"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-63215 → checkpoint-63838}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-63215 → checkpoint-63838}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-63215 → checkpoint-63838}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-63215 → checkpoint-63838}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-63215 → checkpoint-63838}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-63215 → checkpoint-63838}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-63215 → checkpoint-63838}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-63215 → checkpoint-63838}/trainer_state.json +793 -4
- model-bin/finetune/base/{checkpoint-63215 → checkpoint-63838}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629814655.6609554/events.out.tfevents.1629814655.c435e1c5ee04.920.151 +3 -0
- model-bin/finetune/base/log/1629815285.6311035/events.out.tfevents.1629815285.c435e1c5ee04.920.153 +3 -0
- model-bin/finetune/base/log/1629815928.140718/events.out.tfevents.1629815928.c435e1c5ee04.920.155 +3 -0
- model-bin/finetune/base/log/1629816566.4764583/events.out.tfevents.1629816566.c435e1c5ee04.920.157 +3 -0
- model-bin/finetune/base/log/1629817202.370693/events.out.tfevents.1629817202.c435e1c5ee04.920.159 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629814655.c435e1c5ee04.920.150 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629815285.c435e1c5ee04.920.152 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629815928.c435e1c5ee04.920.154 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629816566.c435e1c5ee04.920.156 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629817202.c435e1c5ee04.920.158 +3 -0
model-bin/finetune/base/{checkpoint-63215 → checkpoint-63838}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-63215 → checkpoint-63838}/optimizer.pt
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 722165009
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6ea2aa8286139cec8752846c0fbd9f67354d4c5b6c4ec3eb9ec3361e3d70efdf
|
| 3 |
size 722165009
|
model-bin/finetune/base/{checkpoint-63215 → checkpoint-63838}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-63215 → checkpoint-63838}/pytorch_model.bin
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 377909911
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:fa35a7bb152e4065b5b9e5285075617217ab93611d50621bc49316253b039e46
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-63215 → checkpoint-63838}/rng_state.pth
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 14503
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:0f65e75947d06eb3cf0ff769245304f3090707e720ac49794be88c35ec07ae1f
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-63215 → checkpoint-63838}/scaler.pt
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 559
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6d83a9d88cf493e73fee270ee9c4462215d6e70194fd22738405e5065bc66bd3
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-63215 → checkpoint-63838}/scheduler.pt
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 623
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2432027de2b4adf0b8c6e8581fca8bf85fcd6ebdc13e99599ee0dd5c1dd04e41
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-63215 → checkpoint-63838}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.18588425381903642,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-60729",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -179796,11 +179796,800 @@
|
|
| 179796 |
"eval_steps_per_second": 0.68,
|
| 179797 |
"eval_wer": 0.19423467905896022,
|
| 179798 |
"step": 63215
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 179799 |
}
|
| 179800 |
],
|
| 179801 |
-
"max_steps":
|
| 179802 |
"num_train_epochs": 5000,
|
| 179803 |
-
"total_flos": 1.
|
| 179804 |
"trial_name": null,
|
| 179805 |
"trial_params": null
|
