"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-112750 → checkpoint-113370}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-112750 → checkpoint-113370}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-112750 → checkpoint-113370}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-112750 → checkpoint-113370}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-112750 → checkpoint-113370}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-112750 → checkpoint-113370}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-112750 → checkpoint-113370}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-112750 → checkpoint-113370}/trainer_state.json +793 -4
- model-bin/finetune/base/{checkpoint-112750 → checkpoint-113370}/training_args.bin +0 -0
- model-bin/finetune/base/log/1630082884.969348/events.out.tfevents.1630082884.dea8124df033.1391.7 +3 -0
- model-bin/finetune/base/log/1630083307.1487231/events.out.tfevents.1630083307.dea8124df033.1391.9 +3 -0
- model-bin/finetune/base/log/1630083735.6256015/events.out.tfevents.1630083735.dea8124df033.1391.11 +3 -0
- model-bin/finetune/base/log/1630084174.1170754/events.out.tfevents.1630084174.dea8124df033.1391.13 +3 -0
- model-bin/finetune/base/log/1630084598.2761755/events.out.tfevents.1630084598.dea8124df033.1391.15 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630082884.dea8124df033.1391.6 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630083307.dea8124df033.1391.8 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630083735.dea8124df033.1391.10 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630084174.dea8124df033.1391.12 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630084598.dea8124df033.1391.14 +3 -0
model-bin/finetune/base/{checkpoint-112750 → checkpoint-113370}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-112750 → checkpoint-113370}/optimizer.pt
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 722165393
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1d348482d17ba9e738c11c659795d83ff2bcf3cfd11baf8797570f6c9b26411a
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-112750 → checkpoint-113370}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-112750 → checkpoint-113370}/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:cc2b6659e86d951e27ce212a257a0b916a94a565fd3d84e430bfde3718539cb4
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-112750 → checkpoint-113370}/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:df743e69dcda2e4cf4a9f3f8ef82556dde39a400b2c9ae8b01c8bdc7894d03fb
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-112750 → checkpoint-113370}/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 sha256:7fe07346d923a1da088bea107f5ab09f94c498e2865fabad8db6fc04e90f2d85
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-112750 → checkpoint-113370}/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 sha256:1905b108e5460ced8e245231c20dc93cf00ce38aab3458596eda343809ebdd48
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-112750 → checkpoint-113370}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1743826049391605,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-101551",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -242820,11 +242820,800 @@
|
|
| 242820 |
"eval_steps_per_second": 0.674,
|
| 242821 |
"eval_wer": 0.19159501927067893,
|
| 242822 |
"step": 112750
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 242823 |
}
|
| 242824 |
],
|
| 242825 |
-
"max_steps":
|
| 242826 |
"num_train_epochs": 5000,
|
| 242827 |
-
"total_flos": 3.
|
| 242828 |
"trial_name": null,
|
| 242829 |
"trial_params": null
|
| 242830 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1743826049391605,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-101551",
|
| 4 |
+
"epoch": 913.995983935743,
|
| 5 |
+
"global_step": 113370,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 242820 |
