metadata

dataset_info:
  features:
    - name: layer
      dtype: uint8
    - name: neuron
      dtype: uint16
    - name: rank
      dtype: uint8
    - name: activation
      dtype: float32
    - name: position
      dtype: uint8
    - name: text
      dtype: string
    - name: id0
      dtype: uint16
    - name: id1
      dtype: uint16
    - name: id2
      dtype: uint16
    - name: id3
      dtype: uint16
    - name: id4
      dtype: uint16
    - name: id5
      dtype: uint16
    - name: id6
      dtype: uint16
    - name: id7
      dtype: uint16
    - name: id8
      dtype: uint16
    - name: id9
      dtype: uint16
    - name: id10
      dtype: uint16
    - name: id11
      dtype: uint16
    - name: id12
      dtype: uint16
    - name: id13
      dtype: uint16
    - name: id14
      dtype: uint16
    - name: id15
      dtype: uint16
  splits:
    - name: train
      num_bytes: 4202951650
      num_examples: 47185920
  download_size: 4766191985
  dataset_size: 4202951650

pythia-12b-neuron-dataset-examples

This dataset contains the top 64 highest activating dataset examples for each MLP neuron in Pythia-12b. The dataset examples are all 16 tokens long. See https://confirmlabs.org/posts/dreaming.html for details.

Columns:

layer: the layer of the neuron
neuron: the index of the neuron
rank: the rank of the example
activation: the activation of the neuron on the example
position: the token position for which the neuron is maximally activated.
id#: the token id at position # in 0 to 15.