"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-82758 → checkpoint-83380}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-82758 → checkpoint-83380}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-82758 → checkpoint-83380}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-82758 → checkpoint-83380}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-82758 → checkpoint-83380}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-82758 → checkpoint-83380}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-82758 → checkpoint-83380}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-82758 → checkpoint-83380}/trainer_state.json +799 -4
- model-bin/finetune/base/{checkpoint-82758 → checkpoint-83380}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629913855.2077327/events.out.tfevents.1629913855.7e498afd5545.7645.105 +3 -0
- model-bin/finetune/base/log/1629914338.7905462/events.out.tfevents.1629914338.7e498afd5545.7645.107 +3 -0
- model-bin/finetune/base/log/1629914817.1913228/events.out.tfevents.1629914817.7e498afd5545.7645.109 +3 -0
- model-bin/finetune/base/log/1629915283.024899/events.out.tfevents.1629915283.7e498afd5545.7645.111 +3 -0
- model-bin/finetune/base/log/1629915770.6633656/events.out.tfevents.1629915770.7e498afd5545.7645.113 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629913855.7e498afd5545.7645.104 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629914338.7e498afd5545.7645.106 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629914816.7e498afd5545.7645.108 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629915283.7e498afd5545.7645.110 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629915770.7e498afd5545.7645.112 +3 -0
model-bin/finetune/base/{checkpoint-82758 → checkpoint-83380}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-82758 → checkpoint-83380}/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:b276f2f3a47077f92fc0333d21295019a95e804483061f60147f452cd06796c8
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-82758 → checkpoint-83380}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-82758 → checkpoint-83380}/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:e150d4058b742423d6a3d8905d20bbd46aebc34f4aea638dc192e165384417a0
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-82758 → checkpoint-83380}/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:cf8582deb2ef872921fb1b62c7c3a810ececb4195311b741256563afed965d75
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-82758 → checkpoint-83380}/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:a40d2ee1a779593686cdc965ce844cba10e21121620394ebb1bae81f9e7233d0
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-82758 → checkpoint-83380}/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 sha256:6dd620a2633f240794c55035a2b1205526f555e327ba42741abe3ab03cd07ee4
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-82758 → checkpoint-83380}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1822989349981638,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-79148",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -204657,11 +204657,806 @@
|
|
| 204657 |
"eval_steps_per_second": 0.652,
|
| 204658 |
"eval_wer": 0.18659354931813263,
|
| 204659 |
"step": 82758
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 204660 |
}
|
| 204661 |
],
|
| 204662 |
-
"max_steps":
|
| 204663 |
"num_train_epochs": 5000,
|
| 204664 |
-
"total_flos": 2.
|
| 204665 |
"trial_name": null,
|
| 204666 |
"trial_params": null
|
| 204667 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1822989349981638,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-79148",
|
| 4 |
+
"epoch": 672.0,
|
| 5 |
+
"global_step": 83380,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 204657 |
"eval_steps_per_second": 0.652,
|
| 204658 |
"eval_wer": 0.18659354931813263,