"eval_steps_per_second": 0.674,
|
| 242821 |
"eval_wer": 0.19159501927067893,
|
| 242822 |
"step": 112750
|
| 242823 |
+
},
|
| 242824 |
+
{
|
| 242825 |
+
"epoch": 909.04,
|
| 242826 |
+
"learning_rate": 8.19529886914378e-06,
|
| 242827 |
+
"loss": 0.3194,
|
| 242828 |
+
"step": 112755
|
| 242829 |
+
},
|
| 242830 |
+
{
|
| 242831 |
+
"epoch": 909.08,
|
| 242832 |
+
"learning_rate": 8.195218093699516e-06,
|
| 242833 |
+
"loss": 0.2594,
|
| 242834 |
+
"step": 112760
|
| 242835 |
+
},
|
| 242836 |
+
{
|
| 242837 |
+
"epoch": 909.12,
|
| 242838 |
+
"learning_rate": 8.19513731825525e-06,
|
| 242839 |
+
"loss": 0.323,
|
| 242840 |
+
"step": 112765
|
| 242841 |
+
},
|
| 242842 |
+
{
|
| 242843 |
+
"epoch": 909.16,
|
| 242844 |
+
"learning_rate": 8.195056542810986e-06,
|
| 242845 |
+
"loss": 0.3869,
|
| 242846 |
+
"step": 112770
|
| 242847 |
+
},
|
| 242848 |
+
{
|
| 242849 |
+
"epoch": 909.2,
|
| 242850 |
+
"learning_rate": 8.19497576736672e-06,
|
| 242851 |
+
"loss": 1.2147,
|
| 242852 |
+
"step": 112775
|
| 242853 |
+
},
|
| 242854 |
+
{
|
| 242855 |
+
"epoch": 909.24,
|
| 242856 |
+
"learning_rate": 8.194894991922456e-06,
|
| 242857 |
+
"loss": 0.2554,
|
| 242858 |
+
"step": 112780
|
| 242859 |
+
},
|
| 242860 |
+
{
|
| 242861 |
+
"epoch": 909.28,
|
| 242862 |
+
"learning_rate": 8.19481421647819e-06,
|
| 242863 |
+
"loss": 0.2766,
|
| 242864 |
+
"step": 112785
|
| 242865 |
+
},
|
| 242866 |
+
{
|
| 242867 |
+
"epoch": 909.32,
|
| 242868 |
+
"learning_rate": 8.194733441033926e-06,
|
| 242869 |
+
"loss": 0.3144,
|
| 242870 |
+
"step": 112790
|
| 242871 |
+
},
|
| 242872 |
+
{
|
| 242873 |
+
"epoch": 909.36,
|
| 242874 |
+
"learning_rate": 8.194652665589662e-06,
|
| 242875 |
+
"loss": 0.4586,
|
| 242876 |
+
"step": 112795
|
| 242877 |
+
},
|
| 242878 |
+
{
|
| 242879 |
+
"epoch": 909.4,
|
| 242880 |
+
"learning_rate": 8.194571890145396e-06,
|
| 242881 |
+
"loss": 1.1297,
|
| 242882 |
+
"step": 112800
|
| 242883 |
+
},
|
| 242884 |
+
{
|
| 242885 |
+
"epoch": 909.44,
|
| 242886 |
+
"learning_rate": 8.194491114701132e-06,
|
| 242887 |
+
"loss": 0.3203,
|
| 242888 |
+
"step": 112805
|
| 242889 |
+
},
|
| 242890 |
+
{
|
| 242891 |
+
"epoch": 909.48,
|
| 242892 |
+
"learning_rate": 8.194410339256866e-06,
|
| 242893 |
+
"loss": 0.2947,
|
| 242894 |
+
"step": 112810
|
| 242895 |
+
},
|
| 242896 |
+
{
|
| 242897 |
+
"epoch": 909.52,
|
| 242898 |
+
"learning_rate": 8.194329563812602e-06,
|
| 242899 |
+
"loss": 0.3166,
|
| 242900 |
+
"step": 112815
|
| 242901 |
+
},
|
| 242902 |
+
{
|
| 242903 |
+
"epoch": 909.56,
|
| 242904 |
+
"learning_rate": 8.194248788368336e-06,
|
| 242905 |
+
"loss": 0.4555,
|
| 242906 |
+
"step": 112820
|
| 242907 |
+
},
|
| 242908 |
+
{
|
| 242909 |
+
"epoch": 909.6,
|
| 242910 |
+
"learning_rate": 8.194168012924072e-06,
|
| 242911 |
+
"loss": 1.2846,
|
| 242912 |
+
"step": 112825
|
| 242913 |
+
},
|
| 242914 |
+
{
|
| 242915 |
+
"epoch": 909.64,
|
| 242916 |
+
"learning_rate": 8.194087237479806e-06,
|
| 242917 |
+
"loss": 0.3921,
|
| 242918 |
+
"step": 112830
|
| 242919 |
+
},
|
| 242920 |
+
{
|
| 242921 |
+
"epoch": 909.68,
|
| 242922 |
+
"learning_rate": 8.194006462035542e-06,
|
| 242923 |
+
"loss": 0.3258,
|
| 242924 |
+
"step": 112835
|
| 242925 |
+
},
|
| 242926 |
+
{
|
| 242927 |
+
"epoch": 909.72,
|
| 242928 |
+
"learning_rate": 8.193925686591276e-06,
|
| 242929 |
+
"loss": 0.3415,
|
| 242930 |
+
"step": 112840
|
| 242931 |
+
},
|
| 242932 |
+
{
|
| 242933 |
+
"epoch": 909.76,
|
| 242934 |
+
"learning_rate": 8.193844911147012e-06,
|
| 242935 |
+
"loss": 0.4425,
|
| 242936 |
+
"step": 112845
|
| 242937 |
+
},
|
| 242938 |
+
{
|
| 242939 |
+
"epoch": 909.8,
|
| 242940 |
+
"learning_rate": 8.193764135702748e-06,
|
| 242941 |
+
"loss": 1.1291,
|
| 242942 |
+
"step": 112850
|
| 242943 |
+
},
|
| 242944 |
+
{
|
| 242945 |
+
"epoch": 909.84,
