Commit
·
52af36e
1
Parent(s):
ba32c1d
End of training
Browse files- all_results.json +16 -0
- eval_results.json +10 -0
- train_results.json +9 -0
- trainer_state.json +375 -0
all_results.json
ADDED
|
@@ -0,0 +1,16 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 32.0,
|
| 3 |
+
"eval_cer": 0.932302092811647,
|
| 4 |
+
"eval_loss": 8.99256706237793,
|
| 5 |
+
"eval_runtime": 53.3111,
|
| 6 |
+
"eval_samples": 542,
|
| 7 |
+
"eval_samples_per_second": 10.167,
|
| 8 |
+
"eval_steps_per_second": 1.276,
|
| 9 |
+
"eval_wer": 0.9865908167411621,
|
| 10 |
+
"total_flos": 3.9986309695731635e+18,
|
| 11 |
+
"train_loss": 7.336955785751343,
|
| 12 |
+
"train_runtime": 6516.6943,
|
| 13 |
+
"train_samples": 448,
|
| 14 |
+
"train_samples_per_second": 10.312,
|
| 15 |
+
"train_steps_per_second": 0.322
|
| 16 |
+
}
|
eval_results.json
ADDED
|
@@ -0,0 +1,10 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 32.0,
|
| 3 |
+
"eval_cer": 0.932302092811647,
|
| 4 |
+
"eval_loss": 8.99256706237793,
|
| 5 |
+
"eval_runtime": 53.3111,
|
| 6 |
+
"eval_samples": 542,
|
| 7 |
+
"eval_samples_per_second": 10.167,
|
| 8 |
+
"eval_steps_per_second": 1.276,
|
| 9 |
+
"eval_wer": 0.9865908167411621
|
| 10 |
+
}
|
train_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 32.0,
|
| 3 |
+
"total_flos": 3.9986309695731635e+18,
|
| 4 |
+
"train_loss": 7.336955785751343,
|
| 5 |
+
"train_runtime": 6516.6943,
|
| 6 |
+
"train_samples": 448,
|
| 7 |
+
"train_samples_per_second": 10.312,
|
| 8 |
+
"train_steps_per_second": 0.322
|
| 9 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,375 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": 8.99256706237793,
|
| 3 |
+
"best_model_checkpoint": "wav2vec2-large-xlsr-coraa-exp-11/checkpoint-168",
|
| 4 |
+
"epoch": 32.0,
|
| 5 |
+
"global_step": 448,
|
| 6 |
+
"is_hyper_param_search": false,
|
| 7 |
+
"is_local_process_zero": true,
|
| 8 |
+
"is_world_process_zero": true,
|
| 9 |
+
"log_history": [
|
| 10 |
+
{
|
| 11 |
+
"epoch": 0.07,
|
| 12 |
+
"learning_rate": 3e-05,
|
| 13 |
+
"loss": 38.5161,
|
| 14 |
+
"step": 1
|
| 15 |
+
},
|
| 16 |
+
{
|
| 17 |
+
"epoch": 1.0,
|
| 18 |
+
"eval_cer": 0.9510100090991811,
|
| 19 |
+
"eval_loss": 34.24892807006836,
|
| 20 |
+
"eval_runtime": 52.8709,
|
| 21 |
+
"eval_samples_per_second": 10.251,
|
| 22 |
+
"eval_steps_per_second": 1.286,
|
| 23 |
+
"eval_wer": 1.0,
|
| 24 |
+
"step": 14
|
| 25 |
+
},
|
| 26 |
+
{
|
| 27 |
+
"epoch": 2.0,
|
| 28 |
+
"eval_cer": 0.9510464058234759,
|
| 29 |
+
"eval_loss": 23.386924743652344,
|
| 30 |
+
"eval_runtime": 52.734,
|
| 31 |
+
"eval_samples_per_second": 10.278,
|
| 32 |
+
"eval_steps_per_second": 1.289,
|
| 33 |
+
"eval_wer": 1.0,
|
| 34 |
+
"step": 28
|
| 35 |
+
},
|
| 36 |
+
{
|
| 37 |
+
"epoch": 3.0,
|
| 38 |
+
"eval_cer": 0.9510464058234759,