|
| 204659 |
"step": 82758
|
| 204660 |
+
},
|
| 204661 |
+
{
|
| 204662 |
+
"epoch": 667.02,
|
| 204663 |
+
"learning_rate": 8.679693053311795e-06,
|
| 204664 |
+
"loss": 0.3797,
|
| 204665 |
+
"step": 82760
|
| 204666 |
+
},
|
| 204667 |
+
{
|
| 204668 |
+
"epoch": 667.06,
|
| 204669 |
+
"learning_rate": 8.679612277867529e-06,
|
| 204670 |
+
"loss": 0.2817,
|
| 204671 |
+
"step": 82765
|
| 204672 |
+
},
|
| 204673 |
+
{
|
| 204674 |
+
"epoch": 667.1,
|
| 204675 |
+
"learning_rate": 8.679531502423265e-06,
|
| 204676 |
+
"loss": 0.3225,
|
| 204677 |
+
"step": 82770
|
| 204678 |
+
},
|
| 204679 |
+
{
|
| 204680 |
+
"epoch": 667.14,
|
| 204681 |
+
"learning_rate": 8.679450726978999e-06,
|
| 204682 |
+
"loss": 0.3512,
|
| 204683 |
+
"step": 82775
|
| 204684 |
+
},
|
| 204685 |
+
{
|
| 204686 |
+
"epoch": 667.18,
|
| 204687 |
+
"learning_rate": 8.679369951534735e-06,
|
| 204688 |
+
"loss": 0.7157,
|
| 204689 |
+
"step": 82780
|
| 204690 |
+
},
|
| 204691 |
+
{
|
| 204692 |
+
"epoch": 667.22,
|
| 204693 |
+
"learning_rate": 8.679289176090469e-06,
|
| 204694 |
+
"loss": 1.0196,
|
| 204695 |
+
"step": 82785
|
| 204696 |
+
},
|
| 204697 |
+
{
|
| 204698 |
+
"epoch": 667.26,
|
| 204699 |
+
"learning_rate": 8.679208400646205e-06,
|
| 204700 |
+
"loss": 0.3187,
|
| 204701 |
+
"step": 82790
|
| 204702 |
+
},
|
| 204703 |
+
{
|
| 204704 |
+
"epoch": 667.3,
|
| 204705 |
+
"learning_rate": 8.67912762520194e-06,
|
| 204706 |
+
"loss": 0.3182,
|
| 204707 |
+
"step": 82795
|
| 204708 |
+
},
|
| 204709 |
+
{
|
| 204710 |
+
"epoch": 667.34,
|
| 204711 |
+
"learning_rate": 8.679046849757675e-06,
|
| 204712 |
+
"loss": 0.3719,
|
| 204713 |
+
"step": 82800
|
| 204714 |
+
},
|
| 204715 |
+
{
|
| 204716 |
+
"epoch": 667.38,
|
| 204717 |
+
"learning_rate": 8.67896607431341e-06,
|
| 204718 |
+
"loss": 0.7065,
|
| 204719 |
+
"step": 82805
|
| 204720 |
+
},
|
| 204721 |
+
{
|
| 204722 |
+
"epoch": 667.42,
|
| 204723 |
+
"learning_rate": 8.678885298869145e-06,
|
| 204724 |
+
"loss": 1.0294,
|
| 204725 |
+
"step": 82810
|
| 204726 |
+
},
|
| 204727 |
+
{
|
| 204728 |
+
"epoch": 667.46,
|
| 204729 |
+
"learning_rate": 8.67880452342488e-06,
|
| 204730 |
+
"loss": 0.3363,
|
| 204731 |
+
"step": 82815
|
| 204732 |
+
},
|
| 204733 |
+
{
|
| 204734 |
+
"epoch": 667.5,
|
| 204735 |
+
"learning_rate": 8.678723747980615e-06,
|
| 204736 |
+
"loss": 0.2837,
|
| 204737 |
+
"step": 82820
|
| 204738 |
+
},
|
| 204739 |
+
{
|
| 204740 |
+
"epoch": 667.54,
|
| 204741 |
+
"learning_rate": 8.67864297253635e-06,
|
| 204742 |
+
"loss": 0.3586,
|
| 204743 |
+
"step": 82825
|
| 204744 |
+
},
|
| 204745 |
+
{
|
| 204746 |
+
"epoch": 667.58,
|
| 204747 |
+
"learning_rate": 8.678562197092085e-06,
|
| 204748 |
+
"loss": 0.6857,
|
| 204749 |
+
"step": 82830
|
| 204750 |
+
},
|
| 204751 |
+
{
|
| 204752 |
+
"epoch": 667.62,
|
| 204753 |
+
"learning_rate": 8.67848142164782e-06,
|
| 204754 |
+
"loss": 0.9368,
|
| 204755 |
+
"step": 82835
|
| 204756 |
+
},
|
| 204757 |
+
{
|
| 204758 |
+
"epoch": 667.66,
|
| 204759 |
+
"learning_rate": 8.678400646203555e-06,
|
| 204760 |
+
"loss": 0.327,
|
| 204761 |
+
"step": 82840
|
| 204762 |
+
},
|
| 204763 |
+
{
|
| 204764 |
+
"epoch": 667.7,
|
| 204765 |
+
"learning_rate": 8.67831987075929e-06,
|
| 204766 |
+
"loss": 0.3053,
|
| 204767 |
+
"step": 82845
|
| 204768 |
+
},
|
| 204769 |
+
{
|
| 204770 |
+
"epoch": 667.74,
|
| 204771 |
+
"learning_rate": 8.678239095315024e-06,
|
| 204772 |
+
"loss": 0.3586,
|
| 204773 |
+
"step": 82850
|
| 204774 |
+
},
|
| 204775 |
+
{
|
| 204776 |
+
"epoch": 667.78,
|
| 204777 |
+
"learning_rate": 8.67815831987076e-06,
|
| 204778 |
+
"loss": 0.6174,
|
| 204779 |
+
"step": 82855
|
| 204780 |
+
},
|
| 204781 |
+
{
|
| 204782 |
+
"epoch": 667.82,
|
| 204783 |
+
"learning_rate": 8.678077544426494e-06,
|
| 204784 |
+
"loss": 0.9567,
|
| 204785 |
+
"step": 82860
|
