Upload results.json with huggingface_hub
Browse files- results.json +94 -0
results.json
ADDED
|
@@ -0,0 +1,94 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"d_sae": 4096,
|
| 3 |
+
"actual_l0": 165.8764343,
|
| 4 |
+
"effective_l0": 100,
|
| 5 |
+
"cross_entropy": 464.4294169433594,
|
| 6 |
+
"sae_type": "saebench",
|
| 7 |
+
"sae_release": "canrager/saebench_gemma-2-2b_width-2pow12_date-0107",
|
| 8 |
+
"num_tokens": 10000000,
|
| 9 |
+
"num_epochs": 1,
|
| 10 |
+
"lr": 0.0005,
|
| 11 |
+
"loaded_dag": null,
|
| 12 |
+
"allow_negative_influence": false,
|
| 13 |
+
"history": {
|
| 14 |
+
"train_loss": [
|
| 15 |
+
756.2381443568638,
|
| 16 |
+
571.8783499581473,
|
| 17 |
+
527.0960534232004,
|
| 18 |
+
504.57795976911274,
|
| 19 |
+
491.4717749459403,
|
| 20 |
+
482.7741740635463,
|
| 21 |
+
477.00267617361885,
|
| 22 |
+
472.19471849714006,
|
| 23 |
+
468.3589043753488,
|
| 24 |
+
464.8287867954799,
|
| 25 |
+
464.81527361188614
|
| 26 |
+
],
|
| 27 |
+
"train_steps": [
|
| 28 |
+
140,
|
| 29 |
+
280,
|
| 30 |
+
420,
|
| 31 |
+
560,
|
| 32 |
+
700,
|
| 33 |
+
840,
|
| 34 |
+
980,
|
| 35 |
+
1120,
|
| 36 |
+
1260,
|
| 37 |
+
1400,
|
| 38 |
+
1407
|
| 39 |
+
],
|
| 40 |
+
"val_loss": [
|
| 41 |
+
613.0383723632813,
|
| 42 |
+
543.270251953125,
|
| 43 |
+
513.7303801757813,
|
| 44 |
+
497.2721671386719,
|
| 45 |
+
486.76615244140623,
|
| 46 |
+
479.4650443359375,
|
| 47 |
+
474.1357983886719,
|
| 48 |
+
470.0854748535156,
|
| 49 |
+
466.90237919921873,
|
| 50 |
+
464.3813434570313,
|
| 51 |
+
464.4294169433594
|
| 52 |
+
],
|
| 53 |
+
"val_cross_entropy": [
|
| 54 |
+
613.0383723632813,
|
| 55 |
+
543.270251953125,
|
| 56 |
+
513.7303801757813,
|
| 57 |
+
497.2721671386719,
|
| 58 |
+
486.76615244140623,
|
| 59 |
+
479.4650443359375,
|
| 60 |
+
474.1357983886719,
|
| 61 |
+
470.0854748535156,
|
| 62 |
+
466.90237919921873,
|
| 63 |
+
464.3813434570313,
|
| 64 |
+
464.4294169433594
|
| 65 |
+
],
|
| 66 |
+
"effective_l0": [
|
| 67 |
+
142,
|
| 68 |
+
121,
|
| 69 |
+
113,
|
| 70 |
+
108,
|
| 71 |
+
106,
|
| 72 |
+
104,
|
| 73 |
+
102,
|
| 74 |
+
101,
|
| 75 |
+
100,
|
| 76 |
+
100,
|
| 77 |
+
100
|
| 78 |
+
],
|
| 79 |
+
"val_steps": [
|
| 80 |
+
140,
|
| 81 |
+
280,
|
| 82 |
+
420,
|
| 83 |
+
560,
|
| 84 |
+
700,
|
| 85 |
+
840,
|
| 86 |
+
980,
|
| 87 |
+
1120,
|
| 88 |
+
1260,
|
| 89 |
+
1400,
|
| 90 |
+
1407
|
| 91 |
+
]
|
| 92 |
+
},
|
| 93 |
+
"sae_id": "gemma-2-2b_jump_relu_width-2pow12_date-0107/resid_post_layer_12/trainer_3"
|
| 94 |
+
}
|