{"slug": "suggestions-for-getting-started-with-local-ai", "title": "Suggestions for getting started with local AI", "summary": "A technical guide details a 46-layer hybrid model architecture alternating between Linear Attention (LA) and Full Attention (FA) layers, with LA layers using qkv, z, and out projections plus A_log, dt_bias, conv1d, a/b projections, and state norm, while FA layers use q, k, v, o projections with q_norm and k_norm, and all layers include gate, up, and down projections. The guide offers suggestions for getting started with local AI, focusing on implementing this specific architecture.", "body_md": "| 0 |\nLA |\nqkv, z, out |\nA_log, dt_bias, conv1d, a/b projections, state norm |\ngate, up, down |\n| 1 |\nLA |\nqkv, z, out |\nA_log, dt_bias, conv1d, a/b projections, state norm |\ngate, up, down |\n| 2 |\nLA |\nqkv, z, out |\nA_log, dt_bias, conv1d, a/b projections, state norm |\ngate, up, down |\n| 3 |\nFA |\nq, k, v, o |\nq_norm, k_norm |\ngate, up, down |\n| 4 |\nLA |\nqkv, z, out |\nA_log, dt_bias, conv1d, a/b projections, state norm |\ngate, up, down |\n| 5 |\nLA |\nqkv, z, out |\nA_log, dt_bias, conv1d, a/b projections, state norm |\ngate, up, down |\n| 6 |\nLA |\nqkv, z, out |\nA_log, dt_bias, conv1d, a/b projections, state norm |\ngate, up, down |\n| 7 |\nFA |\nq, k, v, o |\nq_norm, k_norm |\ngate, up, down |\n| 8 |\nLA |\nqkv, z, out |\nA_log, dt_bias, conv1d, a/b projections, state norm |\ngate, up, down |\n| 9 |\nLA |\nqkv, z, out |\nA_log, dt_bias, conv1d, a/b projections, state norm |\ngate, up, down |\n| 10 |\nLA |\nqkv, z, out |\nA_log, dt_bias, conv1d, a/b projections, state norm |\ngate, up, down |\n| 11 |\nFA |\nq, k, v, o |\nq_norm, k_norm |\ngate, up, down |\n| 12 |\nLA |\nqkv, z, out |\nA_log, dt_bias, conv1d, a/b projections, state norm |\ngate, up, down |\n| 13 |\nLA |\nqkv, z, out |\nA_log, dt_bias, conv1d, a/b projections, state norm |\ngate, up, down |\n| 14 |\nLA |\nqkv, z, out |\nA_log, dt_bias, conv1d, a/b projections, state norm |\ngate, up, down |\n| 15 |\nFA |\nq, k, v, o |\nq_norm, k_norm |\ngate, up, down |\n| 16 |\nLA |\nqkv, z, out |\nA_log, dt_bias, conv1d, a/b projections, state norm |\ngate, up, down |\n| 17 |\nLA |\nqkv, z, out |\nA_log, dt_bias, conv1d, a/b projections, state norm |\ngate, up, down |\n| 18 |\nLA |\nqkv, z, out |\nA_log, dt_bias, conv1d, a/b projections, state norm |\ngate, up, down |\n| 19 |\nFA |\nq, k, v, o |\nq_norm, k_norm |\ngate, up, down |\n| 20 |\nLA |\nqkv, z, out |\nA_log, dt_bias, conv1d, a/b projections, state norm |\ngate, up, down |\n| 21 |\nLA |\nqkv, z, out |\nA_log, dt_bias, conv1d, a/b projections, state norm |\ngate, up, down |\n| 22 |\nLA |\nqkv, z, out |\nA_log, dt_bias, conv1d, a/b projections, state norm |\ngate, up, down |\n| 23 |\nFA |\nq, k, v, o |\nq_norm, k_norm |\ngate, up, down |\n| 24 |\nLA |\nqkv, z, out |\nA_log, dt_bias, conv1d, a/b projections, state norm |\ngate, up, down |\n| 25 |\nLA |\nqkv, z, out |\nA_log, dt_bias, conv1d, a/b projections, state norm |\ngate, up, down |\n| 26 |\nLA |\nqkv, z, out |\nA_log, dt_bias, conv1d, a/b projections, state norm |\ngate, up, down |\n| 27 |\nFA |\nq, k, v, o |\nq_norm, k_norm |\ngate, up, down |\n| 28 |\nLA |\nqkv, z, out |\nA_log, dt_bias, conv1d, a/b projections, state norm |\ngate, up, down |\n| 29 |\nLA |\nqkv, z, out |\nA_log, dt_bias, conv1d, a/b projections, state norm |\ngate, up, down |\n| 30 |\nLA |\nqkv, z, out |\nA_log, dt_bias, conv1d, a/b projections, state norm |\ngate, up, down |\n| 31 |\nFA |\nq, k, v, o |\nq_norm, k_norm |\ngate, up, down |\n| 32 |\nLA |\nqkv, z, out |\nA_log, dt_bias, conv1d, a/b projections, state norm |\ngate, up, down |\n| 33 |\nLA |\nqkv, z, out |\nA_log, dt_bias, conv1d, a/b projections, state norm |\ngate, up, down |\n| 34 |\nLA |\nqkv, z, out |\nA_log, dt_bias, conv1d, a/b projections, state norm |\ngate, up, down |\n| 35 |\nFA |\nq, k, v, o |\nq_norm, k_norm |\ngate, up, down |\n| 36 |\nLA |\nqkv, z, out |\nA_log, dt_bias, conv1d, a/b projections, state norm |\ngate, up, down |\n| 37 |\nLA |\nqkv, z, out |\nA_log, dt_bias, conv1d, a/b projections, state norm |\ngate, up, down |\n| 38 |\nLA |\nqkv, z, out |\nA_log, dt_bias, conv1d, a/b projections, state norm |\ngate, up, down |\n| 39 |\nFA |\nq, k, v, o |\nq_norm, k_norm |\ngate, up, down |\n| 40 |\nLA |\nqkv, z, out |\nA_log, dt_bias, conv1d, a/b projections, state norm |\ngate, up, down |\n| 41 |\nLA |\nqkv, z, out |\nA_log, dt_bias, conv1d, a/b projections, state norm |\ngate, up, down |\n| 42 |\nLA |\nqkv, z, out |\nA_log, dt_bias, conv1d, a/b projections, state norm |\ngate, up, down |\n| 43 |\nFA |\nq, k, v, o |\nq_norm, k_norm |\ngate, up, down |\n| 44 |\nLA |\nqkv, z, out |\nA_log, dt_bias, conv1d, a/b projections, state norm |\ngate, up, down |\n| 45 |\nLA |\nqkv, z, out |\nA_log, dt_bias, conv1d, a/b projections, state norm |\ngate, up, down |\n| 46 |\nLA |\nqkv, z, out |\nA_log, dt_bias, conv1d, a/b projections, state norm |\ngate, up, down |\n| 47 |\nFA |\nq, k, v, o |\nq_norm, k_norm |\ngate, up, down |\n| 48 |\nLA |\nqkv, z, out |\nA_log, dt_bias, conv1d, a/b projections, state norm |\ngate, up, down |\n| 49 |\nLA |\nqkv, z, out |\nA_log, dt_bias, conv1d, a/b projections, state norm |\ngate, up, down |\n| 50 |\nLA |\nqkv, z, out |\nA_log, dt_bias, conv1d, a/b projections, state norm |\ngate, up, down |\n| 51 |\nFA |\nq, k, v, o |\nq_norm, k_norm |\ngate, up, down |\n| 52 |\nLA |\nqkv, z, out |\nA_log, dt_bias, conv1d, a/b projections, state norm |\ngate, up, down |\n| 53 |\nLA |\nqkv, z, out |\nA_log, dt_bias, conv1d, a/b projections, state norm |\ngate, up, down |\n| 54 |\nLA |\nqkv, z, out |\nA_log, dt_bias, conv1d, a/b projections, state norm |\ngate, up, down |\n| 55 |\nFA |\nq, k, v, o |\nq_norm, k_norm |\ngate, up, down |\n| 56 |\nLA |\nqkv, z, out |\nA_log, dt_bias, conv1d, a/b projections, state norm |\ngate, up, down |\n| 57 |\nLA |\nqkv, z, out |\nA_log, dt_bias, conv1d, a/b projections, state norm |\ngate, up, down |\n| 58 |\nLA |\nqkv, z, out |\nA_log, dt_bias, conv1d, a/b projections, state norm |\ngate, up, down |\n| 59 |\nFA |\nq, k, v, o |\nq_norm, k_norm |\ngate, up, down |\n| 60 |\nLA |\nqkv, z, out |\nA_log, dt_bias, conv1d, a/b projections, state norm |\ngate, up, down |\n| 61 |\nLA |\nqkv, z, out |\nA_log, dt_bias, conv1d, a/b projections, state norm |\ngate, up, down |\n| 62 |\nLA |\nqkv, z, out |\nA_log, dt_bias, conv1d, a/b projections, state norm |\ngate, up, down |\n| 63 |\nFA |\nq, k, v, o |\nq_norm, k_norm |\ngate, up, down |", "url": "https://wpnews.pro/news/suggestions-for-getting-started-with-local-ai", "canonical_source": "https://forum.level1techs.com/t/suggestions-for-getting-started-with-local-ai/252681?page=2#post_32", "published_at": "2026-08-19 18:38:13+00:00", "updated_at": "2026-08-19 18:56:28.977748+00:00", "lang": "en", "topics": ["artificial-intelligence", "machine-learning", "large-language-models", "ai-research", "ai-infrastructure"], "entities": [], "alternates": {"html": "https://wpnews.pro/news/suggestions-for-getting-started-with-local-ai", "markdown": "https://wpnews.pro/news/suggestions-for-getting-started-with-local-ai.md", "text": "https://wpnews.pro/news/suggestions-for-getting-started-with-local-ai.txt", "jsonld": "https://wpnews.pro/news/suggestions-for-getting-started-with-local-ai.jsonld"}}