YAML Metadata Warning: empty or missing yaml metadata in repo card (https://huggingface.co/docs/hub/model-cards#model-card-metadata)

CLIP Sparse Autoencoder Checkpoint

This model is a sparse autoencoder trained on CLIP's internal representations.

Model Details

Architecture

  • Layer: 3
  • Layer Type: hook_resid_post
  • Model: open-clip:laion/CLIP-ViT-B-32-DataComp.XL-s13B-b90K
  • Dictionary Size: 49152
  • Input Dimension: 768
  • Expansion Factor: 64
  • CLS Token Only: True

Training

  • Training Images: 183087104
  • Learning Rate: 0.0002
  • L1 Coefficient: 0.3000
  • Batch Size: 4096
  • Context Size: 1

Performance Metrics

Sparsity

  • L0 (Active Features): 64
  • Dead Features: 29335
  • Mean Log10 Feature Sparsity: -9.2914
  • Features Below 1e-5: 48859
  • Features Below 1e-6: 41691
  • Mean Passes Since Fired: 16714.2930

Reconstruction

  • Explained Variance: 0.9595
  • Explained Variance Std: 0.0150
  • MSE Loss: 0.0001
  • L1 Loss: 0
  • Overall Loss: 0.0001

Training Details

  • Training Duration: 17942.3141 seconds
  • Final Learning Rate: 0.0002
  • Warm Up Steps: 200
  • Gradient Clipping: 1

Additional Information

Downloads last month
8
Inference API
Unable to determine this model's library. Check the docs .