"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-84623 → checkpoint-85246}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-84623 → checkpoint-85246}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-84623 → checkpoint-85246}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-84623 → checkpoint-85246}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-84623 → checkpoint-85246}/rng_state.pth +2 -2
- model-bin/finetune/base/{checkpoint-84623 → checkpoint-85246}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-84623 → checkpoint-85246}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-84623 → checkpoint-85246}/trainer_state.json +799 -4
- model-bin/finetune/base/{checkpoint-84623 → checkpoint-85246}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629921107.3117447/events.out.tfevents.1629921107.7e498afd5545.7645.135 +3 -0
- model-bin/finetune/base/log/1629921575.424182/events.out.tfevents.1629921576.7e498afd5545.7645.137 +3 -0
- model-bin/finetune/base/log/1629922040.984044/events.out.tfevents.1629922040.7e498afd5545.7645.139 +3 -0
- model-bin/finetune/base/log/1629922507.7783706/events.out.tfevents.1629922507.7e498afd5545.7645.141 +3 -0
- model-bin/finetune/base/log/1629922973.5943346/events.out.tfevents.1629922973.7e498afd5545.7645.143 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629921107.7e498afd5545.7645.134 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629921575.7e498afd5545.7645.136 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629922040.7e498afd5545.7645.138 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629922507.7e498afd5545.7645.140 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629922973.7e498afd5545.7645.142 +3 -0
model-bin/finetune/base/{checkpoint-84623 → checkpoint-85246}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-84623 → checkpoint-85246}/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:d8a4c8a3acad0c35744c7976d66259b8eb2da920d292bbb305a8496c3d4f9938
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-84623 → checkpoint-85246}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-84623 → checkpoint-85246}/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 sha256:8a3e6c5186bce44cecdacbf35ca2aa71a591d8a49e5a647a85e938bc4b792364
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-84623 → checkpoint-85246}/rng_state.pth
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9504d28e16b3bd9f3f61072c69a9fbabae048efac1d3f4677801698df7fe5e4b
|
| 3 |
+
size 14567
|
model-bin/finetune/base/{checkpoint-84623 → checkpoint-85246}/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:a9a813586972eadf09b86db52b053870e4fc9467f90a7405719ca90ab7c8c325
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-84623 → checkpoint-85246}/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 sha256:915c20f8b02a8adcf54622251e3737e8298173f62e62d7c8e785af8ef8724fd8
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-84623 → checkpoint-85246}/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,
|
|
@@ -207030,11 +207030,806 @@
|
|
| 207030 |
"eval_steps_per_second": 0.643,
|
| 207031 |
"eval_wer": 0.1948430329958482,
|
| 207032 |
"step": 84623
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 207033 |
}
|
| 207034 |
],
|
| 207035 |
-
"max_steps":
|
| 207036 |
"num_train_epochs": 5000,
|
| 207037 |
-
"total_flos": 2.
|
| 207038 |
"trial_name": null,
|
| 207039 |
"trial_params": null
|
| 207040 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1822989349981638,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-79148",
|
| 4 |
+
"epoch": 676.0,
|
| 5 |
+
"global_step": 85246,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 207030 |
"eval_steps_per_second": 0.643,
|
| 207031 |
"eval_wer": 0.1948430329958482,
|
| 207032 |
"step": 84623
|
| 207033 |
+
},
|
| 207034 |
+
{
|
| 207035 |
+
