Comments (4)
I noticed that if I change batch size to 32, then the weird "obs: torch.Size([64])" becomes "obs: torch.Size([32])"
so the error is somehow connected to the batch size
from tianshou.
buffer = PrioritizedVectorReplayBuffer(
args.buffer_size,
buffer_num=len(train_env),
# ignore_obs_next=True,
# save_only_last_obs=True,
alpha=args.alpha,
beta=args.beta,
)
when I comment out those two lines, problem solved.
Can someone help me explain what happens here? Really appreciate.
from tianshou.
@WillInvest I can look into it, but can you format your posts above to make the code more readable?
from tianshou.
buffer = PrioritizedVectorReplayBuffer( args.buffer_size, buffer_num=len(train_env), # ignore_obs_next=True, # save_only_last_obs=True, alpha=args.alpha, beta=args.beta, ) when I comment out those two lines, problem solved. Can someone help me explain what happens here? Really appreciate.
The buffer is currently very complicated, I was working on simplifying it and extending documentation in the last months. Will be merged to master soon.
ignore_obs_next
should essentially never be set to True
, it means that obs
rolled forwards will be used instead of obs_next
, which means making errors at the boundaries of episodes and at interrupted collections. It's only an option if one really really needs to save RAM, like maybe with atari envs, where observations are large.
save_only_last_obs
should also not be touched.
from tianshou.
Related Issues (20)
- will add dreamerv3 ? HOT 1
- Documentation for multi-agent needs fixing HOT 1
- No minibatch for computation of logp_old in PPOPolicy HOT 1
- MPO Implementation HOT 1
- Improve interface of BasePolicy.compute_action HOT 1
- Suggestion - Redesign RayEnvWorker for Improved Performance HOT 12
- tianshou v1.0.0 failed to install on python 3.12.4 HOT 2
- How to add a obs with multiple dimensions to the buffer?
- TypeError: isinstance() argument 2 cannot be a parameterized generic HOT 1
- Make Batch.__eq__ work also for scalar values (0-dimensional arrays) HOT 2
- EOFError on training HOT 8
- ModuleNotFoundError: No module named 'tianshou.highlevel' HOT 3
- Cannot unify float64 and array(float64, 1d, C) for 'gae.2'
- Discrete distributions are broken HOT 2
- Actor and distribution function should not be specified independently HOT 1
- Learning rate scheduler confusion HOT 1
- Handling LR scheduling when episodes instead of steps are collected
- Release 1.1.0 to pypi HOT 1
- Refactoring: make Trainer more flexible
- Support custom scores for Trainer in high-level interfaces
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 tianshou.