inference-memory
Will a Hugging Face LLM fit your GPU or DGX? Plan weights, KV cache or recurrent state, context, TP/DP, and measured calibration