YUKE WANG
YUKE WANG
**Hi, All** I am encounter a problem after running the following training command of `ogb_lsc/pcq/` at https://github.com/deepmind/deepmind-research/tree/master/ogb_lsc/pcq#retraining-our-model ``` /bin/bash run_training.sh -r ${HOME}/pcq/ ``` The error is ``` I0126 23:29:54.918133 139951828096768...
https://github.com/xiezhq-hermann/graphiler/blob/ddba5ba1df518d555012ef597ad24c308d752583/examples/GCN/GCN.py#L51-L52 is the order for `hidden_dim ` and `out_dim ` in layer2 correct? or it should be reversed.
For Gloo in Pytorch distributed, as shown in this document https://pytorch.org/docs/stable/distributed.html, will the following code get performance benefits of using CUDA-aware MPI? (e.g., GPU-to-GPU transferring via PCIe while bypassing CPU)...
Is there any CUTLASS support for X*W = Y, where `X` (Row-major) and `Y`(row-major) are dense matrics, `W`is a weight matrix with sparsity like 2:4 that can leverage the GPU...
The `graph_name` in preprocessed graph like `amazon_review` requires a valid format such that the training/inference can be successfully launched. Based on evaluation, graph name like `/amazon_review` will trigger runtime error...