Upload results.json with huggingface_hub
Browse files- results.json +20 -10
results.json
CHANGED
|
@@ -29,7 +29,8 @@
|
|
| 29 |
"overfit_ratio": 1.2903067379716031,
|
| 30 |
"epochs_to_0.95": 2,
|
| 31 |
"epochs_to_0.99": 3,
|
| 32 |
-
"epochs_to_0.995": 4
|
|
|
|
| 33 |
},
|
| 34 |
"C1_BytePairTokenizer_relu": {
|
| 35 |
"char_acc": 0.999338686466217,
|
|
@@ -61,7 +62,8 @@
|
|
| 61 |
"overfit_ratio": 1.0,
|
| 62 |
"epochs_to_0.95": 2,
|
| 63 |
"epochs_to_0.99": 3,
|
| 64 |
-
"epochs_to_0.995": 4
|
|
|
|
| 65 |
},
|
| 66 |
"C2_BytePairTokenizer_gelu": {
|
| 67 |
"char_acc": 0.9938122034072876,
|
|
@@ -93,7 +95,8 @@
|
|
| 93 |
"overfit_ratio": 1.0147775701616166,
|
| 94 |
"epochs_to_0.95": 3,
|
| 95 |
"epochs_to_0.99": 4,
|
| 96 |
-
"epochs_to_0.995": 4
|
|
|
|
| 97 |
},
|
| 98 |
"C2_BytePairTokenizer_relu": {
|
| 99 |
"char_acc": 0.9965039491653442,
|
|
@@ -125,7 +128,8 @@
|
|
| 125 |
"overfit_ratio": 2.3423576406031454,
|
| 126 |
"epochs_to_0.95": 3,
|
| 127 |
"epochs_to_0.99": 4,
|
| 128 |
-
"epochs_to_0.995": 4
|
|
|
|
| 129 |
},
|
| 130 |
"C3_BytePairTokenizer_gelu": {
|
| 131 |
"char_acc": 0.998898983001709,
|
|
@@ -157,7 +161,8 @@
|
|
| 157 |
"overfit_ratio": 1.0,
|
| 158 |
"epochs_to_0.95": 2,
|
| 159 |
"epochs_to_0.99": 3,
|
| 160 |
-
"epochs_to_0.995": 4
|
|
|
|
| 161 |
},
|
| 162 |
"C3_BytePairTokenizer_relu": {
|
| 163 |
"char_acc": 0.999102771282196,
|
|
@@ -189,7 +194,8 @@
|
|
| 189 |
"overfit_ratio": 1.1771358218921808,
|
| 190 |
"epochs_to_0.95": 2,
|
| 191 |
"epochs_to_0.99": 3,
|
| 192 |
-
"epochs_to_0.995": 4
|
|
|
|
| 193 |
},
|
| 194 |
"C4_BytePairTokenizer_gelu": {
|
| 195 |
"char_acc": 0.9991778135299683,
|
|
@@ -221,7 +227,8 @@
|
|
| 221 |
"overfit_ratio": 1.0,
|
| 222 |
"epochs_to_0.95": 2,
|
| 223 |
"epochs_to_0.99": 3,
|
| 224 |
-
"epochs_to_0.995": 4
|
|
|
|
| 225 |
},
|
| 226 |
"C4_BytePairTokenizer_relu": {
|
| 227 |
"char_acc": 0.9991813898086548,
|
|
@@ -253,7 +260,8 @@
|
|
| 253 |
"overfit_ratio": 1.3646601193743577,
|
| 254 |
"epochs_to_0.95": 2,
|
| 255 |
"epochs_to_0.99": 3,
|
| 256 |
-
"epochs_to_0.995": 4
|
|
|
|
| 257 |
},
|
| 258 |
"C5_ByteTokenizer_gelu": {
|
| 259 |
"char_acc": 0.9999035000801086,
|
|
@@ -281,7 +289,8 @@
|
|
| 281 |
"overfit_ratio": 1.143247336259032,
|
| 282 |
"epochs_to_0.95": 1,
|
| 283 |
"epochs_to_0.99": 1,
|
| 284 |
-
"epochs_to_0.995": 2
|
|
|
|
| 285 |
},
|
| 286 |
"C5_ByteTokenizer_relu": {
|
| 287 |
"char_acc": 0.9999392032623291,
|
|
@@ -309,6 +318,7 @@
|
|
| 309 |
"overfit_ratio": 2.918416596386102,
|
| 310 |
"epochs_to_0.95": 1,
|
| 311 |
"epochs_to_0.99": 2,
|
| 312 |
-
"epochs_to_0.995": 2
|
|
|
|
| 313 |
}
|
| 314 |
}
|
|
|
|
| 29 |
"overfit_ratio": 1.2903067379716031,
|
| 30 |
"epochs_to_0.95": 2,
