设置“region_embedding” 运行报错“RuntimeError: Expected all tensors to be on the same device, but found at least two devices, cuda:0 and cpu!”

[CrawlingForward]

你好！ 相关config如下： "task_info":{ "label_type": "multi_label", "hierarchical": false, "hierar_taxonomy": "data/rcv1.taxonomy", "hierar_penalty": 0.000001 }, "device": "cuda", "model_name": "TextCNN", "checkpoint_dir": "checkpoint_dir_spc", "model_dir": "trained_model_ipc",

"embedding": { "type": "region_embedding", "dimension": 300, "region_embedding_type": "context_word", "region_size": 5, "initializer": "uniform", "fan_mode": "FAN_IN", "uniform_bound": 0.25, "random_stddev": 0.01, "dropout": 0.0 }, 使用自建数据，选择textCNN模型，非层次多标签任务，其他参数未动，"type": "embedding", 可正常运行。

问题出在哪，应该如何修改，谢谢？

[liqunhit]

在 model\embedding.py的201行之上，加上两行，试一试： mask = mask.to(context_embedding.device) neg_mask = neg_mask.to(context_embedding.device)