| 179806 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.18588425381903642,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-60729",
|
| 4 |
+
"epoch": 513.995983935743,
|
| 5 |
+
"global_step": 63838,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 179796 |
"eval_steps_per_second": 0.68,
|
| 179797 |
"eval_wer": 0.19423467905896022,
|
| 179798 |
"step": 63215
|
| 179799 |
+
},
|
| 179800 |
+
{
|
| 179801 |
+
"epoch": 505.04,
|
| 179802 |
+
"learning_rate": 9.003253205128205e-06,
|
| 179803 |
+
"loss": 0.3635,
|
| 179804 |
+
"step": 63220
|
| 179805 |
+
},
|
| 179806 |
+
{
|
| 179807 |
+
"epoch": 505.08,
|
| 179808 |
+
"learning_rate": 9.003173076923077e-06,
|
| 179809 |
+
"loss": 0.3196,
|
| 179810 |
+
"step": 63225
|
| 179811 |
+
},
|
| 179812 |
+
{
|
| 179813 |
+
"epoch": 505.12,
|
| 179814 |
+
"learning_rate": 9.00309294871795e-06,
|
| 179815 |
+
"loss": 0.3548,
|
| 179816 |
+
"step": 63230
|
| 179817 |
+
},
|
| 179818 |
+
{
|
| 179819 |
+
"epoch": 505.16,
|
| 179820 |
+
"learning_rate": 9.003012820512821e-06,
|
| 179821 |
+
"loss": 0.5298,
|
| 179822 |
+
"step": 63235
|
| 179823 |
+
},
|
| 179824 |
+
{
|
| 179825 |
+
"epoch": 505.2,
|
| 179826 |
+
"learning_rate": 9.002932692307693e-06,
|
| 179827 |
+
"loss": 1.2469,
|
| 179828 |
+
"step": 63240
|
| 179829 |
+
},
|
| 179830 |
+
{
|
| 179831 |
+
"epoch": 505.24,
|
| 179832 |
+
"learning_rate": 9.002852564102564e-06,
|
| 179833 |
+
"loss": 0.3912,
|
| 179834 |
+
"step": 63245
|
| 179835 |
+
},
|
| 179836 |
+
{
|
| 179837 |
+
"epoch": 505.28,
|
| 179838 |
+
"learning_rate": 9.002772435897437e-06,
|
| 179839 |
+
"loss": 0.2929,
|
| 179840 |
+
"step": 63250
|
| 179841 |
+
},
|
| 179842 |
+
{
|
| 179843 |
+
"epoch": 505.32,
|
| 179844 |
+
"learning_rate": 9.002692307692308e-06,
|
| 179845 |
+
"loss": 0.3742,
|
| 179846 |
+
"step": 63255
|
| 179847 |
+
},
|
| 179848 |
+
{
|
| 179849 |
+
"epoch": 505.36,
|
| 179850 |
+
"learning_rate": 9.00261217948718e-06,
|
| 179851 |
+
"loss": 0.4614,
|
| 179852 |
+
"step": 63260
|
| 179853 |
+
},
|
| 179854 |
+
{
|
| 179855 |
+
"epoch": 505.4,
|
| 179856 |
+
"learning_rate": 9.002532051282053e-06,
|
| 179857 |
+
"loss": 1.161,
|
| 179858 |
+
"step": 63265
|
| 179859 |
+
},
|
| 179860 |
+
{
|
| 179861 |
+
"epoch": 505.44,
|
| 179862 |
+
"learning_rate": 9.002451923076924e-06,
|
| 179863 |
+
"loss": 0.3384,
|
| 179864 |
+
"step": 63270
|
| 179865 |
+
},
|
| 179866 |
+
{
|
| 179867 |
+
"epoch": 505.48,
|
| 179868 |
+
"learning_rate": 9.002371794871795e-06,
|
| 179869 |
+
"loss": 0.3359,
|
| 179870 |
+
"step": 63275
|
| 179871 |
+
},
|
| 179872 |
+
{
|
| 179873 |
+
"epoch": 505.52,
|
| 179874 |
+
"learning_rate": 9.002291666666667e-06,
|
| 179875 |
+
"loss": 0.2962,
|
| 179876 |
+
"step": 63280
|
| 179877 |
+
},
|
| 179878 |
+
{
|
| 179879 |
+
"epoch": 505.56,
|
| 179880 |
+
"learning_rate": 9.00221153846154e-06,
|
| 179881 |
+
"loss": 0.5481,
|
| 179882 |
+
"step": 63285
|
| 179883 |
+
},
|
| 179884 |
+
{
|
| 179885 |
+
"epoch": 505.6,
|
| 179886 |
+
"learning_rate": 9.002131410256411e-06,
|
| 179887 |
+
"loss": 1.2128,
|
| 179888 |
+
"step": 63290
|
| 179889 |
+
},
|
| 179890 |
+
{
|
| 179891 |
+
"epoch": 505.64,
|
| 179892 |
+
"learning_rate": 9.002051282051283e-06,
|
| 179893 |
+
"loss": 0.3452,
|
| 179894 |
+
"step": 63295
|
| 179895 |
+
},
|
| 179896 |
+
{
|
| 179897 |
+
"epoch": 505.68,
|
| 179898 |
+
"learning_rate": 9.001971153846154e-06,
|
| 179899 |
+
"loss": 0.3947,
|
| 179900 |
+
"step": 63300
|
| 179901 |
+
},
|
| 179902 |
+
{
|
| 179903 |
+
"epoch": 505.72,
|
| 179904 |
+
"learning_rate": 9.001891025641027e-06,
|
| 179905 |
+
"loss": 0.324,
|
| 179906 |
+
"step": 63305
|
| 179907 |
+
},
|
| 179908 |
+
{
|
| 179909 |
+
"epoch": 505.76,
|
| 179910 |
+
"learning_rate": 9.001810897435898e-06,
|
| 179911 |
+
"loss": 0.5189,
|
| 179912 |
+
"step": 63310
|
| 179913 |
+
},
|
| 179914 |
+
{
|
