Upload SAE blocks.9.hook_attn_out

blocks.9.hook_attn_out/cfg.json

@@ -0,0 +1 @@
+{"model_name": "gpt2-small", "model_class_name": "HookedTransformer", "hook_name": "blocks.9.hook_attn_out", "hook_eval": "NOT_IN_USE", "hook_layer": 9, "hook_head_index": null, "dataset_path": "apollo-research/roneneldan-TinyStories-tokenizer-gpt2", "dataset_trust_remote_code": true, "streaming": true, "is_dataset_tokenized": true, "context_size": 512, "use_cached_activations": false, "cached_activations_path": null, "significant_neurons": "top_k_significant_neurons.npy", "architecture": "standard", "d_in": 602, "d_sae": 12288, "b_dec_init_method": "zeros", "expansion_factor": 16, "activation_fn": "relu", "activation_fn_kwargs": {}, "normalize_sae_decoder": false, "noise_scale": 0.0, "from_pretrained_path": null, "apply_b_dec_to_input": false, "decoder_orthogonal_init": false, "decoder_heuristic_init": true, "init_encoder_as_decoder_transpose": true, "n_batches_in_buffer": 64, "training_tokens": 122880000, "finetuning_tokens": 0, "store_batch_size_prompts": 16, "train_batch_size_tokens": 4096, "normalize_activations": "expected_average_only_in", "seqpos_slice": [null], "device": "mps", "act_store_device": "mps", "seed": 42, "dtype": "float32", "prepend_bos": true, "autocast": false, "autocast_lm": false, "compile_llm": false, "llm_compilation_mode": null, "compile_sae": false, "sae_compilation_mode": null, "adam_beta1": 0.9, "adam_beta2": 0.999, "mse_loss_normalization": null, "l1_coefficient": 5, "lp_norm": 1.0, "scale_sparsity_penalty_by_decoder_norm": true, "l1_warm_up_steps": 1500, "lr": 5e-05, "lr_scheduler_name": "constant", "lr_warm_up_steps": 0, "lr_end": 5e-06, "lr_decay_steps": 6000, "n_restart_cycles": 1, "finetuning_method": null, "use_ghost_grads": false, "feature_sampling_window": 1000, "dead_feature_window": 1000, "dead_feature_threshold": 0.0001, "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": "sae_train_602_features", "wandb_id": null, "run_name": "12288-L1-5-LR-5e-05-Tokens-1.229e+08", "wandb_entity": null, "wandb_log_frequency": 30, "eval_every_n_wandb_logs": 20, "resume": false, "n_checkpoints": 1, "checkpoint_path": "checkpoints/p0j7959m", "verbose": true, "model_kwargs": {}, "model_from_pretrained_kwargs": {"center_writing_weights": false}, "sae_lens_version": "4.0.9", "sae_lens_training_version": "4.0.9", "tokens_per_buffer": 134217728}

blocks.9.hook_attn_out/sae_weights.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:11da90ded4c041b56a567d56f3171f1a9551043f999ea2929a6c94f1d12613d9
+size 59230880

blocks.9.hook_attn_out/sparsity.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:002aa2310575438d04e12bd8edd3417d8d47074ddda10ddb6bb052f6f917a962
+size 49232