"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-72674 → checkpoint-73294}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-72674 → checkpoint-73294}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-72674 → checkpoint-73294}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-72674 → checkpoint-73294}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-72674 → checkpoint-73294}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-72674 → checkpoint-73294}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-72674 → checkpoint-73294}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-72674 → checkpoint-73294}/trainer_state.json +792 -3
- model-bin/finetune/base/{checkpoint-72674 → checkpoint-73294}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629869061.2637658/events.out.tfevents.1629869061.7e498afd5545.905.53 +3 -0
- model-bin/finetune/base/log/1629869731.455241/events.out.tfevents.1629869731.7e498afd5545.905.55 +3 -0
- model-bin/finetune/base/log/1629870382.7916713/events.out.tfevents.1629870382.7e498afd5545.905.57 +3 -0
- model-bin/finetune/base/log/1629871038.1608121/events.out.tfevents.1629871038.7e498afd5545.905.59 +3 -0
- model-bin/finetune/base/log/1629871680.9885855/events.out.tfevents.1629871680.7e498afd5545.905.61 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629869061.7e498afd5545.905.52 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629869731.7e498afd5545.905.54 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629870382.7e498afd5545.905.56 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629871038.7e498afd5545.905.58 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629871680.7e498afd5545.905.60 +3 -0
model-bin/finetune/base/{checkpoint-72674 → checkpoint-73294}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-72674 → checkpoint-73294}/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 sha256:474f662d85a0ed7f5bc206a95d86114fc46bc01fa819c5ec2777457c77a00761
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-72674 → checkpoint-73294}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-72674 → checkpoint-73294}/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 sha256:2515fce69a86617d9a349a139a5baf90c4eb5c34710ae7f1938eb4041c8573b3
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-72674 → checkpoint-73294}/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:c9e7b3e60ec7f2a8985b5a92be7a019bc55dd61ea1a80f5e0727362168793b07
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-72674 → checkpoint-73294}/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 sha256:0db8edbac0eb024da8220cc21e3842882d8da7392af65962fe8a5134ce0473dd
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-72674 → checkpoint-73294}/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 sha256:6bd477d4820649f1a048e2fc31a66af8ecccd0bbb65e45ea5df7cdf6444fcd34
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-72674 → checkpoint-73294}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.18412114350410416,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-69565",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -191826,11 +191826,800 @@
|
|
| 191826 |
"eval_steps_per_second": 0.669,
|
| 191827 |
"eval_wer": 0.19356979573777744,
|
| 191828 |
"step": 72674
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 191829 |
}
|
| 191830 |
],
|
| 191831 |
"max_steps": 625000,
|
| 191832 |
"num_train_epochs": 5000,
|
| 191833 |
-
"total_flos": 2.
|
| 191834 |
"trial_name": null,
|
| 191835 |
"trial_params": null
|
| 191836 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.18412114350410416,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-69565",
|
| 4 |
+
"epoch": 586.0,
|
| 5 |
+
"global_step": 73294,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 191826 |
"eval_steps_per_second": 0.669,
|
