Upload artifacts/ulv/rift_causal_qwen.json with huggingface_hub
Browse files
artifacts/ulv/rift_causal_qwen.json
ADDED
|
@@ -0,0 +1,126 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"model": "Qwen/Qwen2.5-1.5B-Instruct",
|
| 3 |
+
"baseline_acc": 8,
|
| 4 |
+
"n_hold": 9,
|
| 5 |
+
"results": {
|
| 6 |
+
"L11_a4": {
|
| 7 |
+
"layer": 11,
|
| 8 |
+
"frac": 0.4,
|
| 9 |
+
"alpha_mul": 4,
|
| 10 |
+
"flips": 1,
|
| 11 |
+
"denom": 8
|
| 12 |
+
},
|
| 13 |
+
"L11_a8": {
|
| 14 |
+
"layer": 11,
|
| 15 |
+
"frac": 0.4,
|
| 16 |
+
"alpha_mul": 8,
|
| 17 |
+
"flips": 1,
|
| 18 |
+
"denom": 8
|
| 19 |
+
},
|
| 20 |
+
"L11_a12": {
|
| 21 |
+
"layer": 11,
|
| 22 |
+
"frac": 0.4,
|
| 23 |
+
"alpha_mul": 12,
|
| 24 |
+
"flips": 3,
|
| 25 |
+
"denom": 8
|
| 26 |
+
},
|
| 27 |
+
"L11_a16": {
|
| 28 |
+
"layer": 11,
|
| 29 |
+
"frac": 0.4,
|
| 30 |
+
"alpha_mul": 16,
|
| 31 |
+
"flips": 5,
|
| 32 |
+
"denom": 8
|
| 33 |
+
},
|
| 34 |
+
"L14_a4": {
|
| 35 |
+
"layer": 14,
|
| 36 |
+
"frac": 0.5,
|
| 37 |
+
"alpha_mul": 4,
|
| 38 |
+
"flips": 1,
|
| 39 |
+
"denom": 8
|
| 40 |
+
},
|
| 41 |
+
"L14_a8": {
|
| 42 |
+
"layer": 14,
|
| 43 |
+
"frac": 0.5,
|
| 44 |
+
"alpha_mul": 8,
|
| 45 |
+
"flips": 1,
|
| 46 |
+
"denom": 8
|
| 47 |
+
},
|
| 48 |
+
"L14_a12": {
|
| 49 |
+
"layer": 14,
|
| 50 |
+
"frac": 0.5,
|
| 51 |
+
"alpha_mul": 12,
|
| 52 |
+
"flips": 3,
|
| 53 |
+
"denom": 8
|
| 54 |
+
},
|
| 55 |
+
"L14_a16": {
|
| 56 |
+
"layer": 14,
|
| 57 |
+
"frac": 0.5,
|
| 58 |
+
"alpha_mul": 16,
|
| 59 |
+
"flips": 8,
|
| 60 |
+
"denom": 8
|
| 61 |
+
},
|
| 62 |
+
"L17_a4": {
|
| 63 |
+
"layer": 17,
|
| 64 |
+
"frac": 0.6,
|
| 65 |
+
"alpha_mul": 4,
|
| 66 |
+
"flips": 0,
|
| 67 |
+
"denom": 8
|
| 68 |
+
},
|
| 69 |
+
"L17_a8": {
|
| 70 |
+
"layer": 17,
|
| 71 |
+
"frac": 0.6,
|
| 72 |
+
"alpha_mul": 8,
|
| 73 |
+
"flips": 2,
|
| 74 |
+
"denom": 8
|
| 75 |
+
},
|
| 76 |
+
"L17_a12": {
|
| 77 |
+
"layer": 17,
|
| 78 |
+
"frac": 0.6,
|
| 79 |
+
"alpha_mul": 12,
|
| 80 |
+
"flips": 8,
|
| 81 |
+
"denom": 8
|
| 82 |
+
},
|
| 83 |
+
"L17_a16": {
|
| 84 |
+
"layer": 17,
|
| 85 |
+
"frac": 0.6,
|
| 86 |
+
"alpha_mul": 16,
|
| 87 |
+
"flips": 8,
|
| 88 |
+
"denom": 8
|
| 89 |
+
},
|
| 90 |
+
"L20_a4": {
|
| 91 |
+
"layer": 20,
|
| 92 |
+
"frac": 0.7,
|
| 93 |
+
"alpha_mul": 4,
|
| 94 |
+
"flips": 2,
|
| 95 |
+
"denom": 8
|
| 96 |
+
},
|
| 97 |
+
"L20_a8": {
|
| 98 |
+
"layer": 20,
|
| 99 |
+
"frac": 0.7,
|
| 100 |
+
"alpha_mul": 8,
|
| 101 |
+
"flips": 1,
|
| 102 |
+
"denom": 8
|
| 103 |
+
},
|
| 104 |
+
"L20_a12": {
|
| 105 |
+
"layer": 20,
|
| 106 |
+
"frac": 0.7,
|
| 107 |
+
"alpha_mul": 12,
|
| 108 |
+
"flips": 7,
|
| 109 |
+
"denom": 8
|
| 110 |
+
},
|
| 111 |
+
"L20_a16": {
|
| 112 |
+
"layer": 20,
|
| 113 |
+
"frac": 0.7,
|
| 114 |
+
"alpha_mul": 16,
|
| 115 |
+
"flips": 8,
|
| 116 |
+
"denom": 8
|
| 117 |
+
}
|
| 118 |
+
},
|
| 119 |
+
"best": {
|
| 120 |
+
"layer": 14,
|
| 121 |
+
"frac": 0.5,
|
| 122 |
+
"alpha_mul": 16,
|
| 123 |
+
"flips": 8,
|
| 124 |
+
"denom": 8
|
| 125 |
+
}
|
| 126 |
+
}
|