Join the conversation

Join the community of Machine Learners and AI enthusiasts.

Sign Up
Banaxi-Tech 
posted an update 6 days ago
Post
171
We're excited to release BananaMind 2 Mini the first model in our BananaMind 2 series!


BananaMind 2 Mini features a custom digit-aware BPE tokenizer that keeps every digit isolated, fixing the core arithmetic weakness of our previous models. It's trained on 30B tokens from FineWeb-Edu, DCLM, Cosmopedia-v2 and FineMath-4+, and already outperforms Pythia-31M despite having fewer parameters.
Check it out at
BananaMind/BananaMind-2-Mini

BananaMind 2 Pro is coming next. Apache 2.0.

Wow, this model is officially my favorite for fine-tuning! I just fine-tuned it on an IF (Instruction Following) + chat mix dataset, and it became really good at simple Q&A, chatting, and it even has common sense. Its creative writing is decent now too not amazing, but solid. For now, my fine-tuned version of Bananamind 2 Mini is actually outperforming some ~70M parameter models in certain tasks. Now I'll just wait until the Pro version comes out to check it out. By the way, keep going!