Container: localhost. This could mean that the variable was uninitialized. Not found: Resource localhost/_AnonymousVar409/N10tensorflow3VarE does not exist.
         [[node ReadVariableOp_1191 (defined at /home/work/.conda/envs/py3-tf.2.2-ccf/lib/python3.6/site-packages/keras/backend/tensorflow_backend.py:3009) ]]
0 successful operations.
0 derived errors ignored. [Op:__inference_keras_scratch_graph_150548]

Function call stack:
keras_scratch_graph -> keras_scratch_graph

导入NEZHA时报错，

运行时报错，检查发现下载的预训练模型里面没有model.ckpt这个文件，请问是需要转换一下吗？谢谢

请问训练模型的过程是？

是在nezha 的 Post training 之后加入了对抗、EDA、自蒸馏优化方案进行nezha的 finetune，然后将这几个模型一起融合了么？是没有在baseline上面继续提升么？还是finetune用了baseline的哪个模型？最终是有多少个模型呢？是怎么个融合方式呢？

大神我想问一下你夺冠，到底用到那个方案

或者说用的那个脚本，名称能发下不，方案太多不知道那个是你的冠军方案

请问模型融合策略是什么？

1、对于单个模型，根据不同随机数，训练多个模型；
2、对QA Pair 与 QA Point两类的模型，平均预测的概率作为最终的预测结果？

非常感谢！