NSY
NSY
Would you release the code of EMNLP 2021 paper "Generate & Rank: A Multi-task Framework for Math Word Problems"?
hi, Great Work! I am writing to ask whether I can use naturalcc to construct AST CFG and DFG for python and java. Thanks for your reply.
hi, great work!! I am wonder if the ast can be Visualizatized as a picture by tree-sitter. I do appreciate for your reply~
Hi，thanks for the amazing work! I wank to ask about the detailed steps about generated code solutions post-processing when testing one solution. (e.g. After a code solution was generated, did...
您好，很有趣的文章！ RocketQAv2中提及，正负样本的组成方式为： 融合数据增强构造的样本包括非去噪的和去噪的。使用RocketQA中的检索模型召回top-n的passage作为备选集。非去噪样本由标注正例和对备选集随机采样得到的样本构成；去噪样本由RocketQA的精排模型对备选集内passage打分得到的去噪正负样本构成。 但是，精排模型的损失为：  我想请问，此处的分子上的p+指的是ground truth呢，还是所有正样本呢？（即ground truth 加所有去噪得到的正样本） 十分感谢！
您好！感谢您精彩有趣的工作！ 在看《JOINT GENERATOR-RANKER LEARNING FOR NATURAL LANGUAGE GENERATION》论文代码时，我发现一个问题：在更新generator计算reward时，reward由两部分相加得到，即reranker_rewards和metric_rewards，但是在将两部分相加前，似乎却没有归一化，self.args.normalize_rewards是False。这样reranker_rewards和metric_rewards便相差了几个数量级，但是论文得出了reranker_rewards更加重要的结论。 不知道我是否遗漏了什么？ 感谢您的解惑！ 以下代码位于：JGR/trainer_utils/trainer.py的compute_loss_generator函数中 ``` self.reward_tracker['reranker_rewards'].append(reranker_rewards.detach().cpu().numpy().tolist()) self.reward_tracker['metric_rewards'].append(metric_rewards.detach().cpu().numpy().tolist()) if self.args.normalize_rewards: # rererank_rewards_std = torch.std(reranker_rewards, dim=1, keepdim = True) metric_rewards_std = torch.std(metric_rewards, dim=1, keepdim =...
Hi，thanks for the amazing work! When I try to rename variable name with recode-main/natgen/transformations/var_renaming_transformation.py，I find when extracting variable name，some function name can be extracted even after flitterng withget_not_var_ptype_var_names(self, root, code_string)...
Hi, nice work! I would like to ask whether you will provide the evaluation results in this repository. thanks!