Model save
Browse files- pytorch_model.bin +1 -1
- wandb/run-20220203_135844-2tzexn1o/files/config.yaml +20 -0
- wandb/run-20220203_135844-2tzexn1o/files/output.log +449 -0
- wandb/run-20220203_135844-2tzexn1o/files/wandb-summary.json +0 -0
- wandb/run-20220203_135844-2tzexn1o/logs/debug-internal.log +0 -0
- wandb/run-20220203_135844-2tzexn1o/run-2tzexn1o.wandb +2 -2
pytorch_model.bin
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 1262112241
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:fe30ae12c53d8f58e684006bd1c1927acfc8a0a8ffea0b10a29486e0564b7ed8
|
| 3 |
size 1262112241
|
wandb/run-20220203_135844-2tzexn1o/files/config.yaml
CHANGED
|
@@ -4802,6 +4802,26 @@ _wandb:
|
|
| 4802 |
5: 1
|
| 4803 |
6:
|
| 4804 |
- 1
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 4805 |
python_version: 3.8.8
|
| 4806 |
start_time: 1643896724
|
| 4807 |
t:
|
|
|
|
| 4802 |
5: 1
|
| 4803 |
6:
|
| 4804 |
- 1
|
| 4805 |
+
- 1: train/train_runtime
|
| 4806 |
+
5: 1
|
| 4807 |
+
6:
|
| 4808 |
+
- 1
|
| 4809 |
+
- 1: train/train_samples_per_second
|
| 4810 |
+
5: 1
|
| 4811 |
+
6:
|
| 4812 |
+
- 1
|
| 4813 |
+
- 1: train/train_steps_per_second
|
| 4814 |
+
5: 1
|
| 4815 |
+
6:
|
| 4816 |
+
- 1
|
| 4817 |
+
- 1: train/total_flos
|
| 4818 |
+
5: 1
|
| 4819 |
+
6:
|
| 4820 |
+
- 1
|
| 4821 |
+
- 1: train/train_loss
|
| 4822 |
+
5: 1
|
| 4823 |
+
6:
|
| 4824 |
+
- 1
|
| 4825 |
python_version: 3.8.8
|
| 4826 |
start_time: 1643896724
|
| 4827 |
t:
|
wandb/run-20220203_135844-2tzexn1o/files/output.log
CHANGED
|
@@ -23082,3 +23082,452 @@ Deleting older checkpoint [checkpoint-8000] due to args.save_total_limit
|
|
| 23082 |
{'eval_loss': inf, 'eval_wer': 0.2171683789697946, 'eval_runtime': 698.574, 'eval_samples_per_second': 22.934, 'eval_steps_per_second': 1.434, 'epoch': 4.87}
|
| 23083 |
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23084 |
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 23082 |
{'eval_loss': inf, 'eval_wer': 0.2171683789697946, 'eval_runtime': 698.574, 'eval_samples_per_second': 22.934, 'eval_steps_per_second': 1.434, 'epoch': 4.87}
|
| 23083 |
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23084 |
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23085 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23086 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23087 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23088 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23089 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23090 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23091 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23092 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23093 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23094 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23095 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23096 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23097 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23098 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23099 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23100 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23101 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23102 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23103 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23104 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23105 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23106 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23107 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23108 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23109 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23110 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23111 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23112 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23113 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23114 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23115 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23116 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23117 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23118 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23119 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23120 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23121 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23122 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23123 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23124 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23125 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23126 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23127 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23128 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23129 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23130 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23131 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23132 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23133 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23134 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23135 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23136 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23137 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23138 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23139 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23140 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23141 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23142 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23143 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23144 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23145 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23146 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23147 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23148 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23149 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23150 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23151 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23152 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23153 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23154 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23155 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23156 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23157 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23158 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23159 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23160 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23161 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23162 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23163 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23164 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23165 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23166 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23167 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23168 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23169 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23170 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23171 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23172 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23173 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23174 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23175 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23176 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23177 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23178 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23179 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23180 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23181 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23182 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23183 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23184 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23185 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23186 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23187 |
+
{'loss': 0.8419, 'learning_rate': 1.7001295336787564e-06, 'epoch': 4.9}
|
| 23188 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23189 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23190 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23191 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23192 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23193 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23194 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23195 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23196 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23197 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23198 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23199 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23200 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23201 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23202 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23203 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23204 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23205 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23206 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23207 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23208 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23209 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23210 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23211 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23212 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23213 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23214 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23215 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23216 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23217 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23218 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23219 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23220 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23221 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23222 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23223 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23224 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23225 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23226 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23227 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23228 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23229 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23230 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23231 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23232 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23233 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23234 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23235 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23236 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23237 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23238 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23239 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23240 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23241 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23242 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23243 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23244 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23245 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23246 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23247 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23248 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23249 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23250 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23251 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23252 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23253 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23254 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23255 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23256 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23257 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23258 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23259 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23260 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23261 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23262 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23263 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23264 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23265 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23266 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23267 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23268 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23269 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23270 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23271 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23272 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23273 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23274 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23275 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23276 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23277 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23278 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23279 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23280 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23281 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23282 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23283 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23284 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23285 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23286 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23287 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23288 |
+
{'loss': 0.8392, 'learning_rate': 1.2143782383419686e-06, 'epoch': 4.93}
|
