Deep diving into custom LLM architectures. Always experimenting with fine-tuning techniques. x.com/ro7393735569845