| 191827 |
"eval_wer": 0.19356979573777744,
|
| 191828 |
"step": 72674
|
| 191829 |
+
},
|
| 191830 |
+
{
|
| 191831 |
+
"epoch": 586.01,
|
| 191832 |
+
"learning_rate": 8.851794871794872e-06,
|
| 191833 |
+
"loss": 0.4982,
|
| 191834 |
+
"step": 72675
|
| 191835 |
+
},
|
| 191836 |
+
{
|
| 191837 |
+
"epoch": 586.05,
|
| 191838 |
+
"learning_rate": 8.851714743589745e-06,
|
| 191839 |
+
"loss": 0.3086,
|
| 191840 |
+
"step": 72680
|
| 191841 |
+
},
|
| 191842 |
+
{
|
| 191843 |
+
"epoch": 586.09,
|
| 191844 |
+
"learning_rate": 8.851634615384616e-06,
|
| 191845 |
+
"loss": 0.3143,
|
| 191846 |
+
"step": 72685
|
| 191847 |
+
},
|
| 191848 |
+
{
|
| 191849 |
+
"epoch": 586.13,
|
| 191850 |
+
"learning_rate": 8.851554487179488e-06,
|
| 191851 |
+
"loss": 0.3705,
|
| 191852 |
+
"step": 72690
|
| 191853 |
+
},
|
| 191854 |
+
{
|
| 191855 |
+
"epoch": 586.17,
|
| 191856 |
+
"learning_rate": 8.851474358974361e-06,
|
| 191857 |
+
"loss": 0.5783,
|
| 191858 |
+
"step": 72695
|
| 191859 |
+
},
|
| 191860 |
+
{
|
| 191861 |
+
"epoch": 586.21,
|
| 191862 |
+
"learning_rate": 8.85139423076923e-06,
|
| 191863 |
+
"loss": 1.174,
|
| 191864 |
+
"step": 72700
|
| 191865 |
+
},
|
| 191866 |
+
{
|
| 191867 |
+
"epoch": 586.25,
|
| 191868 |
+
"learning_rate": 8.851314102564104e-06,
|
| 191869 |
+
"loss": 0.4436,
|
| 191870 |
+
"step": 72705
|
| 191871 |
+
},
|
| 191872 |
+
{
|
| 191873 |
+
"epoch": 586.29,
|
| 191874 |
+
"learning_rate": 8.851233974358975e-06,
|
| 191875 |
+
"loss": 0.2785,
|
| 191876 |
+
"step": 72710
|
| 191877 |
+
},
|
| 191878 |
+
{
|
| 191879 |
+
"epoch": 586.33,
|
| 191880 |
+
"learning_rate": 8.851153846153846e-06,
|
| 191881 |
+
"loss": 0.4354,
|
| 191882 |
+
"step": 72715
|
| 191883 |
+
},
|
| 191884 |
+
{
|
| 191885 |
+
"epoch": 586.37,
|
| 191886 |
+
"learning_rate": 8.851073717948718e-06,
|
| 191887 |
+
"loss": 0.674,
|
| 191888 |
+
"step": 72720
|
| 191889 |
+
},
|
| 191890 |
+
{
|
| 191891 |
+
"epoch": 586.41,
|
| 191892 |
+
"learning_rate": 8.85099358974359e-06,
|
| 191893 |
+
"loss": 1.2375,
|
| 191894 |
+
"step": 72725
|
| 191895 |
+
},
|
| 191896 |
+
{
|
| 191897 |
+
"epoch": 586.45,
|
| 191898 |
+
"learning_rate": 8.850913461538462e-06,
|
| 191899 |
+
"loss": 0.4444,
|
| 191900 |
+
"step": 72730
|
| 191901 |
+
},
|
| 191902 |
+
{
|
| 191903 |
+
"epoch": 586.49,
|
| 191904 |
+
"learning_rate": 8.850833333333333e-06,
|
| 191905 |
+
"loss": 0.3199,
|
| 191906 |
+
"step": 72735
|
| 191907 |
+
},
|
| 191908 |
+
{
|
| 191909 |
+
"epoch": 586.53,
|
| 191910 |
+
"learning_rate": 8.850753205128206e-06,
|
| 191911 |
+
"loss": 0.3793,
|
| 191912 |
+
"step": 72740
|
| 191913 |
+
},
|
| 191914 |
+
{
|
| 191915 |
+
"epoch": 586.57,
|
| 191916 |
+
"learning_rate": 8.850673076923078e-06,
|
| 191917 |
+
"loss": 0.5258,
|
| 191918 |
+
"step": 72745
|
| 191919 |
+
},
|
| 191920 |
+
{
|
| 191921 |
+
"epoch": 586.61,
|
| 191922 |
+
"learning_rate": 8.85059294871795e-06,
|
| 191923 |
+
"loss": 1.1779,
|
| 191924 |
+
"step": 72750
|
| 191925 |
+
},
|
| 191926 |
+
{
|
| 191927 |
+
"epoch": 586.65,
|
| 191928 |
+
"learning_rate": 8.85051282051282e-06,
|
| 191929 |
+
"loss": 0.308,
|
| 191930 |
+
"step": 72755
|
| 191931 |
+
},
|
| 191932 |
+
{
|
| 191933 |
+
"epoch": 586.69,
|
| 191934 |
+
"learning_rate": 8.850432692307694e-06,
|
| 191935 |
+
"loss": 0.3619,
|
| 191936 |
+
"step": 72760
|
| 191937 |
+
},
|
| 191938 |
+
{
|
| 191939 |
+
"epoch": 586.73,
|
| 191940 |
+
"learning_rate": 8.850352564102565e-06,
|
| 191941 |
+
"loss": 0.3743,
|
| 191942 |
+
"step": 72765
|
| 191943 |
+
},
|
| 191944 |
+
{
|
| 191945 |
+
"epoch": 586.77,
|
| 191946 |
+
"learning_rate": 8.850272435897436e-06,
|
| 191947 |
+
"loss": 0.6303,
|
| 191948 |
+
"step": 72770
|
| 191949 |
+
},
|
| 191950 |
+
{
|
| 191951 |
+
"epoch": 586.81,
|
| 191952 |
+
"learning_rate": 8.850192307692308e-06,
