4 ms·
are there also experiments around image embedding generation to use in combination with the LLM? maybe for this use-case is it better to execute the vision tow
by joaquincabezas 3y ago
are there also experiments around image embedding generation to use in combination with the LLM?
maybe for this use-case is it better to execute the vision tower on a GPU and leave the LPU for the language part?
- matanyal 3y agoWe are great for image embedding (and audio, with more to come!) There is no reason you should be forced to use graphics cards intended for gaming for any AI workload.