orrav's picture
first commit
9f456e1
metadata
language: en
tags:
  - clip
  - vision
  - transformers
  - interpretability
  - sparse autoencoder
  - sae
  - mechanistic interpretability
library_name: torch
pipeline_tag: feature-extraction
metrics:
  - type: explained_variance
    value: 89.42
    pretty_name: Explained Variance %
    range:
      min: 0
      max: 100
  - type: l0
    value: 655.9
    pretty_name: L0

CLIP-B-32 Sparse Autoencoder x64 vanilla - L1:1e-05

Training Details

  • Base Model: CLIP-ViT-B-32 (LAION DataComp.XL-s13B-b90K)
  • Layer: 10
  • Component: hook_resid_post

Model Architecture

  • Input Dimension: 768
  • SAE Dimension: 49,152
  • Expansion Factor: x64 (vanilla architecture)
  • Activation Function: ReLU
  • Initialization: encoder_transpose_decoder
  • CLS_only: true

Performance Metrics

  • L1 Coefficient: 1e-05
  • L0 Sparsity: 655.9
  • Explained Variance: 89.42%

Training Configuration

  • Learning Rate: 0.01
  • LR Scheduler: Cosine Annealing with Warmup (200 steps)
  • Epochs: 10
  • Gradient Clipping: 1.0