"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-119214 → checkpoint-119837}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-119214 → checkpoint-119837}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-119214 → checkpoint-119837}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-119214 → checkpoint-119837}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-119214 → checkpoint-119837}/rng_state.pth +2 -2
- model-bin/finetune/base/{checkpoint-119214 → checkpoint-119837}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-119214 → checkpoint-119837}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-119214 → checkpoint-119837}/trainer_state.json +799 -4
- model-bin/finetune/base/{checkpoint-119214 → checkpoint-119837}/training_args.bin +0 -0
- model-bin/finetune/base/log/1630126542.1800706/events.out.tfevents.1630126542.86bb0ddabf9b.4092.21 +3 -0
- model-bin/finetune/base/log/1630126929.6951962/events.out.tfevents.1630126929.86bb0ddabf9b.4092.23 +3 -0
- model-bin/finetune/base/log/1630127435.6478999/events.out.tfevents.1630127435.86bb0ddabf9b.4092.25 +3 -0
- model-bin/finetune/base/log/1630127823.8657374/events.out.tfevents.1630127823.86bb0ddabf9b.4092.27 +3 -0
- model-bin/finetune/base/log/1630128219.1986852/events.out.tfevents.1630128219.86bb0ddabf9b.4092.29 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630126542.86bb0ddabf9b.4092.20 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630126929.86bb0ddabf9b.4092.22 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630127435.86bb0ddabf9b.4092.24 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630127823.86bb0ddabf9b.4092.26 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630128219.86bb0ddabf9b.4092.28 +3 -0
model-bin/finetune/base/{checkpoint-119214 → checkpoint-119837}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-119214 → checkpoint-119837}/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 sha256:638b0c02883395eb33d55d5829db4fcbda422f633998b88b3438b42b18a91453
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-119214 → checkpoint-119837}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-119214 → checkpoint-119837}/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 sha256:97ce672d79cf16ba4147f3f6f3a4f9e8ca4e6e5792a9632830450842ae0d13c0
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-119214 → checkpoint-119837}/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 sha256:71106e40a36858b4437fcedbffb6dac6519e3ec60d3b37ba426c382544b5500d
|
| 3 |
+
size 14503
|
model-bin/finetune/base/{checkpoint-119214 → checkpoint-119837}/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:dffbfbd668412697b53d16beb9f72be6858e977e50fa330e54bbf6f7760d9003
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-119214 → checkpoint-119837}/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:a5e9bb37cf3ecefd52b1db89dc11caf1240f1aa5d5e04f5c01ffc6ab472184de
