Commit
·
3e2b3f1
1
Parent(s):
417a654
End of training
Browse files
all_results.json
ADDED
|
@@ -0,0 +1,14 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 10.0,
|
| 3 |
+
"eval_loss": 0.34843093156814575,
|
| 4 |
+
"eval_runtime": 151.9671,
|
| 5 |
+
"eval_samples": 2586,
|
| 6 |
+
"eval_samples_per_second": 17.017,
|
| 7 |
+
"eval_steps_per_second": 2.132,
|
| 8 |
+
"eval_wer": 0.23675718221172767,
|
| 9 |
+
"train_loss": 0.30854525013881573,
|
| 10 |
+
"train_runtime": 13756.3958,
|
| 11 |
+
"train_samples": 8702,
|
| 12 |
+
"train_samples_per_second": 6.326,
|
| 13 |
+
"train_steps_per_second": 0.791
|
| 14 |
+
}
|
eval_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 10.0,
|
| 3 |
+
"eval_loss": 0.34843093156814575,
|
| 4 |
+
"eval_runtime": 151.9671,
|
| 5 |
+
"eval_samples": 2586,
|
| 6 |
+
"eval_samples_per_second": 17.017,
|
| 7 |
+
"eval_steps_per_second": 2.132,
|
| 8 |
+
"eval_wer": 0.23675718221172767
|
| 9 |
+
}
|
runs/May22_21-20-54_DESKTOP-ASUS/events.out.tfevents.1684798184.DESKTOP-ASUS.9415.2
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9a86e1935b586555772a80006481affe3456e60b9ec795aaf2c39c31ba029743
|
| 3 |
+
size 358
|
train_results.json
ADDED
|
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 10.0,
|
| 3 |
+
"train_loss": 0.30854525013881573,
|
| 4 |
+
"train_runtime": 13756.3958,
|
| 5 |
+
"train_samples": 8702,
|
| 6 |
+
"train_samples_per_second": 6.326,
|
| 7 |
+
"train_steps_per_second": 0.791
|
| 8 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,394 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 10.0,
|
| 5 |
+
"global_step": 10880,
|
| 6 |
+
"is_hyper_param_search": false,
|
| 7 |
+
"is_local_process_zero": true,
|
| 8 |
+
"is_world_process_zero": true,
|
| 9 |
+
"log_history": [
|
| 10 |
+
{
|
| 11 |
+
"epoch": 0.37,
|
| 12 |
+
"eval_loss": 0.9124094247817993,
|
| 13 |
+
"eval_runtime": 151.7243,
|
| 14 |
+
"eval_samples_per_second": 17.044,
|
| 15 |
+
"eval_steps_per_second": 2.135,
|
| 16 |
+
"eval_wer": 0.7336481700118064,
|
| 17 |
+
"step": 400
|
| 18 |
+
},
|
| 19 |
+
{
|
| 20 |
+
"epoch": 0.46,
|
| 21 |
+
"learning_rate": 0.0002982,
|
| 22 |
+
"loss": 3.904,
|
| 23 |
+
"step": 500
|
| 24 |
+
},
|
| 25 |
+
{
|
| 26 |
+
"epoch": 0.74,
|
| 27 |
+
"eval_loss": 0.4753111004829407,
|
| 28 |
+
"eval_runtime": 151.0266,
|
| 29 |
+
"eval_samples_per_second": 17.123,
|
| 30 |
+
"eval_steps_per_second": 2.145,
|
| 31 |
+
"eval_wer": 0.5022038567493113,
|
| 32 |
+
"step": 800
|
| 33 |
+
},
|
| 34 |
+
{
|
| 35 |
+
"epoch": 0.92,
|
| 36 |
+
"learning_rate": 0.000285635838150289,
|
| 37 |
+
"loss": 0.4384,
|
| 38 |
+
"step": 1000
|
| 39 |
+
},
|
| 40 |
+
{
|
| 41 |
+
"epoch": 1.1,
|
| 42 |
+
"eval_loss": 0.3940904438495636,
|
| 43 |
