11 Commits

Author	SHA1	Message	Date
Simon Gardling	df04e36b41	llama-cpp: fix vulkan cache	2026-04-06 02:23:29 -04:00
Simon Gardling	0a927ea893	llama-cpp: maybe use vulkan?	2026-04-06 02:12:46 -04:00
Simon Gardling	3e46c5bfa5	llama-cpp: use turbo3 for everything	2026-04-06 01:53:11 -04:00
Simon Gardling	06aee5af77	llama-cpp: gemma 4 E4B -> gemma 4 E2B	2026-04-06 01:24:25 -04:00
Simon Gardling	8fddd3a954	llama-cpp: context: 32768 -> 65536	2026-04-06 01:04:23 -04:00
Simon Gardling	0e4f0d3176	llama-cpp: fix model name	2026-04-06 00:59:20 -04:00
Simon Gardling	8ea96c8b8e	llama-cpp: fix model hash	2026-04-04 00:28:07 -04:00
Simon Gardling	479ec43b8f	llama-cpp: integrate native prometheus /metrics endpoint ... llama.cpp server has a built-in /metrics endpoint exposing prompt_tokens_seconds, predicted_tokens_seconds, tokens_predicted_total, n_decode_total, and n_busy_slots_per_decode. Enable it with --metrics and add a Prometheus scrape target, replacing the need for any external metric collection for LLM inference monitoring.	2026-04-03 15:19:11 -04:00
Simon Gardling	47aeb58f7a	llama-cpp: do logging	2026-04-03 14:39:46 -04:00
Simon Gardling	d4d01d63f1	llama-cpp: update + re-enable + gemma 4 E4B	2026-04-03 14:06:35 -04:00
Simon Gardling	124d33963e	organize	2026-04-03 00:47:12 -04:00