sae-lens 6.24.0__tar.gz → 6.25.0__tar.gz
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- {sae_lens-6.24.0 → sae_lens-6.25.0}/PKG-INFO +1 -1
- {sae_lens-6.24.0 → sae_lens-6.25.0}/pyproject.toml +1 -1
- {sae_lens-6.24.0 → sae_lens-6.25.0}/sae_lens/__init__.py +1 -1
- {sae_lens-6.24.0 → sae_lens-6.25.0}/sae_lens/loading/pretrained_sae_loaders.py +0 -9
- {sae_lens-6.24.0 → sae_lens-6.25.0}/sae_lens/pretrained_saes.yaml +160 -144
- {sae_lens-6.24.0 → sae_lens-6.25.0}/LICENSE +0 -0
- {sae_lens-6.24.0 → sae_lens-6.25.0}/README.md +0 -0
- {sae_lens-6.24.0 → sae_lens-6.25.0}/sae_lens/analysis/__init__.py +0 -0
- {sae_lens-6.24.0 → sae_lens-6.25.0}/sae_lens/analysis/hooked_sae_transformer.py +0 -0
- {sae_lens-6.24.0 → sae_lens-6.25.0}/sae_lens/analysis/neuronpedia_integration.py +0 -0
- {sae_lens-6.24.0 → sae_lens-6.25.0}/sae_lens/cache_activations_runner.py +0 -0
- {sae_lens-6.24.0 → sae_lens-6.25.0}/sae_lens/config.py +0 -0
- {sae_lens-6.24.0 → sae_lens-6.25.0}/sae_lens/constants.py +0 -0
- {sae_lens-6.24.0 → sae_lens-6.25.0}/sae_lens/evals.py +0 -0
- {sae_lens-6.24.0 → sae_lens-6.25.0}/sae_lens/llm_sae_training_runner.py +0 -0
- {sae_lens-6.24.0 → sae_lens-6.25.0}/sae_lens/load_model.py +0 -0
- {sae_lens-6.24.0 → sae_lens-6.25.0}/sae_lens/loading/__init__.py +0 -0
- {sae_lens-6.24.0 → sae_lens-6.25.0}/sae_lens/loading/pretrained_saes_directory.py +0 -0
- {sae_lens-6.24.0 → sae_lens-6.25.0}/sae_lens/pretokenize_runner.py +0 -0
- {sae_lens-6.24.0 → sae_lens-6.25.0}/sae_lens/registry.py +0 -0
- {sae_lens-6.24.0 → sae_lens-6.25.0}/sae_lens/saes/__init__.py +0 -0
- {sae_lens-6.24.0 → sae_lens-6.25.0}/sae_lens/saes/batchtopk_sae.py +0 -0
- {sae_lens-6.24.0 → sae_lens-6.25.0}/sae_lens/saes/gated_sae.py +0 -0
- {sae_lens-6.24.0 → sae_lens-6.25.0}/sae_lens/saes/jumprelu_sae.py +0 -0
- {sae_lens-6.24.0 → sae_lens-6.25.0}/sae_lens/saes/matryoshka_batchtopk_sae.py +0 -0
- {sae_lens-6.24.0 → sae_lens-6.25.0}/sae_lens/saes/sae.py +0 -0
- {sae_lens-6.24.0 → sae_lens-6.25.0}/sae_lens/saes/standard_sae.py +0 -0
- {sae_lens-6.24.0 → sae_lens-6.25.0}/sae_lens/saes/temporal_sae.py +0 -0
- {sae_lens-6.24.0 → sae_lens-6.25.0}/sae_lens/saes/topk_sae.py +0 -0
- {sae_lens-6.24.0 → sae_lens-6.25.0}/sae_lens/saes/transcoder.py +0 -0
- {sae_lens-6.24.0 → sae_lens-6.25.0}/sae_lens/tokenization_and_batching.py +0 -0
- {sae_lens-6.24.0 → sae_lens-6.25.0}/sae_lens/training/__init__.py +0 -0
- {sae_lens-6.24.0 → sae_lens-6.25.0}/sae_lens/training/activation_scaler.py +0 -0
- {sae_lens-6.24.0 → sae_lens-6.25.0}/sae_lens/training/activations_store.py +0 -0
- {sae_lens-6.24.0 → sae_lens-6.25.0}/sae_lens/training/mixing_buffer.py +0 -0
