Upload g2_scalar_s1.json with huggingface_hub
Browse files- g2_scalar_s1.json +36 -12
g2_scalar_s1.json
CHANGED
|
@@ -105,31 +105,55 @@
|
|
| 105 |
"lr": 0.0003,
|
| 106 |
"eta": 0.3333,
|
| 107 |
"s_per_step": 0.2454
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 108 |
}
|
| 109 |
],
|
| 110 |
"final": {
|
| 111 |
"law_rev": "candm-v1",
|
| 112 |
-
"value_acc_d32": 0.
|
| 113 |
"recall_exact_d32": 0.0,
|
| 114 |
"key_acc_d32": 0.2565,
|
| 115 |
-
"value_acc_d128": 0.
|
| 116 |
"recall_exact_d128": 0.0,
|
| 117 |
-
"key_acc_d128": 0.
|
| 118 |
-
"value_acc_d512": 0.
|
| 119 |
"recall_exact_d512": 0.0,
|
| 120 |
-
"key_acc_d512": 0.
|
| 121 |
-
"value_acc_d1024": 0.
|
| 122 |
"recall_exact_d1024": 0.0,
|
| 123 |
"key_acc_d1024": 0.0299,
|
| 124 |
-
"m_fro": 9.
|
| 125 |
"p_fro": 4.899,
|
| 126 |
-
"step":
|
| 127 |
-
"train_bpb": 5.
|
| 128 |
-
"grad_norm":
|
| 129 |
"pairs": 4,
|
| 130 |
"lr": 0.0003,
|
| 131 |
-
"eta": 0.
|
| 132 |
-
"s_per_step": 0.
|
| 133 |
}
|
| 134 |
}
|
| 135 |
],
|
|
|
|
| 105 |
"lr": 0.0003,
|
| 106 |
"eta": 0.3333,
|
| 107 |
"s_per_step": 0.2454
|
| 108 |
+
},
|
| 109 |
+
{
|
| 110 |
+
"law_rev": "candm-v1",
|
| 111 |
+
"value_acc_d32": 0.2422,
|
| 112 |
+
"recall_exact_d32": 0.0,
|
| 113 |
+
"key_acc_d32": 0.2565,
|
| 114 |
+
"value_acc_d128": 0.2591,
|
| 115 |
+
"recall_exact_d128": 0.0,
|
| 116 |
+
"key_acc_d128": 0.1771,
|
| 117 |
+
"value_acc_d512": 0.0234,
|
| 118 |
+
"recall_exact_d512": 0.0,
|
| 119 |
+
"key_acc_d512": 0.0768,
|
| 120 |
+
"value_acc_d1024": 0.0065,
|
| 121 |
+
"recall_exact_d1024": 0.0,
|
| 122 |
+
"key_acc_d1024": 0.0299,
|
| 123 |
+
"m_fro": 9.767,
|
| 124 |
+
"p_fro": 4.899,
|
| 125 |
+
"step": 1500,
|
| 126 |
+
"train_bpb": 5.6321,
|
| 127 |
+
"grad_norm": 6.56362009,
|
| 128 |
+
"pairs": 4,
|
| 129 |
+
"lr": 0.0003,
|
| 130 |
+
"eta": 0.0,
|
| 131 |
+
"s_per_step": 0.243
|
| 132 |
}
|
| 133 |
],
|
| 134 |
"final": {
|
| 135 |
"law_rev": "candm-v1",
|
| 136 |
+
"value_acc_d32": 0.2422,
|
| 137 |
"recall_exact_d32": 0.0,
|
| 138 |
"key_acc_d32": 0.2565,
|
| 139 |
+
"value_acc_d128": 0.2591,
|
| 140 |
"recall_exact_d128": 0.0,
|
| 141 |
+
"key_acc_d128": 0.1771,
|
| 142 |
+
"value_acc_d512": 0.0234,
|
| 143 |
"recall_exact_d512": 0.0,
|
| 144 |
+
"key_acc_d512": 0.0768,
|
| 145 |
+
"value_acc_d1024": 0.0065,
|
| 146 |
"recall_exact_d1024": 0.0,
|
| 147 |
"key_acc_d1024": 0.0299,
|
| 148 |
+
"m_fro": 9.767,
|
| 149 |
"p_fro": 4.899,
|
| 150 |
+
"step": 1500,
|
| 151 |
+
"train_bpb": 5.6321,
|
| 152 |
+
"grad_norm": 6.56362009,
|
| 153 |
"pairs": 4,
|
| 154 |
"lr": 0.0003,
|
| 155 |
+
"eta": 0.0,
|
| 156 |
+
"s_per_step": 0.243
|
| 157 |
}
|
| 158 |
}
|
| 159 |
],
|