|
| 191953 |
+
"loss": 1.0891,
|
| 191954 |
+
"step": 72775
|
| 191955 |
+
},
|
| 191956 |
+
{
|
| 191957 |
+
"epoch": 586.85,
|
| 191958 |
+
"learning_rate": 8.85011217948718e-06,
|
| 191959 |
+
"loss": 0.3039,
|
| 191960 |
+
"step": 72780
|
| 191961 |
+
},
|
| 191962 |
+
{
|
| 191963 |
+
"epoch": 586.89,
|
| 191964 |
+
"learning_rate": 8.850032051282052e-06,
|
| 191965 |
+
"loss": 0.2967,
|
| 191966 |
+
"step": 72785
|
| 191967 |
+
},
|
| 191968 |
+
{
|
| 191969 |
+
"epoch": 586.93,
|
| 191970 |
+
"learning_rate": 8.849951923076923e-06,
|
| 191971 |
+
"loss": 0.3696,
|
| 191972 |
+
"step": 72790
|
| 191973 |
+
},
|
| 191974 |
+
{
|
| 191975 |
+
"epoch": 586.97,
|
| 191976 |
+
"learning_rate": 8.849871794871797e-06,
|
| 191977 |
+
"loss": 0.5783,
|
| 191978 |
+
"step": 72795
|
| 191979 |
+
},
|
| 191980 |
+
{
|
| 191981 |
+
"epoch": 587.0,
|
| 191982 |
+
"eval_loss": 0.37081632018089294,
|
| 191983 |
+
"eval_runtime": 40.9633,
|
| 191984 |
+
"eval_samples_per_second": 20.506,
|
| 191985 |
+
"eval_steps_per_second": 0.659,
|
| 191986 |
+
"eval_wer": 0.1872572990076226,
|
| 191987 |
+
"step": 72798
|
| 191988 |
+
},
|
| 191989 |
+
{
|
| 191990 |
+
"epoch": 591.02,
|
| 191991 |
+
"learning_rate": 8.849791666666668e-06,
|
| 191992 |
+
"loss": 0.4085,
|
| 191993 |
+
"step": 72800
|
| 191994 |
+
},
|
| 191995 |
+
{
|
| 191996 |
+
"epoch": 591.06,
|
| 191997 |
+
"learning_rate": 8.84971153846154e-06,
|
| 191998 |
+
"loss": 0.35,
|
| 191999 |
+
"step": 72805
|
| 192000 |
+
},
|
| 192001 |
+
{
|
| 192002 |
+
"epoch": 591.1,
|
| 192003 |
+
"learning_rate": 8.84963141025641e-06,
|
| 192004 |
+
"loss": 0.3162,
|
| 192005 |
+
"step": 72810
|
| 192006 |
+
},
|
| 192007 |
+
{
|
| 192008 |
+
"epoch": 591.14,
|
| 192009 |
+
"learning_rate": 8.849551282051284e-06,
|
| 192010 |
+
"loss": 0.3517,
|
| 192011 |
+
"step": 72815
|
| 192012 |
+
},
|
| 192013 |
+
{
|
| 192014 |
+
"epoch": 591.18,
|
| 192015 |
+
"learning_rate": 8.849471153846155e-06,
|
| 192016 |
+
"loss": 0.8103,
|
| 192017 |
+
"step": 72820
|
| 192018 |
+
},
|
| 192019 |
+
{
|
| 192020 |
+
"epoch": 591.22,
|
| 192021 |
+
"learning_rate": 8.849391025641026e-06,
|
| 192022 |
+
"loss": 1.0763,
|
| 192023 |
+
"step": 72825
|
| 192024 |
+
},
|
| 192025 |
+
{
|
| 192026 |
+
"epoch": 591.26,
|
| 192027 |
+
"learning_rate": 8.849310897435898e-06,
|
| 192028 |
+
"loss": 0.3132,
|
| 192029 |
+
"step": 72830
|
| 192030 |
+
},
|
| 192031 |
+
{
|
| 192032 |
+
"epoch": 591.3,
|
| 192033 |
+
"learning_rate": 8.84923076923077e-06,
|
| 192034 |
+
"loss": 0.2745,
|
| 192035 |
+
"step": 72835
|
| 192036 |
+
},
|
| 192037 |
+
{
|
| 192038 |
+
"epoch": 591.34,
|
| 192039 |
+
"learning_rate": 8.849150641025642e-06,
|
| 192040 |
+
"loss": 0.3993,
|
| 192041 |
+
"step": 72840
|
| 192042 |
+
},
|
| 192043 |
+
{
|
| 192044 |
+
"epoch": 591.38,
|
| 192045 |
+
"learning_rate": 8.849070512820513e-06,
|
| 192046 |
+
"loss": 0.6536,
|
| 192047 |
+
"step": 72845
|
| 192048 |
+
},
|
| 192049 |
+
{
|
| 192050 |
+
"epoch": 591.42,
|
| 192051 |
+
"learning_rate": 8.848990384615387e-06,
|
| 192052 |
+
"loss": 1.1869,
|
| 192053 |
+
"step": 72850
|
| 192054 |
+
},
|
| 192055 |
+
{
|
| 192056 |
+
"epoch": 591.46,
|
| 192057 |
+
"learning_rate": 8.848910256410256e-06,
|
| 192058 |
+
"loss": 0.2996,
|
| 192059 |
+
"step": 72855
|
| 192060 |
+
},
|
| 192061 |
+
{
|
| 192062 |
+
"epoch": 591.5,
|
| 192063 |
+
"learning_rate": 8.84883012820513e-06,
|
| 192064 |
+
"loss": 0.3045,
|
| 192065 |
+
"step": 72860
|
| 192066 |
+
},
|
| 192067 |
+
{
|
| 192068 |
+
"epoch": 591.54,
|
| 192069 |
+
"learning_rate": 8.84875e-06,
|
| 192070 |
+
"loss": 0.3851,
|
| 192071 |
+
"step": 72865
|
| 192072 |
+
},
|
| 192073 |
+
{
|
| 192074 |
+
"epoch": 591.58,
|
| 192075 |
+
"learning_rate": 8.848669871794872e-06,
|
| 192076 |
+
"loss": 0.6637,
|
| 192077 |
+
"step": 72870