- {sae_lens-6.24.0 → sae_lens-6.25.0}/sae_lens/training/optim.py +0 -0
- {sae_lens-6.24.0 → sae_lens-6.25.0}/sae_lens/training/sae_trainer.py +0 -0
- {sae_lens-6.24.0 → sae_lens-6.25.0}/sae_lens/training/types.py +0 -0
- {sae_lens-6.24.0 → sae_lens-6.25.0}/sae_lens/training/upload_saes_to_huggingface.py +0 -0
- {sae_lens-6.24.0 → sae_lens-6.25.0}/sae_lens/tutorial/tsea.py +0 -0
- {sae_lens-6.24.0 → sae_lens-6.25.0}/sae_lens/util.py +0 -0
|
@@ -682,15 +682,6 @@ def gemma_3_sae_huggingface_loader(
|
|
|
682
682
|
cfg_overrides,
|
|
683
683
|
)
|
|
684
684
|
|
|
685
|
-
# replace folder name of 65k with 64k
|
|
686
|
-
# TODO: remove this workaround once weights are fixed
|
|
687
|
-
if "270m-pt" in repo_id:
|
|
688
|
-
if "65k" in folder_name:
|
|
689
|
-
folder_name = folder_name.replace("65k", "64k")
|
|
690
|
-
# replace folder name of 262k with 250k
|
|
691
|
-
if "262k" in folder_name:
|
|
692
|
-
folder_name = folder_name.replace("262k", "250k")
|
|
693
|
-
|
|
694
685
|
params_file = "params.safetensors"
|
|
695
686
|
if "clt" in folder_name:
|
|
696
687
|
params_file = folder_name.split("/")[-1] + ".safetensors"
|
|
@@ -9072,150 +9072,150 @@ gemma-scope-2-27b-it-transcoders-all:
|
|
|
9072
9072
|
- id: layer_5_width_262k_l0_small_affine
|
|
9073
9073
|
path: transcoder_all/layer_5_width_262k_l0_small_affine
|
|
9074
9074
|
l0: 12
|
|
9075
|
-
- id: layer_60_width_16k_l0_big
|
|
9076
|
-
|
|
9077
|
-
|
|
9078
|
-
- id: layer_60_width_16k_l0_big_affine
|
|
9079
|
-
|
|
9080
|
-
|
|
9081
|
-
- id: layer_60_width_16k_l0_small
|
|
9082
|
-
|
|
9083
|
-
|
|
9084
|
-
- id: layer_60_width_16k_l0_small_affine
|
|
9085
|
-
|
|
9086
|
-
|
|
9087
|
-
- id: layer_60_width_262k_l0_big
|
|
9088
|
-
|
|
9089
|
-
|
|
9090
|
-
- id: layer_60_width_262k_l0_big_affine
|
|
9091
|
-
|
|
9092
|
-
|
|
9093
|
-
- id: layer_60_width_262k_l0_small
|
|
9094
|
-
|
|
9095
|
-
|
|
9096
|
-
- id: layer_60_width_262k_l0_small_affine
|
|
9097
|
-
|
|
9098
|
-
|
|
9099
|
-
- id: layer_61_width_16k_l0_big
|
|
9100
|
-
|
|
9101
|
-
|
|
9102
|
-
- id: layer_61_width_16k_l0_big_affine
|
|
9103
|
-
|
|
9104
|
-
|
|
9105
|
-
- id: layer_61_width_16k_l0_small
|
|
9106
|
-
|
|
9107
|
-
|
|
9108
|
-
- id: layer_61_width_16k_l0_small_affine
|
|
9109
|
-
|
|
9110
|
-
|
|
9111
|
-
- id: layer_61_width_262k_l0_big
|
|
9112
|
-
|
|
9113
|
-
|
|
9114
|
-
- id: layer_61_width_262k_l0_big_affine
|
|
9115
|
-
|
|
9116
|
-
|
|
9117
|
-
- id: layer_61_width_262k_l0_small
|
|
9118
|
-
|
|
9119
|
-
|
|
9120
|
-
- id: layer_61_width_262k_l0_small_affine
|
|
9121
|
-
|
|
9122
|
-
|
|
9123
|
-
- id: layer_6_width_16k_l0_big
