-
## Describe the bug
Training with the Python-based GRU raises the following error, which indicates the current implementation has some in-place operations that prevent correct backward computation:…
-
I am using vllm==0.2.2 and run a ray ppo training using vllm engines, but got this error:
```
(LLMRayActor pid=433508) Exception raised in creation task: The actor died because of an error raised in…
-
![image](https://user-images.githubusercontent.com/1320252/58152555-6f45cb00-7c9f-11e9-89e4-a7cff8d2404b.png)
![image](https://user-images.githubusercontent.com/1320252/58152540-63f29f80-7c9f-11e9-91…
-
The `LossModule.convert_to_functional(...)` method creates a deep copy of the parameters. If I understand correctly, this leads to the parameters being duplicated in memory and a larger memory footpri…
-
## Motivation
It would nice to add a new option to compute the clipped value loss as used in OpenAI's Baselines for PPO:
https://github.com/openai/baselines/blob/ea25b9e8b234e6ee1bca43083f8f3cf9…
-
First of all, thank you for sharing this project in an open source form!
While testing PPO + LSTM, I've identified 2 potential improvements:
- LSTM historization module requires the next state o…
-
Removing a model does not free all of its memory, there is a memory leak related to meshes/textures (material). This introduces limitation to domain randomisation which causes training to eventually f…
-
您好,非常感谢您分享的hppo算法的代码!我从中学到了很多。
但我在阅读代码时有一些地方没有看明白,希望能和您讨论一下。
1.第一个问题是在`PPO_Hybrid`类中的`compute_loss_pi`方法中,计算连续动作网络的loss时,没有用到该网络的交叉熵,请问为什么要这样计算呢?
```
def compute_loss_pi(self, data):
…
-
See here: https://deepmind.com/blog/alphazero-shedding-new-light-grand-games-chess-shogi-and-go/
Not read yet...
-
A [homoiconic](https://en.wikipedia.org/wiki/Homoiconicity) programming language can parse itself to an AST expressed as native data that can be transformed within the language, evaluated, or isomorph…