|
| 242946 |
+
"learning_rate": 8.193683360258482e-06,
|
| 242947 |
+
"loss": 0.3087,
|
| 242948 |
+
"step": 112855
|
| 242949 |
+
},
|
| 242950 |
+
{
|
| 242951 |
+
"epoch": 909.88,
|
| 242952 |
+
"learning_rate": 8.193602584814218e-06,
|
| 242953 |
+
"loss": 0.3314,
|
| 242954 |
+
"step": 112860
|
| 242955 |
+
},
|
| 242956 |
+
{
|
| 242957 |
+
"epoch": 909.92,
|
| 242958 |
+
"learning_rate": 8.193521809369952e-06,
|
| 242959 |
+
"loss": 0.329,
|
| 242960 |
+
"step": 112865
|
| 242961 |
+
},
|
| 242962 |
+
{
|
| 242963 |
+
"epoch": 909.96,
|
| 242964 |
+
"learning_rate": 8.193441033925688e-06,
|
| 242965 |
+
"loss": 0.4559,
|
| 242966 |
+
"step": 112870
|
| 242967 |
+
},
|
| 242968 |
+
{
|
| 242969 |
+
"epoch": 910.0,
|
| 242970 |
+
"eval_loss": 0.3853907585144043,
|
| 242971 |
+
"eval_runtime": 39.1333,
|
| 242972 |
+
"eval_samples_per_second": 21.363,
|
| 242973 |
+
"eval_steps_per_second": 0.69,
|
| 242974 |
+
"eval_wer": 0.18328861061419202,
|
| 242975 |
+
"step": 112874
|
| 242976 |
+
},
|
| 242977 |
+
{
|
| 242978 |
+
"epoch": 910.01,
|
| 242979 |
+
"learning_rate": 8.193360258481422e-06,
|
| 242980 |
+
"loss": 0.3418,
|
| 242981 |
+
"step": 112875
|
| 242982 |
+
},
|
| 242983 |
+
{
|
| 242984 |
+
"epoch": 910.05,
|
| 242985 |
+
"learning_rate": 8.193279483037158e-06,
|
| 242986 |
+
"loss": 0.3218,
|
| 242987 |
+
"step": 112880
|
| 242988 |
+
},
|
| 242989 |
+
{
|
| 242990 |
+
"epoch": 910.09,
|
| 242991 |
+
"learning_rate": 8.193198707592892e-06,
|
| 242992 |
+
"loss": 0.2755,
|
| 242993 |
+
"step": 112885
|
| 242994 |
+
},
|
| 242995 |
+
{
|
| 242996 |
+
"epoch": 910.13,
|
| 242997 |
+
"learning_rate": 8.193117932148628e-06,
|
| 242998 |
+
"loss": 0.3143,
|
| 242999 |
+
"step": 112890
|
| 243000 |
+
},
|
| 243001 |
+
{
|
| 243002 |
+
"epoch": 910.17,
|
| 243003 |
+
"learning_rate": 8.193037156704362e-06,
|
| 243004 |
+
"loss": 0.4879,
|
| 243005 |
+
"step": 112895
|
| 243006 |
+
},
|
| 243007 |
+
{
|
| 243008 |
+
"epoch": 910.21,
|
| 243009 |
+
"learning_rate": 8.192956381260098e-06,
|
| 243010 |
+
"loss": 1.1204,
|
| 243011 |
+
"step": 112900
|
| 243012 |
+
},
|
| 243013 |
+
{
|
| 243014 |
+
"epoch": 910.25,
|
| 243015 |
+
"learning_rate": 8.192875605815834e-06,
|
| 243016 |
+
"loss": 0.3345,
|
| 243017 |
+
"step": 112905
|
| 243018 |
+
},
|
| 243019 |
+
{
|
| 243020 |
+
"epoch": 910.29,
|
| 243021 |
+
"learning_rate": 8.192794830371568e-06,
|
| 243022 |
+
"loss": 0.2979,
|
| 243023 |
+
"step": 112910
|
| 243024 |
+
},
|
| 243025 |
+
{
|
| 243026 |
+
"epoch": 910.33,
|
| 243027 |
+
"learning_rate": 8.192714054927304e-06,
|
| 243028 |
+
"loss": 0.3332,
|
| 243029 |
+
"step": 112915
|
| 243030 |
+
},
|
| 243031 |
+
{
|
| 243032 |
+
"epoch": 910.37,
|
| 243033 |
+
"learning_rate": 8.192633279483038e-06,
|
| 243034 |
+
"loss": 0.5393,
|
| 243035 |
+
"step": 112920
|
| 243036 |
+
},
|
| 243037 |
+
{
|
| 243038 |
+
"epoch": 910.41,
|
| 243039 |
+
"learning_rate": 8.192552504038774e-06,
|
| 243040 |
+
"loss": 1.2886,
|
| 243041 |
+
"step": 112925
|
| 243042 |
+
},
|
| 243043 |
+
{
|
| 243044 |
+
"epoch": 910.45,
|
| 243045 |
+
"learning_rate": 8.192471728594508e-06,
|
| 243046 |
+
"loss": 0.3061,
|
| 243047 |
+
"step": 112930
|
| 243048 |
+
},
|
| 243049 |
+
{
|
| 243050 |
+
"epoch": 910.49,
|
| 243051 |
+
"learning_rate": 8.192390953150244e-06,
|
| 243052 |
+
"loss": 0.2628,
|
| 243053 |
+
"step": 112935
|
| 243054 |
+
},
|
| 243055 |
+
{
|
| 243056 |
+
"epoch": 910.53,
|
| 243057 |
+
"learning_rate": 8.192310177705978e-06,
|
| 243058 |
+
"loss": 0.2704,
|
| 243059 |
+
"step": 112940
|
| 243060 |
+
},
|
| 243061 |
+
{
|
| 243062 |
+
"epoch": 910.57,
|
| 243063 |
+
"learning_rate": 8.192229402261714e-06,
|
| 243064 |
+
"loss": 0.5995,
|
| 243065 |
+
"step": 112945
|
| 243066 |
+
},
|
| 243067 |
+
{
|
| 243068 |
+
"epoch": 910.61,
|
| 243069 |
+
"learning_rate": 8.192148626817448e-06,