|
|
9124
|
-
|
|
9125
|
-
|
|
9126
|
-
- id: layer_6_width_16k_l0_big_affine
|
|
9127
|
-
|
|
9128
|
-
|
|
9129
|
-
- id: layer_6_width_16k_l0_small
|
|
9130
|
-
|
|
9131
|
-
|
|
9132
|
-
- id: layer_6_width_16k_l0_small_affine
|
|
9133
|
-
|
|
9134
|
-
|
|
9135
|
-
- id: layer_6_width_262k_l0_big
|
|
9136
|
-
|
|
9137
|
-
|
|
9138
|
-
- id: layer_6_width_262k_l0_big_affine
|
|
9139
|
-
|
|
9140
|
-
|
|
9141
|
-
- id: layer_6_width_262k_l0_small
|
|
9142
|
-
|
|
9143
|
-
|
|
9144
|
-
- id: layer_6_width_262k_l0_small_affine
|
|
9145
|
-
|
|
9146
|
-
|
|
9147
|
-
- id: layer_7_width_16k_l0_big
|
|
9148
|
-
|
|
9149
|
-
|
|
9150
|
-
- id: layer_7_width_16k_l0_big_affine
|
|
9151
|
-
|
|
9152
|
-
|
|
9153
|
-
- id: layer_7_width_16k_l0_small
|
|
9154
|
-
|
|
9155
|
-
|
|
9156
|
-
- id: layer_7_width_16k_l0_small_affine
|
|
9157
|
-
|
|
9158
|
-
|
|
9159
|
-
- id: layer_7_width_262k_l0_big
|
|
9160
|
-
|
|
9161
|
-
|
|
9162
|
-
- id: layer_7_width_262k_l0_big_affine
|
|
9163
|
-
|
|
9164
|
-
|
|
9165
|
-
- id: layer_7_width_262k_l0_small
|
|
9166
|
-
|
|
9167
|
-
|
|
9168
|
-
- id: layer_7_width_262k_l0_small_affine
|
|
9169
|
-
|
|
9170
|
-
|
|
9171
|
-
- id: layer_8_width_16k_l0_big
|
|
9172
|
-
|
|
9173
|
-
|
|
9174
|
-
- id: layer_8_width_16k_l0_big_affine
|
|
9175
|
-
|
|
9176
|
-
|
|
9177
|
-
- id: layer_8_width_16k_l0_small
|
|
9178
|
-
|
|
9179
|
-
|
|
9180
|
-
- id: layer_8_width_16k_l0_small_affine
|
|
9181
|
-
|
|
9182
|
-
|
|
9183
|
-
- id: layer_8_width_262k_l0_big
|
|
9184
|
-
|
|
9185
|
-
|
|
9186
|
-
- id: layer_8_width_262k_l0_big_affine
|
|
9187
|
-
|
|
9188
|
-
|
|
9189
|
-
- id: layer_8_width_262k_l0_small
|
|
9190
|
-
|
|
9191
|
-
|
|
9192
|
-
- id: layer_8_width_262k_l0_small_affine
|
|
9193
|
-
|
|
9194
|
-
|
|
9195
|
-
- id: layer_9_width_16k_l0_big
|
|
9196
|
-
|
|
9197
|
-
|
|
9198
|
-
- id: layer_9_width_16k_l0_big_affine
|
|
9199
|
-
|
|
9200
|
-
|
|
9201
|
-
- id: layer_9_width_16k_l0_small
|
|
9202
|
-
|
|
9203
|
-
|
|
9204
|
-
- id: layer_9_width_16k_l0_small_affine
|
|
9205
|
-
|
|
9206
|
-
|
|
9207
|
-
- id: layer_9_width_262k_l0_big
|
|
9208
|
-
|
|
9209
|
-
|
|
9210
|
-
- id: layer_9_width_262k_l0_big_affine
|
|
9211
|
-
|
|
9212
|
-
|
|
9213
|
-
- id: layer_9_width_262k_l0_small
|
|
9214
|
-
|
|
9215
|
-
|
|
9216
|
-
- id: layer_9_width_262k_l0_small_affine
|
|
9217
|
-
|
|
9218
|
-
|
|
9075
|
+
# - id: layer_60_width_16k_l0_big
|
|
9076
|
+
# path: transcoder_all/layer_60_width_16k_l0_big
|
|
9077
|
+
# l0: 120
|
|
9078
|
+
# - id: layer_60_width_16k_l0_big_affine
|
|
9079
|
+
# path: transcoder_all/layer_60_width_16k_l0_big_affine
|
|
9080
|
+
