Peng Chen
I also met error when evaluating: MTR/tools/eval_utils/eval_utils.py", line 66, in eval_one_epoch logger.info(f'Total number of samples after merging from multiple GPUs (removing duplicate): {len(pred_dicts)}') TypeError: object of type 'NoneType' has no...
Please look at the 'test_detection.cpp' as an example, though I have not had a try.
@AlexofNTHU Hi, may I ask which script you use to computing the AP? Could you share an example?
阅读代码的时候，可以看到mse还是Fisher-diag是可以配置的。不过好奇使用这两个对精度的影响有多大？ 论文里貌似没这两组的ablation study.
@csyhhu Do you have any update on the accuracy? I got 87% on CIFAR10, still 4 points lower than the paper.
@Yulun-Yao Hi, I'm very sorry that I've gaven up the TTN. I take weeks and tried different optimizer strategy as well as modified the gradient based on empirecal experiences gathered...
Hi, I tried to implement on myself. However, I can not figure out the output format. ``` import torch import torch.nn as nn import pdb class CustomLSTM(nn.LSTM): def __init__(self, input_size,...
if I employ ``` # stack hidden_seq_forward and hidden_seq_reverse to hidden_seq hidden_seq_forward = torch.cat(hidden_seq_forward, dim=0) # [seq, bs, self.hidden_size] hidden_seq_reverse = torch.cat(hidden_seq_reverse, dim=0) # [seq, bs, self.hidden_size] print(hidden_seq_forward.shape, hidden_seq_reverse.shape) hidden_seq...
@pobizheV Have you solved the issue? I also met the problem.