|
| 243070 |
+
"loss": 1.1335,
|
| 243071 |
+
"step": 112950
|
| 243072 |
+
},
|
| 243073 |
+
{
|
| 243074 |
+
"epoch": 910.65,
|
| 243075 |
+
"learning_rate": 8.192067851373183e-06,
|
| 243076 |
+
"loss": 0.3048,
|
| 243077 |
+
"step": 112955
|
| 243078 |
+
},
|
| 243079 |
+
{
|
| 243080 |
+
"epoch": 910.69,
|
| 243081 |
+
"learning_rate": 8.191987075928918e-06,
|
| 243082 |
+
"loss": 0.3209,
|
| 243083 |
+
"step": 112960
|
| 243084 |
+
},
|
| 243085 |
+
{
|
| 243086 |
+
"epoch": 910.73,
|
| 243087 |
+
"learning_rate": 8.191906300484653e-06,
|
| 243088 |
+
"loss": 0.3437,
|
| 243089 |
+
"step": 112965
|
| 243090 |
+
},
|
| 243091 |
+
{
|
| 243092 |
+
"epoch": 910.77,
|
| 243093 |
+
"learning_rate": 8.19182552504039e-06,
|
| 243094 |
+
"loss": 0.6269,
|
| 243095 |
+
"step": 112970
|
| 243096 |
+
},
|
| 243097 |
+
{
|
| 243098 |
+
"epoch": 910.81,
|
| 243099 |
+
"learning_rate": 8.191744749596123e-06,
|
| 243100 |
+
"loss": 1.0891,
|
| 243101 |
+
"step": 112975
|
| 243102 |
+
},
|
| 243103 |
+
{
|
| 243104 |
+
"epoch": 910.85,
|
| 243105 |
+
"learning_rate": 8.19166397415186e-06,
|
| 243106 |
+
"loss": 0.3427,
|
| 243107 |
+
"step": 112980
|
| 243108 |
+
},
|
| 243109 |
+
{
|
| 243110 |
+
"epoch": 910.89,
|
| 243111 |
+
"learning_rate": 8.191583198707593e-06,
|
| 243112 |
+
"loss": 0.2962,
|
| 243113 |
+
"step": 112985
|
| 243114 |
+
},
|
| 243115 |
+
{
|
| 243116 |
+
"epoch": 910.93,
|
| 243117 |
+
"learning_rate": 8.19150242326333e-06,
|
| 243118 |
+
"loss": 0.3852,
|
| 243119 |
+
"step": 112990
|
| 243120 |
+
},
|
| 243121 |
+
{
|
| 243122 |
+
"epoch": 910.97,
|
| 243123 |
+
"learning_rate": 8.191421647819063e-06,
|
| 243124 |
+
"loss": 0.663,
|
| 243125 |
+
"step": 112995
|
| 243126 |
+
},
|
| 243127 |
+
{
|
| 243128 |
+
"epoch": 911.0,
|
| 243129 |
+
"eval_loss": 0.3766191005706787,
|
| 243130 |
+
"eval_runtime": 41.0289,
|
| 243131 |
+
"eval_samples_per_second": 20.376,
|
| 243132 |
+
"eval_steps_per_second": 0.658,
|
| 243133 |
+
"eval_wer": 0.18356084809758932,
|
| 243134 |
+
"step": 112998
|
| 243135 |
+
},
|
| 243136 |
+
{
|
| 243137 |
+
"epoch": 911.02,
|
| 243138 |
+
"learning_rate": 8.1913408723748e-06,
|
| 243139 |
+
"loss": 0.3573,
|
| 243140 |
+
"step": 113000
|
| 243141 |
+
},
|
| 243142 |
+
{
|
| 243143 |
+
"epoch": 911.06,
|
| 243144 |
+
"learning_rate": 8.191260096930533e-06,
|
| 243145 |
+
"loss": 0.2916,
|
| 243146 |
+
"step": 113005
|
| 243147 |
+
},
|
| 243148 |
+
{
|
| 243149 |
+
"epoch": 911.1,
|
| 243150 |
+
"learning_rate": 8.19117932148627e-06,
|
| 243151 |
+
"loss": 0.2908,
|
| 243152 |
+
"step": 113010
|
| 243153 |
+
},
|
| 243154 |
+
{
|
| 243155 |
+
"epoch": 911.14,
|
| 243156 |
+
"learning_rate": 8.191098546042003e-06,
|
| 243157 |
+
"loss": 0.322,
|
| 243158 |
+
"step": 113015
|
| 243159 |
+
},
|
| 243160 |
+
{
|
| 243161 |
+
"epoch": 911.18,
|
| 243162 |
+
"learning_rate": 8.19101777059774e-06,
|
| 243163 |
+
"loss": 0.5831,
|
| 243164 |
+
"step": 113020
|
| 243165 |
+
},
|
| 243166 |
+
{
|
| 243167 |
+
"epoch": 911.22,
|
| 243168 |
+
"learning_rate": 8.190936995153475e-06,
|
| 243169 |
+
"loss": 0.9917,
|
| 243170 |
+
"step": 113025
|
| 243171 |
+
},
|
| 243172 |
+
{
|
| 243173 |
+
"epoch": 911.26,
|
| 243174 |
+
"learning_rate": 8.19085621970921e-06,
|
| 243175 |
+
"loss": 0.2945,
|
| 243176 |
+
"step": 113030
|
| 243177 |
+
},
|
| 243178 |
+
{
|
| 243179 |
+
"epoch": 911.3,
|
| 243180 |
+
"learning_rate": 8.190775444264945e-06,
|
| 243181 |
+
"loss": 0.2512,
|
| 243182 |
+
"step": 113035
|
| 243183 |
+
},
|
| 243184 |
+
{
|
| 243185 |
+
"epoch": 911.34,
|
| 243186 |
+
"learning_rate": 8.190694668820679e-06,
|
| 243187 |
+
"loss": 0.3935,
|
| 243188 |
+
"step": 113040
|
| 243189 |
+
},
|
| 243190 |
+
{
|
| 243191 |
+
"epoch": 911.38,
|
| 243192 |
+
"learning_rate": 8.190613893376415e-06,
|
| 243193 |
+
"loss": 0.6559,
|
| 243194 |
+
