Upload g2_delta_s2.json with huggingface_hub
Browse files- g2_delta_s2.json +44 -18
g2_delta_s2.json
CHANGED
|
@@ -137,33 +137,59 @@
|
|
| 137 |
"lr": 0.0003,
|
| 138 |
"eta": 0.0,
|
| 139 |
"s_per_step": 0.2865
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 140 |
}
|
| 141 |
],
|
| 142 |
"final": {
|
| 143 |
"law_rev": "candm-v1",
|
| 144 |
-
"value_acc_d32": 0.
|
| 145 |
-
"recall_exact_d32": 0.
|
| 146 |
-
"key_acc_d32": 0.
|
| 147 |
-
"value_acc_d128": 0.
|
| 148 |
-
"recall_exact_d128": 0.
|
| 149 |
-
"key_acc_d128": 0.
|
| 150 |
-
"value_acc_d512": 0.
|
| 151 |
-
"recall_exact_d512": 0.
|
| 152 |
-
"key_acc_d512": 0.
|
| 153 |
-
"value_acc_d1024": 0.
|
| 154 |
-
"recall_exact_d1024": 0.
|
| 155 |
-
"key_acc_d1024": 0.
|
| 156 |
"alpha_mean": 1.0,
|
| 157 |
-
"beta_mean": 0.
|
| 158 |
-
"m_fro":
|
| 159 |
"p_fro": 4.899,
|
| 160 |
-
"step":
|
| 161 |
-
"train_bpb": 5.
|
| 162 |
-
"grad_norm":
|
| 163 |
"pairs": 4,
|
| 164 |
"lr": 0.0003,
|
| 165 |
"eta": 0.0,
|
| 166 |
-
"s_per_step": 0.
|
| 167 |
}
|
| 168 |
}
|
| 169 |
],
|
|
|
|
| 137 |
"lr": 0.0003,
|
| 138 |
"eta": 0.0,
|
| 139 |
"s_per_step": 0.2865
|
| 140 |
+
},
|
| 141 |
+
{
|
| 142 |
+
"law_rev": "candm-v1",
|
| 143 |
+
"value_acc_d32": 0.9896,
|
| 144 |
+
"recall_exact_d32": 0.9271,
|
| 145 |
+
"key_acc_d32": 0.8854,
|
| 146 |
+
"value_acc_d128": 0.9661,
|
| 147 |
+
"recall_exact_d128": 0.8333,
|
| 148 |
+
"key_acc_d128": 0.8958,
|
| 149 |
+
"value_acc_d512": 0.9857,
|
| 150 |
+
"recall_exact_d512": 0.9167,
|
| 151 |
+
"key_acc_d512": 0.8815,
|
| 152 |
+
"value_acc_d1024": 0.974,
|
| 153 |
+
"recall_exact_d1024": 0.8646,
|
| 154 |
+
"key_acc_d1024": 0.8776,
|
| 155 |
+
"alpha_mean": 1.0,
|
| 156 |
+
"beta_mean": 0.0672,
|
| 157 |
+
"m_fro": 30.236,
|
| 158 |
+
"p_fro": 4.899,
|
| 159 |
+
"step": 2000,
|
| 160 |
+
"train_bpb": 5.1633,
|
| 161 |
+
"grad_norm": 6.115122795,
|
| 162 |
+
"pairs": 4,
|
| 163 |
+
"lr": 0.0003,
|
| 164 |
+
"eta": 0.0,
|
| 165 |
+
"s_per_step": 0.434
|
| 166 |
}
|
| 167 |
],
|
| 168 |
"final": {
|
| 169 |
"law_rev": "candm-v1",
|
| 170 |
+
"value_acc_d32": 0.9896,
|
| 171 |
+
"recall_exact_d32": 0.9271,
|
| 172 |
+
"key_acc_d32": 0.8854,
|
| 173 |
+
"value_acc_d128": 0.9661,
|
| 174 |
+
"recall_exact_d128": 0.8333,
|
| 175 |
+
"key_acc_d128": 0.8958,
|
| 176 |
+
"value_acc_d512": 0.9857,
|
| 177 |
+
"recall_exact_d512": 0.9167,
|
| 178 |
+
"key_acc_d512": 0.8815,
|
| 179 |
+
"value_acc_d1024": 0.974,
|
| 180 |
+
"recall_exact_d1024": 0.8646,
|
| 181 |
+
"key_acc_d1024": 0.8776,
|
| 182 |
"alpha_mean": 1.0,
|
| 183 |
+
"beta_mean": 0.0672,
|
| 184 |
+
"m_fro": 30.236,
|
| 185 |
"p_fro": 4.899,
|
| 186 |
+
"step": 2000,
|
| 187 |
+
"train_bpb": 5.1633,
|
| 188 |
+
"grad_norm": 6.115122795,
|
| 189 |
"pairs": 4,
|
| 190 |
"lr": 0.0003,
|
| 191 |
"eta": 0.0,
|
| 192 |
+
"s_per_step": 0.434
|
| 193 |
}
|
| 194 |
}
|
| 195 |
],
|