"epoch": 682.02,
|
| 207036 |
+
"learning_rate": 8.64956381260097e-06,
|
| 207037 |
+
"loss": 0.4455,
|
| 207038 |
+
"step": 84625
|
| 207039 |
+
},
|
| 207040 |
+
{
|
| 207041 |
+
"epoch": 682.06,
|
| 207042 |
+
"learning_rate": 8.649483037156705e-06,
|
| 207043 |
+
"loss": 0.33,
|
| 207044 |
+
"step": 84630
|
| 207045 |
+
},
|
| 207046 |
+
{
|
| 207047 |
+
"epoch": 682.1,
|
| 207048 |
+
"learning_rate": 8.64940226171244e-06,
|
| 207049 |
+
"loss": 0.2829,
|
| 207050 |
+
"step": 84635
|
| 207051 |
+
},
|
| 207052 |
+
{
|
| 207053 |
+
"epoch": 682.14,
|
| 207054 |
+
"learning_rate": 8.649321486268175e-06,
|
| 207055 |
+
"loss": 0.398,
|
| 207056 |
+
"step": 84640
|
| 207057 |
+
},
|
| 207058 |
+
{
|
| 207059 |
+
"epoch": 682.18,
|
| 207060 |
+
"learning_rate": 8.649240710823911e-06,
|
| 207061 |
+
"loss": 0.5545,
|
| 207062 |
+
"step": 84645
|
| 207063 |
+
},
|
| 207064 |
+
{
|
| 207065 |
+
"epoch": 682.22,
|
| 207066 |
+
"learning_rate": 8.649159935379645e-06,
|
| 207067 |
+
"loss": 1.0411,
|
| 207068 |
+
"step": 84650
|
| 207069 |
+
},
|
| 207070 |
+
{
|
| 207071 |
+
"epoch": 682.26,
|
| 207072 |
+
"learning_rate": 8.649079159935381e-06,
|
| 207073 |
+
"loss": 0.3272,
|
| 207074 |
+
"step": 84655
|
| 207075 |
+
},
|
| 207076 |
+
{
|
| 207077 |
+
"epoch": 682.3,
|
| 207078 |
+
"learning_rate": 8.648998384491115e-06,
|
| 207079 |
+
"loss": 0.2905,
|
| 207080 |
+
"step": 84660
|
| 207081 |
+
},
|
| 207082 |
+
{
|
| 207083 |
+
"epoch": 682.34,
|
| 207084 |
+
"learning_rate": 8.648917609046851e-06,
|
| 207085 |
+
"loss": 0.3133,
|
| 207086 |
+
"step": 84665
|
| 207087 |
+
},
|
| 207088 |
+
{
|
| 207089 |
+
"epoch": 682.38,
|
| 207090 |
+
"learning_rate": 8.648836833602585e-06,
|
| 207091 |
+
"loss": 0.7321,
|
| 207092 |
+
"step": 84670
|
| 207093 |
+
},
|
| 207094 |
+
{
|
| 207095 |
+
"epoch": 682.42,
|
| 207096 |
+
"learning_rate": 8.648756058158321e-06,
|
| 207097 |
+
"loss": 1.0805,
|
| 207098 |
+
"step": 84675
|
| 207099 |
+
},
|
| 207100 |
+
{
|
| 207101 |
+
"epoch": 682.46,
|
| 207102 |
+
"learning_rate": 8.648675282714055e-06,
|
| 207103 |
+
"loss": 0.395,
|
| 207104 |
+
"step": 84680
|
| 207105 |
+
},
|
| 207106 |
+
{
|
| 207107 |
+
"epoch": 682.5,
|
| 207108 |
+
"learning_rate": 8.648594507269791e-06,
|
| 207109 |
+
"loss": 0.3281,
|
| 207110 |
+
"step": 84685
|
| 207111 |
+
},
|
| 207112 |
+
{
|
| 207113 |
+
"epoch": 682.54,
|
| 207114 |
+
"learning_rate": 8.648513731825525e-06,
|
| 207115 |
+
"loss": 0.4026,
|
| 207116 |
+
"step": 84690
|
| 207117 |
+
},
|
| 207118 |
+
{
|
| 207119 |
+
"epoch": 682.58,
|
| 207120 |
+
"learning_rate": 8.648432956381261e-06,
|
| 207121 |
+
"loss": 0.6815,
|
| 207122 |
+
"step": 84695
|
| 207123 |
+
},
|
| 207124 |
+
{
|
| 207125 |
+
"epoch": 682.62,
|
| 207126 |
+
"learning_rate": 8.648352180936997e-06,
|
| 207127 |
+
"loss": 0.9631,
|
| 207128 |
+
"step": 84700
|
| 207129 |
+
},
|
| 207130 |
+
{
|
| 207131 |
+
"epoch": 682.66,
|
| 207132 |
+
"learning_rate": 8.648271405492731e-06,
|
| 207133 |
+
"loss": 0.2653,
|
| 207134 |
+
"step": 84705
|
| 207135 |
+
},
|
| 207136 |
+
{
|
| 207137 |
+
"epoch": 682.7,
|
| 207138 |
+
"learning_rate": 8.648190630048467e-06,
|
| 207139 |
+
"loss": 0.306,
|
| 207140 |
+
"step": 84710
|
| 207141 |
+
},
|
| 207142 |
+
{
|
| 207143 |
+
"epoch": 682.74,
|
| 207144 |
+
"learning_rate": 8.648109854604201e-06,
|
| 207145 |
+
"loss": 0.354,
|
| 207146 |
+
"step": 84715
|
| 207147 |
+
},
|
| 207148 |
+
{
|
| 207149 |
+
"epoch": 682.78,
|
| 207150 |
+
"learning_rate": 8.648029079159937e-06,
|
| 207151 |
+
"loss": 0.5983,
|
| 207152 |
+
"step": 84720
|
| 207153 |
+
},
|
| 207154 |
+
{
|
| 207155 |
+
"epoch": 682.82,
|
| 207156 |
+
"learning_rate": 8.647948303715671e-06,
|
| 207157 |
+
"loss": 1.1133,
|
| 207158 |
+
"step": 84725
|
| 207159 |
+
},
|
| 207160 |
+
{
|
| 207161 |
+
"epoch": 682.86,
|
| 207162 |