"step": 113045
|
| 243195 |
+
},
|
| 243196 |
+
{
|
| 243197 |
+
"epoch": 911.42,
|
| 243198 |
+
"learning_rate": 8.190533117932149e-06,
|
| 243199 |
+
"loss": 1.014,
|
| 243200 |
+
"step": 113050
|
| 243201 |
+
},
|
| 243202 |
+
{
|
| 243203 |
+
"epoch": 911.46,
|
| 243204 |
+
"learning_rate": 8.190452342487885e-06,
|
| 243205 |
+
"loss": 0.3467,
|
| 243206 |
+
"step": 113055
|
| 243207 |
+
},
|
| 243208 |
+
{
|
| 243209 |
+
"epoch": 911.5,
|
| 243210 |
+
"learning_rate": 8.190371567043619e-06,
|
| 243211 |
+
"loss": 0.3123,
|
| 243212 |
+
"step": 113060
|
| 243213 |
+
},
|
| 243214 |
+
{
|
| 243215 |
+
"epoch": 911.54,
|
| 243216 |
+
"learning_rate": 8.190290791599355e-06,
|
| 243217 |
+
"loss": 0.4067,
|
| 243218 |
+
"step": 113065
|
| 243219 |
+
},
|
| 243220 |
+
{
|
| 243221 |
+
"epoch": 911.58,
|
| 243222 |
+
"learning_rate": 8.190210016155089e-06,
|
| 243223 |
+
"loss": 0.6399,
|
| 243224 |
+
"step": 113070
|
| 243225 |
+
},
|
| 243226 |
+
{
|
| 243227 |
+
"epoch": 911.62,
|
| 243228 |
+
"learning_rate": 8.190129240710825e-06,
|
| 243229 |
+
"loss": 0.9487,
|
| 243230 |
+
"step": 113075
|
| 243231 |
+
},
|
| 243232 |
+
{
|
| 243233 |
+
"epoch": 911.66,
|
| 243234 |
+
"learning_rate": 8.19004846526656e-06,
|
| 243235 |
+
"loss": 0.2608,
|
| 243236 |
+
"step": 113080
|
| 243237 |
+
},
|
| 243238 |
+
{
|
| 243239 |
+
"epoch": 911.7,
|
| 243240 |
+
"learning_rate": 8.189967689822295e-06,
|
| 243241 |
+
"loss": 0.229,
|
| 243242 |
+
"step": 113085
|
| 243243 |
+
},
|
| 243244 |
+
{
|
| 243245 |
+
"epoch": 911.74,
|
| 243246 |
+
"learning_rate": 8.18988691437803e-06,
|
| 243247 |
+
"loss": 0.3772,
|
| 243248 |
+
"step": 113090
|
| 243249 |
+
},
|
| 243250 |
+
{
|
| 243251 |
+
"epoch": 911.78,
|
| 243252 |
+
"learning_rate": 8.189806138933765e-06,
|
| 243253 |
+
"loss": 0.6131,
|
| 243254 |
+
"step": 113095
|
| 243255 |
+
},
|
| 243256 |
+
{
|
| 243257 |
+
"epoch": 911.82,
|
| 243258 |
+
"learning_rate": 8.189741518578352e-06,
|
| 243259 |
+
"loss": 1.0935,
|
| 243260 |
+
"step": 113100
|
| 243261 |
+
},
|
| 243262 |
+
{
|
| 243263 |
+
"epoch": 911.86,
|
| 243264 |
+
"learning_rate": 8.189660743134088e-06,
|
| 243265 |
+
"loss": 0.3042,
|
| 243266 |
+
"step": 113105
|
| 243267 |
+
},
|
| 243268 |
+
{
|
| 243269 |
+
"epoch": 911.9,
|
| 243270 |
+
"learning_rate": 8.189579967689822e-06,
|
| 243271 |
+
"loss": 0.2531,
|
| 243272 |
+
"step": 113110
|
| 243273 |
+
},
|
| 243274 |
+
{
|
| 243275 |
+
"epoch": 911.94,
|
| 243276 |
+
"learning_rate": 8.189499192245558e-06,
|
| 243277 |
+
"loss": 0.4252,
|
| 243278 |
+
"step": 113115
|
| 243279 |
+
},
|
| 243280 |
+
{
|
| 243281 |
+
"epoch": 911.98,
|
| 243282 |
+
"learning_rate": 8.189418416801294e-06,
|
| 243283 |
+
"loss": 0.6596,
|
| 243284 |
+
"step": 113120
|
| 243285 |
+
},
|
| 243286 |
+
{
|
| 243287 |
+
"epoch": 912.0,
|
| 243288 |
+
"eval_loss": 0.444323867559433,
|
| 243289 |
+
"eval_runtime": 40.7673,
|
| 243290 |
+
"eval_samples_per_second": 20.507,
|
| 243291 |
+
"eval_steps_per_second": 0.662,
|
| 243292 |
+
"eval_wer": 0.18970503181029497,
|
| 243293 |
+
"step": 113122
|
| 243294 |
+
},
|
| 243295 |
+
{
|
| 243296 |
+
"epoch": 912.02,
|
| 243297 |
+
"learning_rate": 8.189337641357028e-06,
|
| 243298 |
+
"loss": 0.3245,
|
| 243299 |
+
"step": 113125
|
| 243300 |
+
},
|
| 243301 |
+
{
|
| 243302 |
+
"epoch": 912.06,
|
| 243303 |
+
"learning_rate": 8.189256865912764e-06,
|
| 243304 |
+
"loss": 0.2808,
|
| 243305 |
+
"step": 113130
|
| 243306 |
+
},
|
| 243307 |
+
{
|
| 243308 |
+
"epoch": 912.1,
|
| 243309 |
+
"learning_rate": 8.189176090468498e-06,
|
| 243310 |
+
"loss": 0.3132,
|
| 243311 |
+
"step": 113135
|
| 243312 |
+
},
|
| 243313 |
+
{
|
| 243314 |
+
"epoch": 912.14,
|
| 243315 |
+
"learning_rate": 8.189095315024234e-06,
|
| 243316 |
+
"loss": 0.3814,
|
| 243317 |
+
"step": 113140
|
| 243318 |
+
},
|
| 243319 |
+
{
|
| 243320 |
