🎉 You are looking at Hummingbird 2.6, which uses different and slightly better datasets!

Hummingbird is a Cerebras-GPT derivative trained to be conversational.

The model was trained using 500 steps, and 4 batch size.

The training corpus is made up of a collated qikp/sillychat to be multi-turn.

The Zephyr chat template was used.

The model frequently outputs incorrect information, confirmation with a larger, mature model is advised.

This model was benchmarked and compared using embeddings. See the results here.