| 23289 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23290 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23291 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23292 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23293 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23294 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23295 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23296 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23297 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23298 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23299 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23300 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23301 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23302 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23303 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23304 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23305 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23306 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23307 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23308 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23309 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23310 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23311 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23312 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23313 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23314 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23315 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23316 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23317 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23318 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23319 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23320 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23321 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23322 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23323 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23324 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23325 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23326 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23327 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23328 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23329 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23330 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23331 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23332 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23333 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23334 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23335 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23336 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23337 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23338 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23339 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23340 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23341 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23342 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23343 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23344 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23345 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23346 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23347 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23348 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23349 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23350 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23351 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23352 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23353 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23354 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23355 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23356 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23357 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23358 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23359 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23360 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23361 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23362 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23363 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23364 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23365 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23366 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23367 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23368 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23369 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23370 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23371 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23372 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23373 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23374 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23375 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23376 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23377 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23378 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23379 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23380 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23381 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23382 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23383 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23384 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23385 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23386 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23387 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23388 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23389 |
+
{'loss': 0.8369, 'learning_rate': 7.286269430051813e-07, 'epoch': 4.96}
|
| 23390 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23391 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23392 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23393 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23394 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23395 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23396 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23397 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23398 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23399 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23400 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23401 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23402 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23403 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23404 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23405 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23406 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23407 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23408 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23409 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23410 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23411 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23412 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23413 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23414 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23415 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23416 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23417 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23418 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23419 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23420 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23421 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23422 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23423 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23424 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23425 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23426 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23427 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23428 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23429 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23430 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23431 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23432 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23433 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23434 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23435 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23436 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23437 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23438 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23439 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23440 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23441 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23442 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23443 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23444 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23445 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23446 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23447 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23448 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23449 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23450 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23451 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23452 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23453 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23454 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23455 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23456 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23457 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23458 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23459 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23460 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23461 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23462 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23463 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23464 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23465 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23466 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23467 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23468 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23469 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23470 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23471 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23472 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23473 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23474 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23475 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23476 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23477 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23478 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23479 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23480 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23481 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23482 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23483 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23484 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23485 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23486 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23487 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23488 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23489 |
+
{'loss': 0.8428, 'learning_rate': 2.4287564766839375e-07, 'epoch': 4.99}
|
| 23490 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23491 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23492 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23493 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23494 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23495 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23496 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23497 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23498 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23499 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23500 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23501 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23502 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23503 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23504 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23505 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23506 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23507 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23508 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23509 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23510 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23511 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23512 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23513 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23514 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23515 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23516 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23517 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23518 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23519 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23520 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23521 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23522 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23523 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23524 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23525 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23526 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23527 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23528 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23529 |
+
Batch size = 16▌ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23530 |
+
Saving model checkpoint to ./ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23531 |
+
Saving model checkpoint to ./ | 81/1002 [00:50<10:06, 1.52it/s]
|
| 23532 |
+
{'train_runtime': 143109.8834, 'train_samples_per_second': 15.599, 'train_steps_per_second': 0.122, 'train_loss': 1.187884036554109, 'epoch': 5.0}
|
| 23533 |
+
Saving model checkpoint to ./ | 81/1002 [00:50<10:06, 1.52it/s]
|
wandb/run-20220203_135844-2tzexn1o/files/wandb-summary.json
CHANGED
|
The diff for this file is too large to render.
See raw diff
|
|
|
wandb/run-20220203_135844-2tzexn1o/logs/debug-internal.log
CHANGED
|
The diff for this file is too large to render.
See raw diff
|
|
|
wandb/run-20220203_135844-2tzexn1o/run-2tzexn1o.wandb
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:16778d7647e5649a67b93699fa64a7d5752d469690253f71dd4aecd2ab1af7de
|
| 3 |
+
size 131078471
|