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-119214 → checkpoint-119837}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1743826049391605,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-101551",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -251040,11 +251040,806 @@
|
|
| 251040 |
"eval_steps_per_second": 0.759,
|
| 251041 |
"eval_wer": 0.18514659919630896,
|
| 251042 |
"step": 119214
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 251043 |
}
|
| 251044 |
],
|
| 251045 |
-
"max_steps":
|
| 251046 |
"num_train_epochs": 5000,
|
| 251047 |
-
"total_flos": 3.
|
| 251048 |
"trial_name": null,
|
| 251049 |
"trial_params": null
|
| 251050 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1743826049391605,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-101551",
|
| 4 |
+
"epoch": 958.0,
|
| 5 |
+
"global_step": 119837,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 251040 |
"eval_steps_per_second": 0.759,
|
| 251041 |
"eval_wer": 0.18514659919630896,
|
| 251042 |
"step": 119214
|
| 251043 |
+
},
|
| 251044 |
+
{
|
| 251045 |
+
"epoch": 953.01,
|
| 251046 |
+
"learning_rate": 8.106282051282051e-06,
|
| 251047 |
+
"loss": 0.4502,
|
| 251048 |
+
"step": 119215
|
| 251049 |
+
},
|
| 251050 |
+
{
|
| 251051 |
+
"epoch": 953.05,
|
| 251052 |
+
"learning_rate": 8.106201923076924e-06,
|
| 251053 |
+
"loss": 0.3254,
|
| 251054 |
+
"step": 119220
|
| 251055 |
+
},
|
| 251056 |
+
{
|
| 251057 |
+
"epoch": 953.09,
|
| 251058 |
+
"learning_rate": 8.106121794871795e-06,
|
| 251059 |
+
"loss": 0.2768,
|
| 251060 |
+
"step": 119225
|
| 251061 |
+
},
|
| 251062 |
+
{
|
| 251063 |
+
"epoch": 953.13,
|
| 251064 |
+
"learning_rate": 8.106041666666667e-06,
|
| 251065 |
+
"loss": 0.3259,
|
| 251066 |
+
"step": 119230
|
| 251067 |
+
},
|
| 251068 |
+
{
|
| 251069 |
+
"epoch": 953.17,
|
| 251070 |
+
"learning_rate": 8.105961538461538e-06,
|
| 251071 |
+
"loss": 0.4879,
|
| 251072 |
+
"step": 119235
|
| 251073 |
+
},
|
| 251074 |
+
{
|
| 251075 |
+
"epoch": 953.21,
|
| 251076 |
+
"learning_rate": 8.105881410256411e-06,
|
| 251077 |
+
"loss": 1.3129,
|
| 251078 |
+
"step": 119240
|
| 251079 |
+
},
|
| 251080 |
+
{
|
| 251081 |
+
"epoch": 953.25,
|
| 251082 |
+
"learning_rate": 8.105801282051283e-06,
|
| 251083 |
+
"loss": 0.3143,
|
| 251084 |
+
"step": 119245
|
| 251085 |
+
},
|
| 251086 |
+
{
|
| 251087 |
+
"epoch": 953.29,
|
| 251088 |
+
"learning_rate": 8.105721153846154e-06,
|
| 251089 |
+
"loss": 0.2931,
|
| 251090 |
+
"step": 119250
|
| 251091 |
+
},
|
| 251092 |
+
{
|
| 251093 |
+
"epoch": 953.33,
|
| 251094 |
+
"learning_rate": 8.105641025641027e-06,
|
| 251095 |
+
"loss": 0.294,
|
| 251096 |
+
"step": 119255
|
| 251097 |
+
},
|
| 251098 |
+
{
|
| 251099 |
+
"epoch": 953.37,
|
| 251100 |
+
"learning_rate": 8.105560897435898e-06,
|
| 251101 |
+
"loss": 0.4796,
|
| 251102 |
+
"step": 119260
|
| 251103 |
+
},
|
| 251104 |
+
{
|
| 251105 |
+
"epoch": 953.41,
|
| 251106 |
+
"learning_rate": 8.10548076923077e-06,
|
| 251107 |
+
"loss": 1.1216,
