Hugging Face's logo

Datasets:
juiceb0xc0de
/
gemma-4-e2b-it-SAE

Tasks:
Feature Extraction
Size:
10B<n<100B
Tags:
sparse-autoencoder
mechanistic-interpretability
sae
gemma
brain-atlas
jumprelu
License:
Dataset card Files
xet
 Community
1
juiceb0xc0de commited on
Commit
eae143d
·
verified ·
1 Parent(s): 0f94f9b

Copy SAE layer 18 from model repo

Browse files
Files changed (3) hide show
  1. layer_18_s0/checkpoint_full.pt +3 -0
  2. layer_18_s0/meta.json +135 -0
  3. layer_18_s0/sae.pt +3 -0
layer_18_s0/checkpoint_full.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c0a51db6bc4378291b777bd811738b3b41ba61d51dca7e7442358b1cf5b7e974
3
+ size 1813939659
layer_18_s0/meta.json ADDED
@@ -0,0 +1,135 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "layer": 18,
3
+ "seed": 0,
4
+ "model_id": "google/gemma-4-E2B-it",
5
+ "d_in": 1536,
6
+ "n_features": 49152,
7
+ "k": 50,
8
+ "batch_tokens": 32768,
9
+ "n_steps": 2751,
10
+ "lr": 0.0002,
11
+ "lambda_l0": 0.0029881176667452955,
12
+ "tier": "mid",
13
+ "preflight": {
14
+ "activation_norm_probe": 1.8016763925552368,
15
+ "activation_norm_ref": 0.7526426911354065,
16
+ "initial_l0_probe": 73.6717529296875,
17
+ "initial_lr_multiplier": 1.6243674405029294,
18
+ "early_pulse_multiplier": 1.1178748046875,
19
+ "layer_lr_multiplier": 1.17
20
+ },
21
+ "scheduler_mode": "BASELINE",
22
+ "scheduler_phase": "PIN",
23
+ "scheduler_phase_summary": {
24
+ "phase": "PIN",
25
+ "phase_step": 0,
26
+ "pin_entry_step": 2750,
27
+ "pinned_lambda": 0.0029881176667452955,
28
+ "pin_ev_count": 0,
29
+ "pin_retry_count": 0
30
+ },
31
+ "scheduler_transitions": 0,
32
+ "total_tokens": 90144768,
33
+ "early_stopped": true,
34
+ "path": "rolling",
35
+ "best_ev": 0.8751982779826244,
36
+ "best_ev_step": 2000,
37
+ "final_metrics": {
38
+ "recon_loss": 0.416984,
39
+ "mean_l0": 50.14,
40
+ "dead_pct": 0.01,
41
+ "resampled": 0,
42
+ "ev": 0.874214378594217,
43
+ "nonlinear_err": null,
44
+ "linear_err": null,
45
+ "qo_gap": 0.6595,
46
+ "activation_norm_probe": 1.8016763925552368,
47
+ "activation_norm_ref": 0.7526426911354065,
48
+ "initial_l0_probe": 73.6717529296875,
49
+ "initial_lr_multiplier": 1.6243674405029294,
50
+ "early_pulse_multiplier": 1.1178748046875,
51
+ "layer_lr_multiplier": 1.17
52
+ },
53
+ "training_curve": {
54
+ "recon_loss": [
55
+ 0.606999,
56
+ 0.514466,
57
+ 0.495188,
58
+ 0.489768,
59
+ 0.447178,
60
+ 0.414076,
61
+ 0.413669,
62
+ 0.42892,
63
+ 0.412977,
64
+ 0.416992,
65
+ 0.416984
66
+ ],
67
+ "mean_l0": [
68
+ 27.4,
69
+ 29.08,
70
+ 28.68,
71
+ 32.82,
72
+ 40.73,
73
+ 48.08,
74
+ 52.69,
75
+ 51.72,
76
+ 49.12,
77
+ 49.13,
78
+ 50.14
79
+ ],
80
+ "dead_pct": [
81
+ 0.0,
82
+ 0.0,
83
+ 0.0,
84
+ 0.0,
85
+ 0.0,
86
+ 0.0,
87
+ 0.0,
88
+ 0.0,
89
+ 0.0,
90
+ 0.0,
91
+ 0.01
92
+ ],
93
+ "resampled": [
94
+ 0,
95
+ 0,
96
+ 0,
97
+ 0,
98
+ 0,
99
+ 0,
100
+ 0,
101
+ 0,
102
+ 0,
103
+ 0,
104
+ 0
105
+ ],
106
+ "ev": [
107
+ 0.817935520268152,
108
+ 0.8447622221961634,
109
+ 0.8511420835292034,
110
+ 0.854184440497951,
111
+ 0.8648727764086059,
112
+ 0.8733059813011164,
113
+ 0.8753421783916479,
114
+ 0.8732944800476897,
115
+ 0.8751017683299624,
116
+ 0.8761718937231922,
117
+ 0.874214378594217
118
+ ],
119
+ "nonlinear_err": [],
120
+ "linear_err": [],
121
+ "qo_gap": [
122
+ 0.7971,
123
+ 0.7555,
124
+ 0.7437,
125
+ 0.7197,
126
+ 0.6958,
127
+ 0.6717,
128
+ 0.6613,
129
+ 0.659,
130
+ 0.6646,
131
+ 0.6729,
132
+ 0.6595
133
+ ]
134
+ }
135
+ }
layer_18_s0/sae.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e688bcb4a92cd4fc86b0725441d71046174af7bf0ae3d771339dc3d612b9e439
3
+ size 604381549