+
"eval_runtime": 151.5781,
|
| 44 |
+
"eval_samples_per_second": 17.061,
|
| 45 |
+
"eval_steps_per_second": 2.138,
|
| 46 |
+
"eval_wer": 0.3730814639905549,
|
| 47 |
+
"step": 1200
|
| 48 |
+
},
|
| 49 |
+
{
|
| 50 |
+
"epoch": 1.38,
|
| 51 |
+
"learning_rate": 0.00027118497109826587,
|
| 52 |
+
"loss": 0.2985,
|
| 53 |
+
"step": 1500
|
| 54 |
+
},
|
| 55 |
+
{
|
| 56 |
+
"epoch": 1.47,
|
| 57 |
+
"eval_loss": 0.40066808462142944,
|
| 58 |
+
"eval_runtime": 151.8065,
|
| 59 |
+
"eval_samples_per_second": 17.035,
|
| 60 |
+
"eval_steps_per_second": 2.134,
|
| 61 |
+
"eval_wer": 0.3829988193624557,
|
| 62 |
+
"step": 1600
|
| 63 |
+
},
|
| 64 |
+
{
|
| 65 |
+
"epoch": 1.84,
|
| 66 |
+
"learning_rate": 0.00025673410404624276,
|
| 67 |
+
"loss": 0.2719,
|
| 68 |
+
"step": 2000
|
| 69 |
+
},
|
| 70 |
+
{
|
| 71 |
+
"epoch": 1.84,
|
| 72 |
+
"eval_loss": 0.35764721035957336,
|
| 73 |
+
"eval_runtime": 151.6256,
|
| 74 |
+
"eval_samples_per_second": 17.055,
|
| 75 |
+
"eval_steps_per_second": 2.137,
|
| 76 |
+
"eval_wer": 0.3597009051554506,
|
| 77 |
+
"step": 2000
|
| 78 |
+
},
|
| 79 |
+
{
|
| 80 |
+
"epoch": 2.21,
|
| 81 |
+
"eval_loss": 0.3570806384086609,
|
| 82 |
+
"eval_runtime": 151.5024,
|
| 83 |
+
"eval_samples_per_second": 17.069,
|
| 84 |
+
"eval_steps_per_second": 2.139,
|
| 85 |
+
"eval_wer": 0.32857142857142857,
|
| 86 |
+
"step": 2400
|
| 87 |
+
},
|
| 88 |
+
{
|
| 89 |
+
"epoch": 2.3,
|
| 90 |
+
"learning_rate": 0.00024228323699421964,
|
| 91 |
+
"loss": 0.2158,
|
| 92 |
+
"step": 2500
|
| 93 |
+
},
|
| 94 |
+
{
|
| 95 |
+
"epoch": 2.57,
|
| 96 |
+
"eval_loss": 0.3464970290660858,
|
| 97 |
+
"eval_runtime": 151.87,
|
| 98 |
+
"eval_samples_per_second": 17.028,
|
| 99 |
+
"eval_steps_per_second": 2.133,
|
| 100 |
+
"eval_wer": 0.3197560015741834,
|
| 101 |
+
"step": 2800
|
| 102 |
+
},
|
| 103 |
+
{
|
| 104 |
+
"epoch": 2.76,
|
| 105 |
+
"learning_rate": 0.00022783236994219653,
|
| 106 |
+
"loss": 0.2054,
|
| 107 |
+
"step": 3000
|
| 108 |
+
},
|
| 109 |
+
{
|
| 110 |
+
"epoch": 2.94,
|
| 111 |
+
"eval_loss": 0.3161996603012085,
|
| 112 |
+
"eval_runtime": 151.7631,
|
| 113 |
+
"eval_samples_per_second": 17.04,
|
| 114 |
+
"eval_steps_per_second": 2.135,
|
| 115 |
+
"eval_wer": 0.29818968909878,
|
| 116 |
+
"step": 3200
|
| 117 |
+
},
|
| 118 |
+
{
|
| 119 |
+
"epoch": 3.22,
|
| 120 |
+
"learning_rate": 0.00021338150289017341,
|
| 121 |
+
"loss": 0.1783,
|
| 122 |
+
"step": 3500
|
| 123 |
+
},
|
| 124 |
+
{
|
| 125 |
+
"epoch": 3.31,
|
| 126 |
+
"eval_loss": 0.32945507764816284,
|
| 127 |
+
"eval_runtime": 151.3693,
|
| 128 |
+
"eval_samples_per_second": 17.084,
|
| 129 |
+
"eval_steps_per_second": 2.14,
|
| 130 |
+
"eval_wer": 0.3089334907516726,
|
| 131 |
+
"step": 3600