|
| 251108 |
+
"step": 119265
|
| 251109 |
+
},
|
| 251110 |
+
{
|
| 251111 |
+
"epoch": 953.45,
|
| 251112 |
+
"learning_rate": 8.105400641025641e-06,
|
| 251113 |
+
"loss": 0.3011,
|
| 251114 |
+
"step": 119270
|
| 251115 |
+
},
|
| 251116 |
+
{
|
| 251117 |
+
"epoch": 953.49,
|
| 251118 |
+
"learning_rate": 8.105320512820514e-06,
|
| 251119 |
+
"loss": 0.2957,
|
| 251120 |
+
"step": 119275
|
| 251121 |
+
},
|
| 251122 |
+
{
|
| 251123 |
+
"epoch": 953.53,
|
| 251124 |
+
"learning_rate": 8.105240384615385e-06,
|
| 251125 |
+
"loss": 0.3445,
|
| 251126 |
+
"step": 119280
|
| 251127 |
+
},
|
| 251128 |
+
{
|
| 251129 |
+
"epoch": 953.57,
|
| 251130 |
+
"learning_rate": 8.105160256410257e-06,
|
| 251131 |
+
"loss": 0.5431,
|
| 251132 |
+
"step": 119285
|
| 251133 |
+
},
|
| 251134 |
+
{
|
| 251135 |
+
"epoch": 953.61,
|
| 251136 |
+
"learning_rate": 8.105080128205128e-06,
|
| 251137 |
+
"loss": 1.0237,
|
| 251138 |
+
"step": 119290
|
| 251139 |
+
},
|
| 251140 |
+
{
|
| 251141 |
+
"epoch": 953.65,
|
| 251142 |
+
"learning_rate": 8.105000000000001e-06,
|
| 251143 |
+
"loss": 0.2889,
|
| 251144 |
+
"step": 119295
|
| 251145 |
+
},
|
| 251146 |
+
{
|
| 251147 |
+
"epoch": 953.69,
|
| 251148 |
+
"learning_rate": 8.104919871794873e-06,
|
| 251149 |
+
"loss": 0.2613,
|
| 251150 |
+
"step": 119300
|
| 251151 |
+
},
|
| 251152 |
+
{
|
| 251153 |
+
"epoch": 953.73,
|
| 251154 |
+
"learning_rate": 8.104839743589744e-06,
|
| 251155 |
+
"loss": 0.3774,
|
| 251156 |
+
"step": 119305
|
| 251157 |
+
},
|
| 251158 |
+
{
|
| 251159 |
+
"epoch": 953.77,
|
| 251160 |
+
"learning_rate": 8.104759615384617e-06,
|
| 251161 |
+
"loss": 0.496,
|
| 251162 |
+
"step": 119310
|
| 251163 |
+
},
|
| 251164 |
+
{
|
| 251165 |
+
"epoch": 953.81,
|
| 251166 |
+
"learning_rate": 8.104679487179488e-06,
|
| 251167 |
+
"loss": 1.2313,
|
| 251168 |
+
"step": 119315
|
| 251169 |
+
},
|
| 251170 |
+
{
|
| 251171 |
+
"epoch": 953.85,
|
| 251172 |
+
"learning_rate": 8.10459935897436e-06,
|
| 251173 |
+
"loss": 0.3028,
|
| 251174 |
+
"step": 119320
|
| 251175 |
+
},
|
| 251176 |
+
{
|
| 251177 |
+
"epoch": 953.89,
|
| 251178 |
+
"learning_rate": 8.104519230769231e-06,
|
| 251179 |
+
"loss": 0.2401,
|
| 251180 |
+
"step": 119325
|
| 251181 |
+
},
|
| 251182 |
+
{
|
| 251183 |
+
"epoch": 953.93,
|
| 251184 |
+
"learning_rate": 8.104439102564104e-06,
|
| 251185 |
+
"loss": 0.331,
|
| 251186 |
+
"step": 119330
|
| 251187 |
+
},
|
| 251188 |
+
{
|
| 251189 |
+
"epoch": 953.97,
|
| 251190 |
+
"learning_rate": 8.104358974358974e-06,
|
| 251191 |
+
"loss": 0.5488,
|
| 251192 |
+
"step": 119335
|
| 251193 |
+
},
|
| 251194 |
+
{
|
| 251195 |
+
"epoch": 954.0,
|
| 251196 |
+
"eval_loss": 0.5151069760322571,
|
| 251197 |
+
"eval_runtime": 36.4885,
|
| 251198 |
+
"eval_samples_per_second": 23.021,
|
| 251199 |
+
"eval_steps_per_second": 0.74,
|
| 251200 |
+
"eval_wer": 0.18439151523926736,
|
| 251201 |
+
"step": 119339
|
| 251202 |
+
},
|
| 251203 |