+
"learning_rate": 8.647867528271407e-06,
|
| 207163 |
+
"loss": 0.3002,
|
| 207164 |
+
"step": 84730
|
| 207165 |
+
},
|
| 207166 |
+
{
|
| 207167 |
+
"epoch": 682.9,
|
| 207168 |
+
"learning_rate": 8.647786752827141e-06,
|
| 207169 |
+
"loss": 0.3107,
|
| 207170 |
+
"step": 84735
|
| 207171 |
+
},
|
| 207172 |
+
{
|
| 207173 |
+
"epoch": 682.94,
|
| 207174 |
+
"learning_rate": 8.647705977382877e-06,
|
| 207175 |
+
"loss": 0.3594,
|
| 207176 |
+
"step": 84740
|
| 207177 |
+
},
|
| 207178 |
+
{
|
| 207179 |
+
"epoch": 682.98,
|
| 207180 |
+
"learning_rate": 8.647625201938611e-06,
|
| 207181 |
+
"loss": 0.7494,
|
| 207182 |
+
"step": 84745
|
| 207183 |
+
},
|
| 207184 |
+
{
|
| 207185 |
+
"epoch": 683.0,
|
| 207186 |
+
"eval_loss": 0.3968411087989807,
|
| 207187 |
+
"eval_runtime": 42.8582,
|
| 207188 |
+
"eval_samples_per_second": 19.623,
|
| 207189 |
+
"eval_steps_per_second": 0.63,
|
| 207190 |
+
"eval_wer": 0.18702648098708421,
|
| 207191 |
+
"step": 84747
|
| 207192 |
+
},
|
| 207193 |
+
{
|
| 207194 |
+
"epoch": 677.02,
|
| 207195 |
+
"learning_rate": 8.647544426494347e-06,
|
| 207196 |
+
"loss": 0.4038,
|
| 207197 |
+
"step": 84750
|
| 207198 |
+
},
|
| 207199 |
+
{
|
| 207200 |
+
"epoch": 677.06,
|
| 207201 |
+
"learning_rate": 8.647463651050083e-06,
|
| 207202 |
+
"loss": 0.2545,
|
| 207203 |
+
"step": 84755
|
| 207204 |
+
},
|
| 207205 |
+
{
|
| 207206 |
+
"epoch": 677.1,
|
| 207207 |
+
"learning_rate": 8.647382875605817e-06,
|
| 207208 |
+
"loss": 0.3456,
|
| 207209 |
+
"step": 84760
|
| 207210 |
+
},
|
| 207211 |
+
{
|
| 207212 |
+
"epoch": 677.14,
|
| 207213 |
+
"learning_rate": 8.647302100161553e-06,
|
| 207214 |
+
"loss": 0.3674,
|
| 207215 |
+
"step": 84765
|
| 207216 |
+
},
|
| 207217 |
+
{
|
| 207218 |
+
"epoch": 677.18,
|
| 207219 |
+
"learning_rate": 8.647221324717287e-06,
|
| 207220 |
+
"loss": 0.7989,
|
| 207221 |
+
"step": 84770
|
| 207222 |
+
},
|
| 207223 |
+
{
|
| 207224 |
+
"epoch": 677.22,
|
| 207225 |
+
"learning_rate": 8.647140549273023e-06,
|
| 207226 |
+
"loss": 0.9935,
|
| 207227 |
+
"step": 84775
|
| 207228 |
+
},
|
| 207229 |
+
{
|
| 207230 |
+
"epoch": 677.26,
|
| 207231 |
+
"learning_rate": 8.647059773828757e-06,
|
| 207232 |
+
"loss": 0.2908,
|
| 207233 |
+
"step": 84780
|
| 207234 |
+
},
|
| 207235 |
+
{
|
| 207236 |
+
"epoch": 677.3,
|
| 207237 |
+
"learning_rate": 8.646978998384493e-06,
|
| 207238 |
+
"loss": 0.3214,
|
| 207239 |
+
"step": 84785
|
| 207240 |
+
},
|
| 207241 |
+
{
|
| 207242 |
+
"epoch": 677.34,
|
| 207243 |
+
"learning_rate": 8.646898222940227e-06,
|
| 207244 |
+
"loss": 0.405,
|
| 207245 |
+
"step": 84790
|
| 207246 |
+
},
|
| 207247 |
+
{
|
| 207248 |
+
"epoch": 677.38,
|
| 207249 |
+
"learning_rate": 8.646817447495963e-06,
|
| 207250 |
+
"loss": 0.7544,
|
| 207251 |
+
"step": 84795
|
| 207252 |
+
},
|
| 207253 |
+
{
|
| 207254 |
+
"epoch": 677.42,
|
| 207255 |
+
"learning_rate": 8.646736672051697e-06,
|
| 207256 |
+
"loss": 0.7998,
|
| 207257 |
+
"step": 84800
|
| 207258 |
+
},
|
| 207259 |
+
{
|
| 207260 |
+
"epoch": 677.46,
|
| 207261 |
+
"learning_rate": 8.646655896607433e-06,
|
| 207262 |
+
"loss": 0.3496,
|
| 207263 |
+
"step": 84805
|
| 207264 |
+
},
|
| 207265 |
+
{
|
| 207266 |
+
"epoch": 677.5,
|
| 207267 |
+
"learning_rate": 8.646575121163167e-06,
|
| 207268 |
+
"loss": 0.3057,
|
| 207269 |
+
"step": 84810
|
| 207270 |
+
},
|
| 207271 |
+
{
|
| 207272 |
+
"epoch": 677.54,
|
| 207273 |
+
"learning_rate": 8.646494345718903e-06,
|
| 207274 |
+
"loss": 0.3735,
|
| 207275 |
+
"step": 84815
|
| 207276 |
+
},
|
| 207277 |
+
{
|
| 207278 |
+
"epoch": 677.58,
|
| 207279 |
+
"learning_rate": 8.646413570274638e-06,
|
| 207280 |
+
"loss": 0.767,
|
| 207281 |
+
"step": 84820
|
| 207282 |
+
},
|
| 207283 |
+
{
|
| 207284 |
+
"epoch": 677.62,
|
| 207285 |
+
"learning_rate": 8.646332794830373e-06,
|
| 207286 |