|
| 39 |
+
"eval_loss": 19.672130584716797,
|
| 40 |
+
"eval_runtime": 53.3907,
|
| 41 |
+
"eval_samples_per_second": 10.152,
|
| 42 |
+
"eval_steps_per_second": 1.274,
|
| 43 |
+
"eval_wer": 1.0,
|
| 44 |
+
"step": 42
|
| 45 |
+
},
|
| 46 |
+
{
|
| 47 |
+
"epoch": 4.0,
|
| 48 |
+
"eval_cer": 0.9510464058234759,
|
| 49 |
+
"eval_loss": 18.373493194580078,
|
| 50 |
+
"eval_runtime": 52.736,
|
| 51 |
+
"eval_samples_per_second": 10.278,
|
| 52 |
+
"eval_steps_per_second": 1.289,
|
| 53 |
+
"eval_wer": 1.0,
|
| 54 |
+
"step": 56
|
| 55 |
+
},
|
| 56 |
+
{
|
| 57 |
+
"epoch": 5.0,
|
| 58 |
+
"eval_cer": 0.9495905368516834,
|
| 59 |
+
"eval_loss": 17.55065155029297,
|
| 60 |
+
"eval_runtime": 52.7319,
|
| 61 |
+
"eval_samples_per_second": 10.278,
|
| 62 |
+
"eval_steps_per_second": 1.29,
|
| 63 |
+
"eval_wer": 1.0026412027631044,
|
| 64 |
+
"step": 70
|
| 65 |
+
},
|
| 66 |
+
{
|
| 67 |
+
"epoch": 6.0,
|
| 68 |
+
"eval_cer": 0.9688080072793449,
|
| 69 |
+
"eval_loss": 16.934043884277344,
|
| 70 |
+
"eval_runtime": 52.6762,
|
| 71 |
+
"eval_samples_per_second": 10.289,
|
| 72 |
+
"eval_steps_per_second": 1.291,
|
| 73 |
+
"eval_wer": 1.0737505079236083,
|
| 74 |
+
"step": 84
|
| 75 |
+
},
|
| 76 |
+
{
|
| 77 |
+
"epoch": 7.0,
|
| 78 |
+
"eval_cer": 0.9511191992720656,
|
| 79 |
+
"eval_loss": 17.322906494140625,
|
| 80 |
+
"eval_runtime": 52.7337,
|
| 81 |
+
"eval_samples_per_second": 10.278,
|
| 82 |
+
"eval_steps_per_second": 1.289,
|
| 83 |
+
"eval_wer": 1.0004063388866316,
|
| 84 |
+
"step": 98
|
| 85 |
+
},
|
| 86 |
+
{
|
| 87 |
+
"epoch": 7.14,
|
| 88 |
+
"learning_rate": 2.8614285714285714e-05,
|
| 89 |
+
"loss": 17.5323,
|
| 90 |
+
"step": 100
|
| 91 |
+
},
|
| 92 |
+
{
|
| 93 |
+
"epoch": 8.0,
|
| 94 |
+
"eval_cer": 0.9313921747042766,
|
| 95 |
+
"eval_loss": 16.459413528442383,
|
| 96 |
+
"eval_runtime": 52.7143,
|
| 97 |
+
"eval_samples_per_second": 10.282,
|
| 98 |
+
"eval_steps_per_second": 1.29,
|
| 99 |
+
"eval_wer": 1.0156440471353108,
|
| 100 |
+
"step": 112
|
| 101 |
+
},
|
| 102 |
+
{
|
| 103 |
+
"epoch": 9.0,
|
| 104 |
+
"eval_cer": 0.935213830755232,
|
| 105 |
+
"eval_loss": 12.445086479187012,
|
| 106 |
+
"eval_runtime": 52.7004,
|
| 107 |
+
"eval_samples_per_second": 10.285,
|
| 108 |
+
"eval_steps_per_second": 1.29,
|
| 109 |
+
"eval_wer": 1.0298659081674115,
|
| 110 |
+
"step": 126
|
| 111 |
+
},
|
| 112 |
+
{
|
| 113 |
+
"epoch": 10.0,
|
| 114 |
+
"eval_cer": 0.9618562329390354,
|
| 115 |
+
"eval_loss": 10.092220306396484,
|
| 116 |
+
"eval_runtime": 52.7354,
|
| 117 |
+
"eval_samples_per_second": 10.278,
|
| 118 |
+
"eval_steps_per_second": 1.289,
|
| 119 |
+
"eval_wer": 1.0,
|
| 120 |
+
"step": 140
|
| 121 |
+
},
|
| 122 |
+
{
|
| 123 |
+
"epoch": 11.0,
|
| 124 |
+
"eval_cer": 0.9617834394904459,
|
| 125 |
+