+
{
|
| 251204 |
+
"epoch": 954.01,
|
| 251205 |
+
"learning_rate": 8.104278846153847e-06,
|
| 251206 |
+
"loss": 0.3871,
|
| 251207 |
+
"step": 119340
|
| 251208 |
+
},
|
| 251209 |
+
{
|
| 251210 |
+
"epoch": 954.05,
|
| 251211 |
+
"learning_rate": 8.10419871794872e-06,
|
| 251212 |
+
"loss": 0.2759,
|
| 251213 |
+
"step": 119345
|
| 251214 |
+
},
|
| 251215 |
+
{
|
| 251216 |
+
"epoch": 954.09,
|
| 251217 |
+
"learning_rate": 8.10411858974359e-06,
|
| 251218 |
+
"loss": 0.2756,
|
| 251219 |
+
"step": 119350
|
| 251220 |
+
},
|
| 251221 |
+
{
|
| 251222 |
+
"epoch": 954.13,
|
| 251223 |
+
"learning_rate": 8.104038461538463e-06,
|
| 251224 |
+
"loss": 0.3578,
|
| 251225 |
+
"step": 119355
|
| 251226 |
+
},
|
| 251227 |
+
{
|
| 251228 |
+
"epoch": 954.17,
|
| 251229 |
+
"learning_rate": 8.103958333333334e-06,
|
| 251230 |
+
"loss": 0.4686,
|
| 251231 |
+
"step": 119360
|
| 251232 |
+
},
|
| 251233 |
+
{
|
| 251234 |
+
"epoch": 954.21,
|
| 251235 |
+
"learning_rate": 8.103878205128205e-06,
|
| 251236 |
+
"loss": 1.0711,
|
| 251237 |
+
"step": 119365
|
| 251238 |
+
},
|
| 251239 |
+
{
|
| 251240 |
+
"epoch": 954.25,
|
| 251241 |
+
"learning_rate": 8.103798076923077e-06,
|
| 251242 |
+
"loss": 0.3258,
|
| 251243 |
+
"step": 119370
|
| 251244 |
+
},
|
| 251245 |
+
{
|
| 251246 |
+
"epoch": 954.29,
|
| 251247 |
+
"learning_rate": 8.10371794871795e-06,
|
| 251248 |
+
"loss": 0.2456,
|
| 251249 |
+
"step": 119375
|
| 251250 |
+
},
|
| 251251 |
+
{
|
| 251252 |
+
"epoch": 954.33,
|
| 251253 |
+
"learning_rate": 8.103637820512821e-06,
|
| 251254 |
+
"loss": 0.3783,
|
| 251255 |
+
"step": 119380
|
| 251256 |
+
},
|
| 251257 |
+
{
|
| 251258 |
+
"epoch": 954.37,
|
| 251259 |
+
"learning_rate": 8.103557692307692e-06,
|
| 251260 |
+
"loss": 0.5758,
|
| 251261 |
+
"step": 119385
|
| 251262 |
+
},
|
| 251263 |
+
{
|
| 251264 |
+
"epoch": 954.41,
|
| 251265 |
+
"learning_rate": 8.103477564102564e-06,
|
| 251266 |
+
"loss": 1.053,
|
| 251267 |
+
"step": 119390
|
| 251268 |
+
},
|
| 251269 |
+
{
|
| 251270 |
+
"epoch": 954.45,
|
| 251271 |
+
"learning_rate": 8.103397435897437e-06,
|
| 251272 |
+
"loss": 0.3093,
|
| 251273 |
+
"step": 119395
|
| 251274 |
+
},
|
| 251275 |
+
{
|
| 251276 |
+
"epoch": 954.49,
|
| 251277 |
+
"learning_rate": 8.103317307692308e-06,
|
| 251278 |
+
"loss": 0.341,
|
| 251279 |
+
"step": 119400
|
| 251280 |
+
},
|
| 251281 |
+
{
|
| 251282 |
+
"epoch": 954.53,
|
| 251283 |
+
"learning_rate": 8.10323717948718e-06,
|
| 251284 |
+
"loss": 0.3665,
|
| 251285 |
+
"step": 119405
|
| 251286 |
+
},
|
| 251287 |
+
{
|
| 251288 |
+
"epoch": 954.57,
|
| 251289 |
+
"learning_rate": 8.103157051282053e-06,
|
| 251290 |
+
"loss": 0.5634,
|
| 251291 |
+
"step": 119410
|
| 251292 |
+
},
|
| 251293 |
+
{
|
| 251294 |
+
"epoch": 954.61,
|
| 251295 |
+
"learning_rate": 8.103076923076924e-06,
|
| 251296 |
+
"loss": 1.1929,
|
| 251297 |
+
"step": 119415
|
| 251298 |
+
},
|
| 251299 |
+
{
|
| 251300 |
+
"epoch": 954.65,