+
"epoch": 912.18,
|
| 243321 |
+
"learning_rate": 8.189014539579968e-06,
|
| 243322 |
+
"loss": 0.7656,
|
| 243323 |
+
"step": 113145
|
| 243324 |
+
},
|
| 243325 |
+
{
|
| 243326 |
+
"epoch": 912.22,
|
| 243327 |
+
"learning_rate": 8.188933764135704e-06,
|
| 243328 |
+
"loss": 0.9321,
|
| 243329 |
+
"step": 113150
|
| 243330 |
+
},
|
| 243331 |
+
{
|
| 243332 |
+
"epoch": 912.27,
|
| 243333 |
+
"learning_rate": 8.188852988691438e-06,
|
| 243334 |
+
"loss": 0.2926,
|
| 243335 |
+
"step": 113155
|
| 243336 |
+
},
|
| 243337 |
+
{
|
| 243338 |
+
"epoch": 912.31,
|
| 243339 |
+
"learning_rate": 8.188772213247174e-06,
|
| 243340 |
+
"loss": 0.3008,
|
| 243341 |
+
"step": 113160
|
| 243342 |
+
},
|
| 243343 |
+
{
|
| 243344 |
+
"epoch": 912.35,
|
| 243345 |
+
"learning_rate": 8.188691437802908e-06,
|
| 243346 |
+
"loss": 0.3474,
|
| 243347 |
+
"step": 113165
|
| 243348 |
+
},
|
| 243349 |
+
{
|
| 243350 |
+
"epoch": 912.39,
|
| 243351 |
+
"learning_rate": 8.188610662358644e-06,
|
| 243352 |
+
"loss": 0.8603,
|
| 243353 |
+
"step": 113170
|
| 243354 |
+
},
|
| 243355 |
+
{
|
| 243356 |
+
"epoch": 912.43,
|
| 243357 |
+
"learning_rate": 8.18852988691438e-06,
|
| 243358 |
+
"loss": 0.9023,
|
| 243359 |
+
"step": 113175
|
| 243360 |
+
},
|
| 243361 |
+
{
|
| 243362 |
+
"epoch": 912.47,
|
| 243363 |
+
"learning_rate": 8.188449111470114e-06,
|
| 243364 |
+
"loss": 0.2997,
|
| 243365 |
+
"step": 113180
|
| 243366 |
+
},
|
| 243367 |
+
{
|
| 243368 |
+
"epoch": 912.51,
|
| 243369 |
+
"learning_rate": 8.18836833602585e-06,
|
| 243370 |
+
"loss": 0.3067,
|
| 243371 |
+
"step": 113185
|
| 243372 |
+
},
|
| 243373 |
+
{
|
| 243374 |
+
"epoch": 912.55,
|
| 243375 |
+
"learning_rate": 8.188287560581584e-06,
|
| 243376 |
+
"loss": 0.36,
|
| 243377 |
+
"step": 113190
|
| 243378 |
+
},
|
| 243379 |
+
{
|
| 243380 |
+
"epoch": 912.59,
|
| 243381 |
+
"learning_rate": 8.18820678513732e-06,
|
| 243382 |
+
"loss": 0.7879,
|
| 243383 |
+
"step": 113195
|
| 243384 |
+
},
|
| 243385 |
+
{
|
| 243386 |
+
"epoch": 912.63,
|
| 243387 |
+
"learning_rate": 8.188126009693054e-06,
|
| 243388 |
+
"loss": 0.9157,
|
| 243389 |
+
"step": 113200
|
| 243390 |
+
},
|
| 243391 |
+
{
|
| 243392 |
+
"epoch": 912.67,
|
| 243393 |
+
"learning_rate": 8.18804523424879e-06,
|
| 243394 |
+
"loss": 0.3224,
|
| 243395 |
+
"step": 113205
|
| 243396 |
+
},
|
| 243397 |
+
{
|
| 243398 |
+
"epoch": 912.71,
|
| 243399 |
+
"learning_rate": 8.187964458804524e-06,
|
| 243400 |
+
"loss": 0.2828,
|
| 243401 |
+
"step": 113210
|
| 243402 |
+
},
|
| 243403 |
+
{
|
| 243404 |
+
"epoch": 912.75,
|
| 243405 |
+
"learning_rate": 8.18788368336026e-06,
|
| 243406 |
+
"loss": 0.3505,
|
| 243407 |
+
"step": 113215
|
| 243408 |
+
},
|
| 243409 |
+
{
|
| 243410 |
+
"epoch": 912.79,
|
| 243411 |
+
"learning_rate": 8.187802907915994e-06,
|
| 243412 |
+
"loss": 0.7458,
|
| 243413 |
+
"step": 113220
|
| 243414 |
+
},
|
| 243415 |
+
{
|
| 243416 |
+
"epoch": 912.83,
|
| 243417 |
+
"learning_rate": 8.18772213247173e-06,
|
| 243418 |
+
"loss": 0.793,
|
| 243419 |
+
"step": 113225
|
| 243420 |
+
},
|
| 243421 |
+
{
|
| 243422 |
+
"epoch": 912.87,
|
| 243423 |
+
"learning_rate": 8.187641357027464e-06,
|
| 243424 |
+
"loss": 0.3386,
|
| 243425 |
+
"step": 113230
|
| 243426 |
+
},
|
| 243427 |
+
{
|
| 243428 |
+
"epoch": 912.91,
|
| 243429 |
+
"learning_rate": 8.1875605815832e-06,
|
| 243430 |
+
"loss": 0.3076,
|
| 243431 |
+
"step": 113235
|
| 243432 |
+
},
|
| 243433 |
+
{
|
| 243434 |
+
"epoch": 912.95,
|
| 243435 |
+
"learning_rate": 8.187479806138935e-06,
|
| 243436 |
+
"loss": 0.3701,
|
| 243437 |
+
"step": 113240
|
| 243438 |
+
},
|
| 243439 |
+
{
|
| 243440 |
+
"epoch": 912.99,
|
| 243441 |
+
"learning_rate": 8.18739903069467e-06,
|
| 243442 |
+
"loss": 0.807,
|
| 243443 |
+
"step": 113245
|
| 243444 |
+
},
|
| 243445 |
+
{
|
| 243446 |
+
"epoch": 913.0,
|
| 243447 |
