# Agent Instructions ## Python Environment Use the following Conda environment for this repository: ```bash conda activate lerobot ``` If the named environment is unavailable, use the absolute path: ```bash conda activate /llm_jzm/cache/conda_env/lerobot ``` If using `conda activate lerobot`: - Treat it as personal machine mode. - Do not set shared-machine Hugging Face cache or token overrides. If using `conda activate /llm_jzm/cache/conda_env/lerobot`: - Treat it as shared machine mode. - Hugging Face cache is not in the default location. - Set Hugging Face token to the shared token for runs on this machine. - Set Wandb token to the shared token for runs on this machine Use: ```bash export HF_HOME=/llm_jzm/cache/huggingface/ export HF_TOKEN= export HF_ENDPOINT=https://hf-mirror.com export WANDB_API_KEY= wandb login --relogin "$WANDB_API_KEY" ``` ## User Preferences - During coding, avoid try/except unless explicitly requested; prefer direct and rough failure/debug output. - Reuse existing code/tools first; do not rebuild from scratch unless needed. - Do not preserve forward/backward compatibility unless explicitly requested; prefer direct cleanup and replacement over compatibility shims. - For every experiment run request (e.g., training/evaluation runs), always create a new Notion page with purpose, exact parameters/command, and key results/observations. - Debug/troubleshooting runs do not count as experiments by default, and do not require creating a Notion experiment page unless explicitly requested. - Write complete notes proactively and boldly; user will decide what to keep.