|
| 192078 |
+
},
|
| 192079 |
+
{
|
| 192080 |
+
"epoch": 591.62,
|
| 192081 |
+
"learning_rate": 8.848589743589743e-06,
|
| 192082 |
+
"loss": 1.0274,
|
| 192083 |
+
"step": 72875
|
| 192084 |
+
},
|
| 192085 |
+
{
|
| 192086 |
+
"epoch": 591.66,
|
| 192087 |
+
"learning_rate": 8.848509615384616e-06,
|
| 192088 |
+
"loss": 0.3225,
|
| 192089 |
+
"step": 72880
|
| 192090 |
+
},
|
| 192091 |
+
{
|
| 192092 |
+
"epoch": 591.7,
|
| 192093 |
+
"learning_rate": 8.848429487179488e-06,
|
| 192094 |
+
"loss": 0.3141,
|
| 192095 |
+
"step": 72885
|
| 192096 |
+
},
|
| 192097 |
+
{
|
| 192098 |
+
"epoch": 591.74,
|
| 192099 |
+
"learning_rate": 8.848349358974359e-06,
|
| 192100 |
+
"loss": 0.373,
|
| 192101 |
+
"step": 72890
|
| 192102 |
+
},
|
| 192103 |
+
{
|
| 192104 |
+
"epoch": 591.79,
|
| 192105 |
+
"learning_rate": 8.848269230769232e-06,
|
| 192106 |
+
"loss": 0.9046,
|
| 192107 |
+
"step": 72895
|
| 192108 |
+
},
|
| 192109 |
+
{
|
| 192110 |
+
"epoch": 591.83,
|
| 192111 |
+
"learning_rate": 8.848189102564104e-06,
|
| 192112 |
+
"loss": 1.149,
|
| 192113 |
+
"step": 72900
|
| 192114 |
+
},
|
| 192115 |
+
{
|
| 192116 |
+
"epoch": 591.87,
|
| 192117 |
+
"learning_rate": 8.848108974358975e-06,
|
| 192118 |
+
"loss": 0.3039,
|
| 192119 |
+
"step": 72905
|
| 192120 |
+
},
|
| 192121 |
+
{
|
| 192122 |
+
"epoch": 591.91,
|
| 192123 |
+
"learning_rate": 8.848028846153846e-06,
|
| 192124 |
+
"loss": 0.562,
|
| 192125 |
+
"step": 72910
|
| 192126 |
+
},
|
| 192127 |
+
{
|
| 192128 |
+
"epoch": 591.95,
|
| 192129 |
+
"learning_rate": 8.84794871794872e-06,
|
| 192130 |
+
"loss": 0.4599,
|
| 192131 |
+
"step": 72915
|
| 192132 |
+
},
|
| 192133 |
+
{
|
| 192134 |
+
"epoch": 591.99,
|
| 192135 |
+
"learning_rate": 8.84786858974359e-06,
|
| 192136 |
+
"loss": 0.9062,
|
| 192137 |
+
"step": 72920
|
| 192138 |
+
},
|
| 192139 |
+
{
|
| 192140 |
+
"epoch": 592.0,
|
| 192141 |
+
"eval_loss": 0.39401471614837646,
|
| 192142 |
+
"eval_runtime": 38.9267,
|
| 192143 |
+
"eval_samples_per_second": 21.579,
|
| 192144 |
+
"eval_steps_per_second": 0.694,
|
| 192145 |
+
"eval_wer": 0.18718740806119447,
|
| 192146 |
+
"step": 72921
|
| 192147 |
+
},
|
| 192148 |
+
{
|
| 192149 |
+
"epoch": 588.03,
|
| 192150 |
+
"learning_rate": 8.847788461538462e-06,
|
| 192151 |
+
"loss": 0.3517,
|
| 192152 |
+
"step": 72925
|
| 192153 |
+
},
|
| 192154 |
+
{
|
| 192155 |
+
"epoch": 588.07,
|
| 192156 |
+
"learning_rate": 8.847708333333333e-06,
|
| 192157 |
+
"loss": 0.3238,
|
| 192158 |
+
"step": 72930
|
| 192159 |
+
},
|
| 192160 |
+
{
|
| 192161 |
+
"epoch": 588.11,
|
| 192162 |
+
"learning_rate": 8.847628205128206e-06,
|
| 192163 |
+
"loss": 0.3437,
|
| 192164 |
+
"step": 72935
|
| 192165 |
+
},
|
| 192166 |
+
{
|
| 192167 |
+
"epoch": 588.15,
|
| 192168 |
+
"learning_rate": 8.847548076923078e-06,
|
| 192169 |
+
"loss": 0.4455,
|
| 192170 |
+
"step": 72940
|
| 192171 |
+
},
|
| 192172 |
+
{
|
| 192173 |
+
"epoch": 588.19,
|
| 192174 |
+
"learning_rate": 8.847467948717949e-06,
|
| 192175 |
+
"loss": 0.9492,
|
| 192176 |
+
"step": 72945
|
| 192177 |
+
},
|
| 192178 |
+
{
|
| 192179 |
+
"epoch": 588.23,
|
| 192180 |
+
"learning_rate": 8.847387820512822e-06,
|
| 192181 |
+
"loss": 0.6762,
|
| 192182 |
+
"step": 72950
|
| 192183 |
+
},
|
| 192184 |
+
{
|
| 192185 |
+
"epoch": 588.27,
|
| 192186 |
+
"learning_rate": 8.847307692307694e-06,
|
| 192187 |
+
"loss": 0.3302,
|
| 192188 |
+
"step": 72955
|
| 192189 |
+
},
|
| 192190 |
+
{
|
| 192191 |
+
"epoch": 588.31,
|
| 192192 |
+
"learning_rate": 8.847227564102565e-06,
|
| 192193 |
+
"loss": 0.3051,
|
| 192194 |
+
"step": 72960
|
| 192195 |
+
},
|
| 192196 |
+
{
|
| 192197 |
+
"epoch": 588.35,
|
| 192198 |
+
"learning_rate": 8.847147435897436e-06,
|
| 192199 |
+
"loss": 0.3876,
|
| 192200 |
+
"step": 72965
|
| 192201 |
+
},
|
| 192202 |
+
{
|
