This dataset contains pre-computed embeddings for 77,477 English words from WordNet using the Gemma-2B model across all 27 layers.

metadata.json: Contains dataset metadata (model info, dimensions, word count) embeddings_layer_X.pkl: Pickle files containing embeddings for layer X (0-26)

import pickle from huggingface_hub import hf_hub_download

layer_0_path =… See the full description on the dataset page: https://huggingface.co/datasets/LeeHarrold/gemma-2b-dictionary-embeddings-all-layers.