Upload g2_no_memory_s2.json with huggingface_hub
Browse files- g2_no_memory_s2.json +33 -11
g2_no_memory_s2.json
CHANGED
|
@@ -231,29 +231,51 @@
|
|
| 231 |
"lr": 0.0003,
|
| 232 |
"eta": 0.0,
|
| 233 |
"s_per_step": 0.0546
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 234 |
}
|
| 235 |
],
|
| 236 |
"final": {
|
| 237 |
"law_rev": "candm-v1",
|
| 238 |
-
"value_acc_d32": 0.
|
| 239 |
"recall_exact_d32": 0.0,
|
| 240 |
-
"key_acc_d32": 0.
|
| 241 |
-
"value_acc_d128": 0.
|
| 242 |
"recall_exact_d128": 0.0,
|
| 243 |
-
"key_acc_d128": 0.
|
| 244 |
"value_acc_d512": 0.0039,
|
| 245 |
"recall_exact_d512": 0.0,
|
| 246 |
-
"key_acc_d512": 0.
|
| 247 |
-
"value_acc_d1024": 0.
|
| 248 |
"recall_exact_d1024": 0.0,
|
| 249 |
-
"key_acc_d1024": 0.
|
| 250 |
-
"step":
|
| 251 |
-
"train_bpb": 7.
|
| 252 |
-
"grad_norm":
|
| 253 |
"pairs": 4,
|
| 254 |
"lr": 0.0003,
|
| 255 |
"eta": 0.0,
|
| 256 |
-
"s_per_step": 0.
|
| 257 |
}
|
| 258 |
}
|
| 259 |
],
|
|
|
|
| 231 |
"lr": 0.0003,
|
| 232 |
"eta": 0.0,
|
| 233 |
"s_per_step": 0.0546
|
| 234 |
+
},
|
| 235 |
+
{
|
| 236 |
+
"law_rev": "candm-v1",
|
| 237 |
+
"value_acc_d32": 0.0039,
|
| 238 |
+
"recall_exact_d32": 0.0,
|
| 239 |
+
"key_acc_d32": 0.0026,
|
| 240 |
+
"value_acc_d128": 0.0026,
|
| 241 |
+
"recall_exact_d128": 0.0,
|
| 242 |
+
"key_acc_d128": 0.0013,
|
| 243 |
+
"value_acc_d512": 0.0039,
|
| 244 |
+
"recall_exact_d512": 0.0,
|
| 245 |
+
"key_acc_d512": 0.0039,
|
| 246 |
+
"value_acc_d1024": 0.0052,
|
| 247 |
+
"recall_exact_d1024": 0.0,
|
| 248 |
+
"key_acc_d1024": 0.0,
|
| 249 |
+
"step": 4500,
|
| 250 |
+
"train_bpb": 7.22,
|
| 251 |
+
"grad_norm": 3.930905581,
|
| 252 |
+
"pairs": 4,
|
| 253 |
+
"lr": 0.0003,
|
| 254 |
+
"eta": 0.0,
|
| 255 |
+
"s_per_step": 0.0557
|
| 256 |
}
|
| 257 |
],
|
| 258 |
"final": {
|
| 259 |
"law_rev": "candm-v1",
|
| 260 |
+
"value_acc_d32": 0.0039,
|
| 261 |
"recall_exact_d32": 0.0,
|
| 262 |
+
"key_acc_d32": 0.0026,
|
| 263 |
+
"value_acc_d128": 0.0026,
|
| 264 |
"recall_exact_d128": 0.0,
|
| 265 |
+
"key_acc_d128": 0.0013,
|
| 266 |
"value_acc_d512": 0.0039,
|
| 267 |
"recall_exact_d512": 0.0,
|
| 268 |
+
"key_acc_d512": 0.0039,
|
| 269 |
+
"value_acc_d1024": 0.0052,
|
| 270 |
"recall_exact_d1024": 0.0,
|
| 271 |
+
"key_acc_d1024": 0.0,
|
| 272 |
+
"step": 4500,
|
| 273 |
+
"train_bpb": 7.22,
|
| 274 |
+
"grad_norm": 3.930905581,
|
| 275 |
"pairs": 4,
|
| 276 |
"lr": 0.0003,
|
| 277 |
"eta": 0.0,
|
| 278 |
+
"s_per_step": 0.0557
|
| 279 |
}
|
| 280 |
}
|
| 281 |
],
|