| 192203 |
+
"epoch": 588.39,
|
| 192204 |
+
"learning_rate": 8.84706730769231e-06,
|
| 192205 |
+
"loss": 0.921,
|
| 192206 |
+
"step": 72970
|
| 192207 |
+
},
|
| 192208 |
+
{
|
| 192209 |
+
"epoch": 588.43,
|
| 192210 |
+
"learning_rate": 8.846987179487179e-06,
|
| 192211 |
+
"loss": 0.6965,
|
| 192212 |
+
"step": 72975
|
| 192213 |
+
},
|
| 192214 |
+
{
|
| 192215 |
+
"epoch": 588.47,
|
| 192216 |
+
"learning_rate": 8.846907051282052e-06,
|
| 192217 |
+
"loss": 0.2984,
|
| 192218 |
+
"step": 72980
|
| 192219 |
+
},
|
| 192220 |
+
{
|
| 192221 |
+
"epoch": 588.51,
|
| 192222 |
+
"learning_rate": 8.846826923076923e-06,
|
| 192223 |
+
"loss": 0.3759,
|
| 192224 |
+
"step": 72985
|
| 192225 |
+
},
|
| 192226 |
+
{
|
| 192227 |
+
"epoch": 588.55,
|
| 192228 |
+
"learning_rate": 8.846746794871795e-06,
|
| 192229 |
+
"loss": 0.4898,
|
| 192230 |
+
"step": 72990
|
| 192231 |
+
},
|
| 192232 |
+
{
|
| 192233 |
+
"epoch": 588.59,
|
| 192234 |
+
"learning_rate": 8.846666666666668e-06,
|
| 192235 |
+
"loss": 1.0942,
|
| 192236 |
+
"step": 72995
|
| 192237 |
+
},
|
| 192238 |
+
{
|
| 192239 |
+
"epoch": 588.63,
|
| 192240 |
+
"learning_rate": 8.84658653846154e-06,
|
| 192241 |
+
"loss": 0.6618,
|
| 192242 |
+
"step": 73000
|
| 192243 |
+
},
|
| 192244 |
+
{
|
| 192245 |
+
"epoch": 588.67,
|
| 192246 |
+
"learning_rate": 8.84650641025641e-06,
|
| 192247 |
+
"loss": 0.355,
|
| 192248 |
+
"step": 73005
|
| 192249 |
+
},
|
| 192250 |
+
{
|
| 192251 |
+
"epoch": 588.71,
|
| 192252 |
+
"learning_rate": 8.846426282051282e-06,
|
| 192253 |
+
"loss": 0.3374,
|
| 192254 |
+
"step": 73010
|
| 192255 |
+
},
|
| 192256 |
+
{
|
| 192257 |
+
"epoch": 588.76,
|
| 192258 |
+
"learning_rate": 8.846346153846155e-06,
|
| 192259 |
+
"loss": 0.3638,
|
| 192260 |
+
"step": 73015
|
| 192261 |
+
},
|
| 192262 |
+
{
|
| 192263 |
+
"epoch": 588.8,
|
| 192264 |
+
"learning_rate": 8.846266025641026e-06,
|
| 192265 |
+
"loss": 0.9145,
|
| 192266 |
+
"step": 73020
|
| 192267 |
+
},
|
| 192268 |
+
{
|
| 192269 |
+
"epoch": 588.84,
|
| 192270 |
+
"learning_rate": 8.846185897435898e-06,
|
| 192271 |
+
"loss": 0.6553,
|
| 192272 |
+
"step": 73025
|
| 192273 |
+
},
|
| 192274 |
+
{
|
| 192275 |
+
"epoch": 588.88,
|
| 192276 |
+
"learning_rate": 8.846105769230769e-06,
|
| 192277 |
+
"loss": 0.3028,
|
| 192278 |
+
"step": 73030
|
| 192279 |
+
},
|
| 192280 |
+
{
|
| 192281 |
+
"epoch": 588.92,
|
| 192282 |
+
"learning_rate": 8.846025641025642e-06,
|
| 192283 |
+
"loss": 0.3749,
|
| 192284 |
+
"step": 73035
|
| 192285 |
+
},
|
| 192286 |
+
{
|
| 192287 |
+
"epoch": 588.96,
|
| 192288 |
+
"learning_rate": 8.845945512820513e-06,
|
| 192289 |
+
"loss": 0.4571,
|
| 192290 |
+
"step": 73040
|
| 192291 |
+
},
|
| 192292 |
+
{
|
| 192293 |
+
"epoch": 589.0,
|
| 192294 |
+
"learning_rate": 8.845865384615385e-06,
|
| 192295 |
+
"loss": 1.2306,
|
| 192296 |
+
"step": 73045
|
| 192297 |
+
},
|
| 192298 |
+
{
|
| 192299 |
+
"epoch": 589.0,
|
| 192300 |
+
"eval_loss": 0.4083371162414551,
|
| 192301 |
+
"eval_runtime": 39.5334,
|
| 192302 |
+
"eval_samples_per_second": 21.248,
|
| 192303 |
+
"eval_steps_per_second": 0.683,
|
| 192304 |
+
"eval_wer": 0.19103616596405593,
|
| 192305 |
+
"step": 73045
|
| 192306 |
+
},
|
| 192307 |
+
{
|
| 192308 |
+
"epoch": 589.04,
|
| 192309 |
+
"learning_rate": 8.845785256410258e-06,
|
| 192310 |
+
"loss": 0.3375,
|
| 192311 |
+
"step": 73050
|
| 192312 |
+
},
|
| 192313 |
+
{
|
| 192314 |
+
"epoch": 589.08,
|
| 192315 |
+
"learning_rate": 8.84570512820513e-06,
|
| 192316 |
+
"loss": 0.2793,
|
| 192317 |
+
"step": 73055
|
| 192318 |
+
},
|
| 192319 |
+
{
|
| 192320 |
+
"epoch": 589.12,
|
| 192321 |
+
"learning_rate": 8.845625e-06,
|
| 192322 |
+
"loss": 0.3388,
|
| 192323 |
+
"step": 73060
|
| 192324 |
+
},
|
| 192325 |
+
{
|
| 192326 |
+
"epoch": 589.16,
|
| 192327 |
+