| 179915 |
+
"epoch": 505.8,
|
| 179916 |
+
"learning_rate": 9.001746794871796e-06,
|
| 179917 |
+
"loss": 1.4983,
|
| 179918 |
+
"step": 63315
|
| 179919 |
+
},
|
| 179920 |
+
{
|
| 179921 |
+
"epoch": 505.84,
|
| 179922 |
+
"learning_rate": 9.001666666666667e-06,
|
| 179923 |
+
"loss": 0.3326,
|
| 179924 |
+
"step": 63320
|
| 179925 |
+
},
|
| 179926 |
+
{
|
| 179927 |
+
"epoch": 505.88,
|
| 179928 |
+
"learning_rate": 9.00158653846154e-06,
|
| 179929 |
+
"loss": 0.4192,
|
| 179930 |
+
"step": 63325
|
| 179931 |
+
},
|
| 179932 |
+
{
|
| 179933 |
+
"epoch": 505.92,
|
| 179934 |
+
"learning_rate": 9.001506410256412e-06,
|
| 179935 |
+
"loss": 0.3404,
|
| 179936 |
+
"step": 63330
|
| 179937 |
+
},
|
| 179938 |
+
{
|
| 179939 |
+
"epoch": 505.96,
|
| 179940 |
+
"learning_rate": 9.001426282051283e-06,
|
| 179941 |
+
"loss": 0.5378,
|
| 179942 |
+
"step": 63335
|
| 179943 |
+
},
|
| 179944 |
+
{
|
| 179945 |
+
"epoch": 506.0,
|
| 179946 |
+
"learning_rate": 9.001346153846154e-06,
|
| 179947 |
+
"loss": 1.4378,
|
| 179948 |
+
"step": 63340
|
| 179949 |
+
},
|
| 179950 |
+
{
|
| 179951 |
+
"epoch": 506.0,
|
| 179952 |
+
"eval_loss": 0.4314885139465332,
|
| 179953 |
+
"eval_runtime": 39.1186,
|
| 179954 |
+
"eval_samples_per_second": 21.575,
|
| 179955 |
+
"eval_steps_per_second": 0.69,
|
| 179956 |
+
"eval_wer": 0.1998677831643896,
|
| 179957 |
+
"step": 63340
|
| 179958 |
+
},
|
| 179959 |
+
{
|
| 179960 |
+
"epoch": 506.04,
|
| 179961 |
+
"learning_rate": 9.001266025641027e-06,
|
| 179962 |
+
"loss": 0.352,
|
| 179963 |
+
"step": 63345
|
| 179964 |
+
},
|
| 179965 |
+
{
|
| 179966 |
+
"epoch": 506.08,
|
| 179967 |
+
"learning_rate": 9.001185897435899e-06,
|
| 179968 |
+
"loss": 0.3085,
|
| 179969 |
+
"step": 63350
|
| 179970 |
+
},
|
| 179971 |
+
{
|
| 179972 |
+
"epoch": 506.12,
|
| 179973 |
+
"learning_rate": 9.00110576923077e-06,
|
| 179974 |
+
"loss": 0.3506,
|
| 179975 |
+
"step": 63355
|
| 179976 |
+
},
|
| 179977 |
+
{
|
| 179978 |
+
"epoch": 506.16,
|
| 179979 |
+
"learning_rate": 9.001025641025641e-06,
|
| 179980 |
+
"loss": 0.5612,
|
| 179981 |
+
"step": 63360
|
| 179982 |
+
},
|
| 179983 |
+
{
|
| 179984 |
+
"epoch": 506.2,
|
| 179985 |
+
"learning_rate": 9.000945512820514e-06,
|
| 179986 |
+
"loss": 1.2121,
|
| 179987 |
+
"step": 63365
|
| 179988 |
+
},
|
| 179989 |
+
{
|
| 179990 |
+
"epoch": 506.24,
|
| 179991 |
+
"learning_rate": 9.000865384615384e-06,
|
| 179992 |
+
"loss": 0.3428,
|
| 179993 |
+
"step": 63370
|
| 179994 |
+
},
|
| 179995 |
+
{
|
| 179996 |
+
"epoch": 506.28,
|
| 179997 |
+
"learning_rate": 9.000785256410257e-06,
|
| 179998 |
+
"loss": 0.4635,
|
| 179999 |
+
"step": 63375
|
| 180000 |
+
},
|
| 180001 |
+
{
|
| 180002 |
+
"epoch": 506.32,
|
| 180003 |
+
"learning_rate": 9.00070512820513e-06,
|
| 180004 |
+
"loss": 0.3347,
|
| 180005 |
+
"step": 63380
|
| 180006 |
+
},
|
| 180007 |
+
{
|
| 180008 |
+
"epoch": 506.36,
|
| 180009 |
+
"learning_rate": 9.000625e-06,
|
| 180010 |
+
"loss": 0.4775,
|
| 180011 |
+
"step": 63385
|
| 180012 |
+
},
|
| 180013 |
+
{
|
| 180014 |
+
"epoch": 506.4,
|
| 180015 |
+
"learning_rate": 9.000544871794873e-06,
|
| 180016 |
+
"loss": 1.2133,
|
| 180017 |
+
"step": 63390
|
| 180018 |
+
},
|
| 180019 |
+
{
|
| 180020 |
+
"epoch": 506.44,
|
| 180021 |
+
"learning_rate": 9.000464743589744e-06,
|
| 180022 |
+
"loss": 0.2746,
|
| 180023 |
+
"step": 63395
|
| 180024 |
+
},
|
| 180025 |
+
{
|
| 180026 |
+
"epoch": 506.48,
|
| 180027 |
+
"learning_rate": 9.000384615384616e-06,
|
| 180028 |
+
"loss": 0.3917,
|
| 180029 |
+
"step": 63400
|
| 180030 |
+
},
|
| 180031 |
+
{
|
| 180032 |
+
"epoch": 506.52,
|
| 180033 |
+
"learning_rate": 9.000304487179487e-06,
|
| 180034 |
+
"loss": 0.3857,
|
| 180035 |
+
"step": 63405
|
| 180036 |
+
},
|
| 180037 |
+
{
|
| 180038 |
+
"epoch": 506.56,
|
| 180039 |
+
"learning_rate": 9.00022435897436e-06,
|
| 180040 |
+
"loss": 0.5096,
