liang8qi / data2textwithauxiliarysupervision Goto Github PK

作者你好，我最近跑你的代码，但是在跑train.sh会有如下的报错：
Traceback (most recent call last):
File "train.py", line 544, in
main()
File "train.py", line 520, in main
fields = load_fields(first_dataset, data_type, checkpoint, cal_indices=opt.cal_indices)
File "train.py", line 404, in load_fields
fields = dict([(k, f) for (k, f) in fields
TypeError: 'NoneType' object is not iterable

主要就是变量fields的问题，代码里说fields不能迭代，而我debug的时候fields是空的，这个问题改了好久也没有解决，实在是没有办法了麻烦一下你，提前谢谢了

直接运行会报target.view()出现错误，之后改成.reshape(),或者.contiguous().view(),的时候还会报下面第二个错误

Traceback (most recent call last):
File "/home/qiansh/data2text/Data2TextWithAuxiliarySupervision/train.py", line 556, in
main()
File "/home/qiansh/data2text/Data2TextWithAuxiliarySupervision/train.py", line 548, in main
train_model(model1, model2, fields, optim1, optim2, data_type, model_opt)
File "/home/qiansh/data2text/Data2TextWithAuxiliarySupervision/train.py", line 297, in train_model
train_stats, train_stats2 = trainer.train(train_iter, epoch, report_func)
File "/home/qiansh/data2text/Data2TextWithAuxiliarySupervision/onmt/Trainer.py", line 294, in train
self._gradient_accumulation_basic_encdec(true_batchs, total_stats2, report_stats2, normalization)
File "/home/qiansh/data2text/Data2TextWithAuxiliarySupervision/onmt/Trainer.py", line 513, in _gradient_accumulation_basic_encdec
batch_stats = self.train_loss2.sharded_compute_loss(
File "/home/qiansh/data2text/Data2TextWithAuxiliarySupervision/onmt/Loss.py", line 136, in sharded_compute_loss
loss, stats = self._compute_loss(batch, **shard)
File "/home/qiansh/data2text/Data2TextWithAuxiliarySupervision/onmt/modules/CopyGenerator.py", line 204, in _compute_loss
target = target.view(-1)
RuntimeError: view size is not compatible with input tensor's size and stride (at least one dimension spans across two contiguous subspaces). Use .reshape(...) instead.
改成.reshape(),或者.contiguous().view(),的时候
Traceback (most recent call last):
File "/home/qiansh/data2text/Data2TextWithAuxiliarySupervision/train.py", line 556, in
main()
File "/home/qiansh/data2text/Data2TextWithAuxiliarySupervision/train.py", line 548, in main
train_model(model1, model2, fields, optim1, optim2, data_type, model_opt)
File "/home/qiansh/data2text/Data2TextWithAuxiliarySupervision/train.py", line 297, in train_model
train_stats, train_stats2 = trainer.train(train_iter, epoch, report_func)
File "/home/qiansh/data2text/Data2TextWithAuxiliarySupervision/onmt/Trainer.py", line 294, in train
self._gradient_accumulation_basic_encdec(true_batchs, total_stats2, report_stats2, normalization)
File "/home/qiansh/data2text/Data2TextWithAuxiliarySupervision/onmt/Trainer.py", line 513, in _gradient_accumulation_basic_encdec
batch_stats = self.train_loss2.sharded_compute_loss(
File "/home/qiansh/data2text/Data2TextWithAuxiliarySupervision/onmt/Loss.py", line 135, in sharded_compute_loss
for shard in shards(shard_state, shard_size, retain_graph=retain_graph):
File "/home/qiansh/data2text/Data2TextWithAuxiliarySupervision/onmt/Loss.py", line 299, in shards
torch.autograd.backward(inputs, grads, retain_graph=retain_graph)
File "/home/qiansh/anaconda3/envs/qsh/lib/python3.9/site-packages/torch/autograd/init.py", line 130, in backward
Variable._execution_engine.run_backward(
RuntimeError: one of the variables needed for gradient computation has been modified by an inplace operation: [torch.cuda.FloatTensor [1200, 6
00]], which is output 0 of TBackward, is at version 2; expected version 1 instead. Hint: enable anomaly detection to find the operation that f
ailed to compute its gradient, with torch.autograd.set_detect_anomaly(True).