| 204786 |
+
},
|
| 204787 |
+
{
|
| 204788 |
+
"epoch": 667.86,
|
| 204789 |
+
"learning_rate": 8.67799676898223e-06,
|
| 204790 |
+
"loss": 0.2918,
|
| 204791 |
+
"step": 82865
|
| 204792 |
+
},
|
| 204793 |
+
{
|
| 204794 |
+
"epoch": 667.9,
|
| 204795 |
+
"learning_rate": 8.677915993537966e-06,
|
| 204796 |
+
"loss": 0.3679,
|
| 204797 |
+
"step": 82870
|
| 204798 |
+
},
|
| 204799 |
+
{
|
| 204800 |
+
"epoch": 667.94,
|
| 204801 |
+
"learning_rate": 8.6778352180937e-06,
|
| 204802 |
+
"loss": 0.3969,
|
| 204803 |
+
"step": 82875
|
| 204804 |
+
},
|
| 204805 |
+
{
|
| 204806 |
+
"epoch": 667.98,
|
| 204807 |
+
"learning_rate": 8.677754442649436e-06,
|
| 204808 |
+
"loss": 0.7097,
|
| 204809 |
+
"step": 82880
|
| 204810 |
+
},
|
| 204811 |
+
{
|
| 204812 |
+
"epoch": 668.0,
|
| 204813 |
+
"eval_loss": 0.3980284035205841,
|
| 204814 |
+
"eval_runtime": 43.2095,
|
| 204815 |
+
"eval_samples_per_second": 19.463,
|
| 204816 |
+
"eval_steps_per_second": 0.625,
|
| 204817 |
+
"eval_wer": 0.18870289801637047,
|
| 204818 |
+
"step": 82882
|
| 204819 |
+
},
|
| 204820 |
+
{
|
| 204821 |
+
"epoch": 668.02,
|
| 204822 |
+
"learning_rate": 8.67767366720517e-06,
|
| 204823 |
+
"loss": 0.3909,
|
| 204824 |
+
"step": 82885
|
| 204825 |
+
},
|
| 204826 |
+
{
|
| 204827 |
+
"epoch": 668.06,
|
| 204828 |
+
"learning_rate": 8.677592891760906e-06,
|
| 204829 |
+
"loss": 0.2988,
|
| 204830 |
+
"step": 82890
|
| 204831 |
+
},
|
| 204832 |
+
{
|
| 204833 |
+
"epoch": 668.1,
|
| 204834 |
+
"learning_rate": 8.67751211631664e-06,
|
| 204835 |
+
"loss": 0.4776,
|
| 204836 |
+
"step": 82895
|
| 204837 |
+
},
|
| 204838 |
+
{
|
| 204839 |
+
"epoch": 668.14,
|
| 204840 |
+
"learning_rate": 8.677431340872376e-06,
|
| 204841 |
+
"loss": 0.38,
|
| 204842 |
+
"step": 82900
|
| 204843 |
+
},
|
| 204844 |
+
{
|
| 204845 |
+
"epoch": 668.18,
|
| 204846 |
+
"learning_rate": 8.67735056542811e-06,
|
| 204847 |
+
"loss": 0.7143,
|
| 204848 |
+
"step": 82905
|
| 204849 |
+
},
|
| 204850 |
+
{
|
| 204851 |
+
"epoch": 668.22,
|
| 204852 |
+
"learning_rate": 8.677269789983846e-06,
|
| 204853 |
+
"loss": 0.9613,
|
| 204854 |
+
"step": 82910
|
| 204855 |
+
},
|
| 204856 |
+
{
|
| 204857 |
+
"epoch": 668.27,
|
| 204858 |
+
"learning_rate": 8.67718901453958e-06,
|
| 204859 |
+
"loss": 0.2899,
|
| 204860 |
+
"step": 82915
|
| 204861 |
+
},
|
| 204862 |
+
{
|
| 204863 |
+
"epoch": 668.31,
|
| 204864 |
+
"learning_rate": 8.677108239095316e-06,
|
| 204865 |
+
"loss": 0.3208,
|
| 204866 |
+
"step": 82920
|
| 204867 |
+
},
|
| 204868 |
+
{
|
| 204869 |
+
"epoch": 668.35,
|
| 204870 |
+
"learning_rate": 8.67702746365105e-06,
|
| 204871 |
+
"loss": 0.4151,
|
| 204872 |
+
"step": 82925
|
| 204873 |
+
},
|
| 204874 |
+
{
|
| 204875 |
+
"epoch": 668.39,
|
| 204876 |
+
"learning_rate": 8.676946688206786e-06,
|
| 204877 |
+
"loss": 0.7929,
|
| 204878 |
+
"step": 82930
|
| 204879 |
+
},
|
| 204880 |
+
{
|
| 204881 |
+
"epoch": 668.43,
|
| 204882 |
+
"learning_rate": 8.676865912762522e-06,
|
| 204883 |
+
"loss": 0.8304,
|
| 204884 |
+
"step": 82935
|
| 204885 |
+
},
|
| 204886 |
+
{
|
| 204887 |
+
"epoch": 668.47,
|
| 204888 |
+
"learning_rate": 8.676785137318256e-06,
|
| 204889 |
+
"loss": 0.3072,
|
| 204890 |
+
"step": 82940
|
| 204891 |
+
},
|
| 204892 |
+
{
|
| 204893 |
+
"epoch": 668.51,
|
| 204894 |
+
"learning_rate": 8.676704361873992e-06,
|
| 204895 |
+
"loss": 0.363,
|
| 204896 |
+
"step": 82945
|
| 204897 |
+
},
|
| 204898 |
+
{
|
| 204899 |
+
"epoch": 668.55,
|
| 204900 |
+
"learning_rate": 8.676623586429726e-06,
|
| 204901 |
+
"loss": 0.3762,
|
| 204902 |
+
"step": 82950
|
| 204903 |
+
},
|
| 204904 |
+
{
|
| 204905 |
+
"epoch": 668.59,
|
| 204906 |
+
"learning_rate": 8.676542810985462e-06,
|
| 204907 |
+
"loss": 0.7142,
|
| 204908 |
+
"step": 82955
|
| 204909 |
+
},
|
| 204910 |
+
{
|
| 204911 |
+
"epoch": 668.63,
|
