08 2024 档案

Transformer based Pluralistic Image Completion with Reduced Information Loss
摘要：They quantize 256^3 RGB values to a small number (such as 512) of quantized color values. The indices of quantized pixels are used as tokens for the i 阅读全文
posted @ 2024-08-13 11:13 Trkly 阅读(3) 评论(0) 推荐(0) 编辑
手写Swin Transformer
摘要：1.如何基于图片生成patch_embedding? 方法一： - 基于pytorch unfold的API来将图片进行分块，也就是模仿卷积的思路，设置kernel_size=stride=patch_size,得到分块后的图片 - 得到格式为[bs, num_patch, patch_depth] 阅读全文
posted @ 2024-08-09 15:21 Trkly 阅读(11) 评论(0) 推荐(0) 编辑
手写VIT
摘要：import torch import torch.nn as nn import torch.nn.functional as F from einops import rearrange, repeat from einops.layers.torch import Rearrange clas 阅读全文
posted @ 2024-08-08 17:42 Trkly 阅读(7) 评论(0) 推荐(0) 编辑

点击右上角即可分享
微信分享提示