search

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

[Examples] Update DGL Examples #96

Closed yaox12 closed 2 years ago

yaox12 commented 2 years ago
  1. Update the DGL example to make it work.
  2. Add multiple feature storage locations (cpu|gpu|quiver|unified) and CPU/GPU sampling support. Also attach a Readme for more details.
  3. Modify the training script to have similar outputs and evaluation logic as the PyG version.
  4. Remove examples/dgl/ogbn_products_sage_dgl.py because it's included in the current unified version.

To Do

yaox12 commented 2 years ago

Hi maintainers, any advice on the single-GPU DGL example? If overall good, I will go ahead with the multi-GPU example.

eedalong commented 2 years ago

@yaox12 Hi， Great Job！We will discuss about this!

eedalong commented 2 years ago

@yaox12 Single GPU example is great, please move on to multi-gpu examples! Looking forward to that!

luomai commented 2 years ago

Thanks for the PR. I am thinking if we can make single-GPU DGL and multi-GPU DGL two separated PRs.

yaox12 commented 2 years ago

Sure, I'm supportive of putting the multi-GPU DGL examples in another PR. I'm a little busy this week and haven't got much time to finish it. Feel free to merge this PR if it looks good for you. Thanks.

eedalong commented 2 years ago

Sure, I'm supportive of putting the multi-GPU DGL examples in another PR. I'm a little busy this week and haven't got much time to finish it. Feel free to merge this PR if it looks good for you. Thanks.

Great!, We will merge this pr first

luomai commented 2 years ago

Merged. Thanks a lot for the PR. @yaox12 :)