|
| 180041 |
+
"step": 63410
|
| 180042 |
+
},
|
| 180043 |
+
{
|
| 180044 |
+
"epoch": 506.6,
|
| 180045 |
+
"learning_rate": 9.000144230769231e-06,
|
| 180046 |
+
"loss": 1.2501,
|
| 180047 |
+
"step": 63415
|
| 180048 |
+
},
|
| 180049 |
+
{
|
| 180050 |
+
"epoch": 506.64,
|
| 180051 |
+
"learning_rate": 9.000064102564103e-06,
|
| 180052 |
+
"loss": 0.3473,
|
| 180053 |
+
"step": 63420
|
| 180054 |
+
},
|
| 180055 |
+
{
|
| 180056 |
+
"epoch": 506.68,
|
| 180057 |
+
"learning_rate": 8.999983974358976e-06,
|
| 180058 |
+
"loss": 0.3344,
|
| 180059 |
+
"step": 63425
|
| 180060 |
+
},
|
| 180061 |
+
{
|
| 180062 |
+
"epoch": 506.72,
|
| 180063 |
+
"learning_rate": 8.999903846153847e-06,
|
| 180064 |
+
"loss": 0.4294,
|
| 180065 |
+
"step": 63430
|
| 180066 |
+
},
|
| 180067 |
+
{
|
| 180068 |
+
"epoch": 506.76,
|
| 180069 |
+
"learning_rate": 8.999823717948719e-06,
|
| 180070 |
+
"loss": 0.4635,
|
| 180071 |
+
"step": 63435
|
| 180072 |
+
},
|
| 180073 |
+
{
|
| 180074 |
+
"epoch": 506.8,
|
| 180075 |
+
"learning_rate": 8.99974358974359e-06,
|
| 180076 |
+
"loss": 1.3841,
|
| 180077 |
+
"step": 63440
|
| 180078 |
+
},
|
| 180079 |
+
{
|
| 180080 |
+
"epoch": 506.84,
|
| 180081 |
+
"learning_rate": 8.999663461538463e-06,
|
| 180082 |
+
"loss": 0.3623,
|
| 180083 |
+
"step": 63445
|
| 180084 |
+
},
|
| 180085 |
+
{
|
| 180086 |
+
"epoch": 506.88,
|
| 180087 |
+
"learning_rate": 8.999583333333334e-06,
|
| 180088 |
+
"loss": 0.2916,
|
| 180089 |
+
"step": 63450
|
| 180090 |
+
},
|
| 180091 |
+
{
|
| 180092 |
+
"epoch": 506.92,
|
| 180093 |
+
"learning_rate": 8.999503205128206e-06,
|
| 180094 |
+
"loss": 0.3427,
|
| 180095 |
+
"step": 63455
|
| 180096 |
+
},
|
| 180097 |
+
{
|
| 180098 |
+
"epoch": 506.96,
|
| 180099 |
+
"learning_rate": 8.999423076923077e-06,
|
| 180100 |
+
"loss": 0.5609,
|
| 180101 |
+
"step": 63460
|
| 180102 |
+
},
|
| 180103 |
+
{
|
| 180104 |
+
"epoch": 507.0,
|
| 180105 |
+
"learning_rate": 8.99934294871795e-06,
|
| 180106 |
+
"loss": 1.3925,
|
| 180107 |
+
"step": 63465
|
| 180108 |
+
},
|
| 180109 |
+
{
|
| 180110 |
+
"epoch": 507.0,
|
| 180111 |
+
"eval_loss": 0.4419548511505127,
|
| 180112 |
+
"eval_runtime": 38.8185,
|
| 180113 |
+
"eval_samples_per_second": 21.742,
|
| 180114 |
+
"eval_steps_per_second": 0.696,
|
| 180115 |
+
"eval_wer": 0.2030339261694814,
|
| 180116 |
+
"step": 63465
|
| 180117 |
+
},
|
| 180118 |
+
{
|
| 180119 |
+
"epoch": 507.04,
|
| 180120 |
+
"learning_rate": 8.999262820512821e-06,
|
| 180121 |
+
"loss": 0.3727,
|
| 180122 |
+
"step": 63470
|
| 180123 |
+
},
|
| 180124 |
+
{
|
| 180125 |
+
"epoch": 507.08,
|
| 180126 |
+
"learning_rate": 8.999182692307693e-06,
|
| 180127 |
+
"loss": 0.3517,
|
| 180128 |
+
"step": 63475
|
| 180129 |
+
},
|
| 180130 |
+
{
|
| 180131 |
+
"epoch": 507.12,
|
| 180132 |
+
"learning_rate": 8.999102564102566e-06,
|
| 180133 |
+
"loss": 0.3514,
|
| 180134 |
+
"step": 63480
|
| 180135 |
+
},
|
| 180136 |
+
{
|
| 180137 |
+
"epoch": 507.16,
|
| 180138 |
+
"learning_rate": 8.999022435897437e-06,
|
| 180139 |
+
"loss": 0.4517,
|
| 180140 |
+
"step": 63485
|
| 180141 |
+
},
|
| 180142 |
+
{
|
| 180143 |
+
"epoch": 507.2,
|
| 180144 |
+
"learning_rate": 8.998942307692309e-06,
|
| 180145 |
+
"loss": 1.3289,
|
| 180146 |
+
"step": 63490
|
| 180147 |
+
},
|
| 180148 |
+
{
|
| 180149 |
+
"epoch": 507.24,
|
| 180150 |
+
"learning_rate": 8.99886217948718e-06,
|
| 180151 |
+
"loss": 0.4235,
|
| 180152 |
+
"step": 63495
|
| 180153 |
+
},
|
| 180154 |
+
{
|
| 180155 |
+
"epoch": 507.28,
|
| 180156 |
+
"learning_rate": 8.998782051282053e-06,
|
| 180157 |
+
"loss": 0.3557,
|
| 180158 |
+
"step": 63500
|
| 180159 |
+
},
|
| 180160 |
+
{
|
| 180161 |
+
"epoch": 507.32,
|
| 180162 |
+
"learning_rate": 8.998701923076923e-06,
|
| 180163 |
+
"loss": 0.3481,
|
| 180164 |
+
"step": 63505
|
| 180165 |
+
},
|
| 180166 |
+
{
|
| 180167 |
+