| 204912 |
+
"learning_rate": 8.676462035541196e-06,
|
| 204913 |
+
"loss": 0.9227,
|
| 204914 |
+
"step": 82960
|
| 204915 |
+
},
|
| 204916 |
+
{
|
| 204917 |
+
"epoch": 668.67,
|
| 204918 |
+
"learning_rate": 8.676381260096932e-06,
|
| 204919 |
+
"loss": 0.254,
|
| 204920 |
+
"step": 82965
|
| 204921 |
+
},
|
| 204922 |
+
{
|
| 204923 |
+
"epoch": 668.71,
|
| 204924 |
+
"learning_rate": 8.676300484652666e-06,
|
| 204925 |
+
"loss": 0.2961,
|
| 204926 |
+
"step": 82970
|
| 204927 |
+
},
|
| 204928 |
+
{
|
| 204929 |
+
"epoch": 668.75,
|
| 204930 |
+
"learning_rate": 8.676219709208402e-06,
|
| 204931 |
+
"loss": 0.3838,
|
| 204932 |
+
"step": 82975
|
| 204933 |
+
},
|
| 204934 |
+
{
|
| 204935 |
+
"epoch": 668.79,
|
| 204936 |
+
"learning_rate": 8.676138933764136e-06,
|
| 204937 |
+
"loss": 0.7836,
|
| 204938 |
+
"step": 82980
|
| 204939 |
+
},
|
| 204940 |
+
{
|
| 204941 |
+
"epoch": 668.83,
|
| 204942 |
+
"learning_rate": 8.676058158319872e-06,
|
| 204943 |
+
"loss": 0.908,
|
| 204944 |
+
"step": 82985
|
| 204945 |
+
},
|
| 204946 |
+
{
|
| 204947 |
+
"epoch": 668.87,
|
| 204948 |
+
"learning_rate": 8.675977382875606e-06,
|
| 204949 |
+
"loss": 0.2685,
|
| 204950 |
+
"step": 82990
|
| 204951 |
+
},
|
| 204952 |
+
{
|
| 204953 |
+
"epoch": 668.91,
|
| 204954 |
+
"learning_rate": 8.675896607431342e-06,
|
| 204955 |
+
"loss": 0.3392,
|
| 204956 |
+
"step": 82995
|
| 204957 |
+
},
|
| 204958 |
+
{
|
| 204959 |
+
"epoch": 668.95,
|
| 204960 |
+
"learning_rate": 8.675815831987078e-06,
|
| 204961 |
+
"loss": 0.3687,
|
| 204962 |
+
"step": 83000
|
| 204963 |
+
},
|
| 204964 |
+
{
|
| 204965 |
+
"epoch": 668.99,
|
| 204966 |
+
"learning_rate": 8.675735056542812e-06,
|
| 204967 |
+
"loss": 0.7767,
|
| 204968 |
+
"step": 83005
|
| 204969 |
+
},
|
| 204970 |
+
{
|
| 204971 |
+
"epoch": 669.0,
|
| 204972 |
+
"eval_loss": 0.3812508285045624,
|
| 204973 |
+
"eval_runtime": 43.7044,
|
| 204974 |
+
"eval_samples_per_second": 19.243,
|
| 204975 |
+
"eval_steps_per_second": 0.618,
|
| 204976 |
+
"eval_wer": 0.1953050958664431,
|
| 204977 |
+
"step": 83006
|
| 204978 |
+
},
|
| 204979 |
+
{
|
| 204980 |
+
"epoch": 664.03,
|
| 204981 |
+
"learning_rate": 8.675654281098547e-06,
|
| 204982 |
+
"loss": 0.3748,
|
| 204983 |
+
"step": 83010
|
| 204984 |
+
},
|
| 204985 |
+
{
|
| 204986 |
+
"epoch": 664.07,
|
| 204987 |
+
"learning_rate": 8.675573505654282e-06,
|
| 204988 |
+
"loss": 0.3215,
|
| 204989 |
+
"step": 83015
|
| 204990 |
+
},
|
| 204991 |
+
{
|
| 204992 |
+
"epoch": 664.11,
|
| 204993 |
+
"learning_rate": 8.675492730210017e-06,
|
| 204994 |
+
"loss": 0.3078,
|
| 204995 |
+
"step": 83020
|
| 204996 |
+
},
|
| 204997 |
+
{
|
| 204998 |
+
"epoch": 664.15,
|
| 204999 |
+
"learning_rate": 8.675411954765752e-06,
|
| 205000 |
+
"loss": 0.4927,
|
| 205001 |
+
"step": 83025
|
| 205002 |
+
},
|
| 205003 |
+
{
|
| 205004 |
+
"epoch": 664.19,
|
| 205005 |
+
"learning_rate": 8.675331179321487e-06,
|
| 205006 |
+
"loss": 0.813,
|
| 205007 |
+
"step": 83030
|
| 205008 |
+
},
|
| 205009 |
+
{
|
| 205010 |
+
"epoch": 664.23,
|
| 205011 |
+
"learning_rate": 8.675250403877222e-06,
|
| 205012 |
+
"loss": 0.8072,
|
| 205013 |
+
"step": 83035
|
| 205014 |
+
},
|
| 205015 |
+
{
|
| 205016 |
+
"epoch": 664.27,
|
| 205017 |
+
"learning_rate": 8.675169628432957e-06,
|
| 205018 |
+
"loss": 0.3009,
|
| 205019 |
+
"step": 83040
|
| 205020 |
+
},
|
| 205021 |
+
{
|
| 205022 |
+
"epoch": 664.31,
|
| 205023 |
+
"learning_rate": 8.675088852988692e-06,
|
| 205024 |
+
"loss": 0.3703,
|
| 205025 |
+
"step": 83045
|
| 205026 |
+
},
|
| 205027 |
+
{
|
| 205028 |
+
"epoch": 664.35,
|
| 205029 |
+
"learning_rate": 8.675008077544427e-06,
|
| 205030 |
+
"loss": 0.4237,
|
| 205031 |
+
"step": 83050
|
| 205032 |
+
},
|
| 205033 |
+
{
|
| 205034 |
+
"epoch": 664.39,
|
| 205035 |
+
"learning_rate": 8.674927302100162e-06,
|
| 205036 |
+