# l0: 120
|
|
9081
|
+
# - id: layer_60_width_16k_l0_small
|
|
9082
|
+
# path: transcoder_all/layer_60_width_16k_l0_small
|
|
9083
|
+
# l0: 20
|
|
9084
|
+
# - id: layer_60_width_16k_l0_small_affine
|
|
9085
|
+
# path: transcoder_all/layer_60_width_16k_l0_small_affine
|
|
9086
|
+
# l0: 20
|
|
9087
|
+
# - id: layer_60_width_262k_l0_big
|
|
9088
|
+
# path: transcoder_all/layer_60_width_262k_l0_big
|
|
9089
|
+
# l0: 120
|
|
9090
|
+
# - id: layer_60_width_262k_l0_big_affine
|
|
9091
|
+
# path: transcoder_all/layer_60_width_262k_l0_big_affine
|
|
9092
|
+
# l0: 120
|
|
9093
|
+
# - id: layer_60_width_262k_l0_small
|
|
9094
|
+
# path: transcoder_all/layer_60_width_262k_l0_small
|
|
9095
|
+
# l0: 20
|
|
9096
|
+
# - id: layer_60_width_262k_l0_small_affine
|
|
9097
|
+
# path: transcoder_all/layer_60_width_262k_l0_small_affine
|
|
9098
|
+
# l0: 20
|
|
9099
|
+
# - id: layer_61_width_16k_l0_big
|
|
9100
|
+
# path: transcoder_all/layer_61_width_16k_l0_big
|
|
9101
|
+
# l0: 120
|
|
9102
|
+
# - id: layer_61_width_16k_l0_big_affine
|
|
9103
|
+
# path: transcoder_all/layer_61_width_16k_l0_big_affine
|
|
9104
|
+
# l0: 120
|
|
9105
|
+
# - id: layer_61_width_16k_l0_small
|
|
9106
|
+
# path: transcoder_all/layer_61_width_16k_l0_small
|
|
9107
|
+
# l0: 20
|
|
9108
|
+
# - id: layer_61_width_16k_l0_small_affine
|
|
9109
|
+
# path: transcoder_all/layer_61_width_16k_l0_small_affine
|
|
9110
|
+
# l0: 20
|
|
9111
|
+
# - id: layer_61_width_262k_l0_big
|
|
9112
|
+
# path: transcoder_all/layer_61_width_262k_l0_big
|
|
9113
|
+
# l0: 120
|
|
9114
|
+
# - id: layer_61_width_262k_l0_big_affine
|
|
9115
|
+
# path: transcoder_all/layer_61_width_262k_l0_big_affine
|
|
9116
|
+
# l0: 120
|
|
9117
|
+
# - id: layer_61_width_262k_l0_small
|
|
9118
|
+
# path: transcoder_all/layer_61_width_262k_l0_small
|
|
9119
|
+
# l0: 20
|
|
9120
|
+
# - id: layer_61_width_262k_l0_small_affine
|
|
9121
|
+
# path: transcoder_all/layer_61_width_262k_l0_small_affine
|
|
9122
|
+
# l0: 20
|
|
9123
|
+
# - id: layer_6_width_16k_l0_big
|
|
9124
|
+
# path: transcoder_all/layer_6_width_16k_l0_big
|
|
9125
|
+
# l0: 77
|
|
9126
|
+
# - id: layer_6_width_16k_l0_big_affine
|
|
9127
|
+
# path: transcoder_all/layer_6_width_16k_l0_big_affine
|
|
9128
|
+
# l0: 77
|
|
9129
|
+
# - id: layer_6_width_16k_l0_small
|
|
9130
|
+
# path: transcoder_all/layer_6_width_16k_l0_small
|
|
9131
|
+
# l0: 12
|
|
9132
|
+
# - id: layer_6_width_16k_l0_small_affine
|
|
9133
|
+
# path: transcoder_all/layer_6_width_16k_l0_small_affine
|
|
9134
|
+
# l0: 12
|
|
9135
|
+
# - id: layer_6_width_262k_l0_big
|
|
9136
|
+
# path: transcoder_all/layer_6_width_262k_l0_big
|
