weixin_41318625的博客

mteb下载，pytrec，trec_eval

RuntimeError: one of the variables needed for gradient computation has been modified by an inplace operation: [torch.cuda.LongTensor [896, 1]] is at version 11539; expected version 1 instead. Hint: enable anomaly detection to find the operation that failed

import numpy as npimport torch from transformers import BertTokenizer, BertConfig, BertForMaskedLM, BertForNextSentencePredictionfrom transformers import BertModelmodel_name = 'bert-base-chinese'MODEL_PATH = 'F:/Transformer-Bert/bert-base-chinese/'

https://zhuanlan.zhihu.com/p/462469487

fairseq训练框架，原代码解析，fairseq的训练过程是如何进行的

1. 首先定义的类是整个模型的类GuidedTransformerModel(FairseqEncoderDecoderMoel)通过@register的方式注册模型，模型继承FairseqEncoderDecoderModel@register_model("guided_transformer")class GuidedTransformerModel(FairseqEncoderDecoderModel):1）定义init函数 def __init__(self, args, enc