Upload g2_accum_s1.json with huggingface_hub
Browse files- g2_accum_s1.json +38 -12
g2_accum_s1.json
CHANGED
|
@@ -241,33 +241,59 @@
|
|
| 241 |
"lr": 0.0003,
|
| 242 |
"eta": 0.0,
|
| 243 |
"s_per_step": 0.4558
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 244 |
}
|
| 245 |
],
|
| 246 |
"final": {
|
| 247 |
"law_rev": "candm-v1",
|
| 248 |
"value_acc_d32": 1.0,
|
| 249 |
"recall_exact_d32": 1.0,
|
| 250 |
-
"key_acc_d32": 0.
|
| 251 |
"value_acc_d128": 0.9987,
|
| 252 |
"recall_exact_d128": 0.9896,
|
| 253 |
-
"key_acc_d128": 0.
|
| 254 |
-
"value_acc_d512":
|
| 255 |
-
"recall_exact_d512":
|
| 256 |
-
"key_acc_d512": 0.
|
| 257 |
"value_acc_d1024": 1.0,
|
| 258 |
"recall_exact_d1024": 1.0,
|
| 259 |
-
"key_acc_d1024": 0.
|
| 260 |
"alpha_mean": 1.0,
|
| 261 |
-
"beta_mean": 0.
|
| 262 |
-
"m_fro":
|
| 263 |
"p_fro": 4.899,
|
| 264 |
-
"step":
|
| 265 |
-
"train_bpb": 4.
|
| 266 |
-
"grad_norm":
|
| 267 |
"pairs": 4,
|
| 268 |
"lr": 0.0003,
|
| 269 |
"eta": 0.0,
|
| 270 |
-
"s_per_step": 0.
|
| 271 |
}
|
| 272 |
}
|
| 273 |
],
|
|
|
|
| 241 |
"lr": 0.0003,
|
| 242 |
"eta": 0.0,
|
| 243 |
"s_per_step": 0.4558
|
| 244 |
+
},
|
| 245 |
+
{
|
| 246 |
+
"law_rev": "candm-v1",
|
| 247 |
+
"value_acc_d32": 1.0,
|
| 248 |
+
"recall_exact_d32": 1.0,
|
| 249 |
+
"key_acc_d32": 0.9102,
|
| 250 |
+
"value_acc_d128": 0.9987,
|
| 251 |
+
"recall_exact_d128": 0.9896,
|
| 252 |
+
"key_acc_d128": 0.9076,
|
| 253 |
+
"value_acc_d512": 0.9987,
|
| 254 |
+
"recall_exact_d512": 0.9896,
|
| 255 |
+
"key_acc_d512": 0.9076,
|
| 256 |
+
"value_acc_d1024": 1.0,
|
| 257 |
+
"recall_exact_d1024": 1.0,
|
| 258 |
+
"key_acc_d1024": 0.9128,
|
| 259 |
+
"alpha_mean": 1.0,
|
| 260 |
+
"beta_mean": 0.0633,
|
| 261 |
+
"m_fro": 45.866,
|
| 262 |
+
"p_fro": 4.899,
|
| 263 |
+
"step": 4000,
|
| 264 |
+
"train_bpb": 4.6562,
|
| 265 |
+
"grad_norm": 5.387327194,
|
| 266 |
+
"pairs": 4,
|
| 267 |
+
"lr": 0.0003,
|
| 268 |
+
"eta": 0.0,
|
| 269 |
+
"s_per_step": 0.4783
|
| 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": 0.9102,
|
| 277 |
"value_acc_d128": 0.9987,
|
| 278 |
"recall_exact_d128": 0.9896,
|
| 279 |
+
"key_acc_d128": 0.9076,
|
| 280 |
+
"value_acc_d512": 0.9987,
|
| 281 |
+
"recall_exact_d512": 0.9896,
|
| 282 |
+
"key_acc_d512": 0.9076,
|
| 283 |
"value_acc_d1024": 1.0,
|
| 284 |
"recall_exact_d1024": 1.0,
|
| 285 |
+
"key_acc_d1024": 0.9128,
|
| 286 |
"alpha_mean": 1.0,
|
| 287 |
+
"beta_mean": 0.0633,
|
| 288 |
+
"m_fro": 45.866,
|
| 289 |
"p_fro": 4.899,
|
| 290 |
+
"step": 4000,
|
| 291 |
+
"train_bpb": 4.6562,
|
| 292 |
+
"grad_norm": 5.387327194,
|
| 293 |
"pairs": 4,
|
| 294 |
"lr": 0.0003,
|
| 295 |
"eta": 0.0,
|
| 296 |
+
"s_per_step": 0.4783
|
| 297 |
}
|
| 298 |
}
|
| 299 |
],
|