+
"loss": 0.9717,
|
| 207287 |
+
"step": 84825
|
| 207288 |
+
},
|
| 207289 |
+
{
|
| 207290 |
+
"epoch": 677.66,
|
| 207291 |
+
"learning_rate": 8.646252019386108e-06,
|
| 207292 |
+
"loss": 0.3024,
|
| 207293 |
+
"step": 84830
|
| 207294 |
+
},
|
| 207295 |
+
{
|
| 207296 |
+
"epoch": 677.7,
|
| 207297 |
+
"learning_rate": 8.646171243941842e-06,
|
| 207298 |
+
"loss": 0.3292,
|
| 207299 |
+
"step": 84835
|
| 207300 |
+
},
|
| 207301 |
+
{
|
| 207302 |
+
"epoch": 677.74,
|
| 207303 |
+
"learning_rate": 8.646090468497578e-06,
|
| 207304 |
+
"loss": 0.3997,
|
| 207305 |
+
"step": 84840
|
| 207306 |
+
},
|
| 207307 |
+
{
|
| 207308 |
+
"epoch": 677.78,
|
| 207309 |
+
"learning_rate": 8.646009693053312e-06,
|
| 207310 |
+
"loss": 0.8281,
|
| 207311 |
+
"step": 84845
|
| 207312 |
+
},
|
| 207313 |
+
{
|
| 207314 |
+
"epoch": 677.82,
|
| 207315 |
+
"learning_rate": 8.645928917609048e-06,
|
| 207316 |
+
"loss": 0.7386,
|
| 207317 |
+
"step": 84850
|
| 207318 |
+
},
|
| 207319 |
+
{
|
| 207320 |
+
"epoch": 677.86,
|
| 207321 |
+
"learning_rate": 8.645848142164782e-06,
|
| 207322 |
+
"loss": 0.261,
|
| 207323 |
+
"step": 84855
|
| 207324 |
+
},
|
| 207325 |
+
{
|
| 207326 |
+
"epoch": 677.9,
|
| 207327 |
+
"learning_rate": 8.645767366720518e-06,
|
| 207328 |
+
"loss": 0.2954,
|
| 207329 |
+
"step": 84860
|
| 207330 |
+
},
|
| 207331 |
+
{
|
| 207332 |
+
"epoch": 677.94,
|
| 207333 |
+
"learning_rate": 8.645686591276252e-06,
|
| 207334 |
+
"loss": 0.4011,
|
| 207335 |
+
"step": 84865
|
| 207336 |
+
},
|
| 207337 |
+
{
|
| 207338 |
+
"epoch": 677.98,
|
| 207339 |
+
"learning_rate": 8.645605815831988e-06,
|
| 207340 |
+
"loss": 0.805,
|
| 207341 |
+
"step": 84870
|
| 207342 |
+
},
|
| 207343 |
+
{
|
| 207344 |
+
"epoch": 678.0,
|
| 207345 |
+
"eval_loss": 0.3907510042190552,
|
| 207346 |
+
"eval_runtime": 41.8947,
|
| 207347 |
+
"eval_samples_per_second": 20.074,
|
| 207348 |
+
"eval_steps_per_second": 0.644,
|
| 207349 |
+
"eval_wer": 0.18803923014526952,
|
| 207350 |
+
"step": 84872
|
| 207351 |
+
},
|
| 207352 |
+
{
|
| 207353 |
+
"epoch": 684.02,
|
| 207354 |
+
"learning_rate": 8.645525040387724e-06,
|
| 207355 |
+
"loss": 0.368,
|
| 207356 |
+
"step": 84875
|
| 207357 |
+
},
|
| 207358 |
+
{
|
| 207359 |
+
"epoch": 684.06,
|
| 207360 |
+
"learning_rate": 8.645444264943458e-06,
|
| 207361 |
+
"loss": 0.3252,
|
| 207362 |
+
"step": 84880
|
| 207363 |
+
},
|
| 207364 |
+
{
|
| 207365 |
+
"epoch": 684.1,
|
| 207366 |
+
"learning_rate": 8.645363489499194e-06,
|
| 207367 |
+
"loss": 0.2773,
|
| 207368 |
+
"step": 84885
|
| 207369 |
+
},
|
| 207370 |
+
{
|
| 207371 |
+
"epoch": 684.14,
|
| 207372 |
+
"learning_rate": 8.645282714054928e-06,
|
| 207373 |
+
"loss": 0.4335,
|
| 207374 |
+
"step": 84890
|
| 207375 |
+
},
|
| 207376 |
+
{
|
| 207377 |
+
"epoch": 684.18,
|
| 207378 |
+
"learning_rate": 8.645201938610664e-06,
|
| 207379 |
+
"loss": 0.9202,
|
| 207380 |
+
"step": 84895
|
| 207381 |
+
},
|
| 207382 |
+
{
|
| 207383 |
+
"epoch": 684.22,
|
| 207384 |
+
"learning_rate": 8.645121163166398e-06,
|
| 207385 |
+
"loss": 0.9484,
|
| 207386 |
+
"step": 84900
|
| 207387 |
+
},
|
| 207388 |
+
{
|
| 207389 |
+
"epoch": 684.27,
|
| 207390 |
+
"learning_rate": 8.645040387722134e-06,
|
| 207391 |
+
"loss": 0.3289,
|
| 207392 |
+
"step": 84905
|
| 207393 |
+
},
|
| 207394 |
+
{
|
| 207395 |
+
"epoch": 684.31,
|
| 207396 |
+
"learning_rate": 8.644959612277868e-06,
|
| 207397 |
+
"loss": 0.2782,
|
| 207398 |
+
"step": 84910
|
| 207399 |
+
},
|
| 207400 |
+
{
|
| 207401 |
+
"epoch": 684.35,
|
| 207402 |
+
"learning_rate": 8.644878836833604e-06,
|
| 207403 |
+
"loss": 0.381,
|
| 207404 |
+
"step": 84915
|
| 207405 |
+
},
|
| 207406 |
+
{
|
| 207407 |
+
"epoch": 684.39,
|
| 207408 |
+
"learning_rate": 8.644798061389338e-06,
|
| 207409 |
+
"loss": 0.7576,
|
| 207410 |
+
"step": 84920
|
