Upload g2_no_memory_s1.json with huggingface_hub
Browse files- g2_no_memory_s1.json +34 -12
g2_no_memory_s1.json
CHANGED
|
@@ -143,29 +143,51 @@
|
|
| 143 |
"lr": 0.0003,
|
| 144 |
"eta": 0.0,
|
| 145 |
"s_per_step": 0.0373
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 146 |
}
|
| 147 |
],
|
| 148 |
"final": {
|
| 149 |
"law_rev": "candm-v1",
|
| 150 |
-
"value_acc_d32": 0.
|
| 151 |
"recall_exact_d32": 0.0,
|
| 152 |
-
"key_acc_d32": 0.
|
| 153 |
-
"value_acc_d128": 0.
|
| 154 |
"recall_exact_d128": 0.0,
|
| 155 |
-
"key_acc_d128": 0.
|
| 156 |
-
"value_acc_d512": 0.
|
| 157 |
"recall_exact_d512": 0.0,
|
| 158 |
-
"key_acc_d512": 0.
|
| 159 |
-
"value_acc_d1024": 0.
|
| 160 |
"recall_exact_d1024": 0.0,
|
| 161 |
-
"key_acc_d1024": 0.
|
| 162 |
-
"step":
|
| 163 |
-
"train_bpb": 7.
|
| 164 |
-
"grad_norm":
|
| 165 |
"pairs": 4,
|
| 166 |
"lr": 0.0003,
|
| 167 |
"eta": 0.0,
|
| 168 |
-
"s_per_step": 0.
|
| 169 |
}
|
| 170 |
}
|
| 171 |
],
|
|
|
|
| 143 |
"lr": 0.0003,
|
| 144 |
"eta": 0.0,
|
| 145 |
"s_per_step": 0.0373
|
| 146 |
+
},
|
| 147 |
+
{
|
| 148 |
+
"law_rev": "candm-v1",
|
| 149 |
+
"value_acc_d32": 0.0078,
|
| 150 |
+
"recall_exact_d32": 0.0,
|
| 151 |
+
"key_acc_d32": 0.0052,
|
| 152 |
+
"value_acc_d128": 0.0065,
|
| 153 |
+
"recall_exact_d128": 0.0,
|
| 154 |
+
"key_acc_d128": 0.0078,
|
| 155 |
+
"value_acc_d512": 0.0078,
|
| 156 |
+
"recall_exact_d512": 0.0,
|
| 157 |
+
"key_acc_d512": 0.0013,
|
| 158 |
+
"value_acc_d1024": 0.0052,
|
| 159 |
+
"recall_exact_d1024": 0.0,
|
| 160 |
+
"key_acc_d1024": 0.0,
|
| 161 |
+
"step": 2500,
|
| 162 |
+
"train_bpb": 7.2313,
|
| 163 |
+
"grad_norm": 1.91311574,
|
| 164 |
+
"pairs": 4,
|
| 165 |
+
"lr": 0.0003,
|
| 166 |
+
"eta": 0.0,
|
| 167 |
+
"s_per_step": 0.0418
|
| 168 |
}
|
| 169 |
],
|
| 170 |
"final": {
|
| 171 |
"law_rev": "candm-v1",
|
| 172 |
+
"value_acc_d32": 0.0078,
|
| 173 |
"recall_exact_d32": 0.0,
|
| 174 |
+
"key_acc_d32": 0.0052,
|
| 175 |
+
"value_acc_d128": 0.0065,
|
| 176 |
"recall_exact_d128": 0.0,
|
| 177 |
+
"key_acc_d128": 0.0078,
|
| 178 |
+
"value_acc_d512": 0.0078,
|
| 179 |
"recall_exact_d512": 0.0,
|
| 180 |
+
"key_acc_d512": 0.0013,
|
| 181 |
+
"value_acc_d1024": 0.0052,
|
| 182 |
"recall_exact_d1024": 0.0,
|
| 183 |
+
"key_acc_d1024": 0.0,
|
| 184 |
+
"step": 2500,
|
| 185 |
+
"train_bpb": 7.2313,
|
| 186 |
+
"grad_norm": 1.91311574,
|
| 187 |
"pairs": 4,
|
| 188 |
"lr": 0.0003,
|
| 189 |
"eta": 0.0,
|
| 190 |
+
"s_per_step": 0.0418
|
| 191 |
}
|
| 192 |
}
|
| 193 |
],
|