Upload g2_delta_s0.json with huggingface_hub
Browse files- g2_delta_s0.json +23 -7
g2_delta_s0.json
CHANGED
|
@@ -141,23 +141,39 @@
|
|
| 141 |
"pairs": 4,
|
| 142 |
"lr": 0.0003,
|
| 143 |
"s_per_step": 0.4783
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 144 |
}
|
| 145 |
],
|
| 146 |
"final": {
|
| 147 |
-
"value_acc_d32": 0.
|
| 148 |
"recall_exact_d32": 0.0,
|
| 149 |
-
"value_acc_d128": 0.
|
| 150 |
"recall_exact_d128": 0.0,
|
| 151 |
-
"value_acc_d512": 0.
|
| 152 |
"recall_exact_d512": 0.0,
|
| 153 |
"value_acc_d1024": 0.0065,
|
| 154 |
"recall_exact_d1024": 0.0,
|
| 155 |
-
"step":
|
| 156 |
-
"train_bpb": 7.
|
| 157 |
-
"grad_norm": 1.
|
| 158 |
"pairs": 4,
|
| 159 |
"lr": 0.0003,
|
| 160 |
-
"s_per_step": 0.
|
| 161 |
}
|
| 162 |
}
|
| 163 |
],
|
|
|
|
| 141 |
"pairs": 4,
|
| 142 |
"lr": 0.0003,
|
| 143 |
"s_per_step": 0.4783
|
| 144 |
+
},
|
| 145 |
+
{
|
| 146 |
+
"value_acc_d32": 0.0,
|
| 147 |
+
"recall_exact_d32": 0.0,
|
| 148 |
+
"value_acc_d128": 0.0013,
|
| 149 |
+
"recall_exact_d128": 0.0,
|
| 150 |
+
"value_acc_d512": 0.0039,
|
| 151 |
+
"recall_exact_d512": 0.0,
|
| 152 |
+
"value_acc_d1024": 0.0065,
|
| 153 |
+
"recall_exact_d1024": 0.0,
|
| 154 |
+
"step": 3500,
|
| 155 |
+
"train_bpb": 7.2231,
|
| 156 |
+
"grad_norm": 1.429182649,
|
| 157 |
+
"pairs": 4,
|
| 158 |
+
"lr": 0.0003,
|
| 159 |
+
"s_per_step": 0.478
|
| 160 |
}
|
| 161 |
],
|
| 162 |
"final": {
|
| 163 |
+
"value_acc_d32": 0.0,
|
| 164 |
"recall_exact_d32": 0.0,
|
| 165 |
+
"value_acc_d128": 0.0013,
|
| 166 |
"recall_exact_d128": 0.0,
|
| 167 |
+
"value_acc_d512": 0.0039,
|
| 168 |
"recall_exact_d512": 0.0,
|
| 169 |
"value_acc_d1024": 0.0065,
|
| 170 |
"recall_exact_d1024": 0.0,
|
| 171 |
+
"step": 3500,
|
| 172 |
+
"train_bpb": 7.2231,
|
| 173 |
+
"grad_norm": 1.429182649,
|
| 174 |
"pairs": 4,
|
| 175 |
"lr": 0.0003,
|
| 176 |
+
"s_per_step": 0.478
|
| 177 |
}
|
| 178 |
}
|
| 179 |
],
|