| 207411 |
+
},
|
| 207412 |
+
{
|
| 207413 |
+
"epoch": 684.43,
|
| 207414 |
+
"learning_rate": 8.644717285945074e-06,
|
| 207415 |
+
"loss": 0.9742,
|
| 207416 |
+
"step": 84925
|
| 207417 |
+
},
|
| 207418 |
+
{
|
| 207419 |
+
"epoch": 684.47,
|
| 207420 |
+
"learning_rate": 8.64463651050081e-06,
|
| 207421 |
+
"loss": 0.3536,
|
| 207422 |
+
"step": 84930
|
| 207423 |
+
},
|
| 207424 |
+
{
|
| 207425 |
+
"epoch": 684.51,
|
| 207426 |
+
"learning_rate": 8.644555735056544e-06,
|
| 207427 |
+
"loss": 0.3314,
|
| 207428 |
+
"step": 84935
|
| 207429 |
+
},
|
| 207430 |
+
{
|
| 207431 |
+
"epoch": 684.55,
|
| 207432 |
+
"learning_rate": 8.64447495961228e-06,
|
| 207433 |
+
"loss": 0.3589,
|
| 207434 |
+
"step": 84940
|
| 207435 |
+
},
|
| 207436 |
+
{
|
| 207437 |
+
"epoch": 684.59,
|
| 207438 |
+
"learning_rate": 8.644394184168014e-06,
|
| 207439 |
+
"loss": 0.7003,
|
| 207440 |
+
"step": 84945
|
| 207441 |
+
},
|
| 207442 |
+
{
|
| 207443 |
+
"epoch": 684.63,
|
| 207444 |
+
"learning_rate": 8.64431340872375e-06,
|
| 207445 |
+
"loss": 0.8698,
|
| 207446 |
+
"step": 84950
|
| 207447 |
+
},
|
| 207448 |
+
{
|
| 207449 |
+
"epoch": 684.67,
|
| 207450 |
+
"learning_rate": 8.644232633279484e-06,
|
| 207451 |
+
"loss": 0.3337,
|
| 207452 |
+
"step": 84955
|
| 207453 |
+
},
|
| 207454 |
+
{
|
| 207455 |
+
"epoch": 684.71,
|
| 207456 |
+
"learning_rate": 8.64415185783522e-06,
|
| 207457 |
+
"loss": 0.3363,
|
| 207458 |
+
"step": 84960
|
| 207459 |
+
},
|
| 207460 |
+
{
|
| 207461 |
+
"epoch": 684.75,
|
| 207462 |
+
"learning_rate": 8.644071082390954e-06,
|
| 207463 |
+
"loss": 0.3857,
|
| 207464 |
+
"step": 84965
|
| 207465 |
+
},
|
| 207466 |
+
{
|
| 207467 |
+
"epoch": 684.79,
|
| 207468 |
+
"learning_rate": 8.64399030694669e-06,
|
| 207469 |
+
"loss": 0.7182,
|
| 207470 |
+
"step": 84970
|
| 207471 |
+
},
|
| 207472 |
+
{
|
| 207473 |
+
"epoch": 684.83,
|
| 207474 |
+
"learning_rate": 8.643909531502424e-06,
|
| 207475 |
+
"loss": 0.7883,
|
| 207476 |
+
"step": 84975
|
| 207477 |
+
},
|
| 207478 |
+
{
|
| 207479 |
+
"epoch": 684.87,
|
| 207480 |
+
"learning_rate": 8.64382875605816e-06,
|
| 207481 |
+
"loss": 0.3446,
|
| 207482 |
+
"step": 84980
|
| 207483 |
+
},
|
| 207484 |
+
{
|
| 207485 |
+
"epoch": 684.91,
|
| 207486 |
+
"learning_rate": 8.643747980613894e-06,
|
| 207487 |
+
"loss": 0.2757,
|
| 207488 |
+
"step": 84985
|
| 207489 |
+
},
|
| 207490 |
+
{
|
| 207491 |
+
"epoch": 684.95,
|
| 207492 |
+
"learning_rate": 8.64366720516963e-06,
|
| 207493 |
+
"loss": 0.3882,
|
| 207494 |
+
"step": 84990
|
| 207495 |
+
},
|
| 207496 |
+
{
|
| 207497 |
+
"epoch": 684.99,
|
| 207498 |
+
"learning_rate": 8.643586429725365e-06,
|
| 207499 |
+
"loss": 0.8628,
|
| 207500 |
+
"step": 84995
|
| 207501 |
+
},
|
| 207502 |
+
{
|
| 207503 |
+
"epoch": 685.0,
|
| 207504 |
+
"eval_loss": 0.38257458806037903,
|
| 207505 |
+
"eval_runtime": 41.8758,
|
| 207506 |
+
"eval_samples_per_second": 20.083,
|
| 207507 |
+
"eval_steps_per_second": 0.645,
|
| 207508 |
+
"eval_wer": 0.19552380258073923,
|
| 207509 |
+
"step": 84996
|
| 207510 |
+
},
|
| 207511 |
+
{
|
| 207512 |
+
"epoch": 685.03,
|
| 207513 |
+
"learning_rate": 8.6435056542811e-06,
|
| 207514 |
+
"loss": 0.3104,
|
| 207515 |
+
"step": 85000
|
| 207516 |
+
},
|
| 207517 |
+
{
|
| 207518 |
+
"epoch": 685.07,
|
| 207519 |
+
"learning_rate": 8.643424878836835e-06,
|
| 207520 |
+
"loss": 0.2549,
|
| 207521 |
+
"step": 85005
|
| 207522 |
+
},
|
| 207523 |
+
{
|
| 207524 |
+
"epoch": 685.11,
|
| 207525 |
+
"learning_rate": 8.64334410339257e-06,
|
| 207526 |
+
"loss": 0.2987,
|
| 207527 |
+
"step": 85010
|
| 207528 |
+
},
|
| 207529 |
+
{
|
| 207530 |
+
"epoch": 685.15,
|
| 207531 |
+
"learning_rate": 8.643263327948305e-06,
|
| 207532 |
+
"loss": 0.4116,
|
| 207533 |
+
"step": 85015
|
| 207534 |
+
},
|
| 207535 |
+
{
|
| 207536 |
+
"epoch": 685.19,
