 mckay
|
6c76b66fb4
|
config: change batch and epoch
|
9 months ago |
|
mckay
|
0fa6a3ac3c
|
fix: neg sdf too small, -> shape error
|
9 months ago |
|
mckay
|
6d5c24c842
|
feat: script read npz
|
9 months ago |
|
mckay
|
de231aeb04
|
refactor: config
|
9 months ago |
|
mckay
|
bcc8ea8cf3
|
style: del log
|
9 months ago |
|
mckay
|
03fa4012a4
|
feat: 统计brep数据
|
9 months ago |
|
mckay
|
f81ca6040e
|
feat: SDF值分布log
|
9 months ago |
|
mckay
|
d9ecfbf4c7
|
style: 注释了不必要log
|
9 months ago |
|
mckay
|
959dafc4b1
|
refctor: print_tensor_stats 移入log
|
9 months ago |
|
mckay
|
41eb043d3f
|
fix: 梯度改在train里面加
|
10 months ago |
|
mckay
|
7f1ee89ba5
|
fix: del collate and load_brep
|
10 months ago |
|
mckay
|
ad58e0c12a
|
fix: 文件格式化器中使用了 caller_info 但没有正确设置
|
10 months ago |
|
mckay
|
e8e7f0e607
|
feat: log for param in debug mode
|
10 months ago |
|
mckay
|
530f680657
|
fix: network 测试脚本 tofix: some
norm =0
|
10 months ago |
|
mckay
|
ca2c3734d5
|
tofix： 概率性有梯度
|
10 months ago |
|
mckay
|
e1a76cf0ae
|
fix: can run, but still no grad
|
10 months ago |
|
mckay
|
7bfe8cad44
|
fix: grad escape
|
10 months ago |
|
mckay
|
b955b7d770
|
fix: git ignore 增加checkpoints/
|
10 months ago |
|
mckay
|
d8b5b545e2
|
fix:修复loss参数
|
10 months ago |
|
mckay
|
bf6f473edd
|
refactor: 在networks文件里面移除loss相关的东西
|
10 months ago |
|
mckay
|
9f9b5816eb
|
feat: test脚本
|
10 months ago |
|
mckay
|
608c6b75bf
|
feat: 完善log记录，可以显示报错位置
|
10 months ago |
|
mckay
|
0eb2063e28
|
refactor: 模型相关函数拆分
|
10 months ago |
|
mckay
|
49e0f8e9bc
|
fix: git ignore add *.pth
|
10 months ago |
|
mckay
|
df186deae6
|
fix: can train
|
10 months ago |
|
mckay
|
1110dc220e
|
fix: 由于设置的采样点太小，取消Encoder1D的池化操作
|
10 months ago |
|
mckay
|
d839dcc091
|
todo: 配置修改
|
10 months ago |
|
mckay
|
2f269e2150
|
增加sdf dataloader随机采样
|
10 months ago |
|
mckay
|
66f214790c
|
fix :trian
|
10 months ago |
|
mckay
|
fe4f00a323
|
refactor： config配net
|
10 months ago |
|
mckay
|
fe9a5a68a7
|
TODO: 太大加载不了
|
10 months ago |
|
mckay
|
53c1f51826
|
fix ：BRepFeatureEmbedder
|
10 months ago |
|
mckay
|
6361188e03
|
fix: sdf_xyz和sdf_value 分开加载
|
10 months ago |
|
mckay
|
6c452e1007
|
fix:使用size()而不是shape来获取维度
|
10 months ago |
|
mckay
|
3b494aa6e0
|
fix :train script
|
10 months ago |
|
mckay
|
271d237b66
|
fix: ignore wandb files
|
10 months ago |
|
mckay
|
055a3e6ee6
|
fix: config 路径使用绝对路径
|
10 months ago |
|
mckay
|
3de9a39ef5
|
fix : config refactor log & details separate
|
10 months ago |
|
mckay
|
e5e5034991
|
refactor & style： 用config配置train，同时增加保存checkpoint的信息
|
10 months ago |
|
mckay
|
e12c492d69
|
fix: net dim问题。num*3,变成[num,3],保留空间特征
|
10 months ago |
|
mckay
|
ee598e8340
|
refactor: 用config替代硬编码
|
10 months ago |
|
mckay
|
2aafbc87e3
|
refactor: process_brep，用config替代原来的100边采样
|
10 months ago |
|
mckay
|
b0b93b70ef
|
feat: 增加全局config
|
10 months ago |
|
mckay
|
4d4dbb7c7d
|
feat: train 脚本
|
10 months ago |
|
mckay
|
84fac4660c
|
test： 修改数据集加载测试
|
10 months ago |
|
mckay
|
31fcbecfea
|
fix: 修改数据加载process_brep_data，防止维度超过最大（用截取）
|
10 months ago |
|
mckay
|
bee51c9dc0
|
修复: 改进数据预处理中的数组填充逻辑
1. 修复了pad_zero函数中的负数维度问题
- 添加了对超长输入的截断处理
- 优化了掩码生成逻辑
- 保持多维数组的维度一致性
2. 主要改动：
- 处理数组长度 > max_len：进行截断
- 处理数组长度 < max_len：进行填充
- 处理数组长度 = max_len：直接返回
- 改进布尔掩码的生成方式
|
10 months ago |
|
mckay
|
dfaeca63ba
|
fix: 数据集可以加载
|
10 months ago |
|
mckay
|
295e707e2b
|
fix: 修改了check_data_format函数，正确检查对象数组类型
在parse_solid函数中确保所有数组都有正确的类型
对点云数据使用对象数组类型
对其他数据使用正确的数值类型(float32或int32)
|
10 months ago |
|
mckay
|
7ed86aee29
|
fix: corners维度变成[num_faces, max_edge, 2, 3]
|
10 months ago |