|
| 251301 |
+
"learning_rate": 8.102996794871795e-06,
|
| 251302 |
+
"loss": 0.3694,
|
| 251303 |
+
"step": 119420
|
| 251304 |
+
},
|
| 251305 |
+
{
|
| 251306 |
+
"epoch": 954.69,
|
| 251307 |
+
"learning_rate": 8.102916666666667e-06,
|
| 251308 |
+
"loss": 0.2839,
|
| 251309 |
+
"step": 119425
|
| 251310 |
+
},
|
| 251311 |
+
{
|
| 251312 |
+
"epoch": 954.73,
|
| 251313 |
+
"learning_rate": 8.10283653846154e-06,
|
| 251314 |
+
"loss": 0.3478,
|
| 251315 |
+
"step": 119430
|
| 251316 |
+
},
|
| 251317 |
+
{
|
| 251318 |
+
"epoch": 954.77,
|
| 251319 |
+
"learning_rate": 8.102756410256411e-06,
|
| 251320 |
+
"loss": 0.4601,
|
| 251321 |
+
"step": 119435
|
| 251322 |
+
},
|
| 251323 |
+
{
|
| 251324 |
+
"epoch": 954.81,
|
| 251325 |
+
"learning_rate": 8.102676282051282e-06,
|
| 251326 |
+
"loss": 1.1339,
|
| 251327 |
+
"step": 119440
|
| 251328 |
+
},
|
| 251329 |
+
{
|
| 251330 |
+
"epoch": 954.85,
|
| 251331 |
+
"learning_rate": 8.102596153846155e-06,
|
| 251332 |
+
"loss": 0.2712,
|
| 251333 |
+
"step": 119445
|
| 251334 |
+
},
|
| 251335 |
+
{
|
| 251336 |
+
"epoch": 954.89,
|
| 251337 |
+
"learning_rate": 8.102516025641027e-06,
|
| 251338 |
+
"loss": 0.3081,
|
| 251339 |
+
"step": 119450
|
| 251340 |
+
},
|
| 251341 |
+
{
|
| 251342 |
+
"epoch": 954.93,
|
| 251343 |
+
"learning_rate": 8.102435897435898e-06,
|
| 251344 |
+
"loss": 0.3613,
|
| 251345 |
+
"step": 119455
|
| 251346 |
+
},
|
| 251347 |
+
{
|
| 251348 |
+
"epoch": 954.97,
|
| 251349 |
+
"learning_rate": 8.10235576923077e-06,
|
| 251350 |
+
"loss": 0.7565,
|
| 251351 |
+
"step": 119460
|
| 251352 |
+
},
|
| 251353 |
+
{
|
| 251354 |
+
"epoch": 955.0,
|
| 251355 |
+
"eval_loss": 0.3896730840206146,
|
| 251356 |
+
"eval_runtime": 50.1326,
|
| 251357 |
+
"eval_samples_per_second": 16.756,
|
| 251358 |
+
"eval_steps_per_second": 0.539,
|
| 251359 |
+
"eval_wer": 0.17685986473711002,
|
| 251360 |
+
"step": 119464
|
| 251361 |
+
},
|
| 251362 |
+
{
|
| 251363 |
+
"epoch": 963.01,
|
| 251364 |
+
"learning_rate": 8.102275641025643e-06,
|
| 251365 |
+
"loss": 0.4834,
|
| 251366 |
+
"step": 119465
|
| 251367 |
+
},
|
| 251368 |
+
{
|
| 251369 |
+
"epoch": 963.05,
|
| 251370 |
+
"learning_rate": 8.102195512820514e-06,
|
| 251371 |
+
"loss": 0.2779,
|
| 251372 |
+
"step": 119470
|
| 251373 |
+
},
|
| 251374 |
+
{
|
| 251375 |
+
"epoch": 963.09,
|
| 251376 |
+
"learning_rate": 8.102115384615385e-06,
|
| 251377 |
+
"loss": 0.3101,
|
| 251378 |
+
"step": 119475
|
| 251379 |
+
},
|
| 251380 |
+
{
|
| 251381 |
+
"epoch": 963.13,
|
| 251382 |
+
"learning_rate": 8.102035256410257e-06,
|
| 251383 |
+
"loss": 0.3189,
|
| 251384 |
+
"step": 119480
|
| 251385 |
+
},
|
| 251386 |
+
{
|
| 251387 |
+
"epoch": 963.17,
|
| 251388 |
+
"learning_rate": 8.10195512820513e-06,
|
| 251389 |
+
"loss": 0.5695,
|
| 251390 |
+
"step": 119485
|
| 251391 |
+
},
|
| 251392 |
+
{
|
| 251393 |
+
"epoch": 963.21,
|
| 251394 |
+
"learning_rate": 8.101875e-06,
|
| 251395 |
