Upload 4 files

blocks.6.hook_normalized_resid_post/activations_store_state.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:279ba412a5a8142ac12a19a3dc9959815ab9ed565049c1febe95a58fd3129f87
+size 50331824

blocks.6.hook_normalized_resid_post/cfg.json

@@ -0,0 +1 @@
+{"model_name": "gelu-2l", "model_class_name": "HookedTransformer", "hook_name": "blocks.6.hook_normalized_resid_post", "hook_eval": "NOT_IN_USE", "hook_layer": 6, "hook_head_index": null, "dataset_path": "legacy-datasets/wikipedia", "dataset_trust_remote_code": true, "streaming": true, "is_dataset_tokenized": false, "context_size": 128, "use_cached_activations": false, "cached_activations_path": null, "architecture": "standard", "d_in": 768, "d_sae": 12288, "b_dec_init_method": "geometric_median", "expansion_factor": 16, "activation_fn": "relu", "activation_fn_kwargs": {}, "normalize_sae_decoder": true, "noise_scale": 0.0, "from_pretrained_path": null, "apply_b_dec_to_input": false, "decoder_orthogonal_init": false, "decoder_heuristic_init": false, "init_encoder_as_decoder_transpose": false, "n_batches_in_buffer": 32, "training_tokens": 204800000, "finetuning_tokens": 50000, "store_batch_size_prompts": 8, "train_batch_size_tokens": 256, "normalize_activations": "none", "seqpos_slice": [null], "device": "cuda", "act_store_device": "cuda", "seed": 42, "dtype": "float32", "prepend_bos": false, "jumprelu_init_threshold": 0.001, "jumprelu_bandwidth": 0.001, "autocast": false, "autocast_lm": false, "compile_llm": false, "llm_compilation_mode": null, "compile_sae": false, "sae_compilation_mode": null, "adam_beta1": 0, "adam_beta2": 0.999, "mse_loss_normalization": null, "l1_coefficient": 10.0, "lp_norm": 1, "scale_sparsity_penalty_by_decoder_norm": false, "l1_warm_up_steps": 0, "lr": 5e-07, "lr_scheduler_name": "constant", "lr_warm_up_steps": 5000, "lr_end": 5e-08, "lr_decay_steps": 0, "n_restart_cycles": 1, "finetuning_method": "decoder", "use_ghost_grads": false, "feature_sampling_window": 2500, "dead_feature_window": 5000, "dead_feature_threshold": 1e-08, "n_eval_batches": 10, "eval_batch_size_prompts": null, "log_to_wandb": true, "log_activations_store_to_wandb": false, "log_optimizer_state_to_wandb": false, "wandb_project": "bertTest5", "wandb_id": null, "run_name": "12288-L1-10.0-LR-5e-07-Tokens-2.048e+08", "wandb_entity": null, "wandb_log_frequency": 100, "eval_every_n_wandb_logs": 100, "resume": false, "n_checkpoints": 10, "checkpoint_path": "checkpoints/dn0lhvoy", "verbose": true, "model_kwargs": {}, "model_from_pretrained_kwargs": {"center_writing_weights": false}, "sae_lens_version": "5.3.1", "sae_lens_training_version": "5.3.1", "tokens_per_buffer": 1048576}

blocks.6.hook_normalized_resid_post/sae_weights.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:502cbbe70e5b9fdf338bf6c6babc978fd0d3bc4d06d8888d428e7ce54f75f4ac
+size 75601456

blocks.6.hook_normalized_resid_post/sparsity.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e0bb1791e10b0f3bf918fd3a4c8d9dc91b3b47a157c3eb3e596b44dd9d777aeb
+size 99107