"eval_loss": 9.518609046936035,
|
| 126 |
+
"eval_runtime": 53.2311,
|
| 127 |
+
"eval_samples_per_second": 10.182,
|
| 128 |
+
"eval_steps_per_second": 1.277,
|
| 129 |
+
"eval_wer": 0.9997968305566842,
|
| 130 |
+
"step": 154
|
| 131 |
+
},
|
| 132 |
+
{
|
| 133 |
+
"epoch": 12.0,
|
| 134 |
+
"eval_cer": 0.932302092811647,
|
| 135 |
+
"eval_loss": 8.99256706237793,
|
| 136 |
+
"eval_runtime": 52.6632,
|
| 137 |
+
"eval_samples_per_second": 10.292,
|
| 138 |
+
"eval_steps_per_second": 1.291,
|
| 139 |
+
"eval_wer": 0.9865908167411621,
|
| 140 |
+
"step": 168
|
| 141 |
+
},
|
| 142 |
+
{
|
| 143 |
+
"epoch": 13.0,
|
| 144 |
+
"eval_cer": 0.9166515013648772,
|
| 145 |
+
"eval_loss": 9.018474578857422,
|
| 146 |
+
"eval_runtime": 52.7545,
|
| 147 |
+
"eval_samples_per_second": 10.274,
|
| 148 |
+
"eval_steps_per_second": 1.289,
|
| 149 |
+
"eval_wer": 0.9839496139780577,
|
| 150 |
+
"step": 182
|
| 151 |
+
},
|
| 152 |
+
{
|
| 153 |
+
"epoch": 14.0,
|
| 154 |
+
"eval_cer": 0.9216378525932666,
|
| 155 |
+
"eval_loss": 9.124226570129395,
|
| 156 |
+
"eval_runtime": 53.3159,
|
| 157 |
+
"eval_samples_per_second": 10.166,
|
| 158 |
+
"eval_steps_per_second": 1.275,
|
| 159 |
+
"eval_wer": 0.983746444534742,
|
| 160 |
+
"step": 196
|
| 161 |
+
},
|
| 162 |
+
{
|
| 163 |
+
"epoch": 14.29,
|
| 164 |
+
"learning_rate": 2.7185714285714287e-05,
|
| 165 |
+
"loss": 6.6506,
|
| 166 |
+
"step": 200
|
| 167 |
+
},
|
| 168 |
+
{
|
| 169 |
+
"epoch": 15.0,
|
| 170 |
+
"eval_cer": 0.8844404003639672,
|
| 171 |
+
"eval_loss": 9.050068855285645,
|
| 172 |
+
"eval_runtime": 52.6837,
|
| 173 |
+
"eval_samples_per_second": 10.288,
|
| 174 |
+
"eval_steps_per_second": 1.291,
|
| 175 |
+
"eval_wer": 0.9880130028443722,
|
| 176 |
+
"step": 210
|
| 177 |
+
},
|
| 178 |
+
{
|
| 179 |
+
"epoch": 16.0,
|
| 180 |
+
"eval_cer": 0.902238398544131,
|
| 181 |
+
"eval_loss": 9.189239501953125,
|
| 182 |
+
"eval_runtime": 52.7185,
|
| 183 |
+
"eval_samples_per_second": 10.281,
|
| 184 |
+
"eval_steps_per_second": 1.29,
|
| 185 |
+
"eval_wer": 0.9776513612352702,
|
| 186 |
+
"step": 224
|
| 187 |
+
},
|
| 188 |
+
{
|
| 189 |
+
"epoch": 17.0,
|
| 190 |
+
"eval_cer": 0.8847315741583257,
|
| 191 |
+
"eval_loss": 9.173331260681152,
|
| 192 |
+
"eval_runtime": 52.733,
|
| 193 |
+
"eval_samples_per_second": 10.278,
|
| 194 |
+
"eval_steps_per_second": 1.29,
|
| 195 |
+
"eval_wer": 0.9798862251117432,
|
| 196 |
+
"step": 238
|
| 197 |
+
},
|
| 198 |
+
{
|
| 199 |
+
"epoch": 18.0,
|
| 200 |
+
"eval_cer": 0.8733393994540491,
|
| 201 |
+
"eval_loss": 9.303343772888184,
|
| 202 |
+
"eval_runtime": 52.7609,
|
| 203 |
+
"eval_samples_per_second": 10.273,
|
| 204 |
+
"eval_steps_per_second": 1.289,
|
| 205 |
+
"eval_wer": 0.9798862251117432,
|
| 206 |
+
"step": 252
|
| 207 |
+
},
|
| 208 |
+
