Upload g2_no_memory_s2.json with huggingface_hub
Browse files- g2_no_memory_s2.json +33 -11
g2_no_memory_s2.json
CHANGED
|
@@ -297,29 +297,51 @@
|
|
| 297 |
"lr": 0.0003,
|
| 298 |
"eta": 0.0,
|
| 299 |
"s_per_step": 0.0575
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 300 |
}
|
| 301 |
],
|
| 302 |
"final": {
|
| 303 |
"law_rev": "candm-v1",
|
| 304 |
-
"value_acc_d32": 0.
|
| 305 |
"recall_exact_d32": 0.0,
|
| 306 |
-
"key_acc_d32": 0.
|
| 307 |
-
"value_acc_d128": 0.
|
| 308 |
"recall_exact_d128": 0.0,
|
| 309 |
-
"key_acc_d128": 0.
|
| 310 |
"value_acc_d512": 0.0026,
|
| 311 |
"recall_exact_d512": 0.0,
|
| 312 |
-
"key_acc_d512": 0.
|
| 313 |
-
"value_acc_d1024": 0.
|
| 314 |
"recall_exact_d1024": 0.0,
|
| 315 |
-
"key_acc_d1024": 0.
|
| 316 |
-
"step":
|
| 317 |
-
"train_bpb": 7.
|
| 318 |
-
"grad_norm":
|
| 319 |
"pairs": 4,
|
| 320 |
"lr": 0.0003,
|
| 321 |
"eta": 0.0,
|
| 322 |
-
"s_per_step": 0.
|
| 323 |
}
|
| 324 |
}
|
| 325 |
],
|
|
|
|
| 297 |
"lr": 0.0003,
|
| 298 |
"eta": 0.0,
|
| 299 |
"s_per_step": 0.0575
|
| 300 |
+
},
|
| 301 |
+
{
|
| 302 |
+
"law_rev": "candm-v1",
|
| 303 |
+
"value_acc_d32": 0.0052,
|
| 304 |
+
"recall_exact_d32": 0.0,
|
| 305 |
+
"key_acc_d32": 0.0065,
|
| 306 |
+
"value_acc_d128": 0.0013,
|
| 307 |
+
"recall_exact_d128": 0.0,
|
| 308 |
+
"key_acc_d128": 0.0052,
|
| 309 |
+
"value_acc_d512": 0.0026,
|
| 310 |
+
"recall_exact_d512": 0.0,
|
| 311 |
+
"key_acc_d512": 0.0065,
|
| 312 |
+
"value_acc_d1024": 0.0013,
|
| 313 |
+
"recall_exact_d1024": 0.0,
|
| 314 |
+
"key_acc_d1024": 0.0039,
|
| 315 |
+
"step": 5999,
|
| 316 |
+
"train_bpb": 7.8644,
|
| 317 |
+
"grad_norm": 0.3393445611,
|
| 318 |
+
"pairs": 4,
|
| 319 |
+
"lr": 0.0003,
|
| 320 |
+
"eta": 0.0,
|
| 321 |
+
"s_per_step": 0.0579
|
| 322 |
}
|
| 323 |
],
|
| 324 |
"final": {
|
| 325 |
"law_rev": "candm-v1",
|
| 326 |
+
"value_acc_d32": 0.0052,
|
| 327 |
"recall_exact_d32": 0.0,
|
| 328 |
+
"key_acc_d32": 0.0065,
|
| 329 |
+
"value_acc_d128": 0.0013,
|
| 330 |
"recall_exact_d128": 0.0,
|
| 331 |
+
"key_acc_d128": 0.0052,
|
| 332 |
"value_acc_d512": 0.0026,
|
| 333 |
"recall_exact_d512": 0.0,
|
| 334 |
+
"key_acc_d512": 0.0065,
|
| 335 |
+
"value_acc_d1024": 0.0013,
|
| 336 |
"recall_exact_d1024": 0.0,
|
| 337 |
+
"key_acc_d1024": 0.0039,
|
| 338 |
+
"step": 5999,
|
| 339 |
+
"train_bpb": 7.8644,
|
| 340 |
+
"grad_norm": 0.3393445611,
|
| 341 |
"pairs": 4,
|
| 342 |
"lr": 0.0003,
|
| 343 |
"eta": 0.0,
|
| 344 |
+
"s_per_step": 0.0579
|
| 345 |
}
|
| 346 |
}
|
| 347 |
],
|