"learning_rate": 8.845544871794872e-06,
|
| 192328 |
+
"loss": 0.4248,
|
| 192329 |
+
"step": 73065
|
| 192330 |
+
},
|
| 192331 |
+
{
|
| 192332 |
+
"epoch": 589.2,
|
| 192333 |
+
"learning_rate": 8.845464743589745e-06,
|
| 192334 |
+
"loss": 1.1882,
|
| 192335 |
+
"step": 73070
|
| 192336 |
+
},
|
| 192337 |
+
{
|
| 192338 |
+
"epoch": 589.24,
|
| 192339 |
+
"learning_rate": 8.845384615384616e-06,
|
| 192340 |
+
"loss": 0.2749,
|
| 192341 |
+
"step": 73075
|
| 192342 |
+
},
|
| 192343 |
+
{
|
| 192344 |
+
"epoch": 589.28,
|
| 192345 |
+
"learning_rate": 8.845304487179488e-06,
|
| 192346 |
+
"loss": 0.4075,
|
| 192347 |
+
"step": 73080
|
| 192348 |
+
},
|
| 192349 |
+
{
|
| 192350 |
+
"epoch": 589.32,
|
| 192351 |
+
"learning_rate": 8.845224358974359e-06,
|
| 192352 |
+
"loss": 0.3066,
|
| 192353 |
+
"step": 73085
|
| 192354 |
+
},
|
| 192355 |
+
{
|
| 192356 |
+
"epoch": 589.36,
|
| 192357 |
+
"learning_rate": 8.845144230769232e-06,
|
| 192358 |
+
"loss": 0.5385,
|
| 192359 |
+
"step": 73090
|
| 192360 |
+
},
|
| 192361 |
+
{
|
| 192362 |
+
"epoch": 589.4,
|
| 192363 |
+
"learning_rate": 8.845064102564103e-06,
|
| 192364 |
+
"loss": 1.1468,
|
| 192365 |
+
"step": 73095
|
| 192366 |
+
},
|
| 192367 |
+
{
|
| 192368 |
+
"epoch": 589.44,
|
| 192369 |
+
"learning_rate": 8.844983974358975e-06,
|
| 192370 |
+
"loss": 0.3261,
|
| 192371 |
+
"step": 73100
|
| 192372 |
+
},
|
| 192373 |
+
{
|
| 192374 |
+
"epoch": 589.48,
|
| 192375 |
+
"learning_rate": 8.844903846153848e-06,
|
| 192376 |
+
"loss": 0.2908,
|
| 192377 |
+
"step": 73105
|
| 192378 |
+
},
|
| 192379 |
+
{
|
| 192380 |
+
"epoch": 589.52,
|
| 192381 |
+
"learning_rate": 8.84482371794872e-06,
|
| 192382 |
+
"loss": 0.418,
|
| 192383 |
+
"step": 73110
|
| 192384 |
+
},
|
| 192385 |
+
{
|
| 192386 |
+
"epoch": 589.56,
|
| 192387 |
+
"learning_rate": 8.84474358974359e-06,
|
| 192388 |
+
"loss": 0.4959,
|
| 192389 |
+
"step": 73115
|
| 192390 |
+
},
|
| 192391 |
+
{
|
| 192392 |
+
"epoch": 589.6,
|
| 192393 |
+
"learning_rate": 8.844663461538462e-06,
|
| 192394 |
+
"loss": 1.0528,
|
| 192395 |
+
"step": 73120
|
| 192396 |
+
},
|
| 192397 |
+
{
|
| 192398 |
+
"epoch": 589.64,
|
| 192399 |
+
"learning_rate": 8.844583333333335e-06,
|
| 192400 |
+
"loss": 0.3883,
|
| 192401 |
+
"step": 73125
|
| 192402 |
+
},
|
| 192403 |
+
{
|
| 192404 |
+
"epoch": 589.68,
|
| 192405 |
+
"learning_rate": 8.844503205128205e-06,
|
| 192406 |
+
"loss": 0.3486,
|
| 192407 |
+
"step": 73130
|
| 192408 |
+
},
|
| 192409 |
+
{
|
| 192410 |
+
"epoch": 589.72,
|
| 192411 |
+
"learning_rate": 8.844423076923078e-06,
|
| 192412 |
+
"loss": 0.2976,
|
| 192413 |
+
"step": 73135
|
| 192414 |
+
},
|
| 192415 |
+
{
|
| 192416 |
+
"epoch": 589.76,
|
| 192417 |
+
"learning_rate": 8.84434294871795e-06,
|
| 192418 |
+
"loss": 0.4688,
|
| 192419 |
+
"step": 73140
|
| 192420 |
+
},
|
| 192421 |
+
{
|
| 192422 |
+
"epoch": 589.8,
|
| 192423 |
+
"learning_rate": 8.84426282051282e-06,
|
| 192424 |
+
"loss": 1.2549,
|
| 192425 |
+
"step": 73145
|
| 192426 |
+
},
|
| 192427 |
+
{
|
| 192428 |
+
"epoch": 589.84,
|
| 192429 |
+
"learning_rate": 8.844182692307693e-06,
|
| 192430 |
+
"loss": 0.3377,
|
| 192431 |
+
"step": 73150
|
| 192432 |
+
},
|
| 192433 |
+
{
|
| 192434 |
+
"epoch": 589.88,
|
| 192435 |
+
"learning_rate": 8.844102564102565e-06,
|
| 192436 |
+
"loss": 0.2784,
|
| 192437 |
+
"step": 73155
|
| 192438 |
+
},
|
| 192439 |
+
{
|
| 192440 |
+
"epoch": 589.92,
|
| 192441 |
+
"learning_rate": 8.844022435897436e-06,
|
| 192442 |
+
"loss": 0.3856,
|
| 192443 |
+
"step": 73160
|
| 192444 |
+
},
|
| 192445 |
+
{
|
| 192446 |
+
"epoch": 589.96,
|
| 192447 |
+
"learning_rate": 8.843942307692308e-06,
|
| 192448 |
+
"loss": 0.5648,
|
| 192449 |
+
"step": 73165
|
| 192450 |
+
},
|
| 192451 |
+
{
|
| 192452 |
+
"epoch": 590.0,
|
| 192453 |
+
