zjunlp / deepke Goto Github PK

我想问一下LOSS用的交叉熵什么最后的激活函数用的relu,是为了可以发现多种关系的意思吗

运行GCN报错
========== Start training ========== Traceback (most recent call last): File "main.py", line 81, in <module> train(epoch, device, train_dataloader, model, optimizer, criterion, config) File "/home/mere/deepke/deepke/trainer.py", line 16, in train y_pred = model(x) File "/usr/local/lib/python3.6/dist-packages/torch/nn/modules/module.py", line 541, in __call__ result = self.forward(*input, **kwargs) File "/home/mere/deepke/deepke/model/GCN.py", line 32, in forward x = self.embedding(x) File "/usr/local/lib/python3.6/dist-packages/torch/nn/modules/module.py", line 541, in __call__ result = self.forward(*input, **kwargs) File "/home/mere/deepke/deepke/model/Embedding.py", line 13, in forward words, head_pos, tail_pos = x ValueError: not enough values to unpack (expected 3, got 2)

运行LM报错
========== Start training ========== Traceback (most recent call last): File "main.py", line 81, in <module> train(epoch, device, train_dataloader, model, optimizer, criterion, config) File "/home/mere/deepke/deepke/trainer.py", line 16, in train y_pred = model(x) File "/usr/local/lib/python3.6/dist-packages/torch/nn/modules/module.py", line 541, in __call__ result = self.forward(*input, **kwargs) File "/home/mere/deepke/deepke/model/LM.py", line 19, in forward out = self.fc(out) File "/usr/local/lib/python3.6/dist-packages/torch/nn/modules/module.py", line 541, in __call__ result = self.forward(*input, **kwargs) File "/usr/local/lib/python3.6/dist-packages/torch/nn/modules/linear.py", line 87, in forward return F.linear(input, self.weight, self.bias) File "/usr/local/lib/python3.6/dist-packages/torch/nn/functional.py", line 1370, in linear ret = torch.addmm(bias, input, weight.t()) RuntimeError: size mismatch, m1: [64 x 1024], m2: [768 x 10] at /pytorch/aten/src/TH/generic/THTensorMath.cpp:197

您好，我在使用多个显卡训练的时候开始出现报错，但之前使用单显卡的时候没有报错。
我只在main.py中添加了语句model = nn.DataParallel(model)，训练的模型是transformer。
报错信息如下：
Traceback (most recent call last):
File "main.py", line 144, in
main()
File "/root/anaconda3/envs/NER/lib/python3.7/site-packages/hydra/main.py", line 24, in decorated_main
strict=strict,
File "/root/anaconda3/envs/NER/lib/python3.7/site-packages/hydra/_internal/utils.py", line 174, in run_hydra
overrides=args.overrides,
File "/root/anaconda3/envs/NER/lib/python3.7/site-packages/hydra/_internal/hydra.py", line 86, in run
job_subdir_key=None,
File "/root/anaconda3/envs/NER/lib/python3.7/site-packages/hydra/plugins/common/utils.py", line 109, in run_job
ret.return_value = task_function(task_cfg)
File "main.py", line 93, in main
train_loss = train(epoch, model, train_dataloader, optimizer, criterion, device, writer, cfg)
File "/root/joi/deepke/trainer.py", line 21, in train
y_pred = model(x)
File "/root/anaconda3/envs/NER/lib/python3.7/site-packages/torch/nn/modules/module.py", line 493, in call
result = self.forward(*input, **kwargs)
File "/root/anaconda3/envs/NER/lib/python3.7/site-packages/torch/nn/parallel/data_parallel.py", line 152, in forward
outputs = self.parallel_apply(replicas, inputs, kwargs)
File "/root/anaconda3/envs/NER/lib/python3.7/site-packages/torch/nn/parallel/data_parallel.py", line 162, in parallel_apply
return parallel_apply(replicas, inputs, kwargs, self.device_ids[:len(replicas)])
File "/root/anaconda3/envs/NER/lib/python3.7/site-packages/torch/nn/parallel/parallel_apply.py", line 83, in parallel_apply
raise output
File "/root/anaconda3/envs/NER/lib/python3.7/site-packages/torch/nn/parallel/parallel_apply.py", line 59, in worker
output = module(*input, **kwargs)
File "/root/anaconda3/envs/NER/lib/python3.7/site-packages/torch/nn/modules/module.py", line 493, in call
result = self.forward(*input, **kwargs)
File "/root/joi/deepke/models/Transformer.py", line 25, in forward
last_layer_hidden_state, all_hidden_states, all_attentions = self.transformer(inputs, key_padding_mask=mask)
File "/root/anaconda3/envs/NER/lib/python3.7/site-packages/torch/nn/modules/module.py", line 493, in call
result = self.forward(*input, **kwargs)
File "/root/joi/deepke/module/Transformer.py", line 134, in forward
layer_outputs = layer_module(hidden_states, key_padding_mask, attention_mask, head_mask[i])
File "/root/anaconda3/envs/NER/lib/python3.7/site-packages/torch/nn/modules/module.py", line 493, in call
result = self.forward(*input, **kwargs)
File "/root/joi/deepke/module/Transformer.py", line 97, in forward
attention_outputs = self.attention(hidden_states, key_padding_mask, attention_mask, head_mask)
File "/root/anaconda3/envs/NER/lib/python3.7/site-packages/torch/nn/modules/module.py", line 493, in call
result = self.forward(*input, **kwargs)
File "/root/joi/deepke/module/Transformer.py", line 55, in forward
attention_outputs = self.multihead_attention(x, x, x, key_padding_mask, attention_mask, head_mask)
File "/root/anaconda3/envs/NER/lib/python3.7/site-packages/torch/nn/modules/module.py", line 493, in call
result = self.forward(*input, **kwargs)
File "/root/joi/deepke/module/Attention.py", line 114, in forward
attention_out, attention_weight = self.attention(q, k, v, mask_out=mask_out, head_mask=head_mask)
File "/root/anaconda3/envs/NER/lib/python3.7/site-packages/torch/nn/modules/module.py", line 493, in call
result = self.forward(*input, **kwargs)
File "/root/joi/deepke/module/Attention.py", line 38, in forward
attention_weight.masked_fill(mask_out, -1e8)
RuntimeError: The expanded size of the tensor (270) must match the existing size (96) at non-singleton dimension 3. Target sizes: [1, 4, 270, 270]. Tensor sizes: [1, 1, 1, 96]

