Comments (5)
I encountered this error only when set MCTS_SIMS = 1000. Does that mean, when in such case, I should change maxQU's initial value more smaller, such as -999999 ( one more 9)?
from deepreinforcementlearning.
hi archcra, have you checked your vh_loss or ph_loss?
in my case,
when i face this error, my vh_loss and ph_loss values show NaN,
so, I changed Nan values to 0
i wish you solved this problem
from deepreinforcementlearning.
from deepreinforcementlearning.
hi archcra, have you checked your vh_loss or ph_loss? in my case, when i face this error, my vh_loss and ph_loss values show NaN, so, I changed Nan values to 0
i wish you solved this problem
Can you explain more, I don't find vh_loss or ph_loss in code. I am also facing this issue now.
I tried to comment out if clause, but now it doesn't predict any more, it shows:
MCTS perceived value for O: nan
NN perceived value for O: nan
from deepreinforcementlearning.
hi archcra, have you checked your vh_loss or ph_loss? in my case, when i face this error, my vh_loss and ph_loss values show NaN, so, I changed Nan values to 0
i wish you solved this problemCan you explain more, I don't find vh_loss or ph_loss in code. I am also facing this issue now. I tried to comment out if clause, but now it doesn't predict any more, it shows:
MCTS perceived value for O: nan NN perceived value for O: nan
Here is the training content I wrote down as I don't remember the details of a project I attempted a long time ago. Please understand that this translation is done using GPT.
Case1 - When initially training, errors occurred during the training process. Unsure where the problem originated, I printed out the cost values and observed that the pH values diverged in the negative direction, eventually resulting in NaN values. It appeared to be a gradient exploding problem. While I knew that conventional CNNs typically use techniques like dropout to address this, I was unsure how to handle it in a self-learning scenario. Through searching online, I found that inserting code to replace NaN values with 0 could sometimes resolve the issue. Subsequently, the code proceeded without errors. However, the loss function appeared to fluctuate erratically in the output (loss could decrease to negative values, but NaN values caused it to be converted to 0, resulting in erratic behavior). Further investigation revealed that the problem occurred during the backward pass. It turned out that the issue stemmed from a custom loss function implemented in Keras. Switching to a PyTorch loss function resolved the problem, and training commenced smoothly.
I hope it helped you.
from deepreinforcementlearning.
Related Issues (20)
- ImportError: Failed to import `pydot HOT 4
- Does not work with tensorflow-gpu HOT 5
- ValueError: Initializer for variable conv2d_2/kernel/ is from inside a control-flow construct, such as a loop or conditional. When creating a variable inside a loop or conditional, use a lambda as the initializer. HOT 1
- About learning rate and loss decrease
- Performance against humans?
- HIDDEN_CNN_LAYERS from config.py
- Value can never 1 HOT 2
- What does Game.identities() do? HOT 2
- run.ipynb has issues HOT 1
- Will the code not be much faster if a GPU implementation of MCTS is used? HOT 3
- Incorrect Implementation of Resudial Layer HOT 1
- Interested in implementing other games HOT 5
- environment challenge HOT 2
- How to switch NHWC to NCHW HOT 3
- why you code is not working ? HOT 1
- Error with inference
- help
- working code HOT 2
- Incorrect implementation of adding Dirichlet noise 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 deepreinforcementlearning.