|
9137
|
+
# l0: 77
|
|
9138
|
+
# - id: layer_6_width_262k_l0_big_affine
|
|
9139
|
+
# path: transcoder_all/layer_6_width_262k_l0_big_affine
|
|
9140
|
+
# l0: 77
|
|
9141
|
+
# - id: layer_6_width_262k_l0_small
|
|
9142
|
+
# path: transcoder_all/layer_6_width_262k_l0_small
|
|
9143
|
+
# l0: 12
|
|
9144
|
+
# - id: layer_6_width_262k_l0_small_affine
|
|
9145
|
+
# path: transcoder_all/layer_6_width_262k_l0_small_affine
|
|
9146
|
+
# l0: 12
|
|
9147
|
+
# - id: layer_7_width_16k_l0_big
|
|
9148
|
+
# path: transcoder_all/layer_7_width_16k_l0_big
|
|
9149
|
+
# l0: 80
|
|
9150
|
+
# - id: layer_7_width_16k_l0_big_affine
|
|
9151
|
+
# path: transcoder_all/layer_7_width_16k_l0_big_affine
|
|
9152
|
+
# l0: 80
|
|
9153
|
+
# - id: layer_7_width_16k_l0_small
|
|
9154
|
+
# path: transcoder_all/layer_7_width_16k_l0_small
|
|
9155
|
+
# l0: 13
|
|
9156
|
+
# - id: layer_7_width_16k_l0_small_affine
|
|
9157
|
+
# path: transcoder_all/layer_7_width_16k_l0_small_affine
|
|
9158
|
+
# l0: 13
|
|
9159
|
+
# - id: layer_7_width_262k_l0_big
|
|
9160
|
+
# path: transcoder_all/layer_7_width_262k_l0_big
|
|
9161
|
+
# l0: 80
|
|
9162
|
+
# - id: layer_7_width_262k_l0_big_affine
|
|
9163
|
+
# path: transcoder_all/layer_7_width_262k_l0_big_affine
|
|
9164
|
+
# l0: 80
|
|
9165
|
+
# - id: layer_7_width_262k_l0_small
|
|
9166
|
+
# path: transcoder_all/layer_7_width_262k_l0_small
|
|
9167
|
+
# l0: 13
|
|
9168
|
+
# - id: layer_7_width_262k_l0_small_affine
|
|
9169
|
+
# path: transcoder_all/layer_7_width_262k_l0_small_affine
|
|
9170
|
+
# l0: 13
|
|
9171
|
+
# - id: layer_8_width_16k_l0_big
|
|
9172
|
+
# path: transcoder_all/layer_8_width_16k_l0_big
|
|
9173
|
+
# l0: 83
|
|
9174
|
+
# - id: layer_8_width_16k_l0_big_affine
|
|
9175
|
+
# path: transcoder_all/layer_8_width_16k_l0_big_affine
|
|
9176
|
+
# l0: 83
|
|
9177
|
+
# - id: layer_8_width_16k_l0_small
|
|
9178
|
+
# path: transcoder_all/layer_8_width_16k_l0_small
|
|
9179
|
+
# l0: 13
|
|
9180
|
+
# - id: layer_8_width_16k_l0_small_affine
|
|
9181
|
+
# path: transcoder_all/layer_8_width_16k_l0_small_affine
|
|
9182
|
+
# l0: 13
|
|
9183
|
+
# - id: layer_8_width_262k_l0_big
|
|
9184
|
+
# path: transcoder_all/layer_8_width_262k_l0_big
|
|
9185
|
+
# l0: 83
|
|
9186
|
+
# - id: layer_8_width_262k_l0_big_affine
|
|
9187
|
+
# path: transcoder_all/layer_8_width_262k_l0_big_affine
|
|
9188
|
+
# l0: 83
|
|
9189
|
+
# - id: layer_8_width_262k_l0_small
|
|
9190
|
+
# path: transcoder_all/layer_8_width_262k_l0_small
|
|
9191
|
+
# l0: 13
|
|
9192
|
+
# - id: layer_8_width_262k_l0_small_affine
|
|
9193
|
+