Describe the bug

直接运行显示错误

Environment (please complete the following information):

• OS:windows
• Python 3.7

Screenshots

If applicable, add screenshots to help explain your problem.

Additional context

if torch.cuda.CUDA_ENABLED and use_deterministic_cudnn:
  torch.backends.cudnn.deterministic = True
torch.backends.cudnn.benchmark = False

错误信息

if torch.cuda.CUDA_ENABLED and use_deterministic_cudnn:
AttributeError: module 'torch.cuda' has no attribute 'CUDA_ENABLED'

Set the environment variable HYDRA_FULL_ERROR=1 for a complete stack trace.

您好，在对模型进行训练时，我发现GCN的效果显著低于其他几个模型，查阅代码后在GCN.py中，并没有找到代码关于pyhanlp的相关调用，请问这里是如何通过句法分析构建出词语之间的邻接矩阵来进行关系抽取的呢？如果要添加pyhanlp来做的话，请问应该怎么操作？

你好，你知道这个是什么错误吗？
Traceback (most recent call last):
File "main.py", line 140, in
main()
File "/root/miniconda3/envs/openNRE/lib/python3.7/site-packages/hydra/main.py", line 24, in decorated_main
strict=strict,
File "/root/miniconda3/envs/openNRE/lib/python3.7/site-packages/hydra/_internal/utils.py", line 174, in run_hydra
overrides=args.overrides,
File "/root/miniconda3/envs/openNRE/lib/python3.7/site-packages/hydra/_internal/hydra.py", line 86, in run
job_subdir_key=None,
File "/root/miniconda3/envs/openNRE/lib/python3.7/site-packages/hydra/plugins/common/utils.py", line 109, in run_job
ret.return_value = task_function(task_cfg)
File "main.py", line 46, in main
preprocess(cfg)
File "/root/deepke/deepke/preprocess.py", line 131, in preprocess
_serialize_sentence(train_data, serial, cfg)
File "/root/deepke/deepke/preprocess.py", line 55, in _serialize_sentence
head_idx, tail_idx = d['tokens'].index('head'), d['tokens'].index('tail')
ValueError: 'tail' is not in list

启动lm模型找不到一些文件

好像只有训练，训练好怎么做预测呢？

你好，结果跑出来bert模型的结果比CNN还要差，请问正常吗

Describe the question

A clear and concise description of what the question is.

Environment (please complete the following information):

• OS: [e.g. mac / window]
• Python Version [e.g. 3.6]

Screenshots

If applicable, add screenshots to help explain your problem.

Additional context

Add any other context about the problem here.

