quiver-team / torch-quiver

PyTorch Library for Low-Latency, High-Throughput Graph Learning on GPUs.
https://torch-quiver.readthedocs.io/en/latest/
Apache License 2.0
293 stars 36 forks source link

Does the feature caching module support different GPU cache redundant data? #167

Open Baizx98 opened 10 months ago

Baizx98 commented 10 months ago

I want to develop a quiver-0.1 caching module and apply a new caching strategy on it. In the presence of nvlinks, different GPUs will cache a portion of the same data. I would like to know if quiver's feature module allows me to directly modify the cached data on each GPU?