"epoch": 507.36,
|
| 180168 |
+
"learning_rate": 8.998621794871796e-06,
|
| 180169 |
+
"loss": 0.5474,
|
| 180170 |
+
"step": 63510
|
| 180171 |
+
},
|
| 180172 |
+
{
|
| 180173 |
+
"epoch": 507.4,
|
| 180174 |
+
"learning_rate": 8.998541666666667e-06,
|
| 180175 |
+
"loss": 1.3734,
|
| 180176 |
+
"step": 63515
|
| 180177 |
+
},
|
| 180178 |
+
{
|
| 180179 |
+
"epoch": 507.44,
|
| 180180 |
+
"learning_rate": 8.998461538461538e-06,
|
| 180181 |
+
"loss": 0.3385,
|
| 180182 |
+
"step": 63520
|
| 180183 |
+
},
|
| 180184 |
+
{
|
| 180185 |
+
"epoch": 507.48,
|
| 180186 |
+
"learning_rate": 8.998381410256411e-06,
|
| 180187 |
+
"loss": 0.3331,
|
| 180188 |
+
"step": 63525
|
| 180189 |
+
},
|
| 180190 |
+
{
|
| 180191 |
+
"epoch": 507.52,
|
| 180192 |
+
"learning_rate": 8.998301282051283e-06,
|
| 180193 |
+
"loss": 0.3953,
|
| 180194 |
+
"step": 63530
|
| 180195 |
+
},
|
| 180196 |
+
{
|
| 180197 |
+
"epoch": 507.56,
|
| 180198 |
+
"learning_rate": 8.998221153846154e-06,
|
| 180199 |
+
"loss": 0.5396,
|
| 180200 |
+
"step": 63535
|
| 180201 |
+
},
|
| 180202 |
+
{
|
| 180203 |
+
"epoch": 507.6,
|
| 180204 |
+
"learning_rate": 8.998141025641026e-06,
|
| 180205 |
+
"loss": 1.2358,
|
| 180206 |
+
"step": 63540
|
| 180207 |
+
},
|
| 180208 |
+
{
|
| 180209 |
+
"epoch": 507.64,
|
| 180210 |
+
"learning_rate": 8.998060897435899e-06,
|
| 180211 |
+
"loss": 0.3537,
|
| 180212 |
+
"step": 63545
|
| 180213 |
+
},
|
| 180214 |
+
{
|
| 180215 |
+
"epoch": 507.68,
|
| 180216 |
+
"learning_rate": 8.99798076923077e-06,
|
| 180217 |
+
"loss": 0.3462,
|
| 180218 |
+
"step": 63550
|
| 180219 |
+
},
|
| 180220 |
+
{
|
| 180221 |
+
"epoch": 507.72,
|
| 180222 |
+
"learning_rate": 8.997900641025641e-06,
|
| 180223 |
+
"loss": 0.3443,
|
| 180224 |
+
"step": 63555
|
| 180225 |
+
},
|
| 180226 |
+
{
|
| 180227 |
+
"epoch": 507.76,
|
| 180228 |
+
"learning_rate": 8.997820512820513e-06,
|
| 180229 |
+
"loss": 0.4466,
|
| 180230 |
+
"step": 63560
|
| 180231 |
+
},
|
| 180232 |
+
{
|
| 180233 |
+
"epoch": 507.8,
|
| 180234 |
+
"learning_rate": 8.997740384615386e-06,
|
| 180235 |
+
"loss": 1.2187,
|
| 180236 |
+
"step": 63565
|
| 180237 |
+
},
|
| 180238 |
+
{
|
| 180239 |
+
"epoch": 507.84,
|
| 180240 |
+
"learning_rate": 8.997660256410257e-06,
|
| 180241 |
+
"loss": 0.3439,
|
| 180242 |
+
"step": 63570
|
| 180243 |
+
},
|
| 180244 |
+
{
|
| 180245 |
+
"epoch": 507.88,
|
| 180246 |
+
"learning_rate": 8.997580128205128e-06,
|
| 180247 |
+
"loss": 0.285,
|
| 180248 |
+
"step": 63575
|
| 180249 |
+
},
|
| 180250 |
+
{
|
| 180251 |
+
"epoch": 507.92,
|
| 180252 |
+
"learning_rate": 8.997500000000002e-06,
|
| 180253 |
+
"loss": 0.361,
|
| 180254 |
+
"step": 63580
|
| 180255 |
+
},
|
| 180256 |
+
{
|
| 180257 |
+
"epoch": 507.96,
|
| 180258 |
+
"learning_rate": 8.997419871794873e-06,
|
| 180259 |
+
"loss": 0.4723,
|
| 180260 |
+
"step": 63585
|
| 180261 |
+
},
|
| 180262 |
+
{
|
| 180263 |
+
"epoch": 508.0,
|
| 180264 |
+
"learning_rate": 8.997339743589744e-06,
|
| 180265 |
+
"loss": 1.4911,
|
| 180266 |
+
"step": 63590
|
| 180267 |
+
},
|
| 180268 |
+
{
|
| 180269 |
+
"epoch": 508.0,
|
| 180270 |
+
"eval_loss": 0.38590699434280396,
|
| 180271 |
+
"eval_runtime": 39.7101,
|
| 180272 |
+
"eval_samples_per_second": 21.254,
|
| 180273 |
+
"eval_steps_per_second": 0.68,
|
| 180274 |
+
"eval_wer": 0.19914093164481966,
|
| 180275 |
+
"step": 63590
|
| 180276 |
+
},
|
| 180277 |
+
{
|
| 180278 |
+
"epoch": 512.04,
|
| 180279 |
+
"learning_rate": 8.997259615384616e-06,
|
| 180280 |
+
"loss": 0.3108,
|
| 180281 |
+
"step": 63595
|
| 180282 |
+
},
|
| 180283 |
+
{
|
| 180284 |
+
"epoch": 512.08,
|
| 180285 |
+
"learning_rate": 8.997179487179489e-06,
|
| 180286 |
+
"loss": 0.3144,
|
| 180287 |
+
"step": 63600
|
| 180288 |
+
},
|
| 180289 |
+
{
|
| 180290 |
+
"epoch": 512.12,
|
| 180291 |
+
"learning_rate": 8.99709935897436e-06,
|
| 180292 |
+
"loss": 0.3261,
