Tim Meinhardt

In the current version of the code the following line causes a deprecation warning: https://github.com/kmaninis/OSVOS-PyTorch/blob/36562efeed51dd994b44329ea3cff5e0d76e83ac/train_online.py#L189 As suggested by the warning, this PR replaces `scipy.misc.imsave` with `imageio.imsave`. In addition, this PR...

Based on [this](https://github.com/timmeinhardt/pybm3d_gpu).

Print comments if config from YAML file

2

Is there a way to print the comments in a YAML file when executing the `print_config` command? Currently this YAML file: ``` env_name: 'nn-train-v0' # environment to train on (default:...

Suggestions for the capture decorator

1

I have two suggestions for the `capture` decorator. 1. The `sacred` tool could provide a fail-safe post-definition helper function for decorating. This function could improve could such as this: ```...

Apply learning rate scheduler not to PPO

2

In contrast to A2C and A2C_ACKTR, PPO already includes a learning rate scheduling performed by Adam. In supervised learning it is debatable if one should use manual scheduling in combination...

This issue is rather a RL question in general. Given N environments and a `FixedNormal` distribution with 3 outputs, the log probabilities are getting summed up in the following line:...

VecNormalize only for 1-D observations

5

Is there a particular reason why `VecNormalize` is only applied to 1-D observations? If yes, wouldn't it make sense to apply at least the rewards normalization? https://github.com/ikostrikov/pytorch-a2c-ppo-acktr/blob/47ddcbfab806c37ed19f438100300bd4d58c42f3/main.py#L68-L69

This PR removes the `self.output_proj_box` from `MSDeformAttn` if the mode is `decode` which allows to run `torch.nn.parallel.DistributedDataParallel` without `find_unused_parameters=True`. In theory, this should improve training time as the torch backend...

Missing time_attention_weights from T=1 model files

2

The T=1 pretrained model files do not contain the `time_attention_weights` weights.

Following the README.md, the `ytvis` dataset folder will be in the root directory of this repository. Excecuting the `inference.py` script in the same directory will cause an error as it...