Comments (9)
美女你好,我想问一下,caffe里面怎么配置他这个Python layer,除了取消注释WITH_PYTHON_LAYER := 1,其他还需要设置什么?怎么才能将他写的Python layer加入到caffe里面呢?
from tripletloss.
@louielu1027 没有其它设置,如果有报错,可能是代码里面其它的错误,你看下别人的issue
from tripletloss.
美女,我想再问一下,在作者read me 里面,setup里说的Rebuild your caffe directory and makesure your python could find the added layers.后面半句“确保你的Python能找到这些添加的层”是什么意思?需要对应什么操作,或者在caffe某些文件里添加什么吗?
from tripletloss.
因为我现在出现了一个错误,一直解决不了
File "/home/dl1/lvlu/10000all/tripletloss-master/tripletloss/datalayer.py", line 82, in setup
layer_params = yaml.load(self.param_str_)
AttributeError: 'DataLayer' object has no attribute 'param_str_'
from tripletloss.
@louielu1027 把param_str_ 改为 param_str就可以了
from tripletloss.
@twinsyssy1018 你好~请问最后你效果出来了么?我也是这个情况,觉得是不是batch size太小的原因。 但是按照代码里的方法,每一类图片最小的数字是有要求的,要不就会无限循环。所以batch size我又不能取得太大。想知道你最后是怎么解决的呢?
from tripletloss.
我的LOSS很大,貌似没收敛,有人遇到过这种情况吗?BATCHSIZE=42,1000人左右
speed: 1.262s / iter
I0119 11:29:48.248039 1499 solver.cpp:337] Iteration 1000, Testing net (#0)
I0119 11:29:49.508103 1499 solver.cpp:404] Test net output #0: accuracy = 0
I0119 11:29:49.508152 1499 solver.cpp:404] Test net output #1: loss_cls = 9.70228 (* 1 = 9.70228 loss)
I0119 11:29:50.766808 1499 solver.cpp:228] Iteration 1000, loss = 10.9084
I0119 11:29:50.766860 1499 solver.cpp:244] Train net output #0: loss_cls = 10.9084 (* 1 = 10.9084 loss)
I0119 11:29:50.766870 1499 sgd_solver.cpp:106] Iteration 1000, lr = 0.05
I0119 11:30:15.997308 1499 solver.cpp:228] Iteration 1020, loss = 10.1421
I0119 11:30:15.997361 1499 solver.cpp:244] Train net output #0: loss_cls = 10.1421 (* 1 = 10.1421 loss)
I0119 11:30:15.997371 1499 sgd_solver.cpp:106] Iteration 1020, lr = 0.05
I0119 11:30:41.227283 1499 solver.cpp:228] Iteration 1040, loss = 10.7562
I0119 11:30:41.227332 1499 solver.cpp:244] Train net output #0: loss_cls = 10.7562 (* 1 = 10.7562 loss)
I0119 11:30:41.227342 1499 sgd_solver.cpp:106] Iteration 1040, lr = 0.05
I0119 11:31:06.449278 1499 solver.cpp:228] Iteration 1060, loss = 10.7111
I0119 11:31:06.449329 1499 solver.cpp:244] Train net output #0: loss_cls = 10.7111 (* 1 = 10.7111 loss)
I0119 11:31:06.449338 1499 sgd_solver.cpp:106] Iteration 1060, lr = 0.05
I0119 11:31:31.674744 1499 solver.cpp:228] Iteration 1080, loss = 8.994
I0119 11:31:31.674787 1499 solver.cpp:244] Train net output #0: loss_cls = 8.994 (* 1 = 8.994 loss)
I0119 11:31:31.674796 1499 sgd_solver.cpp:106] Iteration 1080, lr = 0.05
I0119 11:31:56.904220 1499 solver.cpp:228] Iteration 1100, loss = 9.24442
I0119 11:31:56.904271 1499 solver.cpp:244] Train net output #0: loss_cls = 9.24442 (* 1 = 9.24442 loss)
I0119 11:31:56.904281 1499 sgd_solver.cpp:106] Iteration 1100, lr = 0.05
I0119 11:32:22.130637 1499 solver.cpp:228] Iteration 1120, loss = 11.5263
I0119 11:32:22.130688 1499 solver.cpp:244] Train net output #0: loss_cls = 11.5263 (* 1 = 11.5263 loss)
I0119 11:32:22.130698 1499 sgd_solver.cpp:106] Iteration 1120, lr = 0.05
I0119 11:32:47.357022 1499 solver.cpp:228] Iteration 1140, loss = 10.7919
I0119 11:32:47.357074 1499 solver.cpp:244] Train net output #0: loss_cls = 10.7919 (* 1 = 10.7919 loss)
I0119 11:32:47.357084 1499 sgd_solver.cpp:106] Iteration 1140, lr = 0.05
I0119 11:33:12.589174 1499 solver.cpp:228] Iteration 1160, loss = 10.9585
I0119 11:33:12.589224 1499 solver.cpp:244] Train net output #0: loss_cls = 10.9585 (* 1 = 10.9585 loss)
I0119 11:33:12.589234 1499 sgd_solver.cpp:106] Iteration 1160, lr = 0.05
I0119 11:33:37.824996 1499 solver.cpp:228] Iteration 1180, loss = 9.55746
I0119 11:33:37.825044 1499 solver.cpp:244] Train net output #0: loss_cls = 9.55746 (* 1 = 9.55746 loss)
I0119 11:33:37.825054 1499 sgd_solver.cpp:106] Iteration 1180, lr = 0.05
speed: 1.263s / iter
I0119 11:34:03.060225 1499 solver.cpp:228] Iteration 1200, loss = 12.1203
I0119 11:34:03.060277 1499 solver.cpp:244] Train net output #0: loss_cls = 12.1203 (* 1 = 12.1203 loss)
I0119 11:34:03.060286 1499 sgd_solver.cpp:106] Iteration 1200, lr = 0.05
I0119 11:34:28.291772 1499 solver.cpp:228] Iteration 1220, loss = 11.2631
I0119 11:34:28.291823 1499 solver.cpp:244] Train net output #0: loss_cls = 11.2631 (* 1 = 11.2631 loss)
I0119 11:34:28.291833 1499 sgd_solver.cpp:106] Iteration 1220, lr = 0.05
from tripletloss.
@luxiangju @twinsyssy1018 我的loss也是在很小范围内振荡,根本不收敛 请问你们解决了这个问题了吗?谢谢
from tripletloss.
loss在margin附近震荡,你们解决好了吗
from tripletloss.
Related Issues (20)
- the module name is right,but it still failed with boost::python::error_already_set
- 训练到后面 loss都变为0.1 an,ap变为0 HOT 6
- training problem HOT 5
- training problem
- Hard Sample
- the triplet loss architecture is unsupervised, is this correct?
- tipletselectlayer - computing the distance against the anchor image HOT 2
- shuffle
- fc9_1 weights is increase with the train iterations from 0.0x to 40.x HOT 1
- Training your code on custom dataset HOT 1
- how to train it on 2 gpus
- Use of the margin HOT 2
- Negative mining in TripletSelectLayer HOT 1
- what does no_residual_list for?
- 数据集组织需要什么特别处理么? HOT 1
- online triplet sample selection usage?
- 您好,我想用自己的数据集跑这个网络,除了修改config路径,还需要修改哪里? HOT 2
- Online triplet generation HOT 1
- 工作推荐
- ap,an都变得特别大
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from tripletloss.