Comments (24)
已修复
from libmtl.
哥们,还是有报错的,alpha可能是numpy也可能是tensor
from libmtl.
哪个配置下有报错,麻烦提供一下运行命令,谢谢
from libmtl.
恩恩,好的,我明天在跑一下告诉你。我新发现一个nashmtl的问题你能指导我一下吗?我传入2任务的loss,数值分别是700多和600多,结果乘上nashmtl后,总的loss变成了0.5,有什么超参数可以调节的吗,这导致我模型收敛不了。
from libmtl.
哪个配置下有报错,麻烦提供一下运行命令,谢谢
不好意思,是我看错了你的commit,修改错了,你的修改是ok的,已经没有报错了。另外,你能帮我看看我哦的bug吗,nashmtl算出来的weight_loss好小,是10-4级别的,是要调整什么超参数吗?谢谢
from libmtl.
这个问题可能需要问一下Nash-MTL的作者们
from libmtl.
这个问题可能需要问一下Nash-MTL的作者们
问了。但感觉,梯度很大时候,nashmtl就会计算出很小的loss权重(10-4),反传回去时候网络参数的梯度就会非常小,整个网络根本迭代不动。
from libmtl.
我也没想到什么好方法
from libmtl.
下有报错,麻烦提供一下运行命令,谢
这个问题可能需要问一下Nash-MTL的作者们
问了。但感觉,梯度很大时候,nashmtl就会计算出很小的loss权重(10-4),反传回去时候网络参数的梯度就会非常小,整个网络根本迭代不动。
请问,你这边用多任务算法的时候,有什么实际work的算法吗,我这边在实际使用的时候其实都会比单任务差很多
from libmtl.
我这边主要是yolox架构的object detection跟自驾任务中的freespace,这两个任务,每个任务都有若干loss,但是用MTL来优化其实远比我随便调一下loss权重要差
from libmtl.
@MingChaoXu
你都试了哪些方法
from libmtl.
我尝试了pcgrad,uw,nash mtl,cagrad,graddrop,都不好使,不知道你这有没有work的经验?
from libmtl.
@MingChaoXu
和nashmtl的作者聊的,可能要对输出做归一化,AvivNavon/nash-mtl#13
你的nash mtl没问题吗
from libmtl.
我这边nash mtl可以收敛,就是精度变差了,你有尝试别的算法吗?效果怎么样?
from libmtl.
我就是pcgrad,有一点点效果,试了nashmtl不行,现在在尝试gradnorm,我也是用的yolox的主干。感觉可能和你的数据分布相关
from libmtl.
我的两个任务的数据并没有共同标注,就是每份数据只有一个任务的标注,不知道这个会不会影响,因为我看论文里的数据好像是多任务标注都有的?
from libmtl.
这不是非同源数据吗?
from libmtl.
我尝试了pcgrad,uw,nash mtl,cagrad,graddrop,都不好使,不知道你这有没有work的经验?
from libmtl.
是的,可以说非同源数据吧,domain是一样的,只不过只有单个数据的标注,你的任务是两种标注都有是吗?
from libmtl.
shide
from libmtl.
谢谢作者,nashmtl这个方法我先不用了吧。。。。
from libmtl.
是的,可以说非同源数据吧,domain是一样的,只不过只有单个数据的标注,你的任务是两种标注都有是吗?
可以看看这篇paper https://arxiv.org/pdf/2209.11379.pdf
另外,可以加下微信,讨论下吗,DayDayAmazing,备注mtl
from libmtl.
from libmtl.
@yushengjiexy 好的~互相学习一下hhh,加你了
from libmtl.
Related Issues (20)
- Does your project support AMP? HOT 2
- Saving and loading models HOT 3
- 不同weighting方法的rep_grad使用 HOT 14
- How to export saved models to other formats, such as onnx, mnn, etc HOT 2
- Questions about AlignMTL HOT 4
- When running the example code for QM9, the program seems to enter an infinite loop. QM9案例训练代码无响应
- When running the example code for QM9, the program seems to enter an infinite loop. QM9案例训练代码无响应 HOT 8
- Identical result for CAGrad and MoCo HOT 11
- Question about AlignedMTL HOT 2
- 关于rep_grads参数的问题 HOT 5
- 关于tabular数据的训练问题 HOT 4
- Not found the script for testing in examples/* HOT 6
- Image size of NYUv2 dataset should be 3*288*384 HOT 3
- Error while "from torchvision.models.utils import load_state_dict_from_url" HOT 3
- How to implement MTL scenario when each sample has some of the labels available and not for all the tasks. HOT 1
- GradNorm求梯度 HOT 5
- Inconsistency between formula and implementation in count_improvement function HOT 2
- It seems that some functions are not compatible with the latest pytorch HOT 1
- 关于abstract_weighting.py中get_share_params的问题 HOT 2
- MMOE - Replicate the Original paper Chapter 3.2 (Synthetic Data) 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 libmtl.