您好，您项目中所给数据顺利跑通。但换成自己的数据时，疯狂报KeyError，每一个字段都曾报过，最后一个KeyError: 'index'实在是解决不了了。数据格式与您要求的一样。报错信息如下Traceback (most recent call last):
File "main.py", line 140, in
main()
File "/home/wgh/anaconda3/envs/pytorch/lib/python3.6/site-packages/hydra/main.py", line 24, in decorated_main
strict=strict,
File "/home/wgh/anaconda3/envs/pytorch/lib/python3.6/site-packages/hydra/_internal/utils.py", line 174, in run_hydra
overrides=args.overrides,
File "/home/wgh/anaconda3/envs/pytorch/lib/python3.6/site-packages/hydra/_internal/hydra.py", line 86, in run
job_subdir_key=None,
File "/home/wgh/anaconda3/envs/pytorch/lib/python3.6/site-packages/hydra/plugins/common/utils.py", line 109, in run_job
ret.return_value = task_function(task_cfg)
File "main.py", line 46, in main
preprocess(cfg)
File "/home/wgh/deepke/preprocess.py", line 116, in preprocess
rels = _handle_relation_data(relation_data)
File "/home/wgh/deepke/preprocess.py", line 90, in _handle_relation_data
relation_data = sorted(relation_data, key=lambda i: int(i['index']))
File "/home/wgh/deepke/preprocess.py", line 90, in
relation_data = sorted(relation_data, key=lambda i: int(i['index']))
KeyError: 'index'

OSError: file bert_pretrained/config.json not found
为什么已经下载了bert的哈工大模型还是会报这个错误

Describe the question

A clear and concise description of what the question is.

Environment (please complete the following information):

• OS: [e.g. mac / window]
• Python Version [e.g. 3.6]

Screenshots

If applicable, add screenshots to help explain your problem.

Additional context

Add any other context about the problem here.

我使用的是https://github.com/ymcui/Chinese-BERT-wwm#%E4%B8%AD%E6%96%87%E6%A8%A1%E5%9E%8B%E4%B8%8B%E8%BD%BD
上的BERT-wwm, Chinese 放在了 pretrained文件夹下，不知道如何启用，使用LM模型启动时，加载不到其必须的文件。

你好，想请教一下一个问题：
当我用胶囊网络的时候，我发现损失在不断的下降，但是，精度却在不断的降低？这让我很疑惑。
========== Start training ==========
Train Epoch: 1 [640/4000 (16%)] Loss: 0.808406
Train Epoch: 1 [1280/4000 (32%)] Loss: 0.807249
Train Epoch: 1 [1920/4000 (48%)] Loss: 0.805759
Train Epoch: 1 [2560/4000 (63%)] Loss: 0.803572
Train Epoch: 1 [3200/4000 (79%)] Loss: 0.800597
Train Epoch: 1 [3840/4000 (95%)] Loss: 0.796495
Train Epoch: 1 [4000/4000 (100%)] Loss: 0.795066
macro metrics: [p: 0.4085, r:0.2570, f1:0.2254]
micro metrics: [p: 0.2570, r:0.2570, f1:0.2570]
Train Epoch: 2 [640/4000 (16%)] Loss: 0.789447
Train Epoch: 2 [1280/4000 (32%)] Loss: 0.782183
Train Epoch: 2 [1920/4000 (48%)] Loss: 0.773557
Train Epoch: 2 [2560/4000 (63%)] Loss: 0.762742
Train Epoch: 2 [3200/4000 (79%)] Loss: 0.750204
Train Epoch: 2 [3840/4000 (95%)] Loss: 0.735720
Train Epoch: 2 [4000/4000 (100%)] Loss: 0.731115
macro metrics: [p: 0.1102, r:0.1210, f1:0.0532]
micro metrics: [p: 0.1210, r:0.1210, f1:0.1210]
Train Epoch: 3 [640/4000 (16%)] Loss: 0.714159
Train Epoch: 3 [1280/4000 (32%)] Loss: 0.695173
Train Epoch: 3 [1920/4000 (48%)] Loss: 0.674588
Train Epoch: 3 [2560/4000 (63%)] Loss: 0.652236
Train Epoch: 3 [3200/4000 (79%)] Loss: 0.628999
Train Epoch: 3 [3840/4000 (95%)] Loss: 0.607024
Train Epoch: 3 [4000/4000 (100%)] Loss: 0.601029
macro metrics: [p: 0.0100, r:0.1000, f1:0.0182]
micro metrics: [p: 0.1000, r:0.1000, f1:0.1000]
Train Epoch: 4 [640/4000 (16%)] Loss: 0.581706
Train Epoch: 4 [1280/4000 (32%)] Loss: 0.566093
Train Epoch: 4 [1920/4000 (48%)] Loss: 0.553648
Train Epoch: 4 [2560/4000 (63%)] Loss: 0.543705
Train Epoch: 4 [3200/4000 (79%)] Loss: 0.536618
这是什么情况呢？

