mckay
|
2e8dc3b0fe
|
可以直接从 step 开始处理的版本
|
2 months ago |
mckay
|
7f2049c4bd
|
加保存checkpoint
|
2 months ago |
mckay
|
584354bbf7
|
能在八叉树上面训练,但是速度很慢
|
3 months ago |
mckay
|
cabc98090e
|
BREAKPOINT: first attremp based on brepgen encoder and deepsef decoder
|
3 months ago |
mckay
|
0b4df54447
|
feat: filter the data if their face_num or edge_num > max_face or max_edge.
|
6 months ago |
mckay
|
4dddcdcd56
|
feat: diagnose script
|
6 months ago |
mckay
|
7c97b82fdf
|
feat: optimaize loss, by torch.norm and mead
|
6 months ago |
mckay
|
66fa9fcc58
|
fix: zero_grad mv afterwards
|
6 months ago |
mckay
|
86fb4c31e4
|
config: 修改
|
6 months ago |
mckay
|
e34d99a955
|
fix: loss 增加batch归一化
|
6 months ago |
mckay
|
6c76b66fb4
|
config: change batch and epoch
|
6 months ago |
mckay
|
0fa6a3ac3c
|
fix: neg sdf too small, -> shape error
|
6 months ago |
mckay
|
6d5c24c842
|
feat: script read npz
|
6 months ago |
mckay
|
de231aeb04
|
refactor: config
|
6 months ago |
mckay
|
bcc8ea8cf3
|
style: del log
|
6 months ago |
mckay
|
03fa4012a4
|
feat: 统计brep数据
|
6 months ago |
mckay
|
f81ca6040e
|
feat: SDF值分布log
|
6 months ago |
mckay
|
d9ecfbf4c7
|
style: 注释了不必要log
|
6 months ago |
mckay
|
959dafc4b1
|
refctor: print_tensor_stats 移入log
|
6 months ago |
mckay
|
41eb043d3f
|
fix: 梯度改在train里面加
|
6 months ago |
mckay
|
7f1ee89ba5
|
fix: del collate and load_brep
|
6 months ago |
mckay
|
ad58e0c12a
|
fix: 文件格式化器中使用了 caller_info 但没有正确设置
|
6 months ago |
mckay
|
e8e7f0e607
|
feat: log for param in debug mode
|
6 months ago |
mckay
|
530f680657
|
fix: network 测试脚本 tofix: some
norm =0
|
6 months ago |
mckay
|
ca2c3734d5
|
tofix: 概率性有梯度
|
6 months ago |
mckay
|
e1a76cf0ae
|
fix: can run, but still no grad
|
6 months ago |
mckay
|
7bfe8cad44
|
fix: grad escape
|
6 months ago |
mckay
|
b955b7d770
|
fix: git ignore 增加checkpoints/
|
7 months ago |
mckay
|
d8b5b545e2
|
fix:修复loss参数
|
7 months ago |
mckay
|
bf6f473edd
|
refactor: 在networks文件里面移除loss相关的东西
|
7 months ago |
mckay
|
9f9b5816eb
|
feat: test脚本
|
7 months ago |
mckay
|
608c6b75bf
|
feat: 完善log记录,可以显示报错位置
|
7 months ago |
mckay
|
0eb2063e28
|
refactor: 模型相关函数拆分
|
7 months ago |
mckay
|
49e0f8e9bc
|
fix: git ignore add *.pth
|
7 months ago |
mckay
|
df186deae6
|
fix: can train
|
7 months ago |
mckay
|
1110dc220e
|
fix: 由于设置的采样点太小,取消Encoder1D的池化操作
|
7 months ago |
mckay
|
d839dcc091
|
todo: 配置修改
|
7 months ago |
mckay
|
2f269e2150
|
增加sdf dataloader随机采样
|
7 months ago |
mckay
|
66f214790c
|
fix :trian
|
7 months ago |
mckay
|
fe4f00a323
|
refactor: config配net
|
7 months ago |
mckay
|
fe9a5a68a7
|
TODO: 太大加载不了
|
7 months ago |
mckay
|
53c1f51826
|
fix :BRepFeatureEmbedder
|
7 months ago |
mckay
|
6361188e03
|
fix: sdf_xyz和sdf_value 分开加载
|
7 months ago |
mckay
|
6c452e1007
|
fix:使用size()而不是shape来获取维度
|
7 months ago |
王琛涵
|
3b494aa6e0
|
fix :train script
|
7 months ago |
mckay
|
271d237b66
|
fix: ignore wandb files
|
7 months ago |
mckay
|
055a3e6ee6
|
fix: config 路径使用绝对路径
|
7 months ago |
王琛涵
|
3de9a39ef5
|
fix : config refactor log & details separate
|
7 months ago |
王琛涵
|
e5e5034991
|
refactor & style: 用config配置train,同时增加保存checkpoint的信息
|
7 months ago |
王琛涵
|
e12c492d69
|
fix: net dim问题。num*3,变成[num,3],保留空间特征
|
7 months ago |