# path: transcoder_all/layer_8_width_262k_l0_small_affine
|
|
9194
|
+
# l0: 13
|
|
9195
|
+
# - id: layer_9_width_16k_l0_big
|
|
9196
|
+
# path: transcoder_all/layer_9_width_16k_l0_big
|
|
9197
|
+
# l0: 86
|
|
9198
|
+
# - id: layer_9_width_16k_l0_big_affine
|
|
9199
|
+
# path: transcoder_all/layer_9_width_16k_l0_big_affine
|
|
9200
|
+
# l0: 86
|
|
9201
|
+
# - id: layer_9_width_16k_l0_small
|
|
9202
|
+
# path: transcoder_all/layer_9_width_16k_l0_small
|
|
9203
|
+
# l0: 14
|
|
9204
|
+
# - id: layer_9_width_16k_l0_small_affine
|
|
9205
|
+
# path: transcoder_all/layer_9_width_16k_l0_small_affine
|
|
9206
|
+
# l0: 14
|
|
9207
|
+
# - id: layer_9_width_262k_l0_big
|
|
9208
|
+
# path: transcoder_all/layer_9_width_262k_l0_big
|
|
9209
|
+
# l0: 86
|
|
9210
|
+
# - id: layer_9_width_262k_l0_big_affine
|
|
9211
|
+
# path: transcoder_all/layer_9_width_262k_l0_big_affine
|
|
9212
|
+
# l0: 86
|
|
9213
|
+
# - id: layer_9_width_262k_l0_small
|
|
9214
|
+
# path: transcoder_all/layer_9_width_262k_l0_small
|
|
9215
|
+
# l0: 14
|
|
9216
|
+
# - id: layer_9_width_262k_l0_small_affine
|
|
9217
|
+
# path: transcoder_all/layer_9_width_262k_l0_small_affine
|
|
9218
|
+
# l0: 14
|
|
9219
9219
|
gemma-scope-2-27b-it-transcoders:
|
|
9220
9220
|
conversion_func: gemma_3
|
|
9221
9221
|
model: google/gemma-3-27b-it
|
|
@@ -10197,6 +10197,7 @@ gemma-scope-2-27b-it-res:
|
|
|
10197
10197
|
- id: layer_16_width_16k_l0_medium
|
|
10198
10198
|
path: resid_post/layer_16_width_16k_l0_medium
|
|
10199
10199
|
l0: 53
|
|
10200
|
+
neuronpedia: gemma-3-27b-it/16-gemmascope-2-res-16k
|
|
10200
10201
|
- id: layer_16_width_16k_l0_small
|
|
10201
10202
|
path: resid_post/layer_16_width_16k_l0_small
|
|
10202
10203
|
l0: 17
|
|
@@ -10206,6 +10207,7 @@ gemma-scope-2-27b-it-res:
|
|
|
10206
10207
|
- id: layer_16_width_1m_l0_medium
|
|
10207
10208
|
path: resid_post/layer_16_width_1m_l0_medium
|
|
10208
10209
|
l0: 53
|
|
10210
|
+
neuronpedia: gemma-3-27b-it/16-gemmascope-2-res-1m
|
|
10209
10211
|
- id: layer_16_width_1m_l0_small
|
|
10210
10212
|
path: resid_post/layer_16_width_1m_l0_small
|
|
10211
10213
|
l0: 17
|
|
@@ -10215,6 +10217,7 @@ gemma-scope-2-27b-it-res:
|
|
|
10215
10217
|
- id: layer_16_width_262k_l0_medium
|
|
10216
10218
|
path: resid_post/layer_16_width_262k_l0_medium
|
|
10217
10219
|
l0: 53
|
|
10220
|
+
neuronpedia: gemma-3-27b-it/16-gemmascope-2-res-262k
|
|
10218
10221
|
- id: layer_16_width_262k_l0_medium_seed_1
|
|
10219
10222
|
path: resid_post/layer_16_width_262k_l0_medium_seed_1
|
|
10220
10223
|
l0: 53
|
|
@@ -10227,6 +10230,7 @@ gemma-scope-2-27b-it-res:
|
|
|
10227
10230
|