|
| 132 |
+
},
|
| 133 |
+
{
|
| 134 |
+
"epoch": 3.68,
|
| 135 |
+
"learning_rate": 0.0001989306358381503,
|
| 136 |
+
"loss": 0.1495,
|
| 137 |
+
"step": 4000
|
| 138 |
+
},
|
| 139 |
+
{
|
| 140 |
+
"epoch": 3.68,
|
| 141 |
+
"eval_loss": 0.32478460669517517,
|
| 142 |
+
"eval_runtime": 151.7198,
|
| 143 |
+
"eval_samples_per_second": 17.045,
|
| 144 |
+
"eval_steps_per_second": 2.136,
|
| 145 |
+
"eval_wer": 0.3033844942935852,
|
| 146 |
+
"step": 4000
|
| 147 |
+
},
|
| 148 |
+
{
|
| 149 |
+
"epoch": 4.04,
|
| 150 |
+
"eval_loss": 0.31009042263031006,
|
| 151 |
+
"eval_runtime": 152.1775,
|
| 152 |
+
"eval_samples_per_second": 16.993,
|
| 153 |
+
"eval_steps_per_second": 2.129,
|
| 154 |
+
"eval_wer": 0.30275482093663914,
|
| 155 |
+
"step": 4400
|
| 156 |
+
},
|
| 157 |
+
{
|
| 158 |
+
"epoch": 4.14,
|
| 159 |
+
"learning_rate": 0.00018447976878612713,
|
| 160 |
+
"loss": 0.1397,
|
| 161 |
+
"step": 4500
|
| 162 |
+
},
|
| 163 |
+
{
|
| 164 |
+
"epoch": 4.41,
|
| 165 |
+
"eval_loss": 0.3587653338909149,
|
| 166 |
+
"eval_runtime": 154.3439,
|
| 167 |
+
"eval_samples_per_second": 16.755,
|
| 168 |
+
"eval_steps_per_second": 2.099,
|
| 169 |
+
"eval_wer": 0.3005509641873278,
|
| 170 |
+
"step": 4800
|
| 171 |
+
},
|
| 172 |
+
{
|
| 173 |
+
"epoch": 4.6,
|
| 174 |
+
"learning_rate": 0.00017002890173410402,
|
| 175 |
+
"loss": 0.123,
|
| 176 |
+
"step": 5000
|
| 177 |
+
},
|
| 178 |
+
{
|
| 179 |
+
"epoch": 4.78,
|
| 180 |
+
"eval_loss": 0.34509825706481934,
|
| 181 |
+
"eval_runtime": 151.8759,
|
| 182 |
+
"eval_samples_per_second": 17.027,
|
| 183 |
+
"eval_steps_per_second": 2.133,
|
| 184 |
+
"eval_wer": 0.30405352223534043,
|
| 185 |
+
"step": 5200
|
| 186 |
+
},
|
| 187 |
+
{
|
| 188 |
+
"epoch": 5.06,
|
| 189 |
+
"learning_rate": 0.0001555780346820809,
|
| 190 |
+
"loss": 0.115,
|
| 191 |
+
"step": 5500
|
| 192 |
+
},
|
| 193 |
+
{
|
| 194 |
+
"epoch": 5.15,
|
| 195 |
+
"eval_loss": 0.33334314823150635,
|
| 196 |
+
"eval_runtime": 152.7065,
|
| 197 |
+
"eval_samples_per_second": 16.934,
|
| 198 |
+
"eval_steps_per_second": 2.122,
|
| 199 |
+
"eval_wer": 0.2921290830381739,
|
| 200 |
+
"step": 5600
|
| 201 |
+
},
|
| 202 |
+
{
|
| 203 |
+
"epoch": 5.51,
|
| 204 |
+
"learning_rate": 0.0001411271676300578,
|
| 205 |
+
"loss": 0.0947,
|
| 206 |
+
"step": 6000
|
| 207 |
+
},
|
| 208 |
+
{
|
| 209 |
+
"epoch": 5.51,
|
| 210 |
+
"eval_loss": 0.3331167697906494,
|
| 211 |
+
"eval_runtime": 153.2633,
|
| 212 |
+
"eval_samples_per_second": 16.873,
|
| 213 |
+
"eval_steps_per_second": 2.114,
|
| 214 |
+
"eval_wer": 0.285792994883904,
|
| 215 |
+
"step": 6000
|
| 216 |
+
},
|
| 217 |
+
{
|
| 218 |
+
"epoch": 5.88,
|
| 219 |
+
"eval_loss": 0.35364603996276855,
|
| 220 |
+
