LIYUESEN/druggpt ? reverse-engineered prompt

Build me a simple Python tool for designing possible drug like ligands for a target protein using a GPT style model. I want to be able to give it either a raw amino acid sequence or a FASTA file, and optionally a starting ligand prompt if I want it to riff on an existing molecule. It should generate however many candidate molecules I ask for, save them to an output folder, and let me choose things like device, batch size, randomness settings, and atom count limits when no ligand prompt is provided.

Please also include the newer ligand energy minimization workflow, including a way to minimize one batch of generated ligands and a version that can handle multiple ligands efficiently. Keep it usable from the command line, with sensible defaults, clear help text, and an example using the included BCL2L11 FASTA. If there are model weights or setup steps needed, wire that up and document it so I can run it in a local environment or Colab. Look up current docs online if you need to.