Upload g2_no_memory_s1.json with huggingface_hub
Browse files- g2_no_memory_s1.json +33 -11
g2_no_memory_s1.json
CHANGED
|
@@ -187,29 +187,51 @@
|
|
| 187 |
"lr": 0.0003,
|
| 188 |
"eta": 0.0,
|
| 189 |
"s_per_step": 0.0445
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 190 |
}
|
| 191 |
],
|
| 192 |
"final": {
|
| 193 |
"law_rev": "candm-v1",
|
| 194 |
-
"value_acc_d32": 0.
|
| 195 |
"recall_exact_d32": 0.0,
|
| 196 |
-
"key_acc_d32": 0.
|
| 197 |
-
"value_acc_d128": 0.
|
| 198 |
"recall_exact_d128": 0.0,
|
| 199 |
-
"key_acc_d128": 0.
|
| 200 |
"value_acc_d512": 0.0026,
|
| 201 |
"recall_exact_d512": 0.0,
|
| 202 |
-
"key_acc_d512": 0.
|
| 203 |
-
"value_acc_d1024": 0.
|
| 204 |
"recall_exact_d1024": 0.0,
|
| 205 |
-
"key_acc_d1024": 0.
|
| 206 |
-
"step":
|
| 207 |
-
"train_bpb": 7.
|
| 208 |
-
"grad_norm":
|
| 209 |
"pairs": 4,
|
| 210 |
"lr": 0.0003,
|
| 211 |
"eta": 0.0,
|
| 212 |
-
"s_per_step": 0.
|
| 213 |
}
|
| 214 |
}
|
| 215 |
],
|
|
|
|
| 187 |
"lr": 0.0003,
|
| 188 |
"eta": 0.0,
|
| 189 |
"s_per_step": 0.0445
|
| 190 |
+
},
|
| 191 |
+
{
|
| 192 |
+
"law_rev": "candm-v1",
|
| 193 |
+
"value_acc_d32": 0.0065,
|
| 194 |
+
"recall_exact_d32": 0.0,
|
| 195 |
+
"key_acc_d32": 0.0026,
|
| 196 |
+
"value_acc_d128": 0.0052,
|
| 197 |
+
"recall_exact_d128": 0.0,
|
| 198 |
+
"key_acc_d128": 0.0026,
|
| 199 |
+
"value_acc_d512": 0.0026,
|
| 200 |
+
"recall_exact_d512": 0.0,
|
| 201 |
+
"key_acc_d512": 0.0013,
|
| 202 |
+
"value_acc_d1024": 0.0026,
|
| 203 |
+
"recall_exact_d1024": 0.0,
|
| 204 |
+
"key_acc_d1024": 0.0065,
|
| 205 |
+
"step": 3500,
|
| 206 |
+
"train_bpb": 7.2201,
|
| 207 |
+
"grad_norm": 2.277839422,
|
| 208 |
+
"pairs": 4,
|
| 209 |
+
"lr": 0.0003,
|
| 210 |
+
"eta": 0.0,
|
| 211 |
+
"s_per_step": 0.0466
|
| 212 |
}
|
| 213 |
],
|
| 214 |
"final": {
|
| 215 |
"law_rev": "candm-v1",
|
| 216 |
+
"value_acc_d32": 0.0065,
|
| 217 |
"recall_exact_d32": 0.0,
|
| 218 |
+
"key_acc_d32": 0.0026,
|
| 219 |
+
"value_acc_d128": 0.0052,
|
| 220 |
"recall_exact_d128": 0.0,
|
| 221 |
+
"key_acc_d128": 0.0026,
|
| 222 |
"value_acc_d512": 0.0026,
|
| 223 |
"recall_exact_d512": 0.0,
|
| 224 |
+
"key_acc_d512": 0.0013,
|
| 225 |
+
"value_acc_d1024": 0.0026,
|
| 226 |
"recall_exact_d1024": 0.0,
|
| 227 |
+
"key_acc_d1024": 0.0065,
|
| 228 |
+
"step": 3500,
|
| 229 |
+
"train_bpb": 7.2201,
|
| 230 |
+
"grad_norm": 2.277839422,
|
| 231 |
"pairs": 4,
|
| 232 |
"lr": 0.0003,
|
| 233 |
"eta": 0.0,
|
| 234 |
+
"s_per_step": 0.0466
|
| 235 |
}
|
| 236 |
}
|
| 237 |
],
|