Upload SAE blocks.0.hook_resid_post

blocks.0.hook_resid_post/cfg.json

@@ -0,0 +1 @@
+{"architecture": "jumprelu", "d_in": 3584, "d_sae": 32768, "dtype": "float32", "device": "cuda", "model_name": "gemma-2-9b", "hook_name": "blocks.0.hook_resid_post", "hook_layer": 0, "hook_head_index": null, "activation_fn_str": "relu", "activation_fn_kwargs": {"k": 60}, "apply_b_dec_to_input": true, "finetuning_scaling_factor": false, "sae_lens_training_version": "5.5.1", "prepend_bos": true, "dataset_path": "chanind/pile-uncopyrighted-gemma-1024-abbrv-1B", "dataset_trust_remote_code": true, "context_size": 1024, "normalize_activations": "none", "neuronpedia_id": null, "model_from_pretrained_kwargs": {"center_writing_weights": false}, "seqpos_slice": [null], "l1_coefficient": 0.001, "aux_coefficient": 0.03125, "lp_norm": 1.0, "use_ghost_grads": false, "normalize_sae_decoder": false, "noise_scale": 0.0, "decoder_orthogonal_init": false, "init_encoder_as_decoder_transpose": true, "mse_loss_normalization": null, "decoder_heuristic_init": true, "scale_sparsity_penalty_by_decoder_norm": false, "jumprelu_init_threshold": 0.001, "jumprelu_bandwidth": 0.001, "normalize_reconstruction_losses_by_d_in": false, "skip_outer_loss": true, "matryoshka_inner_loss_multipliers": 1.0, "normalize_losses_by_num_matryoshka_steps": true, "outer_loss_multiplier": 1.0, "include_matryoshka_inner_l1_loss": false, "use_matryoshka_aux_loss": true, "matryoshka_reconstruction_loss": "MSE", "matryoshka_inner_l1_loss_multipliers": 1.0, "matryoshka_reconstruction_l2_power": 1.0, "use_delta_loss": false, "include_bdec_loss": false, "matryoshka_steps": [128, 512, 2048, 8192, 32768]}

blocks.0.hook_resid_post/sae_weights.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ac96b764dd0d347dd065cdd141f0ecd87512d1bf7b1994d54506d5f2ecb60805
+size 939800976

blocks.0.hook_resid_post/sparsity.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5a35865a6c5f64e964d96581806e4bfdcc1441ea6c98b7e103c102e9db5ec08d
+size 131152