NER任务中损失函数CrossEntropyLoss()指定weight报错
收藏
使用PaddleNLP进行NER训练，设置的损失函数为：paddle.nn.loss.CrossEntropyLoss(weight=weight)
其中的weight为：weight = paddle.to_tensor(
np.array(['0.000451536', '0.000451536', '0.000645355', '0.000645355', '0.000336313', '0.000336313', '0.003716059',
'0.003716059', '0.003580706', '0.003580706', '0.001656642', '0.001656642', '0.002692495', '0.002692495',
'0.010524314', '0.010524314', '0.012340922', '0.012340922', '0.010868073', '0.010868073', '0.034742916',
'0.034742916', '0.003124228', '0.003124228', '0.013517166', '0.013517166', '0.022825821', '0.022825821',
'0.077240946', '0.077240946', '0.024164765', '0.024164765', '0.262151091', '0.262151091', '0.015420652',
'0.015420652']), dtype='float64')
一共有36个标签，weight的也是36个
然后运行之后报了如下错误：
Error: /paddle/paddle/fluid/operators/gather.cu.h:62 Assertion `index_value >= 0 && index_value < input_dims[j]` failed. The index is out of bounds, please check whether the dimensions of index and input meet the requirements. It should be less than [36] and greater or equal to 0, but received [0]
Traceback (most recent call last):
File "/home/www/.conda/envs/nlp_env/lib/python3.7/site-packages/IPython/core/interactiveshell.py", line 3418, in run_code
exec(code_obj, self.user_global_ns, self.user_ns)
File "", line 1, in
runfile('/data/work/nlp_task/cheng/jz_git/pintu/enent_entity/intelligent_info_second/paddlenlp_version/train.py', wdir='/data/work/nlp_task/cheng/jz_git/pintu/enent_entity/intelligent_info_second/paddlenlp_version')
File "/root/.pycharm_helpers/pydev/_pydev_bundle/pydev_umd.py", line 197, in runfile
pydev_imports.execfile(filename, global_vars, local_vars) # execute the script
File "/root/.pycharm_helpers/pydev/_pydev_imps/_pydev_execfile.py", line 18, in execfile
exec(compile(contents+"\n", file, 'exec'), glob, loc)
File "/data/work/nlp_task/cheng/jz_git/pintu/enent_entity/intelligent_info_second/paddlenlp_version/train.py", line 161, in
loss.backward()
File "", line 2, in backward
File "/home/www/.conda/envs/nlp_env/lib/python3.7/site-packages/paddle/fluid/wrapped_decorator.py", line 25, in __impl__
return wrapped_func(*args, **kwargs)
File "/home/www/.conda/envs/nlp_env/lib/python3.7/site-packages/paddle/fluid/framework.py", line 225, in __impl__
return func(*args, **kwargs)
File "/home/www/.conda/envs/nlp_env/lib/python3.7/site-packages/paddle/fluid/dygraph/varbase_patch_methods.py", line 177, in backward
self._run_backward(framework._dygraph_tracer(), retain_graph)
OSError: (External) Cudnn error, CUDNN_STATUS_EXECUTION_FAILED (at /paddle/paddle/fluid/operators/softmax_cudnn_op.cu:511)
0
收藏
请登录后评论
使用ai studio跑的么，是的话把环境关了重新打开试试
用aistudio是了一下，还是不行
---------------------------------------------------------------------------ValueError Traceback (most recent call last) in 3 for idx, (input_ids, token_type_ids, length, labels) in enumerate(train_loader): 4 logits = model(input_ids, token_type_ids) ----> 5 loss = paddle.mean(loss_fn(logits, labels)) 6 loss.backward() 7 optimizer.step() /opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddle/fluid/dygraph/layers.py in __call__(self, *inputs, **kwargs) 896 self._built = True 897 --> 898 outputs = self.forward(*inputs, **kwargs) 899 900 for forward_post_hook in self._forward_post_hooks.values(): /opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddle/nn/layer/loss.py in forward(self, input, label) 403 axis=self.axis, 404 use_softmax=self.use_softmax, --> 405 name=self.name) 406 407 return ret /opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddle/nn/functional/loss.py in cross_entropy(input, label, weight, ignore_index, reduction, soft_label, axis, use_softmax, name) 1418 'Expected 0 <= label_value < class_dimension({}), but got {} <= label_value <= {} '. 1419 format(input.shape[-1], -> 1420 label_min.numpy(), label_max.numpy())) 1421 weight_gather = core.ops.gather_nd(weight, label) 1422 input_shape = list(label.shape) ValueError: Expected 0 <= label_value < class_dimension(13), but got [-1] <= label_value <= [12]这个是代码有问题了label_value 的值有问题，可以排查一下，你在错误的情况下再去运行代码，就会出现你第一次的问题，这应该是bug
嗯，谢谢你的回复，我提issue了，看怎么回复吧
此问题已经解决，具体请查看，issue为：https://github.com/PaddlePaddle/PaddleNLP/issues/835