{
|
| 209 |
+
"epoch": 19.0,
|
| 210 |
+
"eval_cer": 0.8989626933575978,
|
| 211 |
+
"eval_loss": 9.28529167175293,
|
| 212 |
+
"eval_runtime": 52.7177,
|
| 213 |
+
"eval_samples_per_second": 10.281,
|
| 214 |
+
"eval_steps_per_second": 1.29,
|
| 215 |
+
"eval_wer": 0.9746038195855343,
|
| 216 |
+
"step": 266
|
| 217 |
+
},
|
| 218 |
+
{
|
| 219 |
+
"epoch": 20.0,
|
| 220 |
+
"eval_cer": 0.9086442220200182,
|
| 221 |
+
"eval_loss": 9.438006401062012,
|
| 222 |
+
"eval_runtime": 52.8014,
|
| 223 |
+
"eval_samples_per_second": 10.265,
|
| 224 |
+
"eval_steps_per_second": 1.288,
|
| 225 |
+
"eval_wer": 0.9748069890288501,
|
| 226 |
+
"step": 280
|
| 227 |
+
},
|
| 228 |
+
{
|
| 229 |
+
"epoch": 21.0,
|
| 230 |
+
"eval_cer": 0.8900090991810737,
|
| 231 |
+
"eval_loss": 9.513178825378418,
|
| 232 |
+
"eval_runtime": 52.7023,
|
| 233 |
+
"eval_samples_per_second": 10.284,
|
| 234 |
+
"eval_steps_per_second": 1.29,
|
| 235 |
+
"eval_wer": 0.9750101584721658,
|
| 236 |
+
"step": 294
|
| 237 |
+
},
|
| 238 |
+
{
|
| 239 |
+
"epoch": 21.43,
|
| 240 |
+
"learning_rate": 2.5757142857142857e-05,
|
| 241 |
+
"loss": 3.6568,
|
| 242 |
+
"step": 300
|
| 243 |
+
},
|
| 244 |
+
{
|
| 245 |
+
"epoch": 22.0,
|
| 246 |
+
"eval_cer": 0.8811282984531392,
|
| 247 |
+
"eval_loss": 9.6267728805542,
|
| 248 |
+
"eval_runtime": 53.0646,
|
| 249 |
+
"eval_samples_per_second": 10.214,
|
| 250 |
+
"eval_steps_per_second": 1.281,
|
| 251 |
+
"eval_wer": 0.9817147501015847,
|
| 252 |
+
"step": 308
|
| 253 |
+
},
|
| 254 |
+
{
|
| 255 |
+
"epoch": 23.0,
|
| 256 |
+
"eval_cer": 0.8846587807097361,
|
| 257 |
+
"eval_loss": 9.698858261108398,
|
| 258 |
+
"eval_runtime": 52.7263,
|
| 259 |
+
"eval_samples_per_second": 10.28,
|
| 260 |
+
"eval_steps_per_second": 1.29,
|
| 261 |
+
"eval_wer": 1.0042665583096302,
|
| 262 |
+
"step": 322
|
| 263 |
+
},
|
| 264 |
+
{
|
| 265 |
+
"epoch": 24.0,
|
| 266 |
+
"eval_cer": 0.8963057324840764,
|
| 267 |
+
"eval_loss": 9.611275672912598,
|
| 268 |
+
"eval_runtime": 52.7167,
|
| 269 |
+
"eval_samples_per_second": 10.281,
|
| 270 |
+
"eval_steps_per_second": 1.29,
|
| 271 |
+
"eval_wer": 0.9788703778951645,
|
| 272 |
+
"step": 336
|
| 273 |
+
},
|
| 274 |
+
{
|
| 275 |
+
"epoch": 25.0,
|
| 276 |
+
"eval_cer": 0.8923748862602365,
|
| 277 |
+
"eval_loss": 9.794692039489746,
|
| 278 |
+
"eval_runtime": 52.8702,
|
| 279 |
+
"eval_samples_per_second": 10.252,
|
| 280 |
+
"eval_steps_per_second": 1.286,
|
| 281 |
+
"eval_wer": 0.980698902885006,
|
| 282 |
+
"step": 350
|
| 283 |
+
},
|
| 284 |
+
{
|
| 285 |
+
"epoch": 26.0,
|
| 286 |
+
"eval_cer": 0.8979435850773431,
|
| 287 |
+
"eval_loss": 9.83806324005127,
|
| 288 |
+
"eval_runtime": 52.7321,
|
| 289 |
+
"eval_samples_per_second": 10.278,
|
| 290 |
+
"eval_steps_per_second": 1.29,
|
| 291 |
+
