Upload g2_no_memory_s0.json with huggingface_hub
Browse files- g2_no_memory_s0.json +32 -10
g2_no_memory_s0.json
CHANGED
|
@@ -253,29 +253,51 @@
|
|
| 253 |
"lr": 0.0003,
|
| 254 |
"eta": 0.0,
|
| 255 |
"s_per_step": 0.0524
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 256 |
}
|
| 257 |
],
|
| 258 |
"final": {
|
| 259 |
"law_rev": "candm-v1",
|
| 260 |
-
"value_acc_d32": 0.
|
| 261 |
"recall_exact_d32": 0.0,
|
| 262 |
-
"key_acc_d32": 0.
|
| 263 |
-
"value_acc_d128": 0.
|
| 264 |
"recall_exact_d128": 0.0,
|
| 265 |
-
"key_acc_d128": 0.
|
| 266 |
"value_acc_d512": 0.0052,
|
| 267 |
"recall_exact_d512": 0.0,
|
| 268 |
-
"key_acc_d512": 0.
|
| 269 |
"value_acc_d1024": 0.0039,
|
| 270 |
"recall_exact_d1024": 0.0,
|
| 271 |
-
"key_acc_d1024": 0.
|
| 272 |
-
"step":
|
| 273 |
-
"train_bpb": 7.
|
| 274 |
-
"grad_norm": 1.
|
| 275 |
"pairs": 4,
|
| 276 |
"lr": 0.0003,
|
| 277 |
"eta": 0.0,
|
| 278 |
-
"s_per_step": 0.
|
| 279 |
}
|
| 280 |
}
|
| 281 |
],
|
|
|
|
| 253 |
"lr": 0.0003,
|
| 254 |
"eta": 0.0,
|
| 255 |
"s_per_step": 0.0524
|
| 256 |
+
},
|
| 257 |
+
{
|
| 258 |
+
"law_rev": "candm-v1",
|
| 259 |
+
"value_acc_d32": 0.0013,
|
| 260 |
+
"recall_exact_d32": 0.0,
|
| 261 |
+
"key_acc_d32": 0.0026,
|
| 262 |
+
"value_acc_d128": 0.0026,
|
| 263 |
+
"recall_exact_d128": 0.0,
|
| 264 |
+
"key_acc_d128": 0.0013,
|
| 265 |
+
"value_acc_d512": 0.0052,
|
| 266 |
+
"recall_exact_d512": 0.0,
|
| 267 |
+
"key_acc_d512": 0.0026,
|
| 268 |
+
"value_acc_d1024": 0.0039,
|
| 269 |
+
"recall_exact_d1024": 0.0,
|
| 270 |
+
"key_acc_d1024": 0.0039,
|
| 271 |
+
"step": 5000,
|
| 272 |
+
"train_bpb": 7.2217,
|
| 273 |
+
"grad_norm": 1.172786713,
|
| 274 |
+
"pairs": 4,
|
| 275 |
+
"lr": 0.0003,
|
| 276 |
+
"eta": 0.0,
|
| 277 |
+
"s_per_step": 0.0537
|
| 278 |
}
|
| 279 |
],
|
| 280 |
"final": {
|
| 281 |
"law_rev": "candm-v1",
|
| 282 |
+
"value_acc_d32": 0.0013,
|
| 283 |
"recall_exact_d32": 0.0,
|
| 284 |
+
"key_acc_d32": 0.0026,
|
| 285 |
+
"value_acc_d128": 0.0026,
|
| 286 |
"recall_exact_d128": 0.0,
|
| 287 |
+
"key_acc_d128": 0.0013,
|
| 288 |
"value_acc_d512": 0.0052,
|
| 289 |
"recall_exact_d512": 0.0,
|
| 290 |
+
"key_acc_d512": 0.0026,
|
| 291 |
"value_acc_d1024": 0.0039,
|
| 292 |
"recall_exact_d1024": 0.0,
|
| 293 |
+
"key_acc_d1024": 0.0039,
|
| 294 |
+
"step": 5000,
|
| 295 |
+
"train_bpb": 7.2217,
|
| 296 |
+
"grad_norm": 1.172786713,
|
| 297 |
"pairs": 4,
|
| 298 |
"lr": 0.0003,
|
| 299 |
"eta": 0.0,
|
| 300 |
+
"s_per_step": 0.0537
|
| 301 |
}
|
| 302 |
}
|
| 303 |
],
|