"loss": 0.8886,
|
| 205037 |
+
"step": 83055
|
| 205038 |
+
},
|
| 205039 |
+
{
|
| 205040 |
+
"epoch": 664.43,
|
| 205041 |
+
"learning_rate": 8.674846526655897e-06,
|
| 205042 |
+
"loss": 0.6111,
|
| 205043 |
+
"step": 83060
|
| 205044 |
+
},
|
| 205045 |
+
{
|
| 205046 |
+
"epoch": 664.47,
|
| 205047 |
+
"learning_rate": 8.674765751211632e-06,
|
| 205048 |
+
"loss": 0.3094,
|
| 205049 |
+
"step": 83065
|
| 205050 |
+
},
|
| 205051 |
+
{
|
| 205052 |
+
"epoch": 664.51,
|
| 205053 |
+
"learning_rate": 8.674684975767367e-06,
|
| 205054 |
+
"loss": 0.3109,
|
| 205055 |
+
"step": 83070
|
| 205056 |
+
},
|
| 205057 |
+
{
|
| 205058 |
+
"epoch": 664.55,
|
| 205059 |
+
"learning_rate": 8.674604200323103e-06,
|
| 205060 |
+
"loss": 0.4154,
|
| 205061 |
+
"step": 83075
|
| 205062 |
+
},
|
| 205063 |
+
{
|
| 205064 |
+
"epoch": 664.59,
|
| 205065 |
+
"learning_rate": 8.674523424878837e-06,
|
| 205066 |
+
"loss": 0.9017,
|
| 205067 |
+
"step": 83080
|
| 205068 |
+
},
|
| 205069 |
+
{
|
| 205070 |
+
"epoch": 664.63,
|
| 205071 |
+
"learning_rate": 8.674442649434573e-06,
|
| 205072 |
+
"loss": 0.654,
|
| 205073 |
+
"step": 83085
|
| 205074 |
+
},
|
| 205075 |
+
{
|
| 205076 |
+
"epoch": 664.67,
|
| 205077 |
+
"learning_rate": 8.674361873990307e-06,
|
| 205078 |
+
"loss": 0.3114,
|
| 205079 |
+
"step": 83090
|
| 205080 |
+
},
|
| 205081 |
+
{
|
| 205082 |
+
"epoch": 664.71,
|
| 205083 |
+
"learning_rate": 8.674281098546043e-06,
|
| 205084 |
+
"loss": 0.3278,
|
| 205085 |
+
"step": 83095
|
| 205086 |
+
},
|
| 205087 |
+
{
|
| 205088 |
+
"epoch": 664.75,
|
| 205089 |
+
"learning_rate": 8.674200323101777e-06,
|
| 205090 |
+
"loss": 0.3857,
|
| 205091 |
+
"step": 83100
|
| 205092 |
+
},
|
| 205093 |
+
{
|
| 205094 |
+
"epoch": 664.79,
|
| 205095 |
+
"learning_rate": 8.674119547657513e-06,
|
| 205096 |
+
"loss": 0.9024,
|
| 205097 |
+
"step": 83105
|
| 205098 |
+
},
|
| 205099 |
+
{
|
| 205100 |
+
"epoch": 664.83,
|
| 205101 |
+
"learning_rate": 8.674038772213247e-06,
|
| 205102 |
+
"loss": 0.6743,
|
| 205103 |
+
"step": 83110
|
| 205104 |
+
},
|
| 205105 |
+
{
|
| 205106 |
+
"epoch": 664.87,
|
| 205107 |
+
"learning_rate": 8.673957996768983e-06,
|
| 205108 |
+
"loss": 0.2987,
|
| 205109 |
+
"step": 83115
|
| 205110 |
+
},
|
| 205111 |
+
{
|
| 205112 |
+
"epoch": 664.91,
|
| 205113 |
+
"learning_rate": 8.673877221324717e-06,
|
| 205114 |
+
"loss": 0.4069,
|
| 205115 |
+
"step": 83120
|
| 205116 |
+
},
|
| 205117 |
+
{
|
| 205118 |
+
"epoch": 664.95,
|
| 205119 |
+
"learning_rate": 8.673796445880453e-06,
|
| 205120 |
+
"loss": 0.4953,
|
| 205121 |
+
"step": 83125
|
| 205122 |
+
},
|
| 205123 |
+
{
|
| 205124 |
+
"epoch": 664.99,
|
| 205125 |
+
"learning_rate": 8.673715670436187e-06,
|
| 205126 |
+
"loss": 1.0797,
|
| 205127 |
+
"step": 83130
|
| 205128 |
+
},
|
| 205129 |
+
{
|
| 205130 |
+
"epoch": 665.0,
|
| 205131 |
+
"eval_loss": 0.4016847014427185,
|
| 205132 |
+
"eval_runtime": 41.795,
|
| 205133 |
+
"eval_samples_per_second": 20.122,
|
| 205134 |
+
"eval_steps_per_second": 0.646,
|
| 205135 |
+
"eval_wer": 0.19428870449278612,
|
| 205136 |
+
"step": 83131
|
| 205137 |
+
},
|
| 205138 |
+
{
|
| 205139 |
+
"epoch": 665.03,
|
| 205140 |
+
"learning_rate": 8.673634894991923e-06,
|
| 205141 |
+
"loss": 0.3586,
|
| 205142 |
+
"step": 83135
|
| 205143 |
+
},
|
| 205144 |
+
{
|
| 205145 |
+
"epoch": 665.07,
|
| 205146 |
+
"learning_rate": 8.673554119547659e-06,
|
| 205147 |
+
"loss": 0.2978,
|
| 205148 |
+
"step": 83140
|
| 205149 |
+
},
|
| 205150 |
+
{
|
| 205151 |
+
"epoch": 665.11,
|
| 205152 |
+
"learning_rate": 8.673473344103393e-06,
|
| 205153 |
+
"loss": 0.2852,
|
| 205154 |
+
"step": 83145
|
| 205155 |
+
},
|
| 205156 |
+
{
|
| 205157 |
+
"epoch": 665.15,
|
| 205158 |
+
"learning_rate": 8.673392568659129e-06,
|
| 205159 |
+
"loss": 0.4897,
|
| 205160 |
+
"step": 83150
|
| 205161 |
+
