Copybara import of the project:

--
3f5778b by Yasunori Toshimitsu <yasu313nori@gmail.com>:

add note about TF32

--
4c8ca6e by btaba <btaba@google.com>:

Update README.md

COPYBARA_INTEGRATE_REVIEW=#87 from Yasu31:set-highest-jax-matmul-precision 4c8ca6e
PiperOrigin-RevId: 742371385
Change-Id: Icc6cf28eceed2240fc78fd562a2dbd0448a9439c

Author: Yasu31

README.md

@@ -63,10 +63,16 @@ For vision-based environments, please refer to the installation instructions in
 | [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/google-deepmind/mujoco_playground/blob/main/learning/notebooks/training_vision_1.ipynb) | Training CartPole from Vision |
 | [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/google-deepmind/mujoco_playground/blob/main/learning/notebooks/training_vision_2.ipynb) | Robotic Manipulation from Vision |
 
-## How can I contribute?
+
+## FAQ
+
+### How can I contribute?
 
 Get started by installing the library and exploring its features! Found a bug? Report it in the issue tracker. Interested in contributing? If you are a developer with robotics experience, we would love your help—check out the [contribution guidelines](CONTRIBUTING.md) for more details.
 
+### Reproducibility / GPU Precision Issues
+Users with NVIDIA Ampere architecture GPUs (e.g., RTX 30 and 40 series) may experience reproducibility [issues](https://github.com/google-deepmind/mujoco_playground/issues/86) in mujoco_playground due to JAX’s default use of TF32 for matrix multiplications. This lower precision can adversely affect RL training stability. To ensure consistent behavior with systems using full float32 precision (as on Turing GPUs), please run `export JAX_DEFAULT_MATMUL_PRECISION=highest` in your terminal before starting your experiments (or add it to the end of `~/.bashrc`).
+
 ## Citation
 
 If you use Playground in your scientific works, please cite it as follows: