ChapterPal post on LLM serving tech

- An LLM‑ops curriculum called ChapterPal was shared, covering inference optimizers like FlashAttention, quantization (LLM.int8), and GPTQ techniques for cheaper serving. (x.com) - The post drills into practical stack choices for production: attention kernels, weight quantization and batching strategies. (x.com) - For teams moving small models to production this is an actionable map to cut memory and latency without retraining large weights. (x.com)