+
"loss": 1.0827,
|
| 251396 |
+
"step": 119490
|
| 251397 |
+
},
|
| 251398 |
+
{
|
| 251399 |
+
"epoch": 963.25,
|
| 251400 |
+
"learning_rate": 8.101794871794872e-06,
|
| 251401 |
+
"loss": 0.3008,
|
| 251402 |
+
"step": 119495
|
| 251403 |
+
},
|
| 251404 |
+
{
|
| 251405 |
+
"epoch": 963.29,
|
| 251406 |
+
"learning_rate": 8.101714743589746e-06,
|
| 251407 |
+
"loss": 0.2881,
|
| 251408 |
+
"step": 119500
|
| 251409 |
+
},
|
| 251410 |
+
{
|
| 251411 |
+
"epoch": 963.33,
|
| 251412 |
+
"learning_rate": 8.101634615384615e-06,
|
| 251413 |
+
"loss": 0.3082,
|
| 251414 |
+
"step": 119505
|
| 251415 |
+
},
|
| 251416 |
+
{
|
| 251417 |
+
"epoch": 963.37,
|
| 251418 |
+
"learning_rate": 8.101554487179488e-06,
|
| 251419 |
+
"loss": 0.5139,
|
| 251420 |
+
"step": 119510
|
| 251421 |
+
},
|
| 251422 |
+
{
|
| 251423 |
+
"epoch": 963.41,
|
| 251424 |
+
"learning_rate": 8.10147435897436e-06,
|
| 251425 |
+
"loss": 1.161,
|
| 251426 |
+
"step": 119515
|
| 251427 |
+
},
|
| 251428 |
+
{
|
| 251429 |
+
"epoch": 963.45,
|
| 251430 |
+
"learning_rate": 8.101394230769231e-06,
|
| 251431 |
+
"loss": 0.3094,
|
| 251432 |
+
"step": 119520
|
| 251433 |
+
},
|
| 251434 |
+
{
|
| 251435 |
+
"epoch": 963.49,
|
| 251436 |
+
"learning_rate": 8.101314102564102e-06,
|
| 251437 |
+
"loss": 0.2785,
|
| 251438 |
+
"step": 119525
|
| 251439 |
+
},
|
| 251440 |
+
{
|
| 251441 |
+
"epoch": 963.53,
|
| 251442 |
+
"learning_rate": 8.101233974358975e-06,
|
| 251443 |
+
"loss": 0.3467,
|
| 251444 |
+
"step": 119530
|
| 251445 |
+
},
|
| 251446 |
+
{
|
| 251447 |
+
"epoch": 963.57,
|
| 251448 |
+
"learning_rate": 8.101153846153847e-06,
|
| 251449 |
+
"loss": 0.5172,
|
| 251450 |
+
"step": 119535
|
| 251451 |
+
},
|
| 251452 |
+
{
|
| 251453 |
+
"epoch": 963.61,
|
| 251454 |
+
"learning_rate": 8.101073717948718e-06,
|
| 251455 |
+
"loss": 0.9405,
|
| 251456 |
+
"step": 119540
|
| 251457 |
+
},
|
| 251458 |
+
{
|
| 251459 |
+
"epoch": 963.65,
|
| 251460 |
+
"learning_rate": 8.100993589743591e-06,
|
| 251461 |
+
"loss": 0.2638,
|
| 251462 |
+
"step": 119545
|
| 251463 |
+
},
|
| 251464 |
+
{
|
| 251465 |
+
"epoch": 963.69,
|
| 251466 |
+
"learning_rate": 8.100913461538462e-06,
|
| 251467 |
+
"loss": 0.3327,
|
| 251468 |
+
"step": 119550
|
| 251469 |
+
},
|
| 251470 |
+
{
|
| 251471 |
+
"epoch": 963.73,
|
| 251472 |
+
"learning_rate": 8.100833333333334e-06,
|
| 251473 |
+
"loss": 0.3191,
|
| 251474 |
+
"step": 119555
|
| 251475 |
+
},
|
| 251476 |
+
{
|
| 251477 |
+
"epoch": 963.77,
|
| 251478 |
+
"learning_rate": 8.100753205128205e-06,
|
| 251479 |
+
"loss": 0.5461,
|
| 251480 |
+
"step": 119560
|
| 251481 |
+
},
|
| 251482 |
+
{
|
| 251483 |
+
"epoch": 963.81,
|
| 251484 |
+
"learning_rate": 8.100673076923078e-06,
|
| 251485 |
+
"loss": 1.2435,
|
| 251486 |
+
"step": 119565
|
| 251487 |
+
},
|
| 251488 |
+
{
|
| 251489 |
+
"epoch": 963.85,
|
| 251490 |
+
"learning_rate": 8.10059294871795e-06,
|
| 251491 |
+
"loss": 0.3333,
|
| 251492 |
