Weakly Supervised Automated Language Model Red-Teaming to Identify Likely Toxic Prompts. - View it on GitHub