+
"eval_loss": 0.4701347053050995,
|
| 243448 |
+
"eval_runtime": 41.2757,
|
| 243449 |
+
"eval_samples_per_second": 20.278,
|
| 243450 |
+
"eval_steps_per_second": 0.654,
|
| 243451 |
+
"eval_wer": 0.17971326164874551,
|
| 243452 |
+
"step": 113246
|
| 243453 |
+
},
|
| 243454 |
+
{
|
| 243455 |
+
"epoch": 913.03,
|
| 243456 |
+
"learning_rate": 8.187318255250405e-06,
|
| 243457 |
+
"loss": 0.3049,
|
| 243458 |
+
"step": 113250
|
| 243459 |
+
},
|
| 243460 |
+
{
|
| 243461 |
+
"epoch": 913.07,
|
| 243462 |
+
"learning_rate": 8.18723747980614e-06,
|
| 243463 |
+
"loss": 0.3326,
|
| 243464 |
+
"step": 113255
|
| 243465 |
+
},
|
| 243466 |
+
{
|
| 243467 |
+
"epoch": 913.11,
|
| 243468 |
+
"learning_rate": 8.187156704361875e-06,
|
| 243469 |
+
"loss": 0.2772,
|
| 243470 |
+
"step": 113260
|
| 243471 |
+
},
|
| 243472 |
+
{
|
| 243473 |
+
"epoch": 913.15,
|
| 243474 |
+
"learning_rate": 8.18707592891761e-06,
|
| 243475 |
+
"loss": 0.3754,
|
| 243476 |
+
"step": 113265
|
| 243477 |
+
},
|
| 243478 |
+
{
|
| 243479 |
+
"epoch": 913.19,
|
| 243480 |
+
"learning_rate": 8.186995153473345e-06,
|
| 243481 |
+
"loss": 0.8271,
|
| 243482 |
+
"step": 113270
|
| 243483 |
+
},
|
| 243484 |
+
{
|
| 243485 |
+
"epoch": 913.23,
|
| 243486 |
+
"learning_rate": 8.18691437802908e-06,
|
| 243487 |
+
"loss": 0.7523,
|
| 243488 |
+
"step": 113275
|
| 243489 |
+
},
|
| 243490 |
+
{
|
| 243491 |
+
"epoch": 913.27,
|
| 243492 |
+
"learning_rate": 8.186833602584815e-06,
|
| 243493 |
+
"loss": 0.2855,
|
| 243494 |
+
"step": 113280
|
| 243495 |
+
},
|
| 243496 |
+
{
|
| 243497 |
+
"epoch": 913.31,
|
| 243498 |
+
"learning_rate": 8.18675282714055e-06,
|
| 243499 |
+
"loss": 0.3099,
|
| 243500 |
+
"step": 113285
|
| 243501 |
+
},
|
| 243502 |
+
{
|
| 243503 |
+
"epoch": 913.35,
|
| 243504 |
+
"learning_rate": 8.186672051696285e-06,
|
| 243505 |
+
"loss": 0.3889,
|
| 243506 |
+
"step": 113290
|
| 243507 |
+
},
|
| 243508 |
+
{
|
| 243509 |
+
"epoch": 913.39,
|
| 243510 |
+
"learning_rate": 8.186591276252021e-06,
|
| 243511 |
+
"loss": 0.8476,
|
| 243512 |
+
"step": 113295
|
| 243513 |
+
},
|
| 243514 |
+
{
|
| 243515 |
+
"epoch": 913.43,
|
| 243516 |
+
"learning_rate": 8.186510500807755e-06,
|
| 243517 |
+
"loss": 0.6524,
|
| 243518 |
+
"step": 113300
|
| 243519 |
+
},
|
| 243520 |
+
{
|
| 243521 |
+
"epoch": 913.47,
|
| 243522 |
+
"learning_rate": 8.186429725363491e-06,
|
| 243523 |
+
"loss": 0.2498,
|
| 243524 |
+
"step": 113305
|
| 243525 |
+
},
|
| 243526 |
+
{
|
| 243527 |
+
"epoch": 913.51,
|
| 243528 |
+
"learning_rate": 8.186348949919225e-06,
|
| 243529 |
+
"loss": 0.3264,
|
| 243530 |
+
"step": 113310
|
| 243531 |
+
},
|
| 243532 |
+
{
|
| 243533 |
+
"epoch": 913.55,
|
| 243534 |
+
"learning_rate": 8.186268174474961e-06,
|
| 243535 |
+
"loss": 0.3934,
|
| 243536 |
+
"step": 113315
|
| 243537 |
+
},
|
| 243538 |
+
{
|
| 243539 |
+
"epoch": 913.59,
|
| 243540 |
+
"learning_rate": 8.186187399030695e-06,
|
| 243541 |
+
"loss": 0.9911,
|
| 243542 |
+
"step": 113320
|
| 243543 |
+
},
|
| 243544 |
+
{
|
| 243545 |
+
"epoch": 913.63,
|
| 243546 |
+
"learning_rate": 8.18610662358643e-06,
|
| 243547 |
+
"loss": 0.6645,
|
| 243548 |
+
"step": 113325
|
| 243549 |
+
},
|
| 243550 |
+
{
|
| 243551 |
+
"epoch": 913.67,
|
| 243552 |
+
"learning_rate": 8.186025848142165e-06,
|
| 243553 |
+
"loss": 0.2985,
|
| 243554 |
+
"step": 113330
|
| 243555 |
+
},
|
| 243556 |
+
{
|
| 243557 |
+
"epoch": 913.71,
|
| 243558 |
+
"learning_rate": 8.1859450726979e-06,
|
| 243559 |
+
"loss": 0.3302,
|
| 243560 |
+
"step": 113335
|
| 243561 |
+
},
|
| 243562 |
+
{
|
| 243563 |
+
"epoch": 913.76,
|
| 243564 |
+
"learning_rate": 8.185864297253635e-06,
|
| 243565 |
+
"loss": 0.3933,
|
| 243566 |
+
"step": 113340
|
| 243567 |
+
},
|
| 243568 |
+
{
|
| 243569 |
+
"epoch": 913.8,
|
| 243570 |
+
"learning_rate": 8.18578352180937e-06,
|
| 243571 |