File main.py ,line 7 in
from hydra import utils
ImportError: cannot import name 'utils'

Describe the question

A clear and concise description of what the question is.
我想对比CNN\PCNN\GCN\Transform等模型在自己标注的数据上的准确性，请问代码应该在哪里修改一下代码，谢谢，请问此外模型是否可以输出不同关系的分类准确率、召回率，即导演、国籍、祖籍等类别。

Environment (please complete the following information):

• OS: [e.g. mac / window]
• Python Version [e.g. 3.6]

Screenshots

If applicable, add screenshots to help explain your problem.

Additional context

Add any other context about the problem here.

你好，麻烦更新GCN部分的问题哦，感谢

有没有说明文档，和对接口比较详细的说明呢

Describe the bug

A clear and concise description of what the bug is.
When running the main.py in Colab, I met a error: can't import utils in "from hydra import utils". And I tried to comment the the code "from hydra import utils", the next error---"module 'hydra' has no attribute 'main'" in line 29, which is " @hydra.main(config_path='conf/config.yaml')"

Environment (please complete the following information):

• OS: [e.g. mac / window]
• Python Version [e.g. 3.6]

Screenshots

If applicable, add screenshots to help explain your problem.

Additional context

Add any other context about the problem here.

您好！我在学习您的deepke源代码，并试图复现相关实验，但是在使用Bert的时候（model=lm），一直报无法找到预训练模型的错误，错误代码为：Model name '/deepke-master/deepke-master/pretrained' was not found in tokenizers model name list (bert-base-uncased, bert-large-uncased, bert-base-cased, bert-large-cased, bert-base-multilingual-uncased, bert-base-multilingual-cased, bert-base-chinese, bert-base-german-cased, bert-large-uncased-whole-word-masking, bert-large-cased-whole-word-masking, bert-large-uncased-whole-word-masking-finetuned-squad, bert-large-cased-whole-word-masking-finetuned-squad, bert-base-cased-finetuned-mrpc, bert-base-german-dbmdz-cased, bert-base-german-dbmdz-uncased, bert-base-finnish-cased-v1, bert-base-finnish-uncased-v1, bert-base-dutch-cased). We assumed '/deepke-master/deepke-master/pretrained' was a path, a model identifier, or url to a directory containing vocabulary files named ['vocab.txt'] but couldn't find such vocabulary files at this path or url.
但是我是严格按照readme中的要求，下载了完整的bert模型（pytorch版）并放置在pretrained文件夹内，不知是何错误？
还有一个小问题，源代码中在predict里运行后有“实体类型”输入这一选项，但给出的train data格式中并没有“实体类型”，是否意味着我可以在训练数据中自行添加“实体类型”的训练特征，而后的预测也可以因此提高精确度？
请批评指正！望不吝赐教！谢谢！

你好，请问demo内使用的数据集中各个字段详细的含义或者如何标注的能否简单告知吗，我想针对某一个领域自己标注数据集，但是不太理解demo使用的数据字段的含义，比如tail_offset，head_offset

CNdeepdive$ sudo bash install.sh
DeepDive installer for Ubuntu
curl: (7) Failed to connect to raw.githubusercontent.com port 443: 连接超时

比如说【张小明上高中的时候经常去浙江大学玩】这种先不考虑容易误分到【毕业学校】，只能设置argmax的阈值去处理吗。是因为没有负样本的语料吗，

感谢您开源关系抽取套件，我想请问您如果我想使用NYT数据集，该怎么改动源代码呢？因为源代码中没有注释，还请您耐心回答下，非常感谢、

========= Start training ==========
Traceback (most recent call last):
File "main.py", line 83, in
train(epoch, device, train_dataloader, model, optimizer, criterion, config)
File "/home/mere/deepke/deepke/trainer.py", line 12, in train
for batch_idx, (*x, y) in enumerate(dataloader, 1):
File "/usr/local/lib/python3.6/dist-packages/torch/utils/data/dataloader.py", line 346, in next
data = self._dataset_fetcher.fetch(index) # may raise StopIteration
File "/usr/local/lib/python3.6/dist-packages/torch/utils/data/_utils/fetch.py", line 47, in fetch
return self.collate_fn(data)
File "/home/mere/deepke/deepke/dataset.py", line 20, in collate_fn
batch.sort(key=lambda data: data['seq_len'], reverse=True)
File "/home/mere/deepke/deepke/dataset.py", line 20, in
batch.sort(key=lambda data: data['seq_len'], reverse=True)
KeyError: 'seq_len'

你好，我目前在尝试改写predict部分的代码，看到predict.py中引入了一个serializer库，pip安装之后无法导入，并且看到有一些比较个性化的参数，想问一下这个库是哪里来的？

Describe the question

A clear and concise description of what the question is.

