# Poor performance of offline algorithms tuned examples

**URL:** <https://discuss.ray.io/t/poor-performance-of-offline-algorithms-tuned-examples/13871>\
**Category:** Offline RL\
**Created:** [March 1, 2024, 7:04am UTC](https://discuss.ray.io/t/poor-performance-of-offline-algorithms-tuned-examples/13871 "2024-03-01T07:04:00Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**Author:** ![Morphlng](https://sea2.discourse-cdn.com/flex020/user_avatar/discuss.ray.io/morphlng/32/5785_2.png) [@Morphlng](https://discuss.ray.io/u/Morphlng)\
**Post date:** [March 1, 2024, 7:04am UTC](https://discuss.ray.io/t/poor-performance-of-offline-algorithms-tuned-examples/13871/1 "2024-03-01T07:04:00Z")

</div>

## Description

I’m testing the offline algorithms RLlib provided, especially those using D4RL dataset. However, I’ve found that nearly all tuned\_examples provided in CQL algorithm perform really poor. For example:

1. Hopper-bc.yaml  
 ![image](https://us1.discourse-cdn.com/flex020/uploads/ray/original/2X/c/ceaf5a5128a270b26743a6823482cbe396fbee85.jpeg)

Halfcheetah-bc and Hopper-cql are of the similar results. Both reward and loss curve seems to be oscillating, which I think meaning the algorithm isn’t really learning anything. These examples didn’t come with a result, I wonder is this normal?

## Environment

- Ray 2.2.0 (Since this is the last version support Gym, and D4RL relies on it)
- Python 3.10.13
- Pytorch 2.2.1