awslabs/gluonts

`lagged_sequence_values` helper does not work well with multivariate data

Open

#2,708 创建于 2023年3月6日

在 GitHub 查看
 (7 评论) (0 反应) (0 负责人)Python (3,888 star) (753 fork)batch import
enhancementgood first issuemultivariate

描述

one issue i found with the lagged_sequence_values helper is that with multivariate inputs this helper doesn't return the stacked tensors... e.g.

# prior_input.shape torch.Size([1, 167, 137])
# input.shape torch.Size([1, 24, 137])

lags = lagged_sequence_values([0, 167], prior_input, input, dim=1)
# lags.shape torch.Size([1, 24, 137, 2])

where i would have expected torch.Size([1, 24, 137*2])

Originally posted by @kashif in https://github.com/awslabs/gluonts/issues/2618#issuecomment-1455833300

贡献者指南

技术栈
pythonpytorch
领域
datamachine learning
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。
2
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。
under 1 hour
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。
fresh
清晰度议题是否清楚说明期望改动、验收标准和下一步。
clear
前置要求
PythonPyTorchGluonTS basics
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。
80
研究方向
Examine the implementation of `lagged sequence values` in the gluonts source code (likely in `gluonts/transform/...`). Understand how it handles multivariate inputs and modify it to return stacked tensors along the feature dimension (concatenated) instead of a new dimension. The expected behavior is to flatten the lagged features, resulting in shape `[1, 24, 137*2]` rather than `[1, 24, 137, 2]`. Refer to the discussion in the linked comment for more context.