Upload folder using huggingface_hub

config.json

@@ -0,0 +1 @@
+{"sae": {"expansion_factor": 16, "normalize_decoder": true, "num_latents": 0, "k": -1, "multi_topk": false, "jumprelu": true, "jumprelu_init_threshold": 0.001, "jumprelu_bandwidth": 0.001, "jumprelu_target_l0": null, "init_enc_as_dec_transpose": true}, "batch_size": 4, "max_seq_len": 1024, "num_training_tokens": 1000000000, "cycle_iterator": true, "grad_acc_steps": 1, "micro_acc_steps": 1, "adam_8bit": false, "adam_epsilon": 1e-08, "adam_betas": [0.0, 0.999], "lr": 0.0007, "lr_init": 7e-05, "lr_end": 7e-05, "lr_scheduler_name": "constant", "lr_warmup_steps": 0.01, "lr_decay_steps": 0.2, "l1_coefficient": 0.0003, "l1_warmup_steps": 0.05, "use_l2_loss": true, "auxk_alpha": 0.0, "dead_feature_threshold": 10000000, "hookpoints": ["layers.0", "layers.1", "layers.2", "layers.3", "layers.4", "layers.5", "layers.6", "layers.7", "layers.8", "layers.9", "layers.10"], "layers": [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10], "layer_stride": 1, "distribute_modules": false, "save_every": 50000, "normalize_activations": 1, "num_norm_estimation_tokens": 2000000, "clusters": null, "cluster_hookpoints": null, "hook": null, "log_to_wandb": true, "run_name": "checkpoints/pythia-160m-deduped-1024-lambda-0.0003-target-L0-None-lr-0.0007", "wandb_log_frequency": 1}

l1_scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9e9c3a305a515dd145375b00e49354ea6fd6775d0473c8228467fd9c0e0cd898
+size 1012

layers.0/cfg.json

@@ -0,0 +1 @@
+{"expansion_factor": 16, "normalize_decoder": true, "num_latents": 0, "k": -1, "multi_topk": false, "jumprelu": true, "jumprelu_init_threshold": 0.001, "jumprelu_bandwidth": 0.001, "jumprelu_target_l0": null, "init_enc_as_dec_transpose": true, "d_in": 768}

layers.0/sae.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e46d1ca278cea478d9fab02b68d0ad5b0c1d34faa294b4f27961ca8beec627d2
+size 75599256

layers.1/cfg.json

@@ -0,0 +1 @@
+{"expansion_factor": 16, "normalize_decoder": true, "num_latents": 0, "k": -1, "multi_topk": false, "jumprelu": true, "jumprelu_init_threshold": 0.001, "jumprelu_bandwidth": 0.001, "jumprelu_target_l0": null, "init_enc_as_dec_transpose": true, "d_in": 768}

layers.1/sae.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f51058a5031e22fa71aed5e2730682ced8e6f06deb4aea37562c6de6956820c3
+size 75599256

layers.10/cfg.json

@@ -0,0 +1 @@
+{"expansion_factor": 16, "normalize_decoder": true, "num_latents": 0, "k": -1, "multi_topk": false, "jumprelu": true, "jumprelu_init_threshold": 0.001, "jumprelu_bandwidth": 0.001, "jumprelu_target_l0": null, "init_enc_as_dec_transpose": true, "d_in": 768}

layers.10/sae.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:243605462a4247eb19e563c9754ea6e220bc2b0c026a6ce21b881e10579ee816
+size 75599256

layers.2/cfg.json

@@ -0,0 +1 @@
+{"expansion_factor": 16, "normalize_decoder": true, "num_latents": 0, "k": -1, "multi_topk": false, "jumprelu": true, "jumprelu_init_threshold": 0.001, "jumprelu_bandwidth": 0.001, "jumprelu_target_l0": null, "init_enc_as_dec_transpose": true, "d_in": 768}

layers.2/sae.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b325beed364daf66ea3399164eefc8919069141302b52047f22e39f9c29b7e35
+size 75599256