|
| 180293 |
+
"step": 63605
|
| 180294 |
+
},
|
| 180295 |
+
{
|
| 180296 |
+
"epoch": 512.16,
|
| 180297 |
+
"learning_rate": 8.997019230769231e-06,
|
| 180298 |
+
"loss": 0.4621,
|
| 180299 |
+
"step": 63610
|
| 180300 |
+
},
|
| 180301 |
+
{
|
| 180302 |
+
"epoch": 512.2,
|
| 180303 |
+
"learning_rate": 8.996939102564103e-06,
|
| 180304 |
+
"loss": 1.4322,
|
| 180305 |
+
"step": 63615
|
| 180306 |
+
},
|
| 180307 |
+
{
|
| 180308 |
+
"epoch": 512.24,
|
| 180309 |
+
"learning_rate": 8.996858974358976e-06,
|
| 180310 |
+
"loss": 0.3976,
|
| 180311 |
+
"step": 63620
|
| 180312 |
+
},
|
| 180313 |
+
{
|
| 180314 |
+
"epoch": 512.28,
|
| 180315 |
+
"learning_rate": 8.996778846153847e-06,
|
| 180316 |
+
"loss": 0.3198,
|
| 180317 |
+
"step": 63625
|
| 180318 |
+
},
|
| 180319 |
+
{
|
| 180320 |
+
"epoch": 512.32,
|
| 180321 |
+
"learning_rate": 8.996698717948718e-06,
|
| 180322 |
+
"loss": 0.3835,
|
| 180323 |
+
"step": 63630
|
| 180324 |
+
},
|
| 180325 |
+
{
|
| 180326 |
+
"epoch": 512.36,
|
| 180327 |
+
"learning_rate": 8.996618589743592e-06,
|
| 180328 |
+
"loss": 0.505,
|
| 180329 |
+
"step": 63635
|
| 180330 |
+
},
|
| 180331 |
+
{
|
| 180332 |
+
"epoch": 512.4,
|
| 180333 |
+
"learning_rate": 8.996538461538463e-06,
|
| 180334 |
+
"loss": 1.2777,
|
| 180335 |
+
"step": 63640
|
| 180336 |
+
},
|
| 180337 |
+
{
|
| 180338 |
+
"epoch": 512.44,
|
| 180339 |
+
"learning_rate": 8.996458333333334e-06,
|
| 180340 |
+
"loss": 0.4016,
|
| 180341 |
+
"step": 63645
|
| 180342 |
+
},
|
| 180343 |
+
{
|
| 180344 |
+
"epoch": 512.48,
|
| 180345 |
+
"learning_rate": 8.996378205128206e-06,
|
| 180346 |
+
"loss": 0.3141,
|
| 180347 |
+
"step": 63650
|
| 180348 |
+
},
|
| 180349 |
+
{
|
| 180350 |
+
"epoch": 512.52,
|
| 180351 |
+
"learning_rate": 8.996298076923079e-06,
|
| 180352 |
+
"loss": 0.4872,
|
| 180353 |
+
"step": 63655
|
| 180354 |
+
},
|
| 180355 |
+
{
|
| 180356 |
+
"epoch": 512.56,
|
| 180357 |
+
"learning_rate": 8.996217948717948e-06,
|
| 180358 |
+
"loss": 0.5261,
|
| 180359 |
+
"step": 63660
|
| 180360 |
+
},
|
| 180361 |
+
{
|
| 180362 |
+
"epoch": 512.6,
|
| 180363 |
+
"learning_rate": 8.996137820512821e-06,
|
| 180364 |
+
"loss": 1.2183,
|
| 180365 |
+
"step": 63665
|
| 180366 |
+
},
|
| 180367 |
+
{
|
| 180368 |
+
"epoch": 512.64,
|
| 180369 |
+
"learning_rate": 8.996057692307693e-06,
|
| 180370 |
+
"loss": 0.3727,
|
| 180371 |
+
"step": 63670
|
| 180372 |
+
},
|
| 180373 |
+
{
|
| 180374 |
+
"epoch": 512.68,
|
| 180375 |
+
"learning_rate": 8.995977564102564e-06,
|
| 180376 |
+
"loss": 0.3206,
|
| 180377 |
+
"step": 63675
|
| 180378 |
+
},
|
| 180379 |
+
{
|
| 180380 |
+
"epoch": 512.72,
|
| 180381 |
+
"learning_rate": 8.995897435897437e-06,
|
| 180382 |
+
"loss": 0.3649,
|
| 180383 |
+
"step": 63680
|
| 180384 |
+
},
|
| 180385 |
+
{
|
| 180386 |
+
"epoch": 512.76,
|
| 180387 |
+
"learning_rate": 8.995817307692309e-06,
|
| 180388 |
+
"loss": 0.4285,
|
| 180389 |
+
"step": 63685
|
| 180390 |
+
},
|
| 180391 |
+
{
|
| 180392 |
+
"epoch": 512.8,
|
| 180393 |
+
"learning_rate": 8.99573717948718e-06,
|
| 180394 |
+
"loss": 1.2287,
|
| 180395 |
+
"step": 63690
|
| 180396 |
+
},
|
| 180397 |
+
{
|
| 180398 |
+
"epoch": 512.84,
|
| 180399 |
+
"learning_rate": 8.995657051282051e-06,
|
| 180400 |
+
"loss": 0.3272,
|
| 180401 |
+
"step": 63695
|
| 180402 |
+
},
|
| 180403 |
+
{
|
| 180404 |
+
"epoch": 512.88,
|
| 180405 |
+
"learning_rate": 8.995576923076924e-06,
|
| 180406 |
+
"loss": 0.2947,
|
| 180407 |
+
"step": 63700
|
| 180408 |
+
},
|
| 180409 |
+
{
|
| 180410 |
+
"epoch": 512.92,
|
| 180411 |
+
"learning_rate": 8.995496794871796e-06,
|
| 180412 |
+
"loss": 0.3511,
|
| 180413 |
+
"step": 63705
|
| 180414 |
+
},
|
| 180415 |
+
{
|
| 180416 |
+
"epoch": 512.96,
|
| 180417 |
+
"learning_rate": 8.995416666666667e-06,
|
| 180418 |
+
"loss": 0.5311,
|
| 180419 |
+
"step": 63710
|
| 180420 |
+
},
|
