NegativeBinomialLikelihood - Invalid values during training

描述

贡献者指南

技术栈

pythonpytorch

领域

machine learningdata

议题类型

bug

难度面向新贡献者的预计实现难度，1 表示很小改动，5 表示专家级工作。

3

预计时间有经验贡献者完成调查、实现、测试并准备 pull request 的粗略时间范围。

1-3 hours

活动状态议题当前的可参与程度：新鲜、活跃、陈旧、阻塞或等待维护者输入。

stale

清晰度议题是否清楚说明期望改动、验收标准和下一步。

needs investigation

前置要求

Basic understanding of probabilistic modelsFamiliarity with PyTorch

新手友好度1-100 的估计分数，表示该议题对首次贡献者的友好程度。

20

研究方向

Investigate the NegativeBinomialLikelihood class in the darts codebase, likely in darts/models/likelihood.py or similar. The issue states that parameters assume invalid values during training, so check for parameter constraints (e.g., variance greater than mean) and ensure proper initialization or clamping. Look at the training loop for any gradient issues or unbounded parameter updates. Since there are no comments or PRs, a good starting point is to reproduce the error with a minimal example and trace the parameter values.