CastIronMind 's Collections

Lissa 1.5B

1.45B hybrid linear-attention model trained from scratch for $555 on one GPU. Base, instruct and SFT datasets.