|
| 207537 |
+
"learning_rate": 8.64318255250404e-06,
|
| 207538 |
+
"loss": 0.9415,
|
| 207539 |
+
"step": 85020
|
| 207540 |
+
},
|
| 207541 |
+
{
|
| 207542 |
+
"epoch": 685.23,
|
| 207543 |
+
"learning_rate": 8.643101777059775e-06,
|
| 207544 |
+
"loss": 0.7285,
|
| 207545 |
+
"step": 85025
|
| 207546 |
+
},
|
| 207547 |
+
{
|
| 207548 |
+
"epoch": 685.27,
|
| 207549 |
+
"learning_rate": 8.64302100161551e-06,
|
| 207550 |
+
"loss": 0.3002,
|
| 207551 |
+
"step": 85030
|
| 207552 |
+
},
|
| 207553 |
+
{
|
| 207554 |
+
"epoch": 685.31,
|
| 207555 |
+
"learning_rate": 8.642940226171245e-06,
|
| 207556 |
+
"loss": 0.384,
|
| 207557 |
+
"step": 85035
|
| 207558 |
+
},
|
| 207559 |
+
{
|
| 207560 |
+
"epoch": 685.35,
|
| 207561 |
+
"learning_rate": 8.64285945072698e-06,
|
| 207562 |
+
"loss": 0.4491,
|
| 207563 |
+
"step": 85040
|
| 207564 |
+
},
|
| 207565 |
+
{
|
| 207566 |
+
"epoch": 685.39,
|
| 207567 |
+
"learning_rate": 8.642778675282715e-06,
|
| 207568 |
+
"loss": 0.9211,
|
| 207569 |
+
"step": 85045
|
| 207570 |
+
},
|
| 207571 |
+
{
|
| 207572 |
+
"epoch": 685.43,
|
| 207573 |
+
"learning_rate": 8.64269789983845e-06,
|
| 207574 |
+
"loss": 0.7062,
|
| 207575 |
+
"step": 85050
|
| 207576 |
+
},
|
| 207577 |
+
{
|
| 207578 |
+
"epoch": 685.47,
|
| 207579 |
+
"learning_rate": 8.642617124394185e-06,
|
| 207580 |
+
"loss": 0.2923,
|
| 207581 |
+
"step": 85055
|
| 207582 |
+
},
|
| 207583 |
+
{
|
| 207584 |
+
"epoch": 685.51,
|
| 207585 |
+
"learning_rate": 8.642536348949921e-06,
|
| 207586 |
+
"loss": 0.3318,
|
| 207587 |
+
"step": 85060
|
| 207588 |
+
},
|
| 207589 |
+
{
|
| 207590 |
+
"epoch": 685.55,
|
| 207591 |
+
"learning_rate": 8.642455573505655e-06,
|
| 207592 |
+
"loss": 0.3553,
|
| 207593 |
+
"step": 85065
|
| 207594 |
+
},
|
| 207595 |
+
{
|
| 207596 |
+
"epoch": 685.59,
|
| 207597 |
+
"learning_rate": 8.642374798061391e-06,
|
| 207598 |
+
"loss": 0.9505,
|
| 207599 |
+
"step": 85070
|
| 207600 |
+
},
|
| 207601 |
+
{
|
| 207602 |
+
"epoch": 685.63,
|
| 207603 |
+
"learning_rate": 8.642294022617125e-06,
|
| 207604 |
+
"loss": 0.6356,
|
| 207605 |
+
"step": 85075
|
| 207606 |
+
},
|
| 207607 |
+
{
|
| 207608 |
+
"epoch": 685.67,
|
| 207609 |
+
"learning_rate": 8.642213247172861e-06,
|
| 207610 |
+
"loss": 0.316,
|
| 207611 |
+
"step": 85080
|
| 207612 |
+
},
|
| 207613 |
+
{
|
| 207614 |
+
"epoch": 685.71,
|
| 207615 |
+
"learning_rate": 8.642132471728595e-06,
|
| 207616 |
+
"loss": 0.3511,
|
| 207617 |
+
"step": 85085
|
| 207618 |
+
},
|
| 207619 |
+
{
|
| 207620 |
+
"epoch": 685.76,
|
| 207621 |
+
"learning_rate": 8.642051696284331e-06,
|
| 207622 |
+
"loss": 0.3934,
|
| 207623 |
+
"step": 85090
|
| 207624 |
+
},
|
| 207625 |
+
{
|
| 207626 |
+
"epoch": 685.8,
|
| 207627 |
+
"learning_rate": 8.641970920840065e-06,
|
| 207628 |
+
"loss": 0.9989,
|
| 207629 |
+
"step": 85095
|
| 207630 |
+
},
|
| 207631 |
+
{
|
| 207632 |
+
"epoch": 685.84,
|
| 207633 |
+
"learning_rate": 8.641890145395801e-06,
|
| 207634 |
+
"loss": 0.7279,
|
| 207635 |
+
"step": 85100
|
| 207636 |
+
},
|
| 207637 |
+
{
|
| 207638 |
+
"epoch": 685.88,
|
| 207639 |
+
"learning_rate": 8.641809369951535e-06,
|
| 207640 |
+
"loss": 0.2866,
|
| 207641 |
+
"step": 85105
|
| 207642 |
+
},
|
| 207643 |
+
{
|
| 207644 |
+
"epoch": 685.92,
|
| 207645 |
+
"learning_rate": 8.641728594507271e-06,
|
| 207646 |
+
"loss": 0.3167,
|
| 207647 |
+
"step": 85110
|
| 207648 |
+
},
|
| 207649 |
+
{
|
| 207650 |
+
"epoch": 685.96,
|
| 207651 |
+
"learning_rate": 8.641647819063005e-06,
|
| 207652 |
+
"loss": 0.4247,
|
| 207653 |
+
"step": 85115
|
| 207654 |
+
},
|
| 207655 |
+
{
|
| 207656 |
+
"epoch": 686.0,
|
| 207657 |
+
"learning_rate": 8.641567043618741e-06,
|
| 207658 |
+
"loss": 1.0682,
|
| 207659 |
+
"step": 85120
|
| 207660 |
+
},
|
| 207661 |
+
{
|
| 207662 |
+
"epoch": 686.0,
|
| 207663 |