- id: layer_16_width_65k_l0_medium
|
|
10228
10231
|
path: resid_post/layer_16_width_65k_l0_medium
|
|
10229
10232
|
l0: 53
|
|
10233
|
+
neuronpedia: gemma-3-27b-it/16-gemmascope-2-res-65k
|
|
10230
10234
|
- id: layer_16_width_65k_l0_small
|
|
10231
10235
|
path: resid_post/layer_16_width_65k_l0_small
|
|
10232
10236
|
l0: 17
|
|
@@ -10236,6 +10240,7 @@ gemma-scope-2-27b-it-res:
|
|
|
10236
10240
|
- id: layer_31_width_16k_l0_medium
|
|
10237
10241
|
path: resid_post/layer_31_width_16k_l0_medium
|
|
10238
10242
|
l0: 60
|
|
10243
|
+
neuronpedia: gemma-3-27b-it/31-gemmascope-2-res-16k
|
|
10239
10244
|
- id: layer_31_width_16k_l0_small
|
|
10240
10245
|
path: resid_post/layer_31_width_16k_l0_small
|
|
10241
10246
|
l0: 20
|
|
@@ -10245,6 +10250,7 @@ gemma-scope-2-27b-it-res:
|
|
|
10245
10250
|
- id: layer_31_width_1m_l0_medium
|
|
10246
10251
|
path: resid_post/layer_31_width_1m_l0_medium
|
|
10247
10252
|
l0: 60
|
|
10253
|
+
neuronpedia: gemma-3-27b-it/31-gemmascope-2-res-1m
|
|
10248
10254
|
- id: layer_31_width_1m_l0_small
|
|
10249
10255
|
path: resid_post/layer_31_width_1m_l0_small
|
|
10250
10256
|
l0: 20
|
|
@@ -10254,6 +10260,7 @@ gemma-scope-2-27b-it-res:
|
|
|
10254
10260
|
- id: layer_31_width_262k_l0_medium
|
|
10255
10261
|
path: resid_post/layer_31_width_262k_l0_medium
|
|
10256
10262
|
l0: 60
|
|
10263
|
+
neuronpedia: gemma-3-27b-it/31-gemmascope-2-res-262k
|
|
10257
10264
|
- id: layer_31_width_262k_l0_medium_seed_1
|
|
10258
10265
|
path: resid_post/layer_31_width_262k_l0_medium_seed_1
|
|
10259
10266
|
l0: 60
|
|
@@ -10266,6 +10273,7 @@ gemma-scope-2-27b-it-res:
|
|
|
10266
10273
|
- id: layer_31_width_65k_l0_medium
|
|
10267
10274
|
path: resid_post/layer_31_width_65k_l0_medium
|
|
10268
10275
|
l0: 60
|
|
10276
|
+
neuronpedia: gemma-3-27b-it/31-gemmascope-2-res-65k
|
|
10269
10277
|
- id: layer_31_width_65k_l0_small
|
|
10270
10278
|
path: resid_post/layer_31_width_65k_l0_small
|
|
10271
10279
|
l0: 20
|
|
@@ -10275,6 +10283,7 @@ gemma-scope-2-27b-it-res:
|
|
|
10275
10283
|
- id: layer_40_width_16k_l0_medium
|
|
10276
10284
|
path: resid_post/layer_40_width_16k_l0_medium
|
|
10277
10285
|
l0: 60
|
|
10286
|
+
neuronpedia: gemma-3-27b-it/40-gemmascope-2-res-16k
|
|
10278
10287
|
- id: layer_40_width_16k_l0_small
|
|
10279
10288
|
path: resid_post/layer_40_width_16k_l0_small
|
|
10280
10289
|
l0: 20
|
|
@@ -10284,6 +10293,7 @@ gemma-scope-2-27b-it-res:
|
|
|
10284
10293
|
- id: layer_40_width_1m_l0_medium
|
|
10285
10294
|
path: resid_post/layer_40_width_1m_l0_medium
|
|
10286
10295
|
l0: 60
|
|
10296
|
+
neuronpedia: gemma-3-27b-it/40-gemmascope-2-res-1m
|
|
10287
10297
|
