Onrup guidesGuides on fine-tuning and serving open-weight language models.https://www.onrup.com/en-gbKeeping a fine-tuned model portablehttps://www.onrup.com/guides/keeping-a-fine-tuned-model-portable/https://www.onrup.com/guides/keeping-a-fine-tuned-model-portable/What portability actually requires, the pairing people forget to record, and how to test it before you need it.Thu, 06 Aug 2026 00:00:00 GMTControlling spend on a fine-tuning platformhttps://www.onrup.com/guides/controlling-fine-tuning-spend/https://www.onrup.com/guides/controlling-fine-tuning-spend/Four controls that actually work, and why post-hoc alerting is not one of them.Thu, 06 Aug 2026 00:00:00 GMTServing many fine-tuned models without paying for each onehttps://www.onrup.com/guides/serving-many-fine-tuned-models/https://www.onrup.com/guides/serving-many-fine-tuned-models/Why adapters change the arithmetic of multi-model serving, and the constraint that decides whether you can use it.Thu, 06 Aug 2026 00:00:00 GMTUsing reinforcement learning to improve reasoninghttps://www.onrup.com/guides/reinforcement-learning-for-reasoning/https://www.onrup.com/guides/reinforcement-learning-for-reasoning/The asymmetry that makes reinforcement learning worth its extra cost, and the failure modes that come with a sloppy reward.Thu, 06 Aug 2026 00:00:00 GMTWhen to use preference tuning instead of supervised fine-tuninghttps://www.onrup.com/guides/when-to-use-preference-tuning/https://www.onrup.com/guides/when-to-use-preference-tuning/The shape of problem preference tuning solves, where the pairs come from, and the memory cost people discover too late.Thu, 06 Aug 2026 00:00:00 GMTWhy fine-tuned models fail, and how to tell which failure you havehttps://www.onrup.com/guides/why-fine-tuned-models-fail/https://www.onrup.com/guides/why-fine-tuned-models-fail/Five failure modes with the symptom that identifies each, in the order they are worth checking.Thu, 06 Aug 2026 00:00:00 GMTMigrating from a closed model API to an open-weight fine-tunehttps://www.onrup.com/guides/migrating-from-a-closed-model-api/https://www.onrup.com/guides/migrating-from-a-closed-model-api/A migration sequence that produces a measurement rather than a hope, including the filtering step most teams skip.Thu, 06 Aug 2026 00:00:00 GMTScale to zero or always warm: choosing a serving modehttps://www.onrup.com/guides/scale-to-zero-or-always-warm/https://www.onrup.com/guides/scale-to-zero-or-always-warm/The one serving decision that determines both your cost profile and your worst-case latency, and how to set cooldown from real traffic.Thu, 06 Aug 2026 00:00:00 GMTHow to estimate what a fine-tuning run will costhttps://www.onrup.com/guides/estimating-fine-tuning-cost/https://www.onrup.com/guides/estimating-fine-tuning-cost/The arithmetic behind a run’s cost, why the cheapest class is not always the cheapest run, and where budgets actually go wrong.Thu, 06 Aug 2026 00:00:00 GMTDesigning an evaluation set that actually decides thingshttps://www.onrup.com/guides/designing-an-evaluation-set/https://www.onrup.com/guides/designing-an-evaluation-set/Sizing, sampling and the critical-case set — plus why most evaluation sets are too small to support the decisions made from them.Thu, 06 Aug 2026 00:00:00 GMTHow to choose a base modelhttps://www.onrup.com/guides/choosing-a-base-model/https://www.onrup.com/guides/choosing-a-base-model/A filtering order that avoids the most expensive mistake — choosing a size before measuring whether you need it.Thu, 06 Aug 2026 00:00:00 GMTShareGPT, ChatML and Alpaca: which dataset format to usehttps://www.onrup.com/guides/sharegpt-vs-chatml-vs-alpaca/https://www.onrup.com/guides/sharegpt-vs-chatml-vs-alpaca/What each format is for, how they convert, and the three conversion mistakes that produce a model that trains cleanly and behaves oddly.Thu, 06 Aug 2026 00:00:00 GMTPreparing a fine-tuning datasethttps://www.onrup.com/guides/preparing-a-fine-tuning-dataset/https://www.onrup.com/guides/preparing-a-fine-tuning-dataset/From raw material to a validated dataset, including the four checks worth running before any compute is leased.Thu, 06 Aug 2026 00:00:00 GMTHow much VRAM does fine-tuning actually need?https://www.onrup.com/guides/how-much-vram-do-i-need/https://www.onrup.com/guides/how-much-vram-do-i-need/Where the memory actually goes, why training needs several times what inference does, and how to make a run fit without changing GPU class.Thu, 06 Aug 2026 00:00:00 GMTLoRA, QLoRA or full fine-tuning: how to choosehttps://www.onrup.com/guides/lora-vs-qlora-vs-full-fine-tuning/https://www.onrup.com/guides/lora-vs-qlora-vs-full-fine-tuning/The memory, quality and speed trade between the three approaches, and a rule that resolves nearly every case.Thu, 06 Aug 2026 00:00:00 GMTShould you fine-tune, or is it a prompt problem?https://www.onrup.com/guides/should-you-fine-tune/https://www.onrup.com/guides/should-you-fine-tune/A decision procedure for the question everyone asks first, including the three cases where fine-tuning is the wrong answer.Thu, 06 Aug 2026 00:00:00 GMT