ValueError: Shape of variable bert/embeddings/LayerNorm/beta:0 ((768,)) doesn't match with shape of tensor bert/embeddings/LayerNorm/beta ([312]) from checkpoint reader.

[learnpythontheew]

运行第四步的时候遇到了这样的报错： ValueError: Shape of variable bert/embeddings/LayerNorm/beta:0 ((768,)) doesn't match with shape of tensor bert/embeddings/LayerNorm/beta ([312]) from checkpoint reader.

执行run_lasertagger.py 文件 model config 用的lasertagger 用了你提供的链接下载的RoBERTa-tiny-clue模型

请问是哪里出现了问题呢？

[tongchangD]

训练模型维度不一样，抱歉，我上传的文件夹 中lasertagger_config.json， 是有内容的，GitHub中没了 你可以将 lasertagger_config.json 修改如下 { "attention_probs_dropout_prob": 0.1, "hidden_act": "gelu", "hidden_dropout_prob": 0.1, "hidden_size": 312, "initializer_range": 0.02, "intermediate_size": 1248, "max_position_embeddings": 512, "num_attention_heads": 12, "num_hidden_layers": 4, "type_vocab_size": 2, "vocab_size": 8021, "use_t2t_decoder": true, "decoder_num_hidden_layers": 1, "decoder_hidden_size": 768, "decoder_num_attention_heads": 4, "decoder_filter_size": 3072, "use_full_attention": false }