"eval_loss": 0.3960420787334442,
|
| 192454 |
+
"eval_runtime": 39.2969,
|
| 192455 |
+
"eval_samples_per_second": 21.376,
|
| 192456 |
+
"eval_steps_per_second": 0.687,
|
| 192457 |
+
"eval_wer": 0.19401051040241885,
|
| 192458 |
+
"step": 73169
|
| 192459 |
+
},
|
| 192460 |
+
{
|
| 192461 |
+
"epoch": 585.01,
|
| 192462 |
+
"learning_rate": 8.84386217948718e-06,
|
| 192463 |
+
"loss": 0.4584,
|
| 192464 |
+
"step": 73170
|
| 192465 |
+
},
|
| 192466 |
+
{
|
| 192467 |
+
"epoch": 585.05,
|
| 192468 |
+
"learning_rate": 8.843782051282052e-06,
|
| 192469 |
+
"loss": 0.3261,
|
| 192470 |
+
"step": 73175
|
| 192471 |
+
},
|
| 192472 |
+
{
|
| 192473 |
+
"epoch": 585.09,
|
| 192474 |
+
"learning_rate": 8.843701923076923e-06,
|
| 192475 |
+
"loss": 0.3012,
|
| 192476 |
+
"step": 73180
|
| 192477 |
+
},
|
| 192478 |
+
{
|
| 192479 |
+
"epoch": 585.13,
|
| 192480 |
+
"learning_rate": 8.843621794871795e-06,
|
| 192481 |
+
"loss": 0.3411,
|
| 192482 |
+
"step": 73185
|
| 192483 |
+
},
|
| 192484 |
+
{
|
| 192485 |
+
"epoch": 585.17,
|
| 192486 |
+
"learning_rate": 8.843541666666668e-06,
|
| 192487 |
+
"loss": 0.5135,
|
| 192488 |
+
"step": 73190
|
| 192489 |
+
},
|
| 192490 |
+
{
|
| 192491 |
+
"epoch": 585.21,
|
| 192492 |
+
"learning_rate": 8.843461538461539e-06,
|
| 192493 |
+
"loss": 1.1694,
|
| 192494 |
+
"step": 73195
|
| 192495 |
+
},
|
| 192496 |
+
{
|
| 192497 |
+
"epoch": 585.25,
|
| 192498 |
+
"learning_rate": 8.84338141025641e-06,
|
| 192499 |
+
"loss": 0.3068,
|
| 192500 |
+
"step": 73200
|
| 192501 |
+
},
|
| 192502 |
+
{
|
| 192503 |
+
"epoch": 585.29,
|
| 192504 |
+
"learning_rate": 8.843301282051283e-06,
|
| 192505 |
+
"loss": 0.3249,
|
| 192506 |
+
"step": 73205
|
| 192507 |
+
},
|
| 192508 |
+
{
|
| 192509 |
+
"epoch": 585.33,
|
| 192510 |
+
"learning_rate": 8.843221153846155e-06,
|
| 192511 |
+
"loss": 0.3034,
|
| 192512 |
+
"step": 73210
|
| 192513 |
+
},
|
| 192514 |
+
{
|
| 192515 |
+
"epoch": 585.37,
|
| 192516 |
+
"learning_rate": 8.843141025641026e-06,
|
| 192517 |
+
"loss": 0.5366,
|
| 192518 |
+
"step": 73215
|
| 192519 |
+
},
|
| 192520 |
+
{
|
| 192521 |
+
"epoch": 585.41,
|
| 192522 |
+
"learning_rate": 8.843060897435898e-06,
|
| 192523 |
+
"loss": 1.1144,
|
| 192524 |
+
"step": 73220
|
| 192525 |
+
},
|
| 192526 |
+
{
|
| 192527 |
+
"epoch": 585.45,
|
| 192528 |
+
"learning_rate": 8.84298076923077e-06,
|
| 192529 |
+
"loss": 0.3196,
|
| 192530 |
+
"step": 73225
|
| 192531 |
+
},
|
| 192532 |
+
{
|
| 192533 |
+
"epoch": 585.49,
|
| 192534 |
+
"learning_rate": 8.842900641025642e-06,
|
| 192535 |
+
"loss": 0.2776,
|
| 192536 |
+
"step": 73230
|
| 192537 |
+
},
|
| 192538 |
+
{
|
| 192539 |
+
"epoch": 585.53,
|
| 192540 |
+
"learning_rate": 8.842820512820513e-06,
|
| 192541 |
+
"loss": 0.3681,
|
| 192542 |
+
"step": 73235
|
| 192543 |
+
},
|
| 192544 |
+
{
|
| 192545 |
+
"epoch": 585.57,
|
| 192546 |
+
"learning_rate": 8.842740384615386e-06,
|
| 192547 |
+
"loss": 0.5304,
|
| 192548 |
+
"step": 73240
|
| 192549 |
+
},
|
| 192550 |
+
{
|
| 192551 |
+
"epoch": 585.61,
|
| 192552 |
+
"learning_rate": 8.842660256410258e-06,
|
| 192553 |
+
"loss": 1.0968,
|
| 192554 |
+
"step": 73245
|
| 192555 |
+
},
|
| 192556 |
+
{
|
| 192557 |
+
"epoch": 585.65,
|
| 192558 |
+
"learning_rate": 8.842580128205129e-06,
|
| 192559 |
+
"loss": 0.3225,
|
| 192560 |
+
"step": 73250
|
| 192561 |
+
},
|
| 192562 |
+
{
|
| 192563 |
+
"epoch": 585.69,
|
| 192564 |
+
"learning_rate": 8.8425e-06,
|
| 192565 |
+
"loss": 0.3974,
|
| 192566 |
+
"step": 73255
|
| 192567 |
+
},
|
| 192568 |
+
{
|
| 192569 |
+
"epoch": 585.73,
|
| 192570 |
+
"learning_rate": 8.842419871794874e-06,
|
| 192571 |
+
"loss": 0.3627,
|
| 192572 |
+
"step": 73260
|
| 192573 |
+
},
|
| 192574 |
+
{
|
| 192575 |
+
"epoch": 585.77,
|
| 192576 |
+
"learning_rate": 8.842339743589743e-06,
|
| 192577 |
+
"loss": 0.6047,
