Enhance README with version compatibility and training info

README.md

@@ -31,6 +31,17 @@ Deploy Code: https://github.com/Hellod035/LeggedLabDeploy
 
 LeggedLab is built against the latest version of Isaacsim/IsaacLab. It is recommended to follow the latest updates of legged lab.
 
+### Version Compatibility
+
+The following version combinations have been tested and verified to work:
+
+- ✅ **Isaac Sim 4.5.0** + **Isaac Lab 2.1.0**
+- ✅ **Isaac Sim 5.0.0** + **Isaac Lab 2.2.1**
+
+> **Note**: While other version combinations may work, these are the officially tested configurations. Ensure that your Isaac Sim and Isaac Lab versions are compatible with each other.
+
+### Setup Steps
+
 - Install Isaac Lab by following the [installation guide](https://isaac-sim.github.io/IsaacLab/main/source/setup/installation/index.html). We recommend using the conda installation as it simplifies calling Python scripts from the terminal.
 
 - Clone this repository separately from the Isaac Lab installation (i.e. outside the `IsaacLab` directory):
@@ -56,6 +67,29 @@ pip install -e .
 python legged_lab/scripts/train.py --task=g1_flat --headless --logger=tensorboard --num_envs=64
 ```
 
+### Training
+
+To train a model, use the following command:
+
+```bash
+python legged_lab/scripts/train.py --task=g1_flat --headless --logger=tensorboard --num_envs=8192
+```
+
+**Training Notes:**
+- For a 12GB GPU, training with 8192 environments typically converges after approximately 2000 iterations.
+- Adjust `--num_envs` based on your GPU memory capacity.
+- Use `--logger=tensorboard` or `--logger=wandb` to track training progress.
+
+### Supported Tasks
+
+The following tasks are currently supported:
+
+- `g1_flat` - Unitree G1 on flat terrain
+- `g1_rough` - Unitree G1 on rough terrain
+- `h1_flat` - Unitree H1 on flat terrain
+- `h1_rough` - Unitree H1 on rough terrain
+- `gr2_flat` - FFTai GR2 on flat terrain
+- `gr2_rough` - FFTai GR2 on rough terrain
 
 ## Use Your Own Robot
 
@@ -66,6 +100,25 @@ Assets must be converted into USD format to be compatible with Legged Lab/IsaacL
 
 Legged Lab supports multi-GPU and multi-node reinforcement learning using rsl_rl, the usage is exactly the same as IsaacLab. [Detailed information](https://isaac-sim.github.io/IsaacLab/main/source/features/multi_gpu.html)
 
+## Testing and Evaluation
+
+### Running Trained Models
+
+To test a trained model, use the `play.py` script:
+
+```bash
+# Basic usage - test with default settings
+<path-to-IsaacLab>/isaaclab.sh -p legged_lab/scripts/play.py --task=g1_flat --num_envs=64
+```
+
+
+The script will:
+- Load the latest checkpoint from `logs/<task_name>/` directory
+- Export the policy as JIT and ONNX formats to `logs/<task_name>/<run_id>/exported/`
+- Run the policy in simulation
+
+
+
 ## Troubleshooting
 
 ### Pylance Missing Indexing of Extensions
@@ -86,6 +139,25 @@ In some VsCode versions, the indexing of part of the extensions is missing. In t
 }
 ```
 
+### RSL-RL Version Mismatch
+
+In some cases, you may encounter a version mismatch with RSL-RL, resulting in the following error:
+
+```
+ValueError: The observation configuration dictionary 'obs_groups' must contain the 'policy' key. Found keys: []
+```
+
+**Solution:**
+
+Reinstall RSL-RL using the Isaac Lab installation script:
+
+```bash
+<path-to-IsaacLab>/isaaclab.sh -i rsl-rl
+```
+
+This will ensure that RSL-RL is properly installed and compatible with your Isaac Lab version.
+
+
 ## References and Thanks
 This project repository builds upon the shoulders of giants.
 * [IsaacLab](https://github.com/isaac-sim/IsaacLab)   The various reusable practical components in IsaacLab greatly simplify the complexity of LeggedLab.