A newer version of this model is available: qikp/kite-3.1-20m
Kite
🎉 You are looking at Kite 3, which is now using FineWeb-Edu and pika 3, lower learning rate, and is now public domain-like!
Kite is a small, trained, 15 million parameter language model, without any special optimizations.
Training
It was trained on 50K rows of FineWeb Edu using 1 epoch, 4 batch size, 1.5e-4 learning rate, and the pika 3 tokenizer.
Limitations
Due to its size, the model is not suitable for production workloads.
- Downloads last month
- 477