wzzzd / lm_ner_span

基于实体首尾指针SPAN的序列标注框架

Nice to meet U 👋

lm_ner_span's Issues

DataManager.py中的DataManager:padding()里面的实现有点问题

DataManager.py中的DataManager:padding()里面的实现有点问题，如下：
src = [self.tokenizer.cls_token_id] + src + [self.tokenizer.pad_token_id]*pad_size + [self.tokenizer.sep_token_id]

[sep]应该是在[pad]的前面。这样会导致 attention_mask中的mask=1覆盖[pad]。从而在求loss的时候，会计算[pad]的损失。
@wzzzd @wzzzd

求解！！！

你这个把模型换成bert-large-uncased 可以训练英文的模型吗

data/CNER/里的label为空

原来是自动生成的哈哈

Traceback (most recent call last): File "run.py", line 46, in <module> train_loader = dm.get_dataset(data_type='train') File "/user-data/lm_ner_span-main/DataManager.py", line 34, in get_dataset dataloader = self.data_process(file) File "/user-data/lm_ner_span-main/DataManager.py", line 68, in data_process sampler = RandomSampler(tokenized_datasets) if not torch.cuda.device_count() > 1 else DistributedSampler(tokenized_datasets) File "/opt/conda/lib/python3.8/site-packages/torch/utils/data/sampler.py", line 103, in __init__ raise ValueError("num_samples should be a positive integer " ValueError: num_samples should be a positive integer value, but got num_samples=0

ValueError: num_samples should be a positive integer value, but got num_samples=0 请问此句报错怎么解决啊

请问将序列标注任务转化为预测首尾指针的任务具体在哪儿体现的，是bert编码过程中只对首尾部分进行编码，其他部分mask掉吗？

对抗学习参数问题

对抗学习中：emb_name='word_embeddings.' 这个代码，他的参数是怎么确定的？

Traceback (most recent call last): File "run.py", line 46, in <module> train_loader = dm.get_dataset(data_type='train') File "/user-data/lm_ner_span-main/DataManager.py", line 34, in get_dataset dataloader = self.data_process(file) File "/user-data/lm_ner_span-main/DataManager.py", line 68, in data_process sampler = RandomSampler(tokenized_datasets) if not torch.cuda.device_count() > 1 else DistributedSampler(tokenized_datasets) File "/opt/conda/lib/python3.8/site-packages/torch/utils/data/sampler.py", line 103, in __init__ raise ValueError("num_samples should be a positive integer " ValueError: num_samples should be a positive integer value, but got num_samples=0 请问这个报错怎么解决啊 好像是dataloader出错了 但是按照网上的解决办法将参数shuffle设置为FALSE还是没办法解决 请问大佬有什么解决办法吗