Learn about the new Artificial Hippocampus Networks (AHN) with my pure-PyTorch Qwen2-AHN model
https://huggingface.co/MartialTerran/Toy_Qwen2-AHN_ByteDance-Seed_AHN
Today I published a simplified educational version of the very complex Qwen2-AHN model. [I am developing a competing model architecture and wanted to study this one] The educational version Qwen2-AHN script is able to train on your custom dataset, and will generate a minimal vocab.json that fits your dataset, to reduce computational load. It runs on CPU or one GPU, in Google colab.
Top comments (0)