layers.3/cfg.json

@@ -0,0 +1 @@
+{"expansion_factor": 16, "normalize_decoder": true, "num_latents": 0, "k": -1, "multi_topk": false, "jumprelu": true, "jumprelu_init_threshold": 0.001, "jumprelu_bandwidth": 0.001, "jumprelu_target_l0": null, "init_enc_as_dec_transpose": true, "d_in": 768}

layers.3/sae.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:48f72bf6ed9b7ceb45feec4e8622a1a0060dc3c8888992a8f374972cff17f92f
+size 75599256

layers.4/cfg.json

@@ -0,0 +1 @@
+{"expansion_factor": 16, "normalize_decoder": true, "num_latents": 0, "k": -1, "multi_topk": false, "jumprelu": true, "jumprelu_init_threshold": 0.001, "jumprelu_bandwidth": 0.001, "jumprelu_target_l0": null, "init_enc_as_dec_transpose": true, "d_in": 768}

layers.4/sae.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0aec6d2cadcfa34036497335cdfde2d42d76aea618532053e2fb64e54944749b
+size 75599256

layers.5/cfg.json

@@ -0,0 +1 @@
+{"expansion_factor": 16, "normalize_decoder": true, "num_latents": 0, "k": -1, "multi_topk": false, "jumprelu": true, "jumprelu_init_threshold": 0.001, "jumprelu_bandwidth": 0.001, "jumprelu_target_l0": null, "init_enc_as_dec_transpose": true, "d_in": 768}

layers.5/sae.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:24f8f616e1fd8e481987af931d07cbcac7ffe6a4b87bced53136dac0ce35fa24
+size 75599256

layers.6/cfg.json

@@ -0,0 +1 @@
+{"expansion_factor": 16, "normalize_decoder": true, "num_latents": 0, "k": -1, "multi_topk": false, "jumprelu": true, "jumprelu_init_threshold": 0.001, "jumprelu_bandwidth": 0.001, "jumprelu_target_l0": null, "init_enc_as_dec_transpose": true, "d_in": 768}

layers.6/sae.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:96fc2f89fd4ac8f57406454219922fb269cd97e8da259dd9200e830c4a3e3463
+size 75599256

layers.7/cfg.json

@@ -0,0 +1 @@
+{"expansion_factor": 16, "normalize_decoder": true, "num_latents": 0, "k": -1, "multi_topk": false, "jumprelu": true, "jumprelu_init_threshold": 0.001, "jumprelu_bandwidth": 0.001, "jumprelu_target_l0": null, "init_enc_as_dec_transpose": true, "d_in": 768}

layers.7/sae.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bb5410343745c6b0b6abaf922b50804300acd7c876e9a104dc4c2f285ee606e1
+size 75599256

layers.8/cfg.json

@@ -0,0 +1 @@
+{"expansion_factor": 16, "normalize_decoder": true, "num_latents": 0, "k": -1, "multi_topk": false, "jumprelu": true, "jumprelu_init_threshold": 0.001, "jumprelu_bandwidth": 0.001, "jumprelu_target_l0": null, "init_enc_as_dec_transpose": true, "d_in": 768}

layers.8/sae.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f7195534205189497900b538b78703987bb07276b2895cbbe2d392f30891d450
+size 75599256

layers.9/cfg.json

@@ -0,0 +1 @@
+{"expansion_factor": 16, "normalize_decoder": true, "num_latents": 0, "k": -1, "multi_topk": false, "jumprelu": true, "jumprelu_init_threshold": 0.001, "jumprelu_bandwidth": 0.001, "jumprelu_target_l0": null, "init_enc_as_dec_transpose": true, "d_in": 768}

layers.9/sae.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0bdd961400fe6bfba9240b2ebb256252fa06e975adabb4eaebbb6636206d69ea
+size 75599256

lr_scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2a6dd13de1a28341679fc9a821abb7239b69577501ee7890f363464e17475bf1
+size 1844

optimizer.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e8c9976e55b03ec509bb3aef066f0898e6962522d323ad4b725e38bc596f99c6
+size 1663222074

scaling_factors.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1e3db8777a9549d8a63009b1cccd97ca981a4c8e9142d95c788587e11810a01c
+size 1152

state.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:081c0b8822f8375322740d0b5e5bb09fd654eeac0bd8ad9d366ddb82292769f3
+size 1084503