| 180421 |
+
{
|
| 180422 |
+
"epoch": 513.0,
|
| 180423 |
+
"eval_loss": 0.3599279820919037,
|
| 180424 |
+
"eval_runtime": 39.9542,
|
| 180425 |
+
"eval_samples_per_second": 21.124,
|
| 180426 |
+
"eval_steps_per_second": 0.676,
|
| 180427 |
+
"eval_wer": 0.1937600233270156,
|
| 180428 |
+
"step": 63714
|
| 180429 |
+
},
|
| 180430 |
+
{
|
| 180431 |
+
"epoch": 513.01,
|
| 180432 |
+
"learning_rate": 8.995336538461538e-06,
|
| 180433 |
+
"loss": 0.3977,
|
| 180434 |
+
"step": 63715
|
| 180435 |
+
},
|
| 180436 |
+
{
|
| 180437 |
+
"epoch": 513.05,
|
| 180438 |
+
"learning_rate": 8.995256410256411e-06,
|
| 180439 |
+
"loss": 0.3816,
|
| 180440 |
+
"step": 63720
|
| 180441 |
+
},
|
| 180442 |
+
{
|
| 180443 |
+
"epoch": 513.09,
|
| 180444 |
+
"learning_rate": 8.995176282051283e-06,
|
| 180445 |
+
"loss": 0.3246,
|
| 180446 |
+
"step": 63725
|
| 180447 |
+
},
|
| 180448 |
+
{
|
| 180449 |
+
"epoch": 513.13,
|
| 180450 |
+
"learning_rate": 8.995096153846154e-06,
|
| 180451 |
+
"loss": 0.334,
|
| 180452 |
+
"step": 63730
|
| 180453 |
+
},
|
| 180454 |
+
{
|
| 180455 |
+
"epoch": 513.17,
|
| 180456 |
+
"learning_rate": 8.995016025641027e-06,
|
| 180457 |
+
"loss": 0.6139,
|
| 180458 |
+
"step": 63735
|
| 180459 |
+
},
|
| 180460 |
+
{
|
| 180461 |
+
"epoch": 513.21,
|
| 180462 |
+
"learning_rate": 8.994935897435899e-06,
|
| 180463 |
+
"loss": 1.1952,
|
| 180464 |
+
"step": 63740
|
| 180465 |
+
},
|
| 180466 |
+
{
|
| 180467 |
+
"epoch": 513.25,
|
| 180468 |
+
"learning_rate": 8.99485576923077e-06,
|
| 180469 |
+
"loss": 0.2921,
|
| 180470 |
+
"step": 63745
|
| 180471 |
+
},
|
| 180472 |
+
{
|
| 180473 |
+
"epoch": 513.29,
|
| 180474 |
+
"learning_rate": 8.994775641025641e-06,
|
| 180475 |
+
"loss": 0.3822,
|
| 180476 |
+
"step": 63750
|
| 180477 |
+
},
|
| 180478 |
+
{
|
| 180479 |
+
"epoch": 513.33,
|
| 180480 |
+
"learning_rate": 8.994695512820514e-06,
|
| 180481 |
+
"loss": 0.4243,
|
| 180482 |
+
"step": 63755
|
| 180483 |
+
},
|
| 180484 |
+
{
|
| 180485 |
+
"epoch": 513.37,
|
| 180486 |
+
"learning_rate": 8.994615384615386e-06,
|
| 180487 |
+
"loss": 0.5498,
|
| 180488 |
+
"step": 63760
|
| 180489 |
+
},
|
| 180490 |
+
{
|
| 180491 |
+
"epoch": 513.41,
|
| 180492 |
+
"learning_rate": 8.994535256410257e-06,
|
| 180493 |
+
"loss": 1.2928,
|
| 180494 |
+
"step": 63765
|
| 180495 |
+
},
|
| 180496 |
+
{
|
| 180497 |
+
"epoch": 513.45,
|
| 180498 |
+
"learning_rate": 8.994455128205128e-06,
|
| 180499 |
+
"loss": 0.3345,
|
| 180500 |
+
"step": 63770
|
| 180501 |
+
},
|
| 180502 |
+
{
|
| 180503 |
+
"epoch": 513.49,
|
| 180504 |
+
"learning_rate": 8.994375000000001e-06,
|
| 180505 |
+
"loss": 0.3265,
|
| 180506 |
+
"step": 63775
|
| 180507 |
+
},
|
| 180508 |
+
{
|
| 180509 |
+
"epoch": 513.53,
|
| 180510 |
+
"learning_rate": 8.994294871794873e-06,
|
| 180511 |
+
"loss": 0.3152,
|
| 180512 |
+
"step": 63780
|
| 180513 |
+
},
|
| 180514 |
+
{
|
| 180515 |
+
"epoch": 513.57,
|
| 180516 |
+
"learning_rate": 8.994214743589744e-06,
|
| 180517 |
+
"loss": 0.6282,
|
| 180518 |
+
"step": 63785
|
| 180519 |
+
},
|
| 180520 |
+
{
|
| 180521 |
+
"epoch": 513.61,
|
| 180522 |
+
"learning_rate": 8.994134615384617e-06,
|
| 180523 |
+
"loss": 1.2397,
|
| 180524 |
+
"step": 63790
|
| 180525 |
+
},
|
| 180526 |
+
{
|
| 180527 |
+
"epoch": 513.65,
|
| 180528 |
+
"learning_rate": 8.994054487179487e-06,
|
| 180529 |
+
"loss": 0.3037,
|
| 180530 |
+
"step": 63795
|
| 180531 |
+
},
|
| 180532 |
+
{
|
| 180533 |
+
"epoch": 513.69,
|
| 180534 |
+
"learning_rate": 8.99397435897436e-06,
|
| 180535 |
+
"loss": 0.2941,
|
| 180536 |
+
"step": 63800
|
| 180537 |
+
},
|
| 180538 |
+
{
|
| 180539 |
+
"epoch": 513.73,
|
| 180540 |
+
"learning_rate": 8.993894230769231e-06,
|
| 180541 |
+
"loss": 0.319,
|
| 180542 |
+
"step": 63805
|
| 180543 |
+
},
|
| 180544 |
+
{
|
| 180545 |
+
"epoch": 513.77,
|
| 180546 |
+
"learning_rate": 8.993814102564103e-06,
