kristery / elastic-dt Goto Github PK
View Code? Open in Web Editor NEW[NeurIPS 2023] Implementation of Elastic Decision Transformer
License: MIT License
[NeurIPS 2023] Implementation of Elastic Decision Transformer
License: MIT License
Hello, authors. Elastic Decision Transformer provides a good view to investigate the stitch problem. I am now trying to reproduce the experimental result.
However, there is a problem about evaluation period, and it is not explicitly illustrated in the Decision Transformer paper.
How do you evaluate your final results? For example, do you test the final model after training for 100k steps, or do you use another method?
Dear, authors!
I appreciate your paper and your code. I am currently trying to reproduce the experimental result.
However, I have a few queries regarding evaluating the results, specifically for hopper_medium_replay and walker_medium_replay. From my end, it seems like these 2 medium_replay datasets might get very high variance (I am testing with 5 random seeds and evaluate after 500 epochs) --> hopper: std=12 and walker2d: std=9. I wonder if you also experienced this from your end?
Thank you so much for your attention!
Hi,sir
When I try to eval EDT in "kitchen-partial-v0" dataset,I have these wrong:
Traceback (most recent call last): File "eval_edt.py", line 259, in <module> test(cfg) File "eval_edt.py", line 204, in test heuristic_delta=args.heuristic_delta, File "C:\Users\33908\Desktop\Elastic-DT\decision_transformer\utils.py", line 344, in edt_evaluate real_rtg=real_rtg File "C:\Users\33908\Desktop\Elastic-DT\decision_transformer\utils.py", line 440, in _return_search expert_weight=expert_weight, File "C:\Users\33908\Desktop\Elastic-DT\decision_transformer\utils.py", line 211, in expert_sampling logits + expert_weight * expert_logits, temperature, top_percentile File "C:\Users\33908\Desktop\Elastic-DT\decision_transformer\utils.py", line 196, in sample_from_logits m = Categorical(logits=temperature * logits) File "C:\Users\33908\mambaforge\envs\dt\lib\site-packages\torch\distributions\categorical.py", line 64, in __init__ super(Categorical, self).__init__(batch_shape, validate_args=validate_args) File "C:\Users\33908\mambaforge\envs\dt\lib\site-packages\torch\distributions\distribution.py", line 56, in __init__ f"Expected parameter {param} " ValueError: Expected parameter logits (Tensor of shape (1, 20, 60)) of distribution Categorical(logits: torch.Size([1, 20, 60])) to satisfy the constraint IndependentConstraint(Real(), 1), but found invalid values: tensor([[[nan, nan, nan, ..., nan, nan, nan], [nan, nan, nan, ..., nan, nan, nan], [nan, nan, nan, ..., nan, nan, nan], ..., [nan, nan, nan, ..., nan, nan, nan], [nan, nan, nan, ..., nan, nan, nan], [nan, nan, nan, ..., nan, nan, nan]]], device='cuda:0')
How I can do?
Thanks!
A declarative, efficient, and flexible JavaScript library for building user interfaces.
๐ Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
The Web framework for perfectionists with deadlines.
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. ๐๐๐
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
We are working to build community through open source technology. NB: members must have two-factor auth.
Open source projects and samples from Microsoft.
Google โค๏ธ Open Source for everyone.
Alibaba Open Source for everyone
Data-Driven Documents codes.
China tencent open source team.