yanwii / chinesener Goto Github PK

首先感谢大佬开源了这么好的项目。
有一个小疑问，存放bert模型的文件夹是‘model’还是‘bert_model’，按照说明里是‘bert_model’，但我跑的时候会出现如下错误：

Traceback (most recent call last):
  File "model.py", line 502, in <module>
    model.train()
  File "model.py", line 328, in train
    ARGS.init_checkpoint)
  File "D:\ProgramData\Anaconda3\lib\site-packages\bert_base-0.0.9-py3.6.egg\bert_base\bert\modeling.py", line 331, in get_assignment_map_from_checkpoint
    init_vars = tf.train.list_variables(init_checkpoint)
  File "D:\ProgramData\Anaconda3\lib\site-packages\tensorflow\python\training\checkpoint_utils.py", line 97, in list_variables
    reader = load_checkpoint(ckpt_dir_or_file)
  File "D:\ProgramData\Anaconda3\lib\site-packages\tensorflow\python\training\checkpoint_utils.py", line 65, in load_checkpoint
    "given directory %s" % ckpt_dir_or_file)
ValueError: Couldn't find 'checkpoint' file or checkpoints in given directory bert_model/

改成‘model’之后运行就出现python停止运行的问题。
小白一枚，还请大神指正！

大佬，这个是什么导致的呢，我该怎么改

如图，下载的chinese_L-12_H-768_A-12.zip缺少checkpoint文件，是怎么回事

Traceback (most recent call last):
File "model.py", line 505, in
model.train()
File "model.py", line 284, in train
self.train_data = BertDataUtils(tokenizer, batch_size=5)
File "C:\Users\admin\Desktop\ChineseNER\bert_data_utils.py", line 28, in init
self.prepare_batch()
File "C:\Users\admin\Desktop\ChineseNER\bert_data_utils.py", line 77, in prepare_batch
pad_data = self.pad_data(self.data[-self.batch_size:])
File "C:\Users\admin\Desktop\ChineseNER\bert_data_utils.py", line 87, in pad_data
max_length = max([len(i[0]) for i in c_data] )
ValueError: max() arg is an empty sequence

这个问题怎么解决

你好，可以提供预训练模型吗

                    self.bert_evaluate(sess, "ORG")
                    self.bert_evaluate(sess, "PER")

如果我想识别出时间、地名等信息，除了改标注信息，这里要改动吗

你好，咨询一下，人名，地名识别正确率怎么样呢？

在实际运行代码的过程中，发现batch_size设置为8时就会OOM，使用的是1080Ti显卡，11G显存。请问这种情况是否正常？
如果batch_size只能设置成这么小，训练过程会非常缓慢，1.5w数据的话，一个epoch需要40分钟左右，这个训练速度也是正常的吗？请问有什么可以提速的方法吗？
谢谢~

ub16c9@ub16c9-gpu:~/ub16_prj/ChineseNER$ python3 model.py -e train -m bert

train data: 2757
nums of tags: 9
/usr/local/lib/python3.5/dist-packages/tensorflow/python/ops/gradients_impl.py:112: UserWarning: Converting sparse IndexedSlices to a dense Tensor of unknown shape. This may consume a large amount of memory.
"Converting sparse IndexedSlices to a dense Tensor of unknown shape. "
2019-03-12 17:25:23.711894: I tensorflow/core/platform/cpu_feature_guard.cc:141] Your CPU supports instructions that this TensorFlow binary was not compiled to use: AVX2 FMA
2019-03-12 17:25:23.825882: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:964] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-03-12 17:25:23.826463: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1432] Found device 0 with properties:
name: GeForce GTX 1080 Ti major: 6 minor: 1 memoryClockRate(GHz): 1.6575
pciBusID: 0000:01:00.0
totalMemory: 10.92GiB freeMemory: 10.11GiB
2019-03-12 17:25:23.826481: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1511] Adding visible gpu devices: 0
2019-03-12 17:25:24.069500: I tensorflow/core/common_runtime/gpu/gpu_device.cc:982] Device interconnect StreamExecutor with strength 1 edge matrix:
2019-03-12 17:25:24.069536: I tensorflow/core/common_runtime/gpu/gpu_device.cc:988] 0
2019-03-12 17:25:24.069542: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1001] 0: N
2019-03-12 17:25:24.069766: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1115] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 9769 MB memory) -> physical GPU (device: 0, name: GeForce GTX 1080 Ti, pci bus id: 0000:01:00.0, compute capability: 6.1)
Traceback (most recent call last):
File "model.py", line 502, in
model.train()
File "model.py", line 328, in train
ARGS.init_checkpoint)
File "/usr/local/lib/python3.5/dist-packages/bert_base/bert/modeling.py", line 331, in get_assignment_map_from_checkpoint
init_vars = tf.train.list_variables(init_checkpoint)
File "/usr/local/lib/python3.5/dist-packages/tensorflow/python/training/checkpoint_utils.py", line 95, in list_variables
reader = load_checkpoint(ckpt_dir_or_file)
File "/usr/local/lib/python3.5/dist-packages/tensorflow/python/training/checkpoint_utils.py", line 63, in load_checkpoint
"given directory %s" % ckpt_dir_or_file)
ValueError: Couldn't find 'checkpoint' file or checkpoints in given directory bert_model/
ub16c9@ub16c9-gpu:/ub16_prj/ChineseNER$ ll bert_
bert_data_utils.py bert_model/
ub16c9@ub16c9-gpu:/ub16_prj/ChineseNER$ ll bert_model/
total 804848
drwxrwxr-x 3 ub16c9 ub16c9 4096 三月 12 17:25 ./
drwxrwxr-x 7 ub16c9 ub16c9 4096 三月 12 17:19 ../
-rw-r--r-- 1 ub16c9 ub16c9 520 三月 12 17:19 bert_config.json
-rw-r--r-- 1 ub16c9 ub16c9 411529768 三月 12 17:19 bert_model.ckpt.data-00000-of-00001
-rw-r--r-- 1 ub16c9 ub16c9 8512 三月 12 17:19 bert_model.ckpt.index
-rw-r--r-- 1 ub16c9 ub16c9 905069 三月 12 17:19 bert_model.ckpt.meta
drwxr-xr-x 2 ub16c9 ub16c9 4096 三月 12 17:16 chinese_L-12_H-768_A-12/
-rw-rw-r-- 1 ub16c9 ub16c9 411575980 三月 12 17:19 pytorch_model.bin
-rw-r--r-- 1 ub16c9 ub16c9 109540 三月 12 17:19 vocab.txt
ub16c9@ub16c9-gpu:~/ub16_prj/ChineseNER$

