double Eric
double Eric
谢谢您！ 我自己再de一下 yi-jh16 ***@***.*** 签名由 网易邮箱大师 定制 在2021年06月09日 13:28，Gongfan Fang 写道： 训练的时候可以用裁剪的方式，具体裁剪大小根据你的计算资源决定。 inference阶段按照paper是需要在原始大小上测试的。 — You are receiving this because you authored the thread. Reply to this email directly, view it...
i get the same question
pip install albumentations==0.4.6就可以了
i got the same problem!
> @shutttttdown 我尝试了这段代码，能够产出attn_weights。 `if not output_attentions: attn_weights = None else: attn_weights = torch.matmul( query_states, key_states.transpose(2, 3) ) / math.sqrt(self.head_dim) > > ``` > if attention_mask is not None: > if...
> 我通过使用open3d的方法，写了可视化的代码 import open3d as o3d import numpy as np import json import os > > # 使用绝对路径 > pointcloud_file_path = 'E:/mmdetection3d-main/demo/data/kitti/000008.bin' result_file_path = 'E:/mmdetection3d-main/outputs/preds/000008.json' # 更新路径 > > if...
have you solved the problem?
@narges-tk hello, have you solved the issue? it's bothering me now
have you solved the problem?