},
|
| 205162 |
+
{
|
| 205163 |
+
"epoch": 665.19,
|
| 205164 |
+
"learning_rate": 8.673311793214863e-06,
|
| 205165 |
+
"loss": 0.9942,
|
| 205166 |
+
"step": 83155
|
| 205167 |
+
},
|
| 205168 |
+
{
|
| 205169 |
+
"epoch": 665.23,
|
| 205170 |
+
"learning_rate": 8.673231017770599e-06,
|
| 205171 |
+
"loss": 0.6437,
|
| 205172 |
+
"step": 83160
|
| 205173 |
+
},
|
| 205174 |
+
{
|
| 205175 |
+
"epoch": 665.27,
|
| 205176 |
+
"learning_rate": 8.673150242326333e-06,
|
| 205177 |
+
"loss": 0.3044,
|
| 205178 |
+
"step": 83165
|
| 205179 |
+
},
|
| 205180 |
+
{
|
| 205181 |
+
"epoch": 665.31,
|
| 205182 |
+
"learning_rate": 8.673069466882069e-06,
|
| 205183 |
+
"loss": 0.3659,
|
| 205184 |
+
"step": 83170
|
| 205185 |
+
},
|
| 205186 |
+
{
|
| 205187 |
+
"epoch": 665.35,
|
| 205188 |
+
"learning_rate": 8.672988691437803e-06,
|
| 205189 |
+
"loss": 0.3964,
|
| 205190 |
+
"step": 83175
|
| 205191 |
+
},
|
| 205192 |
+
{
|
| 205193 |
+
"epoch": 665.39,
|
| 205194 |
+
"learning_rate": 8.672907915993539e-06,
|
| 205195 |
+
"loss": 1.1322,
|
| 205196 |
+
"step": 83180
|
| 205197 |
+
},
|
| 205198 |
+
{
|
| 205199 |
+
"epoch": 665.43,
|
| 205200 |
+
"learning_rate": 8.672827140549273e-06,
|
| 205201 |
+
"loss": 0.6599,
|
| 205202 |
+
"step": 83185
|
| 205203 |
+
},
|
| 205204 |
+
{
|
| 205205 |
+
"epoch": 665.47,
|
| 205206 |
+
"learning_rate": 8.672746365105009e-06,
|
| 205207 |
+
"loss": 0.2781,
|
| 205208 |
+
"step": 83190
|
| 205209 |
+
},
|
| 205210 |
+
{
|
| 205211 |
+
"epoch": 665.51,
|
| 205212 |
+
"learning_rate": 8.672665589660743e-06,
|
| 205213 |
+
"loss": 0.3381,
|
| 205214 |
+
"step": 83195
|
| 205215 |
+
},
|
| 205216 |
+
{
|
| 205217 |
+
"epoch": 665.55,
|
| 205218 |
+
"learning_rate": 8.672584814216479e-06,
|
| 205219 |
+
"loss": 0.3906,
|
| 205220 |
+
"step": 83200
|
| 205221 |
+
},
|
| 205222 |
+
{
|
| 205223 |
+
"epoch": 665.59,
|
| 205224 |
+
"learning_rate": 8.672504038772215e-06,
|
| 205225 |
+
"loss": 0.8723,
|
| 205226 |
+
"step": 83205
|
| 205227 |
+
},
|
| 205228 |
+
{
|
| 205229 |
+
"epoch": 665.63,
|
| 205230 |
+
"learning_rate": 8.672423263327949e-06,
|
| 205231 |
+
"loss": 0.5531,
|
| 205232 |
+
"step": 83210
|
| 205233 |
+
},
|
| 205234 |
+
{
|
| 205235 |
+
"epoch": 665.67,
|
| 205236 |
+
"learning_rate": 8.672342487883685e-06,
|
| 205237 |
+
"loss": 0.3006,
|
| 205238 |
+
"step": 83215
|
| 205239 |
+
},
|
| 205240 |
+
{
|
| 205241 |
+
"epoch": 665.71,
|
| 205242 |
+
"learning_rate": 8.672261712439419e-06,
|
| 205243 |
+
"loss": 0.2956,
|
| 205244 |
+
"step": 83220
|
| 205245 |
+
},
|
| 205246 |
+
{
|
| 205247 |
+
"epoch": 665.75,
|
| 205248 |
+
"learning_rate": 8.672180936995154e-06,
|
| 205249 |
+
"loss": 0.4412,
|
| 205250 |
+
"step": 83225
|
| 205251 |
+
},
|
| 205252 |
+
{
|
| 205253 |
+
"epoch": 665.79,
|
| 205254 |
+
"learning_rate": 8.672100161550889e-06,
|
| 205255 |
+
"loss": 1.1329,
|
| 205256 |
+
"step": 83230
|
| 205257 |
+
},
|
| 205258 |
+
{
|
| 205259 |
+
"epoch": 665.83,
|
| 205260 |
+
"learning_rate": 8.672019386106624e-06,
|
| 205261 |
+
"loss": 0.6726,
|
| 205262 |
+
"step": 83235
|
| 205263 |
+
},
|
| 205264 |
+
{
|
| 205265 |
+
"epoch": 665.87,
|
| 205266 |
+
"learning_rate": 8.671938610662359e-06,
|
| 205267 |
+
"loss": 0.3499,
|
| 205268 |
+
"step": 83240
|
| 205269 |
+
},
|
| 205270 |
+
{
|
| 205271 |
+
"epoch": 665.91,
|
| 205272 |
+
"learning_rate": 8.671857835218094e-06,
|
| 205273 |
+
"loss": 0.3695,
|
| 205274 |
+
"step": 83245
|
| 205275 |
+
},
|
| 205276 |
+
{
|
| 205277 |
+
"epoch": 665.95,
|
| 205278 |
+
"learning_rate": 8.671777059773829e-06,
|
| 205279 |
+
"loss": 0.4271,
|
| 205280 |
+
"step": 83250
|
| 205281 |
+
},
|
| 205282 |
+
{
|
| 205283 |
+
"epoch": 665.99,
|
| 205284 |
+
"learning_rate": 8.671696284329564e-06,
|
| 205285 |
+
"loss": 1.0571,
|
| 205286 |
+
"step": 83255
|
| 205287 |
+
},
|
| 205288 |
+
{
|
