imitation

imitation copied to clipboard

Expected all tensors to be on the same device, but found at least two devices, cuda:0 and cpu!

2

## Bug description ``` Traceback (most recent call last): File "/home/kavin/Documents/PycharmProjects/RL/Imitation/example.py", line 150, in bc_trainer.train(n_epochs=1) File "/home/kavin/anaconda3/envs/PythonEnv/lib/python3.8/site-packages/imitation/algorithms/bc.py", line 495, in train training_metrics = self.loss_calculator(self.policy, obs_tensor, acts) File "/home/kavin/anaconda3/envs/PythonEnv/lib/python3.8/site-packages/imitation/algorithms/bc.py", line 130,...

kavinwkp

SyntheticGatherer often gives nearly deterministic feedback

1

## Bug description The current implementation of the `SyntheticGatherer` in the preference comparisons module often chooses the trajectory with the higher reward nearly deterministically. This is because the Boltzmann-rational policy...

timokau

Run time Error when run quickstart.py

1

When run examples/quickstart.py, i'm getting error with RuntimeError: Expected all tensors to be on the same device, but found at least two devices, cuda:0 and cpu! (when checking argument for...

Charles-Lim93

Utilizing Expert Data (.npz) in format of SB3.

5

Is it possible to use the trajectories (.npz) collected as expert response which is compatible for SB3 IRL models. ? I made an env which take user input to move...

azafar1991

Indexing a TrajectoryDatasetSequence instance when observations are images is extremely slow.

11

## Problem If the observations for a given task are images and stored using the `TrajectoryDatasetSequence ` class, indexing is extremely slow. For instance, indexing one trajectory can take upwards...

Bpoole908

## Problem In the last month we doubled our median pipeline runtime and we also spend a lot of CirlceCI credits on failed pipelines ## Solution 1. Make the pipeline...

ernestum

Document running the entire benchmarking suite

1

This adds some instructions on running the benchmarking suite. It is still missing the baseline benchmark values and instructions to update them, which I'll include in a later PR. I'm...

hacobe

## Description Add MCE IRL training script for #392

ZiyueWang25

Add rgb observation to dagger

3

## Description 1. Add an environment wrapper to keep the original observation and rgb version together for interactive policy 2. Remove the rgb observation and its space in the bc...

ZiyueWang25

## Problem The issue is based on efforts from #776 and it works as a last step for #701. For more detail, please see the discussion [here](https://github.com/HumanCompatibleAI/imitation/issues/701#issuecomment-1712411254) ## Solution -...

ZiyueWang25