|
| 31 |
"epochs_to_0.99": 3,
|
| 32 |
+
"epochs_to_0.995": 4,
|
| 33 |
+
"total_train_sec": 1134.476150751114
|
| 34 |
},
|
| 35 |
"C1_BytePairTokenizer_relu": {
|
| 36 |
"char_acc": 0.999338686466217,
|
|
|
|
| 62 |
"overfit_ratio": 1.0,
|
| 63 |
"epochs_to_0.95": 2,
|
| 64 |
"epochs_to_0.99": 3,
|
| 65 |
+
"epochs_to_0.995": 4,
|
| 66 |
+
"total_train_sec": 1122.1742413043976
|
| 67 |
},
|
| 68 |
"C2_BytePairTokenizer_gelu": {
|
| 69 |
"char_acc": 0.9938122034072876,
|
|
|
|
| 95 |
"overfit_ratio": 1.0147775701616166,
|
| 96 |
"epochs_to_0.95": 3,
|
| 97 |
"epochs_to_0.99": 4,
|
| 98 |
+
"epochs_to_0.995": 4,
|
| 99 |
+
"total_train_sec": 1173.8035802841187
|
| 100 |
},
|
| 101 |
"C2_BytePairTokenizer_relu": {
|
| 102 |
"char_acc": 0.9965039491653442,
|
|
|
|
| 128 |
"overfit_ratio": 2.3423576406031454,
|
| 129 |
"epochs_to_0.95": 3,
|
| 130 |
"epochs_to_0.99": 4,
|
| 131 |
+
"epochs_to_0.995": 4,
|
| 132 |
+
"total_train_sec": 1159.4249465465546
|
| 133 |
},
|
| 134 |
"C3_BytePairTokenizer_gelu": {
|
| 135 |
"char_acc": 0.998898983001709,
|
|
|
|
| 161 |
"overfit_ratio": 1.0,
|
| 162 |
"epochs_to_0.95": 2,
|
| 163 |
"epochs_to_0.99": 3,
|
| 164 |
+
"epochs_to_0.995": 4,
|
| 165 |
+
"total_train_sec": 1047.230438709259
|
| 166 |
},
|
| 167 |
"C3_BytePairTokenizer_relu": {
|
| 168 |
"char_acc": 0.999102771282196,
|
|
|
|
| 194 |
"overfit_ratio": 1.1771358218921808,
|
| 195 |
"epochs_to_0.95": 2,
|
| 196 |
"epochs_to_0.99": 3,
|
| 197 |
+
"epochs_to_0.995": 4,
|
| 198 |
+
"total_train_sec": 1030.456927537918
|
| 199 |
},
|
| 200 |
"C4_BytePairTokenizer_gelu": {
|
| 201 |
"char_acc": 0.9991778135299683,
|
|
|
|
| 227 |
"overfit_ratio": 1.0,
|
| 228 |
"epochs_to_0.95": 2,
|
| 229 |
"epochs_to_0.99": 3,
|
| 230 |
+
"epochs_to_0.995": 4,
|
| 231 |
+
"total_train_sec": 1101.650398492813
|
| 232 |
},
|
| 233 |
"C4_BytePairTokenizer_relu": {
|
| 234 |
"char_acc": 0.9991813898086548,
|
|
|
|
| 260 |
"overfit_ratio": 1.3646601193743577,
|
| 261 |
"epochs_to_0.95": 2,
|
| 262 |
"epochs_to_0.99": 3,
|
| 263 |
+
"epochs_to_0.995": 4,
|
| 264 |
+
"total_train_sec": 1086.190969467163
|
| 265 |
},
|
| 266 |
"C5_ByteTokenizer_gelu": {
|
| 267 |
"char_acc": 0.9999035000801086,
|
|
|
|
| 289 |
"overfit_ratio": 1.143247336259032,
|
| 290 |
"epochs_to_0.95": 1,
|
| 291 |
"epochs_to_0.99": 1,
|
| 292 |
+
"epochs_to_0.995": 2,
|
| 293 |
+
"total_train_sec": 1002.6134798526764
|
| 294 |
},
|
| 295 |
"C5_ByteTokenizer_relu": {
|
| 296 |
"char_acc": 0.9999392032623291,
|
|
|
|
| 318 |
"overfit_ratio": 2.918416596386102,
|
| 319 |
"epochs_to_0.95": 1,
|
| 320 |
"epochs_to_0.99": 2,
|
| 321 |
+
"epochs_to_0.995": 2,
|
| 322 |
+
"total_train_sec": 985.0749268531799
|
| 323 |
}
|
| 324 |
}
|