Traceback (most recent call last):
File "model.py", line 13, in
from bert_base.train import tf_metrics
ModuleNotFoundError: No module named 'bert_base.train'
请问是不是缺少什么文件啊

您好，我用自己的数据跑出来各项指标都是0，但loss挺低，acc挺高，是我哪里print出来的不对吗

InvalidArgumentError (see above for traceback): Cannot assign a device for operation init: node init (defined at model.py:323) was explicitly assigned to /device:GPU:0 but available devices are [ /job:localhost/replica:0/task:0/device:CPU:0, /job:localhost/replica:0/task:0/device:XLA_CPU:0 ]. Make sure the device specification refers to a valid device. The requested device appears to be a GPU, but CUDA is not enabled.
[[node init (defined at model.py:323) ]]

请教大神：如果想训练其他实体类别，需要在data_utils里增加tag_map吗？还是只要在训练数据里增加就可以？
我看在 data_utils 的 DataBatch 类下面有一行 tag_map 被注掉了，是不是需要在这里增加其他实体的类别才能识别出来呢？
先谢过啦

我都是按照同样的数据训练的，但是结果只有NER 没有ORG

epoch 0 Traceback (most recent call last): File "D:\ProgramFiles\anaconda\envs\tf\lib\site-packages\tensorflow\python\client\session.py", line 300, in __init__ fetch, allow_tensor=True, allow_operation=True)) File "D:\ProgramFiles\anaconda\envs\tf\lib\site-packages\tensorflow\python\framework\ops.py", line 3478, in as_graph_element return self._as_graph_element_locked(obj, allow_tensor, allow_operation) File "D:\ProgramFiles\anaconda\envs\tf\lib\site-packages\tensorflow\python\framework\ops.py", line 3567, in _as_graph_element_locked types_str)) TypeError: Can not convert a float into a Tensor or Operation.

`TypeError: Fetch argument 5e-05 has invalid type <class 'float'>, must be a string or Tensor. (Can not convert a float into a Tensor or Operation.)

`

这个地方该怎么设置，求解决办法，感谢

Traceback (most recent call last):
not enough values to unpack (expected 2, got 1)
File "D:/dogtime/mission/BiGRU_crf/bert_data_utils.py", line 41, in load_data
word, tag = line.split()
ValueError: not enough values to unpack (expected 2, got 1)

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
File "D:/dogtime/mission/BiGRU_crf/bert_data_utils.py", line 119, in
bert_data_util = BertDataUtils(tokenizer)
File "D:/dogtime/mission/BiGRU_crf/bert_data_utils.py", line 26, in init
self.load_data()
File "D:/dogtime/mission/BiGRU_crf/bert_data_utils.py", line 49, in load_data
inputs_ids = self.tokenizer.convert_tokens_to_ids(ntokens)
File "D:\dogtime\mission\BiGRU_crf\bert_base\bert\tokenization.py", line 179, in convert_tokens_to_ids
return convert_by_vocab(self.vocab, tokens)
File "D:\dogtime\mission\BiGRU_crf\bert_base\bert\tokenization.py", line 140, in convert_by_vocab
output.append(vocab[item])
KeyError: 'D'

博主好，感谢提供了这一份嵌入bert预训练模型的python实现。最近有在看你的代码，发现一个事情，每次再向bert_layer层提供input_ids时，都是重新初始化了embedding_table，那么不同句中的同一个字极有可能是不同的初始化向量，这里是不是规定下种子号更为妥当？bert小白一枚，还请指点下

大佬，请问下该模型的输入输出项都有哪些？
最近在看将模型做成服务的形式，需要这几个参数

            labeling_signature = (
                tf.saved_model.signature_def_utils.build_signature_def(
                    inputs={
                        "input_ids":
                            bert_input_ids,
                        "segment_ids":
                            bert_segment_ids,
                        "input_mask":
                            bert_input_mask,
                        "dropout":
                            bert_dropout,
                    },
                    outputs={
                        "targets":
                            bert_targets,
                    },
                    method_name=tf.saved_model.signature_constants.PREDICT_METHOD_NAME))

按照网上的教程，改写成这个样子，这些输入输出参数是对的吗？望指点下，感谢！

很高兴发现这么棒的项目，学习了几天～
有一个问题想请教一下：
面对像通讯录那种“大舅、杨三姐、王欢处长”这种类型的文本可以利用你这种方案解决吗？
想了挺久的，文本长度太短了，原子化的文本查找有效姓名、关系等信息
期盼探讨～