Comments (7)
Hi,
The printed results at test time (precision@1) are not multiplied with 100. (test.py line 91)
The printed results during training (top 1 scores) are multiplied with 100. (semi_learner.py line 151)
So actually you need to multiple 100 if you are measuring as percentage (%).
from semi-memory.
Hi, the accuracy I talked about was in percent for both.
from semi-memory.
Hi,
Could you paste the results here?
Ideally, the printed results are something as follows during training/testing:
(1) training:
epoch 1, L = 3.84, Ls = 1.63, Lm = 2.05, top_1 = 42.23, lr = 0.0200 (time_per_iter: 0.0623 s)
(2) testing:
2018-09-17 20:54:19.643729: precision @ 1 = 0.9607 recall @ 5 = 0.9921
As said before, top_1 = 42.23 means an accuracy of 42.23%.
precision @ 1 = 0.9607 means an accuracy of 96.07%.
from semi-memory.
testing: precision @ 1 = 0.0062 recall @ 5 = 0.4145
from semi-memory.
These are the results obtained for the first three epochs on training
epoch 0, L = 3.76, Ls = 1.59, Lm = 2.02, top_1 = 44.25, lr = 0.0199 (time_per_iter: 0.3763 s)
epoch 1, L = 1.42, Ls = 0.36, Lm = 0.90, top_1 = 91.30, lr = 0.0198 (time_per_iter: 0.3668 s)
epoch 2, L = 0.82, Ls = 0.20, Lm = 0.46, top_1 = 96.93, lr = 0.0197 (time_per_iter: 0.3637 s)
from semi-memory.
The training acc looks fine as it reflects only a small batch of labelled data.
The testing acc is weird.
We double-check and re-run this code which gives
2019-01-23 23:45:18.043114: precision @ 1 = 0.9575 recall @ 5 = 0.9886 [26032 examples]
(tested with tensorflow 1.5.0 and python 2.7.12.)
Nothing turns out to cost a weird result of 0%.
from semi-memory.
Thanks for your quick reply. I spotted the issue but I havent found a good fix for it. I was solving the models and data on another machine so results/
and tfrecords/
were symlinked. It seems like it might cause an issue in this case. Im closing the issue since it is not related to your code.
from semi-memory.
Related Issues (7)
- Only working with python2.X? HOT 1
- model_checkpoint_path = ckpt.model_checkpoint_path AttributeError: 'NoneType' object has no attribute 'model_checkpoint_path' HOT 4
- key-value updates HOT 5
- Why not use the memory module during test? HOT 1
- Have you run on cifar10 with fewer labeled data?
- data argumentation HOT 1
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 semi-memory.