Upload g2_no_memory_s0.json with huggingface_hub
Browse files- g2_no_memory_s0.json +33 -11
g2_no_memory_s0.json
CHANGED
|
@@ -121,29 +121,51 @@
|
|
| 121 |
"lr": 0.0003,
|
| 122 |
"eta": 0.0,
|
| 123 |
"s_per_step": 0.0344
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 124 |
}
|
| 125 |
],
|
| 126 |
"final": {
|
| 127 |
"law_rev": "candm-v1",
|
| 128 |
-
"value_acc_d32": 0.
|
| 129 |
"recall_exact_d32": 0.0,
|
| 130 |
-
"key_acc_d32": 0.
|
| 131 |
-
"value_acc_d128": 0.
|
| 132 |
"recall_exact_d128": 0.0,
|
| 133 |
"key_acc_d128": 0.0026,
|
| 134 |
-
"value_acc_d512": 0.
|
| 135 |
"recall_exact_d512": 0.0,
|
| 136 |
-
"key_acc_d512": 0.
|
| 137 |
-
"value_acc_d1024": 0.
|
| 138 |
"recall_exact_d1024": 0.0,
|
| 139 |
-
"key_acc_d1024": 0.
|
| 140 |
-
"step":
|
| 141 |
-
"train_bpb": 7.
|
| 142 |
-
"grad_norm":
|
| 143 |
"pairs": 4,
|
| 144 |
"lr": 0.0003,
|
| 145 |
"eta": 0.0,
|
| 146 |
-
"s_per_step": 0.
|
| 147 |
}
|
| 148 |
}
|
| 149 |
],
|
|
|
|
| 121 |
"lr": 0.0003,
|
| 122 |
"eta": 0.0,
|
| 123 |
"s_per_step": 0.0344
|
| 124 |
+
},
|
| 125 |
+
{
|
| 126 |
+
"law_rev": "candm-v1",
|
| 127 |
+
"value_acc_d32": 0.0026,
|
| 128 |
+
"recall_exact_d32": 0.0,
|
| 129 |
+
"key_acc_d32": 0.0026,
|
| 130 |
+
"value_acc_d128": 0.0078,
|
| 131 |
+
"recall_exact_d128": 0.0,
|
| 132 |
+
"key_acc_d128": 0.0026,
|
| 133 |
+
"value_acc_d512": 0.0026,
|
| 134 |
+
"recall_exact_d512": 0.0,
|
| 135 |
+
"key_acc_d512": 0.0052,
|
| 136 |
+
"value_acc_d1024": 0.0052,
|
| 137 |
+
"recall_exact_d1024": 0.0,
|
| 138 |
+
"key_acc_d1024": 0.0026,
|
| 139 |
+
"step": 2000,
|
| 140 |
+
"train_bpb": 7.229,
|
| 141 |
+
"grad_norm": 2.888474464,
|
| 142 |
+
"pairs": 4,
|
| 143 |
+
"lr": 0.0003,
|
| 144 |
+
"eta": 0.0,
|
| 145 |
+
"s_per_step": 0.0413
|
| 146 |
}
|
| 147 |
],
|
| 148 |
"final": {
|
| 149 |
"law_rev": "candm-v1",
|
| 150 |
+
"value_acc_d32": 0.0026,
|
| 151 |
"recall_exact_d32": 0.0,
|
| 152 |
+
"key_acc_d32": 0.0026,
|
| 153 |
+
"value_acc_d128": 0.0078,
|
| 154 |
"recall_exact_d128": 0.0,
|
| 155 |
"key_acc_d128": 0.0026,
|
| 156 |
+
"value_acc_d512": 0.0026,
|
| 157 |
"recall_exact_d512": 0.0,
|
| 158 |
+
"key_acc_d512": 0.0052,
|
| 159 |
+
"value_acc_d1024": 0.0052,
|
| 160 |
"recall_exact_d1024": 0.0,
|
| 161 |
+
"key_acc_d1024": 0.0026,
|
| 162 |
+
"step": 2000,
|
| 163 |
+
"train_bpb": 7.229,
|
| 164 |
+
"grad_norm": 2.888474464,
|
| 165 |
"pairs": 4,
|
| 166 |
"lr": 0.0003,
|
| 167 |
"eta": 0.0,
|
| 168 |
+
"s_per_step": 0.0413
|
| 169 |
}
|
| 170 |
}
|
| 171 |
],
|