Lab 8 HPC Starter Pack
======================

This archive contains the SLURM scripts used in Lab 8:
Reinforcement Learning for Quadruped Locomotion with Isaac Gym.

It does NOT contain the large files:
- pytorch_isaacgym.sif
- Isaac Gym package
- leggedgym repository
- rsl_rl repository
- pyuser_isaac Python user base
- conda_tools

Those must already exist on the HPC cluster in:

    ~/pupper_lab8

Expected final structure:

    ~/pupper_lab8/
    ├── isaacgym/
    ├── leggedgym/
    ├── rsl_rl/
    ├── pytorch_isaacgym.sif
    ├── pyuser_isaac/
    ├── conda_tools/
    ├── local_include/
    ├── torch_extensions/
    ├── logs/
    ├── lab8_gpu_check.slurm
    ├── lab8_import_check.slurm
    ├── lab8_train_isaacgym.slurm
    └── lab8_reward_check.sh

Quick start:

    cd ~/pupper_lab8
    mkdir -p logs

    sbatch lab8_gpu_check.slurm
    sbatch lab8_import_check.slurm

    RUN_NAME="debug_128_${USER}" NUM_ENVS=128 MAX_ITERATIONS=10 sbatch lab8_train_isaacgym.slurm
    RUN_NAME="baseline_zero_reward_${USER}" NUM_ENVS=512 MAX_ITERATIONS=50 sbatch lab8_train_isaacgym.slurm

The main student TODO is in:

    ~/pupper_lab8/leggedgym/legged_gym/envs/pupper/pupper.py

Students should inspect and implement:

    _reward_base_height
    _reward_forward_velocity
    _reward_torques

Important:
- The script assumes the lab folder is ~/pupper_lab8.
- To use another folder, submit jobs with LAB_DIR=/path/to/folder.