+
"eval_loss": 0.39686840772628784,
|
| 207664 |
+
"eval_runtime": 40.9884,
|
| 207665 |
+
"eval_samples_per_second": 20.518,
|
| 207666 |
+
"eval_steps_per_second": 0.659,
|
| 207667 |
+
"eval_wer": 0.19421394931599012,
|
| 207668 |
+
"step": 85120
|
| 207669 |
+
},
|
| 207670 |
+
{
|
| 207671 |
+
"epoch": 675.04,
|
| 207672 |
+
"learning_rate": 8.641486268174475e-06,
|
| 207673 |
+
"loss": 0.2997,
|
| 207674 |
+
"step": 85125
|
| 207675 |
+
},
|
| 207676 |
+
{
|
| 207677 |
+
"epoch": 675.08,
|
| 207678 |
+
"learning_rate": 8.641405492730211e-06,
|
| 207679 |
+
"loss": 0.2814,
|
| 207680 |
+
"step": 85130
|
| 207681 |
+
},
|
| 207682 |
+
{
|
| 207683 |
+
"epoch": 675.12,
|
| 207684 |
+
"learning_rate": 8.641324717285947e-06,
|
| 207685 |
+
"loss": 0.3091,
|
| 207686 |
+
"step": 85135
|
| 207687 |
+
},
|
| 207688 |
+
{
|
| 207689 |
+
"epoch": 675.16,
|
| 207690 |
+
"learning_rate": 8.641243941841681e-06,
|
| 207691 |
+
"loss": 0.4994,
|
| 207692 |
+
"step": 85140
|
| 207693 |
+
},
|
| 207694 |
+
{
|
| 207695 |
+
"epoch": 675.2,
|
| 207696 |
+
"learning_rate": 8.641163166397417e-06,
|
| 207697 |
+
"loss": 1.1782,
|
| 207698 |
+
"step": 85145
|
| 207699 |
+
},
|
| 207700 |
+
{
|
| 207701 |
+
"epoch": 675.24,
|
| 207702 |
+
"learning_rate": 8.641082390953151e-06,
|
| 207703 |
+
"loss": 0.3617,
|
| 207704 |
+
"step": 85150
|
| 207705 |
+
},
|
| 207706 |
+
{
|
| 207707 |
+
"epoch": 675.28,
|
| 207708 |
+
"learning_rate": 8.641001615508887e-06,
|
| 207709 |
+
"loss": 0.2818,
|
| 207710 |
+
"step": 85155
|
| 207711 |
+
},
|
| 207712 |
+
{
|
| 207713 |
+
"epoch": 675.32,
|
| 207714 |
+
"learning_rate": 8.640920840064621e-06,
|
| 207715 |
+
"loss": 0.3871,
|
| 207716 |
+
"step": 85160
|
| 207717 |
+
},
|
| 207718 |
+
{
|
| 207719 |
+
"epoch": 675.36,
|
| 207720 |
+
"learning_rate": 8.640840064620357e-06,
|
| 207721 |
+
"loss": 0.4942,
|
| 207722 |
+
"step": 85165
|
| 207723 |
+
},
|
| 207724 |
+
{
|
| 207725 |
+
"epoch": 675.4,
|
| 207726 |
+
"learning_rate": 8.640759289176091e-06,
|
| 207727 |
+
"loss": 1.5803,
|
| 207728 |
+
"step": 85170
|
| 207729 |
+
},
|
| 207730 |
+
{
|
| 207731 |
+
"epoch": 675.44,
|
| 207732 |
+
"learning_rate": 8.640678513731827e-06,
|
| 207733 |
+
"loss": 0.3471,
|
| 207734 |
+
"step": 85175
|
| 207735 |
+
},
|
| 207736 |
+
{
|
| 207737 |
+
"epoch": 675.48,
|
| 207738 |
+
"learning_rate": 8.640597738287561e-06,
|
| 207739 |
+
"loss": 0.3001,
|
| 207740 |
+
"step": 85180
|
| 207741 |
+
},
|
| 207742 |
+
{
|
| 207743 |
+
"epoch": 675.52,
|
| 207744 |
+
"learning_rate": 8.640516962843297e-06,
|
| 207745 |
+
"loss": 0.289,
|
| 207746 |
+
"step": 85185
|
| 207747 |
+
},
|
| 207748 |
+
{
|
| 207749 |
+
"epoch": 675.56,
|
| 207750 |
+
"learning_rate": 8.64043618739903e-06,
|
| 207751 |
+
"loss": 0.4537,
|
| 207752 |
+
"step": 85190
|
| 207753 |
+
},
|
| 207754 |
+
{
|
| 207755 |
+
"epoch": 675.6,
|
| 207756 |
+
"learning_rate": 8.640355411954767e-06,
|
| 207757 |
+
"loss": 1.4158,
|
| 207758 |
+
"step": 85195
|
| 207759 |
+
},
|
| 207760 |
+
{
|
| 207761 |
+
"epoch": 675.63,
|
| 207762 |
+
"learning_rate": 8.640274636510503e-06,
|
| 207763 |
+
"loss": 0.3806,
|
| 207764 |
+
"step": 85200
|
| 207765 |
+
},
|
| 207766 |
+
{
|
| 207767 |
+
"epoch": 675.67,
|
| 207768 |
+
"learning_rate": 8.640193861066237e-06,
|
| 207769 |
+
"loss": 0.3229,
|
| 207770 |
+
"step": 85205
|
| 207771 |
+
},
|
| 207772 |
+
{
|
| 207773 |
+
"epoch": 675.71,
|
| 207774 |
+
"learning_rate": 8.640113085621972e-06,
|
| 207775 |
+
"loss": 0.3436,
|
| 207776 |
+
"step": 85210
|
| 207777 |
+
},
|
| 207778 |
+
{
|
| 207779 |
+
"epoch": 675.75,
|
| 207780 |
+
"learning_rate": 8.640032310177707e-06,
|
| 207781 |
+
"loss": 0.5279,
|
| 207782 |
+
"step": 85215
|
| 207783 |
+
},
|
| 207784 |
+
{
|
| 207785 |
+
"epoch": 675.79,
|
| 207786 |
+
"learning_rate": 8.639967689822296e-06,
|
| 207787 |
