Upload g2_no_memory_s2.json with huggingface_hub
Browse files- g2_no_memory_s2.json +33 -11
g2_no_memory_s2.json
CHANGED
|
@@ -77,29 +77,51 @@
|
|
| 77 |
"lr": 0.0003,
|
| 78 |
"eta": 0.6667,
|
| 79 |
"s_per_step": 0.0375
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 80 |
}
|
| 81 |
],
|
| 82 |
"final": {
|
| 83 |
"law_rev": "candm-v1",
|
| 84 |
-
"value_acc_d32": 0.
|
| 85 |
"recall_exact_d32": 0.0,
|
| 86 |
-
"key_acc_d32": 0.
|
| 87 |
-
"value_acc_d128": 0.
|
| 88 |
"recall_exact_d128": 0.0,
|
| 89 |
"key_acc_d128": 0.0026,
|
| 90 |
-
"value_acc_d512": 0.
|
| 91 |
"recall_exact_d512": 0.0,
|
| 92 |
-
"key_acc_d512": 0.
|
| 93 |
"value_acc_d1024": 0.0,
|
| 94 |
"recall_exact_d1024": 0.0,
|
| 95 |
-
"key_acc_d1024": 0.
|
| 96 |
-
"step":
|
| 97 |
-
"train_bpb": 7.
|
| 98 |
-
"grad_norm":
|
| 99 |
"pairs": 4,
|
| 100 |
"lr": 0.0003,
|
| 101 |
-
"eta": 0.
|
| 102 |
-
"s_per_step": 0.
|
| 103 |
}
|
| 104 |
}
|
| 105 |
],
|
|
|
|
| 77 |
"lr": 0.0003,
|
| 78 |
"eta": 0.6667,
|
| 79 |
"s_per_step": 0.0375
|
| 80 |
+
},
|
| 81 |
+
{
|
| 82 |
+
"law_rev": "candm-v1",
|
| 83 |
+
"value_acc_d32": 0.0026,
|
| 84 |
+
"recall_exact_d32": 0.0,
|
| 85 |
+
"key_acc_d32": 0.0052,
|
| 86 |
+
"value_acc_d128": 0.0013,
|
| 87 |
+
"recall_exact_d128": 0.0,
|
| 88 |
+
"key_acc_d128": 0.0026,
|
| 89 |
+
"value_acc_d512": 0.0013,
|
| 90 |
+
"recall_exact_d512": 0.0,
|
| 91 |
+
"key_acc_d512": 0.0039,
|
| 92 |
+
"value_acc_d1024": 0.0,
|
| 93 |
+
"recall_exact_d1024": 0.0,
|
| 94 |
+
"key_acc_d1024": 0.0013,
|
| 95 |
+
"step": 1000,
|
| 96 |
+
"train_bpb": 7.2327,
|
| 97 |
+
"grad_norm": 6.870772362,
|
| 98 |
+
"pairs": 4,
|
| 99 |
+
"lr": 0.0003,
|
| 100 |
+
"eta": 0.3333,
|
| 101 |
+
"s_per_step": 0.0371
|
| 102 |
}
|
| 103 |
],
|
| 104 |
"final": {
|
| 105 |
"law_rev": "candm-v1",
|
| 106 |
+
"value_acc_d32": 0.0026,
|
| 107 |
"recall_exact_d32": 0.0,
|
| 108 |
+
"key_acc_d32": 0.0052,
|
| 109 |
+
"value_acc_d128": 0.0013,
|
| 110 |
"recall_exact_d128": 0.0,
|
| 111 |
"key_acc_d128": 0.0026,
|
| 112 |
+
"value_acc_d512": 0.0013,
|
| 113 |
"recall_exact_d512": 0.0,
|
| 114 |
+
"key_acc_d512": 0.0039,
|
| 115 |
"value_acc_d1024": 0.0,
|
| 116 |
"recall_exact_d1024": 0.0,
|
| 117 |
+
"key_acc_d1024": 0.0013,
|
| 118 |
+
"step": 1000,
|
| 119 |
+
"train_bpb": 7.2327,
|
| 120 |
+
"grad_norm": 6.870772362,
|
| 121 |
"pairs": 4,
|
| 122 |
"lr": 0.0003,
|
| 123 |
+
"eta": 0.3333,
|
| 124 |
+
"s_per_step": 0.0371
|
| 125 |
}
|
| 126 |
}
|
| 127 |
],
|