| 205289 |
+
"epoch": 666.0,
|
| 205290 |
+
"eval_loss": 0.3401903510093689,
|
| 205291 |
+
"eval_runtime": 43.4814,
|
| 205292 |
+
"eval_samples_per_second": 19.342,
|
| 205293 |
+
"eval_steps_per_second": 0.621,
|
| 205294 |
+
"eval_wer": 0.1920457042408262,
|
| 205295 |
+
"step": 83256
|
| 205296 |
+
},
|
| 205297 |
+
{
|
| 205298 |
+
"epoch": 671.03,
|
| 205299 |
+
"learning_rate": 8.671615508885299e-06,
|
| 205300 |
+
"loss": 0.36,
|
| 205301 |
+
"step": 83260
|
| 205302 |
+
},
|
| 205303 |
+
{
|
| 205304 |
+
"epoch": 671.07,
|
| 205305 |
+
"learning_rate": 8.671534733441034e-06,
|
| 205306 |
+
"loss": 0.3648,
|
| 205307 |
+
"step": 83265
|
| 205308 |
+
},
|
| 205309 |
+
{
|
| 205310 |
+
"epoch": 671.11,
|
| 205311 |
+
"learning_rate": 8.671453957996769e-06,
|
| 205312 |
+
"loss": 0.2804,
|
| 205313 |
+
"step": 83270
|
| 205314 |
+
},
|
| 205315 |
+
{
|
| 205316 |
+
"epoch": 671.15,
|
| 205317 |
+
"learning_rate": 8.671373182552504e-06,
|
| 205318 |
+
"loss": 0.3486,
|
| 205319 |
+
"step": 83275
|
| 205320 |
+
},
|
| 205321 |
+
{
|
| 205322 |
+
"epoch": 671.19,
|
| 205323 |
+
"learning_rate": 8.67129240710824e-06,
|
| 205324 |
+
"loss": 0.8378,
|
| 205325 |
+
"step": 83280
|
| 205326 |
+
},
|
| 205327 |
+
{
|
| 205328 |
+
"epoch": 671.23,
|
| 205329 |
+
"learning_rate": 8.671211631663974e-06,
|
| 205330 |
+
"loss": 0.656,
|
| 205331 |
+
"step": 83285
|
| 205332 |
+
},
|
| 205333 |
+
{
|
| 205334 |
+
"epoch": 671.27,
|
| 205335 |
+
"learning_rate": 8.67113085621971e-06,
|
| 205336 |
+
"loss": 0.3132,
|
| 205337 |
+
"step": 83290
|
| 205338 |
+
},
|
| 205339 |
+
{
|
| 205340 |
+
"epoch": 671.31,
|
| 205341 |
+
"learning_rate": 8.671050080775444e-06,
|
| 205342 |
+
"loss": 0.329,
|
| 205343 |
+
"step": 83295
|
| 205344 |
+
},
|
| 205345 |
+
{
|
| 205346 |
+
"epoch": 671.35,
|
| 205347 |
+
"learning_rate": 8.67096930533118e-06,
|
| 205348 |
+
"loss": 0.3911,
|
| 205349 |
+
"step": 83300
|
| 205350 |
+
},
|
| 205351 |
+
{
|
| 205352 |
+
"epoch": 671.4,
|
| 205353 |
+
"learning_rate": 8.670888529886914e-06,
|
| 205354 |
+
"loss": 0.92,
|
| 205355 |
+
"step": 83305
|
| 205356 |
+
},
|
| 205357 |
+
{
|
| 205358 |
+
"epoch": 671.44,
|
| 205359 |
+
"learning_rate": 8.67080775444265e-06,
|
| 205360 |
+
"loss": 0.7216,
|
| 205361 |
+
"step": 83310
|
| 205362 |
+
},
|
| 205363 |
+
{
|
| 205364 |
+
"epoch": 671.48,
|
| 205365 |
+
"learning_rate": 8.670726978998384e-06,
|
| 205366 |
+
"loss": 0.3082,
|
| 205367 |
+
"step": 83315
|
| 205368 |
+
},
|
| 205369 |
+
{
|
| 205370 |
+
"epoch": 671.52,
|
| 205371 |
+
"learning_rate": 8.67064620355412e-06,
|
| 205372 |
+
"loss": 0.2757,
|
| 205373 |
+
"step": 83320
|
| 205374 |
+
},
|
| 205375 |
+
{
|
| 205376 |
+
"epoch": 671.56,
|
| 205377 |
+
"learning_rate": 8.670565428109854e-06,
|
| 205378 |
+
"loss": 0.3469,
|
| 205379 |
+
"step": 83325
|
| 205380 |
+
},
|
| 205381 |
+
{
|
| 205382 |
+
"epoch": 671.6,
|
| 205383 |
+
"learning_rate": 8.67048465266559e-06,
|
| 205384 |
+
"loss": 0.927,
|
| 205385 |
+
"step": 83330
|
| 205386 |
+
},
|
| 205387 |
+
{
|
| 205388 |
+
"epoch": 671.64,
|
| 205389 |
+
"learning_rate": 8.670403877221324e-06,
|
| 205390 |
+
"loss": 0.6574,
|
| 205391 |
+
"step": 83335
|
| 205392 |
+
},
|
| 205393 |
+
{
|
| 205394 |
+
"epoch": 671.68,
|
| 205395 |
+
"learning_rate": 8.67032310177706e-06,
|
| 205396 |
+
"loss": 0.2722,
|
| 205397 |
+
"step": 83340
|
| 205398 |
+
},
|
| 205399 |
+
{
|
| 205400 |
+
"epoch": 671.72,
|
| 205401 |
+
"learning_rate": 8.670242326332796e-06,
|
| 205402 |
+
"loss": 0.3259,
|
| 205403 |
+
"step": 83345
|
| 205404 |
+
},
|
| 205405 |
+
{
|
| 205406 |
+
"epoch": 671.76,
|
| 205407 |
+
"learning_rate": 8.67016155088853e-06,
|
| 205408 |
+
"loss": 0.5611,
|
| 205409 |
+
"step": 83350
|
| 205410 |
+
},
|
| 205411 |
+
{
|
| 205412 |
+
"epoch": 671.8,
|
| 205413 |
+
"learning_rate": 8.670080775444266e-06,
