Upload g2_accum_s0.json with huggingface_hub
Browse files- g2_accum_s0.json +37 -11
g2_accum_s0.json
CHANGED
|
@@ -267,33 +267,59 @@
|
|
| 267 |
"lr": 0.0003,
|
| 268 |
"eta": 0.0,
|
| 269 |
"s_per_step": 0.4758
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 270 |
}
|
| 271 |
],
|
| 272 |
"final": {
|
| 273 |
"law_rev": "candm-v1",
|
| 274 |
"value_acc_d32": 1.0,
|
| 275 |
"recall_exact_d32": 1.0,
|
| 276 |
-
"key_acc_d32": 0.
|
| 277 |
-
"value_acc_d128":
|
| 278 |
-
"recall_exact_d128":
|
| 279 |
"key_acc_d128": 0.9076,
|
| 280 |
"value_acc_d512": 1.0,
|
| 281 |
"recall_exact_d512": 1.0,
|
| 282 |
-
"key_acc_d512": 0.
|
| 283 |
"value_acc_d1024": 1.0,
|
| 284 |
"recall_exact_d1024": 1.0,
|
| 285 |
-
"key_acc_d1024": 0.
|
| 286 |
"alpha_mean": 1.0,
|
| 287 |
-
"beta_mean": 0.
|
| 288 |
-
"m_fro":
|
| 289 |
"p_fro": 4.899,
|
| 290 |
-
"step":
|
| 291 |
-
"train_bpb": 4.
|
| 292 |
-
"grad_norm":
|
| 293 |
"pairs": 4,
|
| 294 |
"lr": 0.0003,
|
| 295 |
"eta": 0.0,
|
| 296 |
-
"s_per_step": 0.
|
| 297 |
}
|
| 298 |
}
|
| 299 |
],
|
|
|
|
| 267 |
"lr": 0.0003,
|
| 268 |
"eta": 0.0,
|
| 269 |
"s_per_step": 0.4758
|
| 270 |
+
},
|
| 271 |
+
{
|
| 272 |
+
"law_rev": "candm-v1",
|
| 273 |
+
"value_acc_d32": 1.0,
|
| 274 |
+
"recall_exact_d32": 1.0,
|
| 275 |
+
"key_acc_d32": 0.901,
|
| 276 |
+
"value_acc_d128": 0.9974,
|
| 277 |
+
"recall_exact_d128": 0.9792,
|
| 278 |
+
"key_acc_d128": 0.9076,
|
| 279 |
+
"value_acc_d512": 1.0,
|
| 280 |
+
"recall_exact_d512": 1.0,
|
| 281 |
+
"key_acc_d512": 0.9062,
|
| 282 |
+
"value_acc_d1024": 1.0,
|
| 283 |
+
"recall_exact_d1024": 1.0,
|
| 284 |
+
"key_acc_d1024": 0.9049,
|
| 285 |
+
"alpha_mean": 1.0,
|
| 286 |
+
"beta_mean": 0.0631,
|
| 287 |
+
"m_fro": 46.249,
|
| 288 |
+
"p_fro": 4.899,
|
| 289 |
+
"step": 4500,
|
| 290 |
+
"train_bpb": 4.6145,
|
| 291 |
+
"grad_norm": 3.860359669,
|
| 292 |
+
"pairs": 4,
|
| 293 |
+
"lr": 0.0003,
|
| 294 |
+
"eta": 0.0,
|
| 295 |
+
"s_per_step": 0.4957
|
| 296 |
}
|
| 297 |
],
|
| 298 |
"final": {
|
| 299 |
"law_rev": "candm-v1",
|
| 300 |
"value_acc_d32": 1.0,
|
| 301 |
"recall_exact_d32": 1.0,
|
| 302 |
+
"key_acc_d32": 0.901,
|
| 303 |
+
"value_acc_d128": 0.9974,
|
| 304 |
+
"recall_exact_d128": 0.9792,
|
| 305 |
"key_acc_d128": 0.9076,
|
| 306 |
"value_acc_d512": 1.0,
|
| 307 |
"recall_exact_d512": 1.0,
|
| 308 |
+
"key_acc_d512": 0.9062,
|
| 309 |
"value_acc_d1024": 1.0,
|
| 310 |
"recall_exact_d1024": 1.0,
|
| 311 |
+
"key_acc_d1024": 0.9049,
|
| 312 |
"alpha_mean": 1.0,
|
| 313 |
+
"beta_mean": 0.0631,
|
| 314 |
+
"m_fro": 46.249,
|
| 315 |
"p_fro": 4.899,
|
| 316 |
+
"step": 4500,
|
| 317 |
+
"train_bpb": 4.6145,
|
| 318 |
+
"grad_norm": 3.860359669,
|
| 319 |
"pairs": 4,
|
| 320 |
"lr": 0.0003,
|
| 321 |
"eta": 0.0,
|
| 322 |
+
"s_per_step": 0.4957
|
| 323 |
}
|
| 324 |
}
|
| 325 |
],
|