|
| 192578 |
+
"step": 73265
|
| 192579 |
+
},
|
| 192580 |
+
{
|
| 192581 |
+
"epoch": 585.81,
|
| 192582 |
+
"learning_rate": 8.842259615384616e-06,
|
| 192583 |
+
"loss": 1.4464,
|
| 192584 |
+
"step": 73270
|
| 192585 |
+
},
|
| 192586 |
+
{
|
| 192587 |
+
"epoch": 585.85,
|
| 192588 |
+
"learning_rate": 8.842179487179488e-06,
|
| 192589 |
+
"loss": 0.3017,
|
| 192590 |
+
"step": 73275
|
| 192591 |
+
},
|
| 192592 |
+
{
|
| 192593 |
+
"epoch": 585.89,
|
| 192594 |
+
"learning_rate": 8.842099358974359e-06,
|
| 192595 |
+
"loss": 0.3094,
|
| 192596 |
+
"step": 73280
|
| 192597 |
+
},
|
| 192598 |
+
{
|
| 192599 |
+
"epoch": 585.93,
|
| 192600 |
+
"learning_rate": 8.84201923076923e-06,
|
| 192601 |
+
"loss": 0.3192,
|
| 192602 |
+
"step": 73285
|
| 192603 |
+
},
|
| 192604 |
+
{
|
| 192605 |
+
"epoch": 585.97,
|
| 192606 |
+
"learning_rate": 8.841939102564103e-06,
|
| 192607 |
+
"loss": 0.592,
|
| 192608 |
+
"step": 73290
|
| 192609 |
+
},
|
| 192610 |
+
{
|
| 192611 |
+
"epoch": 586.0,
|
| 192612 |
+
"eval_loss": 0.44540178775787354,
|
| 192613 |
+
"eval_runtime": 41.2543,
|
| 192614 |
+
"eval_samples_per_second": 20.362,
|
| 192615 |
+
"eval_steps_per_second": 0.654,
|
| 192616 |
+
"eval_wer": 0.19331724793448857,
|
| 192617 |
+
"step": 73294
|
| 192618 |
}
|
| 192619 |
],
|
| 192620 |
"max_steps": 625000,
|
| 192621 |
"num_train_epochs": 5000,
|
| 192622 |
+
"total_flos": 2.0626163435744595e+20,
|
| 192623 |
"trial_name": null,
|
| 192624 |
"trial_params": null
|
| 192625 |
}
|
model-bin/finetune/base/{checkpoint-72674 → checkpoint-73294}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629869061.2637658/events.out.tfevents.1629869061.7e498afd5545.905.53
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6ae8b52d1cd0d21233e15daa0c3fd1f10410692d55db06adb0bad90d9bd14f04
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629869731.455241/events.out.tfevents.1629869731.7e498afd5545.905.55
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f7b7bba2db6b9aa26631c81409a7a206463883d621c7b84c66643a8fa0a16ac4
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629870382.7916713/events.out.tfevents.1629870382.7e498afd5545.905.57
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d6e6a23d33d554930a0d80fc0746a3000a6f3f14c8bb3842ebee55a1d8b1ac80
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629871038.1608121/events.out.tfevents.1629871038.7e498afd5545.905.59
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:97f08523b6dc5e662dc16ebc30317ef82b0be6dfd390188c36e6c2fbe39ccb62
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629871680.9885855/events.out.tfevents.1629871680.7e498afd5545.905.61
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:eb95ec1ce9f723a2f319980d8803678ca13d76e0d9bf34a758a9cf0fbff7f4c8
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629869061.7e498afd5545.905.52
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:db1917f0e53b299d476406dc2c4e430220decc4c355ed43929b974e548f41a5c
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629869731.7e498afd5545.905.54
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:142e02e858d8dafb0636532ea92a864b0da56c0bfe2dbd84abfd6af74ec9493c
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629870382.7e498afd5545.905.56
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e64d8a2f979a06e5f77a62fafba02e43e3c683e6fd3b4277f7583f5ff75d66b3
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629871038.7e498afd5545.905.58
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d783b37f9b96d3ce9ed3ff18f0a35d766c1a1e41750eca9b79e9ccda15e23af7
|
| 3 |
+
size 8462
|
model-bin/finetune/base/log/events.out.tfevents.1629871680.7e498afd5545.905.60
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1d523aa3e926f22f482c94d89c4d63a07b628e2e1859a7970fd874cfb8e35232
|
| 3 |
+
size 8622
|