|
| 205414 |
+
"loss": 0.9038,
|
| 205415 |
+
"step": 83355
|
| 205416 |
+
},
|
| 205417 |
+
{
|
| 205418 |
+
"epoch": 671.84,
|
| 205419 |
+
"learning_rate": 8.67e-06,
|
| 205420 |
+
"loss": 0.6725,
|
| 205421 |
+
"step": 83360
|
| 205422 |
+
},
|
| 205423 |
+
{
|
| 205424 |
+
"epoch": 671.88,
|
| 205425 |
+
"learning_rate": 8.669919224555736e-06,
|
| 205426 |
+
"loss": 0.318,
|
| 205427 |
+
"step": 83365
|
| 205428 |
+
},
|
| 205429 |
+
{
|
| 205430 |
+
"epoch": 671.92,
|
| 205431 |
+
"learning_rate": 8.66983844911147e-06,
|
| 205432 |
+
"loss": 0.3709,
|
| 205433 |
+
"step": 83370
|
| 205434 |
+
},
|
| 205435 |
+
{
|
| 205436 |
+
"epoch": 671.96,
|
| 205437 |
+
"learning_rate": 8.669757673667206e-06,
|
| 205438 |
+
"loss": 0.6603,
|
| 205439 |
+
"step": 83375
|
| 205440 |
+
},
|
| 205441 |
+
{
|
| 205442 |
+
"epoch": 672.0,
|
| 205443 |
+
"learning_rate": 8.66967689822294e-06,
|
| 205444 |
+
"loss": 1.2931,
|
| 205445 |
+
"step": 83380
|
| 205446 |
+
},
|
| 205447 |
+
{
|
| 205448 |
+
"epoch": 672.0,
|
| 205449 |
+
"eval_loss": 0.38747358322143555,
|
| 205450 |
+
"eval_runtime": 42.6332,
|
| 205451 |
+
"eval_samples_per_second": 19.703,
|
| 205452 |
+
"eval_steps_per_second": 0.633,
|
| 205453 |
+
"eval_wer": 0.18727738994158713,
|
| 205454 |
+
"step": 83380
|
| 205455 |
}
|
| 205456 |
],
|
| 205457 |
+
"max_steps": 620000,
|
| 205458 |
"num_train_epochs": 5000,
|
| 205459 |
+
"total_flos": 2.3464586746857964e+20,
|
| 205460 |
"trial_name": null,
|
| 205461 |
"trial_params": null
|
| 205462 |
}
|
model-bin/finetune/base/{checkpoint-82758 → checkpoint-83380}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629913855.2077327/events.out.tfevents.1629913855.7e498afd5545.7645.105
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:bd7d57bb99bc036e6f910a2a349e76f01870eb581ffa553d533c1b2e971dffa1
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629914338.7905462/events.out.tfevents.1629914338.7e498afd5545.7645.107
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:56e7ace4228ed9bf7953062f877a3d48c91a91f5630876155f12d2bcb71c4ba7
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629914817.1913228/events.out.tfevents.1629914817.7e498afd5545.7645.109
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:10c6e843871b610798af4beaad6b0443f882357bb9efada1287c8d8243e1cd8a
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629915283.024899/events.out.tfevents.1629915283.7e498afd5545.7645.111
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:998caaf5beab3f7cc008e81af97ae13c035150b89aee4c50040ed932378a9b55
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629915770.6633656/events.out.tfevents.1629915770.7e498afd5545.7645.113
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8d551aa6282537b1b318d24d128a5df23c122c5016777cfa4353b23058ba11ef
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629913855.7e498afd5545.7645.104
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:40df1b23c8a809b79cd6b2dac3ff0e82a734683264727e88ac111c3ad78737aa
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629914338.7e498afd5545.7645.106
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a78c94ba0e3baf3e72d2b507fa842766ec7cfdadff3c778677d850012c29b650
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629914816.7e498afd5545.7645.108
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:cc915d901d810c5ada7d95a63aef974d0bb23a2f85d7e8f2076d891fae27b195
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629915283.7e498afd5545.7645.110
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:5b6eb499448bb00dfacefdebdfade2dc20bbe458e8efcda0a91ad022edc3593d
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629915770.7e498afd5545.7645.112
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2a4217420f5492b96def545e38ec7a8d16b473477f3e6143bb9f9afa2bba969c
|
| 3 |
+
size 8622
|