Yubin Huang

RuntimeError: invalid hash value

3

RuntimeError: invalid hash value (expected "ce0d4300", got "a0a8c5ad919066a582f73746eb7f2ee742e282baab9130860ca03fe1f17ead75") the code I used was model = pretrainedmodels.__dict__['se_resnet50'](num_classes=1000, pretrained='imagenet')

Some question about the label preprocess

1

Hi ,Julian, Thanks for your sharing! I have some question about the label preprocess, there might be some nodules were marked repeatly by different doctors, and these nodules have multiple...

how to run RL using multi-nodes in cluster

2

How to use RayVecEnv in cluster? I want to run my rl code using multi-nodes training, I'm new to ray, is there some demos scripts?

Can you provide details about the experiment on the IDRiD dataset and corresponding model parameters? Thank you!

Hi! Nice work! Did you split the validation set when training in IDRiD data?

get_env_attr not working in SubprocVectorEnv?

4

I want to obtain some attributions from my custom environ, when I use DummyVectorEnv, get_env_attr is working. But for SubprocVectorEnv, it seem not working? It report the following errors: >...

Some question about the implementation of meta-sac

1

Hello, thanks for your nice work. I have some questions about the details of implementations. I noticed that in your code implementation, alpha, critic, policy are updated in a different...

Question about n-step TD in SAC

1

In [test_sac_with_il.py](https://github.com/thu-ml/tianshou/blob/master/test/continuous/test_sac_with_il.py), the estimate-step parameter of SACPolicy is 3, which means that it use 3-step TD to update Q value. But I think the SAC can be considered as off-policy...