2 comments

ramigb8 days ago
That's awesome I came from the jev in python thread. is the model shareable? do you have any more writing on this? would love to read more about it! cool demo anyways.
faangguyindia8 days ago
Here you can see a 26B model hacked using a trained KV cache bank, responding like Gemma. The latency I am seeing is <127 ms.

Read the full thread on Hacker News →

Related stories