huypn16 commited on
Commit
a6632f8
1 Parent(s): dd99f2b

Upload folder using huggingface_hub

Browse files
layer4/cfg.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"model_name": "meta-llama/Llama-3.2-1B-Instruct", "model_class_name": "HookedTransformer", "hook_name": "blocks.4.hook_mlp_out", "hook_eval": "NOT_IN_USE", "hook_layer": 4, "hook_head_index": null, "dataset_path": "open-web-math/open-web-math", "dataset_trust_remote_code": true, "streaming": true, "is_dataset_tokenized": false, "context_size": 512, "use_cached_activations": false, "cached_activations_path": null, "architecture": "standard", "d_in": 2048, "d_sae": 32768, "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": 409600000, "finetuning_tokens": 0, "store_batch_size_prompts": 16, "train_batch_size_tokens": 4096, "normalize_activations": "expected_average_only_in", "device": "cuda:0", "act_store_device": "cuda:0", "seed": 42, "dtype": "bfloat16", "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": 2, "lp_norm": 1.0, "scale_sparsity_penalty_by_decoder_norm": true, "l1_warm_up_steps": 0, "lr": 5e-05, "lr_scheduler_name": "constant", "lr_warm_up_steps": 0, "lr_end": 5e-06, "lr_decay_steps": 20000, "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_lens_tutorial", "wandb_id": null, "run_name": "32768-L1-2-LR-5e-05-Tokens-4.096e+08", "wandb_entity": null, "wandb_log_frequency": 30, "eval_every_n_wandb_logs": 20, "resume": false, "n_checkpoints": 0, "checkpoint_path": "checkpoints/rmed4ics", "verbose": true, "model_kwargs": {}, "model_from_pretrained_kwargs": {"center_writing_weights": false}, "sae_lens_version": "3.23.3", "sae_lens_training_version": "3.23.3", "tokens_per_buffer": 134217728}
layer4/sae_weights.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:22f6294abd3ed0020938300919c0da30312f57d35aa1f7486c425ed33622eed3
3
+ size 268505416
layer4/sparsity.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:104871b65aee1a72eff4acbcd3c1bacd5e0376e46dfaddf85806f935cd8f26a8
3
+ size 131152