+
"loss": 0.9581,
|
| 243572 |
+
"step": 113345
|
| 243573 |
+
},
|
| 243574 |
+
{
|
| 243575 |
+
"epoch": 913.84,
|
| 243576 |
+
"learning_rate": 8.185702746365107e-06,
|
| 243577 |
+
"loss": 0.7659,
|
| 243578 |
+
"step": 113350
|
| 243579 |
+
},
|
| 243580 |
+
{
|
| 243581 |
+
"epoch": 913.88,
|
| 243582 |
+
"learning_rate": 8.18562197092084e-06,
|
| 243583 |
+
"loss": 0.2852,
|
| 243584 |
+
"step": 113355
|
| 243585 |
+
},
|
| 243586 |
+
{
|
| 243587 |
+
"epoch": 913.92,
|
| 243588 |
+
"learning_rate": 8.185541195476577e-06,
|
| 243589 |
+
"loss": 0.2891,
|
| 243590 |
+
"step": 113360
|
| 243591 |
+
},
|
| 243592 |
+
{
|
| 243593 |
+
"epoch": 913.96,
|
| 243594 |
+
"learning_rate": 8.18546042003231e-06,
|
| 243595 |
+
"loss": 0.4714,
|
| 243596 |
+
"step": 113365
|
| 243597 |
+
},
|
| 243598 |
+
{
|
| 243599 |
+
"epoch": 914.0,
|
| 243600 |
+
"learning_rate": 8.185379644588047e-06,
|
| 243601 |
+
"loss": 1.1086,
|
| 243602 |
+
"step": 113370
|
| 243603 |
+
},
|
| 243604 |
+
{
|
| 243605 |
+
"epoch": 914.0,
|
| 243606 |
+
"eval_loss": 0.39893263578414917,
|
| 243607 |
+
"eval_runtime": 39.8088,
|
| 243608 |
+
"eval_samples_per_second": 21.025,
|
| 243609 |
+
"eval_steps_per_second": 0.678,
|
| 243610 |
+
"eval_wer": 0.18508635386553998,
|
| 243611 |
+
"step": 113370
|
| 243612 |
}
|
| 243613 |
],
|
| 243614 |
+
"max_steps": 620000,
|
| 243615 |
"num_train_epochs": 5000,
|
| 243616 |
+
"total_flos": 3.190555567482095e+20,
|
| 243617 |
"trial_name": null,
|
| 243618 |
"trial_params": null
|
| 243619 |
}
|
model-bin/finetune/base/{checkpoint-112750 → checkpoint-113370}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1630082884.969348/events.out.tfevents.1630082884.dea8124df033.1391.7
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:41fd86e6e7c1a5c77593c5bbda325bd5c975686ea34b46f2f3a6e958dbca7861
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630083307.1487231/events.out.tfevents.1630083307.dea8124df033.1391.9
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:11391acdb58354ca2cac2c60fd34dff94d1f18b0bcbdfb0572682728e5a86432
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630083735.6256015/events.out.tfevents.1630083735.dea8124df033.1391.11
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2e25a50c37d80004082e7bc62597d5bf40a647914e074c4a409928874ead615a
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630084174.1170754/events.out.tfevents.1630084174.dea8124df033.1391.13
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6063e6cb2ea6a249fc018a71d2bc14207d6991ca21c0e26094c96758b8d8827a
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630084598.2761755/events.out.tfevents.1630084598.dea8124df033.1391.15
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:0f620b324894ddc530420ee660ac32ef67a80a90e8eaf65637db8ee2548cf5cd
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1630082884.dea8124df033.1391.6
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:3f9dc3f5f57f07b97c5848164fa8a44cf6407370c74df678c5318e16f5e9bd6e
|
| 3 |
+
size 8462
|
model-bin/finetune/base/log/events.out.tfevents.1630083307.dea8124df033.1391.8
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:81129d21b671588c0030205e0ce3ee50b437dcd581b30365dc937c7d279d6f2b
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630083735.dea8124df033.1391.10
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9203203993faf14718631ba75e93c0624f6aa871e966f57dfaa38274522d3aea
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630084174.dea8124df033.1391.12
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e1059774e54b65b2bea6167c90283c2debf4d7672d04144c264476154ca1eedc
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630084598.dea8124df033.1391.14
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:7c1ce76c497b5a7a4d19715c80fc0b5e0948b520e68b45153bb07b93b8e9a7f4
|
| 3 |
+
size 8622
|