Upload g2_delta_s1.json with huggingface_hub
Browse files- g2_delta_s1.json +44 -18
g2_delta_s1.json
CHANGED
|
@@ -163,33 +163,59 @@
|
|
| 163 |
"lr": 0.0003,
|
| 164 |
"eta": 0.0,
|
| 165 |
"s_per_step": 0.4877
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 166 |
}
|
| 167 |
],
|
| 168 |
"final": {
|
| 169 |
"law_rev": "candm-v1",
|
| 170 |
-
"value_acc_d32": 0.
|
| 171 |
-
"recall_exact_d32": 0.
|
| 172 |
-
"key_acc_d32": 0.
|
| 173 |
-
"value_acc_d128": 0.
|
| 174 |
-
"recall_exact_d128": 0.
|
| 175 |
-
"key_acc_d128": 0.
|
| 176 |
-
"value_acc_d512": 0.
|
| 177 |
-
"recall_exact_d512": 0.
|
| 178 |
-
"key_acc_d512": 0.
|
| 179 |
-
"value_acc_d1024": 0.
|
| 180 |
-
"recall_exact_d1024": 0.
|
| 181 |
-
"key_acc_d1024": 0.
|
| 182 |
"alpha_mean": 1.0,
|
| 183 |
-
"beta_mean": 0.
|
| 184 |
-
"m_fro": 30.
|
| 185 |
"p_fro": 4.899,
|
| 186 |
-
"step":
|
| 187 |
-
"train_bpb": 5.
|
| 188 |
-
"grad_norm":
|
| 189 |
"pairs": 4,
|
| 190 |
"lr": 0.0003,
|
| 191 |
"eta": 0.0,
|
| 192 |
-
"s_per_step": 0.
|
| 193 |
}
|
| 194 |
}
|
| 195 |
],
|
|
|
|
| 163 |
"lr": 0.0003,
|
| 164 |
"eta": 0.0,
|
| 165 |
"s_per_step": 0.4877
|
| 166 |
+
},
|
| 167 |
+
{
|
| 168 |
+
"law_rev": "candm-v1",
|
| 169 |
+
"value_acc_d32": 0.9609,
|
| 170 |
+
"recall_exact_d32": 0.8229,
|
| 171 |
+
"key_acc_d32": 0.8763,
|
| 172 |
+
"value_acc_d128": 0.9753,
|
| 173 |
+
"recall_exact_d128": 0.875,
|
| 174 |
+
"key_acc_d128": 0.8802,
|
| 175 |
+
"value_acc_d512": 0.9622,
|
| 176 |
+
"recall_exact_d512": 0.8021,
|
| 177 |
+
"key_acc_d512": 0.8789,
|
| 178 |
+
"value_acc_d1024": 0.9674,
|
| 179 |
+
"recall_exact_d1024": 0.8125,
|
| 180 |
+
"key_acc_d1024": 0.8867,
|
| 181 |
+
"alpha_mean": 1.0,
|
| 182 |
+
"beta_mean": 0.0692,
|
| 183 |
+
"m_fro": 30.517,
|
| 184 |
+
"p_fro": 4.899,
|
| 185 |
+
"step": 2500,
|
| 186 |
+
"train_bpb": 5.1214,
|
| 187 |
+
"grad_norm": 4.495575428,
|
| 188 |
+
"pairs": 4,
|
| 189 |
+
"lr": 0.0003,
|
| 190 |
+
"eta": 0.0,
|
| 191 |
+
"s_per_step": 0.5893
|
| 192 |
}
|
| 193 |
],
|
| 194 |
"final": {
|
| 195 |
"law_rev": "candm-v1",
|
| 196 |
+
"value_acc_d32": 0.9609,
|
| 197 |
+
"recall_exact_d32": 0.8229,
|
| 198 |
+
"key_acc_d32": 0.8763,
|
| 199 |
+
"value_acc_d128": 0.9753,
|
| 200 |
+
"recall_exact_d128": 0.875,
|
| 201 |
+
"key_acc_d128": 0.8802,
|
| 202 |
+
"value_acc_d512": 0.9622,
|
| 203 |
+
"recall_exact_d512": 0.8021,
|
| 204 |
+
"key_acc_d512": 0.8789,
|
| 205 |
+
"value_acc_d1024": 0.9674,
|
| 206 |
+
"recall_exact_d1024": 0.8125,
|
| 207 |
+
"key_acc_d1024": 0.8867,
|
| 208 |
"alpha_mean": 1.0,
|
| 209 |
+
"beta_mean": 0.0692,
|
| 210 |
+
"m_fro": 30.517,
|
| 211 |
"p_fro": 4.899,
|
| 212 |
+
"step": 2500,
|
| 213 |
+
"train_bpb": 5.1214,
|
| 214 |
+
"grad_norm": 4.495575428,
|
| 215 |
"pairs": 4,
|
| 216 |
"lr": 0.0003,
|
| 217 |
"eta": 0.0,
|
| 218 |
+
"s_per_step": 0.5893
|
| 219 |
}
|
| 220 |
}
|
| 221 |
],
|