+
"step": 119570
|
| 251493 |
+
},
|
| 251494 |
+
{
|
| 251495 |
+
"epoch": 963.89,
|
| 251496 |
+
"learning_rate": 8.100512820512821e-06,
|
| 251497 |
+
"loss": 0.3307,
|
| 251498 |
+
"step": 119575
|
| 251499 |
+
},
|
| 251500 |
+
{
|
| 251501 |
+
"epoch": 963.93,
|
| 251502 |
+
"learning_rate": 8.100432692307692e-06,
|
| 251503 |
+
"loss": 0.3158,
|
| 251504 |
+
"step": 119580
|
| 251505 |
+
},
|
| 251506 |
+
{
|
| 251507 |
+
"epoch": 963.97,
|
| 251508 |
+
"learning_rate": 8.100352564102565e-06,
|
| 251509 |
+
"loss": 0.5599,
|
| 251510 |
+
"step": 119585
|
| 251511 |
+
},
|
| 251512 |
+
{
|
| 251513 |
+
"epoch": 964.0,
|
| 251514 |
+
"eval_loss": 0.3675064742565155,
|
| 251515 |
+
"eval_runtime": 35.8223,
|
| 251516 |
+
"eval_samples_per_second": 23.505,
|
| 251517 |
+
"eval_steps_per_second": 0.754,
|
| 251518 |
+
"eval_wer": 0.1800690313578615,
|
| 251519 |
+
"step": 119588
|
| 251520 |
+
},
|
| 251521 |
+
{
|
| 251522 |
+
"epoch": 964.02,
|
| 251523 |
+
"learning_rate": 8.100272435897437e-06,
|
| 251524 |
+
"loss": 0.2881,
|
| 251525 |
+
"step": 119590
|
| 251526 |
+
},
|
| 251527 |
+
{
|
| 251528 |
+
"epoch": 964.06,
|
| 251529 |
+
"learning_rate": 8.100192307692308e-06,
|
| 251530 |
+
"loss": 0.2996,
|
| 251531 |
+
"step": 119595
|
| 251532 |
+
},
|
| 251533 |
+
{
|
| 251534 |
+
"epoch": 964.1,
|
| 251535 |
+
"learning_rate": 8.100112179487181e-06,
|
| 251536 |
+
"loss": 0.2788,
|
| 251537 |
+
"step": 119600
|
| 251538 |
+
},
|
| 251539 |
+
{
|
| 251540 |
+
"epoch": 964.14,
|
| 251541 |
+
"learning_rate": 8.100032051282053e-06,
|
| 251542 |
+
"loss": 0.3476,
|
| 251543 |
+
"step": 119605
|
| 251544 |
+
},
|
| 251545 |
+
{
|
| 251546 |
+
"epoch": 964.18,
|
| 251547 |
+
"learning_rate": 8.099951923076924e-06,
|
| 251548 |
+
"loss": 0.7042,
|
| 251549 |
+
"step": 119610
|
| 251550 |
+
},
|
| 251551 |
+
{
|
| 251552 |
+
"epoch": 964.22,
|
| 251553 |
+
"learning_rate": 8.099871794871795e-06,
|
| 251554 |
+
"loss": 1.0577,
|
| 251555 |
+
"step": 119615
|
| 251556 |
+
},
|
| 251557 |
+
{
|
| 251558 |
+
"epoch": 964.26,
|
| 251559 |
+
"learning_rate": 8.099791666666668e-06,
|
| 251560 |
+
"loss": 0.2967,
|
| 251561 |
+
"step": 119620
|
| 251562 |
+
},
|
| 251563 |
+
{
|
| 251564 |
+
"epoch": 964.3,
|
| 251565 |
+
"learning_rate": 8.09971153846154e-06,
|
| 251566 |
+
"loss": 0.2906,
|
| 251567 |
+
"step": 119625
|
| 251568 |
+
},
|
| 251569 |
+
{
|
| 251570 |
+
"epoch": 964.34,
|
| 251571 |
+
"learning_rate": 8.099631410256411e-06,
|
| 251572 |
+
"loss": 0.296,
|
| 251573 |
+
"step": 119630
|
| 251574 |
+
},
|
| 251575 |
+
{
|
| 251576 |
+
"epoch": 964.38,
|
| 251577 |
+
"learning_rate": 8.099551282051282e-06,
|
| 251578 |
+
"loss": 0.6545,
|
| 251579 |
+
"step": 119635
|
| 251580 |
+
},
|
| 251581 |
+
{
|
| 251582 |
+
"epoch": 964.42,
|
| 251583 |
+
"learning_rate": 8.099471153846155e-06,
|
| 251584 |
+
"loss": 0.9743,
|
| 251585 |
+
"step": 119640
|
| 251586 |
+
},
|
| 251587 |
+
{
|
| 251588 |
