├── README.md ├── sae ├── __init__.py ├── activation_store.py ├── analysis.py ├── config.py ├── metrics.py ├── sparse_autoencoder.py ├── train.py └── utils.py └── sae_training_templates ├── cache_activations.ipynb ├── generate_feature_dashboards.ipynb ├── load_sae_from_hf.ipynb ├── pythia2.8b_run.ipynb ├── test_anthropic_implementation.ipynb ├── test_custom_loss.ipynb ├── test_gated_sae.ipynb ├── test_l0_approx.ipynb ├── test_multiple.ipynb ├── test_openai_topk.ipynb └── test_transcoder.ipynb /README.md: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/efarrell1/train_sparse_autoencoder/HEAD/README.md -------------------------------------------------------------------------------- /sae/__init__.py: -------------------------------------------------------------------------------- 1 | -------------------------------------------------------------------------------- /sae/activation_store.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/efarrell1/train_sparse_autoencoder/HEAD/sae/activation_store.py -------------------------------------------------------------------------------- /sae/analysis.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/efarrell1/train_sparse_autoencoder/HEAD/sae/analysis.py -------------------------------------------------------------------------------- /sae/config.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/efarrell1/train_sparse_autoencoder/HEAD/sae/config.py -------------------------------------------------------------------------------- /sae/metrics.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/efarrell1/train_sparse_autoencoder/HEAD/sae/metrics.py -------------------------------------------------------------------------------- /sae/sparse_autoencoder.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/efarrell1/train_sparse_autoencoder/HEAD/sae/sparse_autoencoder.py -------------------------------------------------------------------------------- /sae/train.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/efarrell1/train_sparse_autoencoder/HEAD/sae/train.py -------------------------------------------------------------------------------- /sae/utils.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/efarrell1/train_sparse_autoencoder/HEAD/sae/utils.py -------------------------------------------------------------------------------- /sae_training_templates/cache_activations.ipynb: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/efarrell1/train_sparse_autoencoder/HEAD/sae_training_templates/cache_activations.ipynb -------------------------------------------------------------------------------- /sae_training_templates/generate_feature_dashboards.ipynb: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/efarrell1/train_sparse_autoencoder/HEAD/sae_training_templates/generate_feature_dashboards.ipynb -------------------------------------------------------------------------------- /sae_training_templates/load_sae_from_hf.ipynb: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/efarrell1/train_sparse_autoencoder/HEAD/sae_training_templates/load_sae_from_hf.ipynb -------------------------------------------------------------------------------- /sae_training_templates/pythia2.8b_run.ipynb: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/efarrell1/train_sparse_autoencoder/HEAD/sae_training_templates/pythia2.8b_run.ipynb -------------------------------------------------------------------------------- /sae_training_templates/test_anthropic_implementation.ipynb: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/efarrell1/train_sparse_autoencoder/HEAD/sae_training_templates/test_anthropic_implementation.ipynb -------------------------------------------------------------------------------- /sae_training_templates/test_custom_loss.ipynb: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/efarrell1/train_sparse_autoencoder/HEAD/sae_training_templates/test_custom_loss.ipynb -------------------------------------------------------------------------------- /sae_training_templates/test_gated_sae.ipynb: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/efarrell1/train_sparse_autoencoder/HEAD/sae_training_templates/test_gated_sae.ipynb -------------------------------------------------------------------------------- /sae_training_templates/test_l0_approx.ipynb: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/efarrell1/train_sparse_autoencoder/HEAD/sae_training_templates/test_l0_approx.ipynb -------------------------------------------------------------------------------- /sae_training_templates/test_multiple.ipynb: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/efarrell1/train_sparse_autoencoder/HEAD/sae_training_templates/test_multiple.ipynb -------------------------------------------------------------------------------- /sae_training_templates/test_openai_topk.ipynb: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/efarrell1/train_sparse_autoencoder/HEAD/sae_training_templates/test_openai_topk.ipynb -------------------------------------------------------------------------------- /sae_training_templates/test_transcoder.ipynb: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/efarrell1/train_sparse_autoencoder/HEAD/sae_training_templates/test_transcoder.ipynb --------------------------------------------------------------------------------