Upload g2_no_memory_s2.json with huggingface_hub
Browse files- g2_no_memory_s2.json +34 -12
g2_no_memory_s2.json
CHANGED
|
@@ -55,29 +55,51 @@
|
|
| 55 |
"lr": 0.0003,
|
| 56 |
"eta": 1.0,
|
| 57 |
"s_per_step": 0.8031
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 58 |
}
|
| 59 |
],
|
| 60 |
"final": {
|
| 61 |
"law_rev": "candm-v1",
|
| 62 |
-
"value_acc_d32": 0.
|
| 63 |
"recall_exact_d32": 0.0,
|
| 64 |
-
"key_acc_d32": 0.
|
| 65 |
-
"value_acc_d128": 0.
|
| 66 |
"recall_exact_d128": 0.0,
|
| 67 |
-
"key_acc_d128": 0.
|
| 68 |
-
"value_acc_d512": 0.
|
| 69 |
"recall_exact_d512": 0.0,
|
| 70 |
-
"key_acc_d512": 0.
|
| 71 |
"value_acc_d1024": 0.0,
|
| 72 |
"recall_exact_d1024": 0.0,
|
| 73 |
-
"key_acc_d1024": 0.
|
| 74 |
-
"step":
|
| 75 |
-
"train_bpb":
|
| 76 |
-
"grad_norm":
|
| 77 |
"pairs": 4,
|
| 78 |
"lr": 0.0003,
|
| 79 |
-
"eta":
|
| 80 |
-
"s_per_step": 0.
|
| 81 |
}
|
| 82 |
}
|
| 83 |
],
|
|
|
|
| 55 |
"lr": 0.0003,
|
| 56 |
"eta": 1.0,
|
| 57 |
"s_per_step": 0.8031
|
| 58 |
+
},
|
| 59 |
+
{
|
| 60 |
+
"law_rev": "candm-v1",
|
| 61 |
+
"value_acc_d32": 0.0065,
|
| 62 |
+
"recall_exact_d32": 0.0,
|
| 63 |
+
"key_acc_d32": 0.0039,
|
| 64 |
+
"value_acc_d128": 0.0026,
|
| 65 |
+
"recall_exact_d128": 0.0,
|
| 66 |
+
"key_acc_d128": 0.0026,
|
| 67 |
+
"value_acc_d512": 0.0,
|
| 68 |
+
"recall_exact_d512": 0.0,
|
| 69 |
+
"key_acc_d512": 0.0078,
|
| 70 |
+
"value_acc_d1024": 0.0,
|
| 71 |
+
"recall_exact_d1024": 0.0,
|
| 72 |
+
"key_acc_d1024": 0.0026,
|
| 73 |
+
"step": 500,
|
| 74 |
+
"train_bpb": 7.2527,
|
| 75 |
+
"grad_norm": 2.579776287,
|
| 76 |
+
"pairs": 4,
|
| 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": 0.0065,
|
| 85 |
"recall_exact_d32": 0.0,
|
| 86 |
+
"key_acc_d32": 0.0039,
|
| 87 |
+
"value_acc_d128": 0.0026,
|
| 88 |
"recall_exact_d128": 0.0,
|
| 89 |
+
"key_acc_d128": 0.0026,
|
| 90 |
+
"value_acc_d512": 0.0,
|
| 91 |
"recall_exact_d512": 0.0,
|
| 92 |
+
"key_acc_d512": 0.0078,
|
| 93 |
"value_acc_d1024": 0.0,
|
| 94 |
"recall_exact_d1024": 0.0,
|
| 95 |
+
"key_acc_d1024": 0.0026,
|
| 96 |
+
"step": 500,
|
| 97 |
+
"train_bpb": 7.2527,
|
| 98 |
+
"grad_norm": 2.579776287,
|
| 99 |
"pairs": 4,
|
| 100 |
"lr": 0.0003,
|
| 101 |
+
"eta": 0.6667,
|
| 102 |
+
"s_per_step": 0.0375
|
| 103 |
}
|
| 104 |
}
|
| 105 |
],
|