- id: layer_40_width_1m_l0_small
|
|
10288
10298
|
path: resid_post/layer_40_width_1m_l0_small
|
|
10289
10299
|
l0: 20
|
|
@@ -10293,6 +10303,7 @@ gemma-scope-2-27b-it-res:
|
|
|
10293
10303
|
- id: layer_40_width_262k_l0_medium
|
|
10294
10304
|
path: resid_post/layer_40_width_262k_l0_medium
|
|
10295
10305
|
l0: 60
|
|
10306
|
+
neuronpedia: gemma-3-27b-it/40-gemmascope-2-res-262k
|
|
10296
10307
|
- id: layer_40_width_262k_l0_medium_seed_1
|
|
10297
10308
|
path: resid_post/layer_40_width_262k_l0_medium_seed_1
|
|
10298
10309
|
l0: 60
|
|
@@ -10305,6 +10316,7 @@ gemma-scope-2-27b-it-res:
|
|
|
10305
10316
|
- id: layer_40_width_65k_l0_medium
|
|
10306
10317
|
path: resid_post/layer_40_width_65k_l0_medium
|
|
10307
10318
|
l0: 60
|
|
10319
|
+
neuronpedia: gemma-3-27b-it/40-gemmascope-2-res-65k
|
|
10308
10320
|
- id: layer_40_width_65k_l0_small
|
|
10309
10321
|
path: resid_post/layer_40_width_65k_l0_small
|
|
10310
10322
|
l0: 20
|
|
@@ -10314,6 +10326,7 @@ gemma-scope-2-27b-it-res:
|
|
|
10314
10326
|
- id: layer_53_width_16k_l0_medium
|
|
10315
10327
|
path: resid_post/layer_53_width_16k_l0_medium
|
|
10316
10328
|
l0: 60
|
|
10329
|
+
neuronpedia: gemma-3-27b-it/53-gemmascope-2-res-16k
|
|
10317
10330
|
- id: layer_53_width_16k_l0_small
|
|
10318
10331
|
path: resid_post/layer_53_width_16k_l0_small
|
|
10319
10332
|
l0: 20
|
|
@@ -10323,6 +10336,7 @@ gemma-scope-2-27b-it-res:
|
|
|
10323
10336
|
- id: layer_53_width_1m_l0_medium
|
|
10324
10337
|
path: resid_post/layer_53_width_1m_l0_medium
|
|
10325
10338
|
l0: 60
|
|
10339
|
+
neuronpedia: gemma-3-27b-it/53-gemmascope-2-res-1m
|
|
10326
10340
|
- id: layer_53_width_1m_l0_small
|
|
10327
10341
|
path: resid_post/layer_53_width_1m_l0_small
|
|
10328
10342
|
l0: 20
|
|
@@ -10332,6 +10346,7 @@ gemma-scope-2-27b-it-res:
|
|
|
10332
10346
|
- id: layer_53_width_262k_l0_medium
|
|
10333
10347
|
path: resid_post/layer_53_width_262k_l0_medium
|
|
10334
10348
|
l0: 60
|
|
10349
|
+
neuronpedia: gemma-3-27b-it/53-gemmascope-2-res-262k
|
|
10335
10350
|
- id: layer_53_width_262k_l0_medium_seed_1
|
|
10336
10351
|
path: resid_post/layer_53_width_262k_l0_medium_seed_1
|
|
10337
10352
|
l0: 60
|
|
@@ -10344,6 +10359,7 @@ gemma-scope-2-27b-it-res:
|
|
|
10344
10359
|
- id: layer_53_width_65k_l0_medium
|
|
10345
10360
|
path: resid_post/layer_53_width_65k_l0_medium
|
|
10346
10361
|
l0: 60
|
|
10362
|
+
neuronpedia: gemma-3-27b-it/53-gemmascope-2-res-65k
|
|
10347
10363
|
- id: layer_53_width_65k_l0_small
|
|
10348
10364
|
path: resid_post/layer_53_width_65k_l0_small
|
|
10349
10365
|
l0: 20
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|