"eval_runtime": 152.0713,
|
| 221 |
+
"eval_samples_per_second": 17.005,
|
| 222 |
+
"eval_steps_per_second": 2.131,
|
| 223 |
+
"eval_wer": 0.2950019677292405,
|
| 224 |
+
"step": 6400
|
| 225 |
+
},
|
| 226 |
+
{
|
| 227 |
+
"epoch": 5.97,
|
| 228 |
+
"learning_rate": 0.00012667630057803468,
|
| 229 |
+
"loss": 0.0952,
|
| 230 |
+
"step": 6500
|
| 231 |
+
},
|
| 232 |
+
{
|
| 233 |
+
"epoch": 6.25,
|
| 234 |
+
"eval_loss": 0.334403395652771,
|
| 235 |
+
"eval_runtime": 153.3648,
|
| 236 |
+
"eval_samples_per_second": 16.862,
|
| 237 |
+
"eval_steps_per_second": 2.113,
|
| 238 |
+
"eval_wer": 0.2785911058638331,
|
| 239 |
+
"step": 6800
|
| 240 |
+
},
|
| 241 |
+
{
|
| 242 |
+
"epoch": 6.43,
|
| 243 |
+
"learning_rate": 0.00011222543352601155,
|
| 244 |
+
"loss": 0.0778,
|
| 245 |
+
"step": 7000
|
| 246 |
+
},
|
| 247 |
+
{
|
| 248 |
+
"epoch": 6.62,
|
| 249 |
+
"eval_loss": 0.3363187313079834,
|
| 250 |
+
"eval_runtime": 153.0496,
|
| 251 |
+
"eval_samples_per_second": 16.896,
|
| 252 |
+
"eval_steps_per_second": 2.117,
|
| 253 |
+
"eval_wer": 0.2698543880362062,
|
| 254 |
+
"step": 7200
|
| 255 |
+
},
|
| 256 |
+
{
|
| 257 |
+
"epoch": 6.89,
|
| 258 |
+
"learning_rate": 9.777456647398842e-05,
|
| 259 |
+
"loss": 0.0744,
|
| 260 |
+
"step": 7500
|
| 261 |
+
},
|
| 262 |
+
{
|
| 263 |
+
"epoch": 6.99,
|
| 264 |
+
"eval_loss": 0.3245686888694763,
|
| 265 |
+
"eval_runtime": 151.5563,
|
| 266 |
+
"eval_samples_per_second": 17.063,
|
| 267 |
+
"eval_steps_per_second": 2.138,
|
| 268 |
+
"eval_wer": 0.2655253837072019,
|
| 269 |
+
"step": 7600
|
| 270 |
+
},
|
| 271 |
+
{
|
| 272 |
+
"epoch": 7.35,
|
| 273 |
+
"learning_rate": 8.33236994219653e-05,
|
| 274 |
+
"loss": 0.0648,
|
| 275 |
+
"step": 8000
|
| 276 |
+
},
|
| 277 |
+
{
|
| 278 |
+
"epoch": 7.35,
|
| 279 |
+
"eval_loss": 0.33904263377189636,
|
| 280 |
+
"eval_runtime": 152.0758,
|
| 281 |
+
"eval_samples_per_second": 17.005,
|
| 282 |
+
"eval_steps_per_second": 2.131,
|
| 283 |
+
"eval_wer": 0.2626918536009445,
|
| 284 |
+
"step": 8000
|
| 285 |
+
},
|
| 286 |
+
{
|
| 287 |
+
"epoch": 7.72,
|
| 288 |
+
"eval_loss": 0.3405352532863617,
|
| 289 |
+
"eval_runtime": 151.6808,
|
| 290 |
+
"eval_samples_per_second": 17.049,
|
| 291 |
+
"eval_steps_per_second": 2.136,
|
| 292 |
+
"eval_wer": 0.26300669027941753,
|
| 293 |
+
"step": 8400
|
| 294 |
+
},
|
| 295 |
+
{
|
| 296 |
+
"epoch": 7.81,
|
| 297 |
+
"learning_rate": 6.887283236994219e-05,
|
| 298 |
+
"loss": 0.0591,
|
| 299 |
+
"step": 8500
|
| 300 |
+
},
|
| 301 |
+
{
|
| 302 |
+
"epoch": 8.09,
|
| 303 |
+
"eval_loss": 0.3366956114768982,
|
| 304 |
+
"eval_runtime": 152.0044,
|
| 305 |
+
"eval_samples_per_second": 17.013,
|
| 306 |
+
"eval_steps_per_second": 2.132,
|
| 307 |
+