+
"loss": 1.3692,
|
| 207788 |
+
"step": 85220
|
| 207789 |
+
},
|
| 207790 |
+
{
|
| 207791 |
+
"epoch": 675.83,
|
| 207792 |
+
"learning_rate": 8.63988691437803e-06,
|
| 207793 |
+
"loss": 0.3193,
|
| 207794 |
+
"step": 85225
|
| 207795 |
+
},
|
| 207796 |
+
{
|
| 207797 |
+
"epoch": 675.87,
|
| 207798 |
+
"learning_rate": 8.639806138933766e-06,
|
| 207799 |
+
"loss": 0.3122,
|
| 207800 |
+
"step": 85230
|
| 207801 |
+
},
|
| 207802 |
+
{
|
| 207803 |
+
"epoch": 675.91,
|
| 207804 |
+
"learning_rate": 8.6397253634895e-06,
|
| 207805 |
+
"loss": 0.3759,
|
| 207806 |
+
"step": 85235
|
| 207807 |
+
},
|
| 207808 |
+
{
|
| 207809 |
+
"epoch": 675.95,
|
| 207810 |
+
"learning_rate": 8.639644588045236e-06,
|
| 207811 |
+
"loss": 0.4346,
|
| 207812 |
+
"step": 85240
|
| 207813 |
+
},
|
| 207814 |
+
{
|
| 207815 |
+
"epoch": 675.99,
|
| 207816 |
+
"learning_rate": 8.63956381260097e-06,
|
| 207817 |
+
"loss": 1.1819,
|
| 207818 |
+
"step": 85245
|
| 207819 |
+
},
|
| 207820 |
+
{
|
| 207821 |
+
"epoch": 676.0,
|
| 207822 |
+
"eval_loss": 0.3436339199542999,
|
| 207823 |
+
"eval_runtime": 41.391,
|
| 207824 |
+
"eval_samples_per_second": 20.318,
|
| 207825 |
+
"eval_steps_per_second": 0.652,
|
| 207826 |
+
"eval_wer": 0.1900681168973852,
|
| 207827 |
+
"step": 85246
|
| 207828 |
}
|
| 207829 |
],
|
| 207830 |
+
"max_steps": 630000,
|
| 207831 |
"num_train_epochs": 5000,
|
| 207832 |
+
"total_flos": 2.3989749933394523e+20,
|
| 207833 |
"trial_name": null,
|
| 207834 |
"trial_params": null
|
| 207835 |
}
|
model-bin/finetune/base/{checkpoint-84623 → checkpoint-85246}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629921107.3117447/events.out.tfevents.1629921107.7e498afd5545.7645.135
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ac0e35d76c6bb1fdc19ab11f76cb6c3f8b2f17d776620b3c264b7d2d04f85b24
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629921575.424182/events.out.tfevents.1629921576.7e498afd5545.7645.137
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ddd6ca0dbb236be24949c7b10365743126824264314aab847be0ecffa18dfe6d
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629922040.984044/events.out.tfevents.1629922040.7e498afd5545.7645.139
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ac2eb7a98a697d1b6d3f19e33c86311d76d74d2c564aba2cec2e1481e7f95ef9
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629922507.7783706/events.out.tfevents.1629922507.7e498afd5545.7645.141
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:23369d4e87b9111172b87933c457827efe20294d2d1f8e61dd2fbf1d23399c0b
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629922973.5943346/events.out.tfevents.1629922973.7e498afd5545.7645.143
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:380d9db5b898bf953721da9d1e243f3b81e1a47258f46061782ae6e1c587b1cc
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629921107.7e498afd5545.7645.134
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9443141accc25d87528d7e4ac1e0af0083dc7742444ab9f62544329298f15e5c
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629921575.7e498afd5545.7645.136
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9aee3daf2ae11921f9767b492aa0588d5f0b57f01051147b666d870d101bef39
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629922040.7e498afd5545.7645.138
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6c1bed51fc7c6f081acbf7cb5b3581f7aab55d461f49864d77713f19dfbded2f
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629922507.7e498afd5545.7645.140
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1e3c9ad07816f3195ea340f598e0b43090d04011ee238dc93ad2176a512b5954
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629922973.7e498afd5545.7645.142
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:029a1472a49bda91e6b68a5da588558e9fb1c51e9566bf009d375755f94c3ea0
|
| 3 |
+
size 8622
|