Before starting, the Estimate button on the Params block produces a real calculation for your model and dataset:
- fits / doesn't fit: whether the model and LoRA fit in memory (QLoRA widens the range).
- time & cost: from the tokens actually seen, reference throughput and model load time.
Tip: unknown models are assumed 7B and the note under the table says so — enter org/name precisely for accurate numbers.
