2 ms·
Maybe not multi-gigabyte, but in a new system/phone in a year, you're basically guaranteed to find at least one tiny model. We may even get some "standard" mode
by viraptor 3y ago
Maybe not multi-gigabyte, but in a new system/phone in a year, you're basically guaranteed to find at least one tiny model. We may even get some "standard" model everyone can reliably use as a reference.
- Filligree 3y agoAt that point it would be useful. Although, I wonder if it wouldn’t make more sense to train one specifically for the job. Current LLMs can predict HTML, sure, but they’re large and slow for the task.