Environment (please complete the following information):

• OS: [e.g. mac / window]
• Python Version [e.g. 3.6]

Screenshots

If applicable, add screenshots to help explain your problem.

Additional context

Add any other context about the problem here.

你好，能不能提供相应的数据，主要是为了跑通模型试一试，小规模的也行

你好，请问您那边啥时候能更新一下预测代码呢

mask

我想问一下在module中的cnn源码里面的mask是作为工程功能吗？因为我在读论文的时候并没有发现类似mask之类的操作

#param mask: [batch_size, max_len], 句长部分为0，padding部分为1。不影响卷积运算，max-pool一定不会pool到pad为0的位置

上面为mask在源码的注释，并不是看的很懂，只是理解为工程中的一个trick，可否帮忙解释一下？谢谢了

你好，想请问一下你们的训练数据是怎样生成的，是用软件标注的么，还是人工整理的，因为我想做其它领域的关系抽取研究。

Environment (please complete the following information):

• OS: [e.g. mac / window]
• Python Version [e.g. 3.6]

Screenshots

If applicable, add screenshots to help explain your problem.

Additional context

Add any other context about the problem here.

如果待抽取的句子是一个问句，比如"姚明的老婆是谁?"，此时是否支持抽取出 实体 “姚明” ，还有 关系属性 “老婆”？

Traceback (most recent call last):
File "main.py", line 83, in
model_name = model.save(epoch=epoch)
File "C:\Users\asus\Desktop\deepke\deepke\model\BasicModule.py", line 33, in save
torch.save(self.state_dict(), name)
File "D:\Python3.6.4\lib\site-packages\torch\serialization.py", line 224, in save
return _with_file_like(f, "wb", lambda f: _save(obj, f, pickle_module, pickle_protocol))
File "D:\Python3.6.4\lib\site-packages\torch\serialization.py", line 147, in _with_file_like
f = open(f, mode)
OSError: [Errno 22] Invalid argument: 'checkpoints/CNN_epoch1_1107_19:31:57.pth'

Describe the question

A clear and concise description of what the question is.
您好，如果我想做一个接口用于测试的话，应该怎么应用呢？因为对hydra不熟，麻烦大神指点指点！

Environment (please complete the following information):

• OS: [e.g. mac / window]
• Python Version [e.g. 3.6]

Screenshots

If applicable, add screenshots to help explain your problem.

Additional context

Add any other context about the problem here.

你好，我重构了百度的数据集，重构后大概又30多w条。其中某些数据在其他模型预处理的时候会报错，但舍弃这些数据后最终跑通了，但是BERT模型遇到了如下问题，看上去应该是维度不太匹配，想请教一下问题可能会出在哪里，谢谢
>, TensorInfo<T, IndexType>, TensorInfo<long, IndexType>, int, int, IndexType, IndexType, long) [with T = float, IndexType = unsigned int, DstDim = 2, SrcDim = 2, IdxDim = -2, IndexIsMajor = true]: block: [195,0,0], thread: [31,0,0] AssertionsrcIndex < srcSelectDimSize failed. Traceback (most recent call last): File "main.py", line 82, in <module> macro_f1, micro_f1 = validate(test_dataloader, model, device, config) File "/home/***/deepke/deepke/trainer.py", line 53, in validate x = [i.to(device) for i in x] File "/home/***/deepke/deepke/trainer.py", line 53, in <listcomp> x = [i.to(device) for i in x] RuntimeError: CUDA error: device-side assert triggered

你好，请问您用的什么方法替换wordpiece处理输入呢

预测的代码是没有在项目中吗

你好~如果是想通过输入句子，然后输出类似{实体A，实体B，关系}的话， 是否需要在相应的模型中加入predict的代码？然后通过model.load_state_dict(torch.load('./checkpoints/xx.pth'))，然后再去用model.predict()来输出结果？

Describe the bug

A clear and concise description of what the bug is.

output,pooler_output = self.lm (word)

Environment (please complete the following information):

• OS: [ window]
• Python Version [3.7]
• torch 1.7.1

Additional context
讲道理会自动匹配元组，但是没有，只返回了字符串，如下

你好，在data中添加了自己的数据集，运行main.py报KeyError 是为什么呢？
是需要在哪个词典添加吗？