I'm currently exploring the implementation provided for your paper, "Large Pre-trained Time Series Models for Cross-domain Time Series Analysis Tasks," and I have encountered some confusion.
Specifically:
- In the paper, it mentions using a GRU to obtain hidden embeddings; however, I cannot find this implementation within the provided code.
- The paper describes an "Adaptive Segmentation" method, but the segmentation logic in
Samay/src/samay/models/lptm/model/backbone.py seems to employ fixed-size patches.
- Additionally, the function
select_segments in Samay/src/samay/models/lptm/segment/selection.py does not appear to be used anywhere in the provided codebase.
- The method
LPTMPipeline.from_pretrained is referenced, but I can't locate its definition in the repository.
Could you please clarify whether there is an error in the uploaded code or direct me to the correct implementation and usage of these components?
I'm currently exploring the implementation provided for your paper, "Large Pre-trained Time Series Models for Cross-domain Time Series Analysis Tasks," and I have encountered some confusion.
Specifically:
Samay/src/samay/models/lptm/model/backbone.pyseems to employ fixed-size patches.select_segmentsinSamay/src/samay/models/lptm/segment/selection.pydoes not appear to be used anywhere in the provided codebase.LPTMPipeline.from_pretrainedis referenced, but I can't locate its definition in the repository.Could you please clarify whether there is an error in the uploaded code or direct me to the correct implementation and usage of these components?