"eval_wer": 0.25340417158598977,
|
| 308 |
+
"step": 8800
|
| 309 |
+
},
|
| 310 |
+
{
|
| 311 |
+
"epoch": 8.27,
|
| 312 |
+
"learning_rate": 5.445086705202312e-05,
|
| 313 |
+
"loss": 0.0527,
|
| 314 |
+
"step": 9000
|
| 315 |
+
},
|
| 316 |
+
{
|
| 317 |
+
"epoch": 8.46,
|
| 318 |
+
"eval_loss": 0.3448215126991272,
|
| 319 |
+
"eval_runtime": 151.6214,
|
| 320 |
+
"eval_samples_per_second": 17.056,
|
| 321 |
+
"eval_steps_per_second": 2.137,
|
| 322 |
+
"eval_wer": 0.2509248327430146,
|
| 323 |
+
"step": 9200
|
| 324 |
+
},
|
| 325 |
+
{
|
| 326 |
+
"epoch": 8.73,
|
| 327 |
+
"learning_rate": 3.9999999999999996e-05,
|
| 328 |
+
"loss": 0.0461,
|
| 329 |
+
"step": 9500
|
| 330 |
+
},
|
| 331 |
+
{
|
| 332 |
+
"epoch": 8.82,
|
| 333 |
+
"eval_loss": 0.3378887474536896,
|
| 334 |
+
"eval_runtime": 151.4481,
|
| 335 |
+
"eval_samples_per_second": 17.075,
|
| 336 |
+
"eval_steps_per_second": 2.139,
|
| 337 |
+
"eval_wer": 0.24254230617866981,
|
| 338 |
+
"step": 9600
|
| 339 |
+
},
|
| 340 |
+
{
|
| 341 |
+
"epoch": 9.19,
|
| 342 |
+
"learning_rate": 2.554913294797688e-05,
|
| 343 |
+
"loss": 0.0408,
|
| 344 |
+
"step": 10000
|
| 345 |
+
},
|
| 346 |
+
{
|
| 347 |
+
"epoch": 9.19,
|
| 348 |
+
"eval_loss": 0.34909504652023315,
|
| 349 |
+
"eval_runtime": 151.2414,
|
| 350 |
+
"eval_samples_per_second": 17.098,
|
| 351 |
+
"eval_steps_per_second": 2.142,
|
| 352 |
+
"eval_wer": 0.24092876820149547,
|
| 353 |
+
"step": 10000
|
| 354 |
+
},
|
| 355 |
+
{
|
| 356 |
+
"epoch": 9.56,
|
| 357 |
+
"eval_loss": 0.3456341028213501,
|
| 358 |
+
"eval_runtime": 151.2228,
|
| 359 |
+
"eval_samples_per_second": 17.101,
|
| 360 |
+
"eval_steps_per_second": 2.143,
|
| 361 |
+
"eval_wer": 0.23774104683195593,
|
| 362 |
+
"step": 10400
|
| 363 |
+
},
|
| 364 |
+
{
|
| 365 |
+
"epoch": 9.65,
|
| 366 |
+
"learning_rate": 1.1098265895953756e-05,
|
| 367 |
+
"loss": 0.0393,
|
| 368 |
+
"step": 10500
|
| 369 |
+
},
|
| 370 |
+
{
|
| 371 |
+
"epoch": 9.93,
|
| 372 |
+
"eval_loss": 0.34881335496902466,
|
| 373 |
+
"eval_runtime": 151.3922,
|
| 374 |
+
"eval_samples_per_second": 17.081,
|
| 375 |
+
"eval_steps_per_second": 2.14,
|
| 376 |
+
"eval_wer": 0.23695395513577333,
|
| 377 |
+
"step": 10800
|
| 378 |
+
},
|
| 379 |
+
{
|
| 380 |
+
"epoch": 10.0,
|
| 381 |
+
"step": 10880,
|
| 382 |
+
"total_flos": 1.3810821820032008e+19,
|
| 383 |
+
"train_loss": 0.30854525013881573,
|
| 384 |
+
"train_runtime": 13756.3958,
|
| 385 |
+
"train_samples_per_second": 6.326,
|
| 386 |
+
"train_steps_per_second": 0.791
|
| 387 |
+
}
|
| 388 |
+
],
|
| 389 |
+
"max_steps": 10880,
|
| 390 |
+
"num_train_epochs": 10,
|
| 391 |
+
"total_flos": 1.3810821820032008e+19,
|
| 392 |
+
"trial_name": null,
|
| 393 |
+
"trial_params": null
|
| 394 |
+
}
|