|
| 180547 |
+
"loss": 0.6175,
|
| 180548 |
+
"step": 63810
|
| 180549 |
+
},
|
| 180550 |
+
{
|
| 180551 |
+
"epoch": 513.81,
|
| 180552 |
+
"learning_rate": 8.993733974358974e-06,
|
| 180553 |
+
"loss": 1.1451,
|
| 180554 |
+
"step": 63815
|
| 180555 |
+
},
|
| 180556 |
+
{
|
| 180557 |
+
"epoch": 513.85,
|
| 180558 |
+
"learning_rate": 8.993653846153847e-06,
|
| 180559 |
+
"loss": 0.345,
|
| 180560 |
+
"step": 63820
|
| 180561 |
+
},
|
| 180562 |
+
{
|
| 180563 |
+
"epoch": 513.89,
|
| 180564 |
+
"learning_rate": 8.993573717948718e-06,
|
| 180565 |
+
"loss": 0.3264,
|
| 180566 |
+
"step": 63825
|
| 180567 |
+
},
|
| 180568 |
+
{
|
| 180569 |
+
"epoch": 513.93,
|
| 180570 |
+
"learning_rate": 8.99349358974359e-06,
|
| 180571 |
+
"loss": 0.4052,
|
| 180572 |
+
"step": 63830
|
| 180573 |
+
},
|
| 180574 |
+
{
|
| 180575 |
+
"epoch": 513.97,
|
| 180576 |
+
"learning_rate": 8.993413461538463e-06,
|
| 180577 |
+
"loss": 0.6459,
|
| 180578 |
+
"step": 63835
|
| 180579 |
+
},
|
| 180580 |
+
{
|
| 180581 |
+
"epoch": 514.0,
|
| 180582 |
+
"eval_loss": 0.42560651898384094,
|
| 180583 |
+
"eval_runtime": 39.1375,
|
| 180584 |
+
"eval_samples_per_second": 21.565,
|
| 180585 |
+
"eval_steps_per_second": 0.69,
|
| 180586 |
+
"eval_wer": 0.20920346062052506,
|
| 180587 |
+
"step": 63838
|
| 180588 |
}
|
| 180589 |
],
|
| 180590 |
+
"max_steps": 620000,
|
| 180591 |
"num_train_epochs": 5000,
|
| 180592 |
+
"total_flos": 1.7963900765073526e+20,
|
| 180593 |
"trial_name": null,
|
| 180594 |
"trial_params": null
|
| 180595 |
}
|
model-bin/finetune/base/{checkpoint-63215 → checkpoint-63838}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629814655.6609554/events.out.tfevents.1629814655.c435e1c5ee04.920.151
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1ff3d063d92adc54f07e1359e045d7e9a1db8008da6b0de4c05eb63e2f10f426
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629815285.6311035/events.out.tfevents.1629815285.c435e1c5ee04.920.153
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:5e314cdaf681ec09249a6fcf60d1f0786a0ef9323ae6f9d3bd406c1ae6e64bc8
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629815928.140718/events.out.tfevents.1629815928.c435e1c5ee04.920.155
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f576748ecb27726186471936c0f6a66605c071eb90fc04c112e42b5f2bc50f67
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629816566.4764583/events.out.tfevents.1629816566.c435e1c5ee04.920.157
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b21482a1644352a021347dbb77895a37dc32081c02e6e7b720f647878564bf34
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629817202.370693/events.out.tfevents.1629817202.c435e1c5ee04.920.159
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:82c616e6021e1bc21cc7f122e984dc3d7d2e42b2439930623f23831718b5cc7e
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629814655.c435e1c5ee04.920.150
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:95f84faad25ee84e66d4ae93c0e4149d64f5a463f7b1ab2dbe54e063f68e7059
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629815285.c435e1c5ee04.920.152
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2af4bd2286435c41ce34b9c99ad93a8ad8fa9a7a1eba9e2dca6b306b1d4f6385
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629815928.c435e1c5ee04.920.154
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ca9e5348fe715a41d303c53d1b95d1e71b85b749174fad0996bba7f7e11c273d
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629816566.c435e1c5ee04.920.156
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:89319c6fafe33e17c0ad21f567c0ec4beea82c86caaba613f0c88bf7056e8f54
|
| 3 |
+
size 8462
|
model-bin/finetune/base/log/events.out.tfevents.1629817202.c435e1c5ee04.920.158
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e935e5f74af3f6cdaf1f1dd40ed3d83a45e709938171a76e72bd2b540cfd2847
|
| 3 |
+
size 8622
|