"eval_wer": 0.9794798862251117,
|
| 292 |
+
"step": 364
|
| 293 |
+
},
|
| 294 |
+
{
|
| 295 |
+
"epoch": 27.0,
|
| 296 |
+
"eval_cer": 0.8951774340309372,
|
| 297 |
+
"eval_loss": 10.03061580657959,
|
| 298 |
+
"eval_runtime": 52.6816,
|
| 299 |
+
"eval_samples_per_second": 10.288,
|
| 300 |
+
"eval_steps_per_second": 1.291,
|
| 301 |
+
"eval_wer": 0.9788703778951645,
|
| 302 |
+
"step": 378
|
| 303 |
+
},
|
| 304 |
+
{
|
| 305 |
+
"epoch": 28.0,
|
| 306 |
+
"eval_cer": 0.8947042766151047,
|
| 307 |
+
"eval_loss": 9.994989395141602,
|
| 308 |
+
"eval_runtime": 53.3774,
|
| 309 |
+
"eval_samples_per_second": 10.154,
|
| 310 |
+
"eval_steps_per_second": 1.274,
|
| 311 |
+
"eval_wer": 0.9792767167817961,
|
| 312 |
+
"step": 392
|
| 313 |
+
},
|
| 314 |
+
{
|
| 315 |
+
"epoch": 28.57,
|
| 316 |
+
"learning_rate": 2.432857142857143e-05,
|
| 317 |
+
"loss": 3.316,
|
| 318 |
+
"step": 400
|
| 319 |
+
},
|
| 320 |
+
{
|
| 321 |
+
"epoch": 29.0,
|
| 322 |
+
"eval_cer": 0.8978707916287534,
|
| 323 |
+
"eval_loss": 10.14880084991455,
|
| 324 |
+
"eval_runtime": 52.6791,
|
| 325 |
+
"eval_samples_per_second": 10.289,
|
| 326 |
+
"eval_steps_per_second": 1.291,
|
| 327 |
+
"eval_wer": 0.9780577001219016,
|
| 328 |
+
"step": 406
|
| 329 |
+
},
|
| 330 |
+
{
|
| 331 |
+
"epoch": 30.0,
|
| 332 |
+
"eval_cer": 0.9092265696087353,
|
| 333 |
+
"eval_loss": 10.193445205688477,
|
| 334 |
+
"eval_runtime": 52.6484,
|
| 335 |
+
"eval_samples_per_second": 10.295,
|
| 336 |
+
"eval_steps_per_second": 1.292,
|
| 337 |
+
"eval_wer": 0.9809020723283218,
|
| 338 |
+
"step": 420
|
| 339 |
+
},
|
| 340 |
+
{
|
| 341 |
+
"epoch": 31.0,
|
| 342 |
+
"eval_cer": 0.9298999090081893,
|
| 343 |
+
"eval_loss": 10.21463680267334,
|
| 344 |
+
"eval_runtime": 52.7062,
|
| 345 |
+
"eval_samples_per_second": 10.283,
|
| 346 |
+
"eval_steps_per_second": 1.29,
|
| 347 |
+
"eval_wer": 0.9880130028443722,
|
| 348 |
+
"step": 434
|
| 349 |
+
},
|
| 350 |
+
{
|
| 351 |
+
"epoch": 32.0,
|
| 352 |
+
"eval_cer": 0.9593084622383985,
|
| 353 |
+
"eval_loss": 10.298548698425293,
|
| 354 |
+
"eval_runtime": 52.7136,
|
| 355 |
+
"eval_samples_per_second": 10.282,
|
| 356 |
+
"eval_steps_per_second": 1.29,
|
| 357 |
+
"eval_wer": 0.9997968305566842,
|
| 358 |
+
"step": 448
|
| 359 |
+
},
|
| 360 |
+
{
|
| 361 |
+
"epoch": 32.0,
|
| 362 |
+
"step": 448,
|
| 363 |
+
"total_flos": 3.9986309695731635e+18,
|
| 364 |
+
"train_loss": 7.336955785751343,
|
| 365 |
+
"train_runtime": 6516.6943,
|
| 366 |
+
"train_samples_per_second": 10.312,
|
| 367 |
+
"train_steps_per_second": 0.322
|
| 368 |
+
}
|
| 369 |
+
],
|
| 370 |
+
"max_steps": 2100,
|
| 371 |
+
"num_train_epochs": 150,
|
| 372 |
+
"total_flos": 3.9986309695731635e+18,
|
| 373 |
+
"trial_name": null,
|
| 374 |
+
"trial_params": null
|
| 375 |
+
}
|