MNIST-SAE Log 6

Work Done

  • cleaned up training pipeline for mnist saes
  • fixed dimension error where I was passing in the wrong channle (3 channel as opposed to 1) and fixed it so the MNIST was 3 channel input
    • standardization of images
  • added caching code to the EnhancedSAE since it seems to perform better
  • attempted to visualize meta-sae but ran into a bug

Confusions

  • How is a model like an autoencoder capable of learning? Like what mathematically allows for it?

Next Steps

  • figure out shape error bug (this is like 90% of the bugs I face) [x]
  • run the image visualization on the meta-sae [x]
  • fully establish the experiment pipelin to actually test whether the meta-sae has different representations
    • (1) track the amount of 10+ image pairs
      • as a future step make it unbounded rather than picking only top 10 or play with thresholds
    • (2) set up a VLM do to auto-interp
      • open source it?
      • Ask Louka (from slack) how he did interp for his vision sae
  • Once I have a working analysis portion set up infinite sae training
    • need a bigger dataset and/or (probably a train test split weighted 80% towards test side for MNIST)
  • Look at Top-K Sae and maybe use it since current training trends are pretty bad (>= 7.0 loss)

links

social