| { |
| "model_type": "sparse_autoencoder", |
| "base_model": "google/gemma-3-27b-it", |
| "architecture": "batch_top_k", |
| "framework": "dictionary_learning", |
| "layers": [ |
| 45, |
| 47 |
| ], |
| "saes": [ |
| { |
| "layer": 45, |
| "name": "dict_16k_k80", |
| "dict_size": 16384, |
| "k": 80, |
| "activation_dim": 5376, |
| "total_params": 176182528, |
| "path": "layer_45/dict_16k_k80" |
| }, |
| { |
| "layer": 45, |
| "name": "dict_16k_k160", |
| "dict_size": 16384, |
| "k": 160, |
| "activation_dim": 5376, |
| "total_params": 176182528, |
| "path": "layer_45/dict_16k_k160" |
| }, |
| { |
| "layer": 45, |
| "name": "dict_65k_k80", |
| "dict_size": 65536, |
| "k": 80, |
| "activation_dim": 5376, |
| "total_params": 704713984, |
| "path": "layer_45/dict_65k_k80" |
| }, |
| { |
| "layer": 45, |
| "name": "dict_65k_k160", |
| "dict_size": 65536, |
| "k": 160, |
| "activation_dim": 5376, |
| "total_params": 704713984, |
| "path": "layer_45/dict_65k_k160" |
| }, |
| { |
| "layer": 47, |
| "name": "dict_16k_k80", |
| "dict_size": 16384, |
| "k": 80, |
| "activation_dim": 5376, |
| "total_params": 176182528, |
| "path": "layer_47/dict_16k_k80" |
| }, |
| { |
| "layer": 47, |
| "name": "dict_16k_k160", |
| "dict_size": 16384, |
| "k": 160, |
| "activation_dim": 5376, |
| "total_params": 176182528, |
| "path": "layer_47/dict_16k_k160" |
| }, |
| { |
| "layer": 47, |
| "name": "dict_65k_k80", |
| "dict_size": 65536, |
| "k": 80, |
| "activation_dim": 5376, |
| "total_params": 704713984, |
| "path": "layer_47/dict_65k_k80" |
| }, |
| { |
| "layer": 47, |
| "name": "dict_65k_k160", |
| "dict_size": 65536, |
| "k": 160, |
| "activation_dim": 5376, |
| "total_params": 704713984, |
| "path": "layer_47/dict_65k_k160" |
| } |
| ], |
| "paper": { |
| "title": "Scaling and evaluating sparse autoencoders", |
| "authors": "Gao et al.", |
| "year": 2024, |
| "arxiv": "2406.04093", |
| "url": "https://arxiv.org/abs/2406.04093" |
| } |
| } |
