File size: 256 Bytes
b066889 |
1 |
This is a [tied-crosscoder](https://github.com/ARBORproject/arborproject.github.io/discussions/23) of 16384*3 latents trained on the 14th layers of Gemma 2 2b and Gemma 2 2b it. We trained for 250M tokens on a mixed dataset of the pile and LMSys chat data. |