yja1
yja1
I have try change but also has error def load_params(self, iteration): cache_file = system_configs.snapshot_file.format(iteration) print("loading model from {}".format(cache_file)) with open(cache_file, "rb") as f: from collections import OrderedDict new_state_dict = OrderedDict()...
dets, center = decode_func(nnet, images, K, ae_threshold=ae_threshold, kernel=nms_kernel) run here, Segmentation fault (core dumped) ???
[err.txt](https://github.com/Duankaiwen/CenterNet/files/3235487/err.txt) @Duankaiwen
I test on wider face(input src image), only 0.17 easy
> 请把测试sample发上来试一下？ 猜测是角度比较大 https://github.com/harlanhong/CVPR2022-DaGAN/assets/24217235/80ffbc4f-cfee-4d55-9f43-51494a2f6a76
> 这是关于自定义mask，传统实现attention和xformer实现的差异。 > > ```python > import xformers.ops as xops > import torch > from torch import nn > from torch.nn import functional as F > import pdb > >...
maybe first check out your anchors_numbers, then do as BOBrown said [train_center.txt](https://github.com/BOBrown/SSD-Centerloss/files/2964355/train_center.txt)
> In next verision, we will release a new vae to figure out this problem. （What time is it approximately） the reason for this is conv3d input size too big.
> In next verision, we will release a new vae to figure out this problem. can open-sora-plan train one-step generation like ADD or LCM