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": 337.2315061,
|
| 4 |
+
"effective_l0": 216,
|
| 5 |
+
"cross_entropy": 845.2914995117187,
|
| 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 |
+
1304.8249581473215,
|
| 16 |
+
979.5266653878348,
|
| 17 |
+
918.5828390938896,
|
| 18 |
+
891.2496141706195,
|
| 19 |
+
875.4288151332311,
|
| 20 |
+
865.2919886997768,
|
| 21 |
+
858.7045811244419,
|
| 22 |
+
852.3984619140625,
|
| 23 |
+
849.0059517996651,
|
| 24 |
+
846.5715981619699,
|
| 25 |
+
846.4956909179688
|
| 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 |
+
1037.0020205078124,
|
| 42 |
+
939.7755484375,
|
| 43 |
+
901.7446291015625,
|
| 44 |
+
881.6231953125,
|
| 45 |
+
869.3407673828125,
|
| 46 |
+
861.0552083984375,
|
| 47 |
+
855.195570703125,
|
| 48 |
+
850.843380859375,
|
| 49 |
+
847.5050969726562,
|
| 50 |
+
844.8984658203125,
|
| 51 |
+
845.2914995117187
|
| 52 |
+
],
|
| 53 |
+
"val_cross_entropy": [
|
| 54 |
+
1037.0020205078124,
|
| 55 |
+
939.7755484375,
|
| 56 |
+
901.7446291015625,
|
| 57 |
+
881.6231953125,
|
| 58 |
+
869.3407673828125,
|
| 59 |
+
861.0552083984375,
|
| 60 |
+
855.195570703125,
|
| 61 |
+
850.843380859375,
|
| 62 |
+
847.5050969726562,
|
| 63 |
+
844.8984658203125,
|
| 64 |
+
845.2914995117187
|
| 65 |
+
],
|
| 66 |
+
"effective_l0": [
|
| 67 |
+
284,
|
| 68 |
+
248,
|
| 69 |
+
235,
|
| 70 |
+
228,
|
| 71 |
+
224,
|
| 72 |
+
221,
|
| 73 |
+
219,
|
| 74 |
+
217,
|
| 75 |
+
216,
|
| 76 |
+
215,
|
| 77 |
+
216
|
| 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_4"
|
| 94 |
+
}
|