lcmeng

Training logs for Swin-B/S/L

12

Hi authors, Following the official training command below, I observed unstable training loss and accuracies around epoch # 20. `python -m torch.distributed.launch --nproc_per_node 8 --master_port 12345 main.py \ --cfg configs/swin_base_patch4_window7_224.yaml...

Thanks for sharing the good work. I have a couple questions about the constant `TOKEN_SELF_ATTN_VALUE ` and how it is used. `TOKEN_SELF_ATTN_VALUE` is first defined [here](https://github.com/lucidrains/reformer-pytorch/blob/365d5c8918dc643a4a606da021253936d82577fc/reformer_pytorch/reformer_pytorch.py#L18) in reformer_pytorch.py with a...

RuntimeError with GCN predictor in installation test

2

Command: `python naszilla/run_experiments.py --search_space nasbench_101 --algo_params all_algos --queries 30 --trials 1` Log: ``` ... Finished GP-BayesOpt query 0 Finished GP-BayesOpt query 10 * Running NAS algorithm: {'algo_name': 'gcn_predictor', 'total_queries': 30}...

Hi, In reference to issue #36, according to the README: > After searching is done, the top1 encoding vector will be shown in the log. By simply copying the encoding...

Any plan to release the configurations that can reproduce the ImageNet/COCO results?

3

It'll be great to try out something other than CIFAR10. TIA.