advNLG

advNLG copied to clipboard

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 =...

MrBlack0220

Hello: I found this GitHub link in the paper `Metric-guided Distillation: Distilling Knowledge from the Metric to Ranker and Retriever for Generative Commonsense Reasoning`, but it doesn't seem to be...

QQQrookie