+
"epoch": 964.46,
|
| 251589 |
+
"learning_rate": 8.099391025641025e-06,
|
| 251590 |
+
"loss": 0.2923,
|
| 251591 |
+
"step": 119645
|
| 251592 |
+
},
|
| 251593 |
+
{
|
| 251594 |
+
"epoch": 964.5,
|
| 251595 |
+
"learning_rate": 8.099310897435898e-06,
|
| 251596 |
+
"loss": 0.2904,
|
| 251597 |
+
"step": 119650
|
| 251598 |
+
},
|
| 251599 |
+
{
|
| 251600 |
+
"epoch": 964.54,
|
| 251601 |
+
"learning_rate": 8.099230769230771e-06,
|
| 251602 |
+
"loss": 0.3179,
|
| 251603 |
+
"step": 119655
|
| 251604 |
+
},
|
| 251605 |
+
{
|
| 251606 |
+
"epoch": 964.58,
|
| 251607 |
+
"learning_rate": 8.09915064102564e-06,
|
| 251608 |
+
"loss": 0.5888,
|
| 251609 |
+
"step": 119660
|
| 251610 |
+
},
|
| 251611 |
+
{
|
| 251612 |
+
"epoch": 964.62,
|
| 251613 |
+
"learning_rate": 8.099070512820514e-06,
|
| 251614 |
+
"loss": 0.9353,
|
| 251615 |
+
"step": 119665
|
| 251616 |
+
},
|
| 251617 |
+
{
|
| 251618 |
+
"epoch": 964.66,
|
| 251619 |
+
"learning_rate": 8.098990384615385e-06,
|
| 251620 |
+
"loss": 0.2513,
|
| 251621 |
+
"step": 119670
|
| 251622 |
+
},
|
| 251623 |
+
{
|
| 251624 |
+
"epoch": 964.7,
|
| 251625 |
+
"learning_rate": 8.098910256410257e-06,
|
| 251626 |
+
"loss": 0.2941,
|
| 251627 |
+
"step": 119675
|
| 251628 |
+
},
|
| 251629 |
+
{
|
| 251630 |
+
"epoch": 964.74,
|
| 251631 |
+
"learning_rate": 8.098830128205128e-06,
|
| 251632 |
+
"loss": 0.3197,
|
| 251633 |
+
"step": 119680
|
| 251634 |
+
},
|
| 251635 |
+
{
|
| 251636 |
+
"epoch": 964.78,
|
| 251637 |
+
"learning_rate": 8.098750000000001e-06,
|
| 251638 |
+
"loss": 0.6669,
|
| 251639 |
+
"step": 119685
|
| 251640 |
+
},
|
| 251641 |
+
{
|
| 251642 |
+
"epoch": 964.82,
|
| 251643 |
+
"learning_rate": 8.098669871794872e-06,
|
| 251644 |
+
"loss": 0.9799,
|
| 251645 |
+
"step": 119690
|
| 251646 |
+
},
|
| 251647 |
+
{
|
| 251648 |
+
"epoch": 964.86,
|
| 251649 |
+
"learning_rate": 8.098589743589744e-06,
|
| 251650 |
+
"loss": 0.2526,
|
| 251651 |
+
"step": 119695
|
| 251652 |
+
},
|
| 251653 |
+
{
|
| 251654 |
+
"epoch": 964.9,
|
| 251655 |
+
"learning_rate": 8.098509615384617e-06,
|
| 251656 |
+
"loss": 0.3663,
|
| 251657 |
+
"step": 119700
|
| 251658 |
+
},
|
| 251659 |
+
{
|
| 251660 |
+
"epoch": 964.94,
|
| 251661 |
+
"learning_rate": 8.098429487179488e-06,
|
| 251662 |
+
"loss": 0.34,
|
| 251663 |
+
"step": 119705
|
| 251664 |
+
},
|
| 251665 |
+
{
|
| 251666 |
+
"epoch": 964.98,
|
| 251667 |
+
"learning_rate": 8.09834935897436e-06,
|
| 251668 |
+
"loss": 0.8057,
|
| 251669 |
+
"step": 119710
|
| 251670 |
+
},
|
| 251671 |
+
{
|
| 251672 |
+
"epoch": 965.0,
|
| 251673 |
+
"eval_loss": 0.36692875623703003,
|
| 251674 |
+
"eval_runtime": 37.1277,
|
| 251675 |
+
"eval_samples_per_second": 22.678,
|
| 251676 |
+
"eval_steps_per_second": 0.727,
|
| 251677 |
+
"eval_wer": 0.18642407277816656,
|
| 251678 |
+
"step": 119712
|
| 251679 |
+
},
|
| 251680 |
+
{
|
| 251681 |
+
"epoch": 957.02,
|
| 251682 |
+
"learning_rate": 8.098269230769231e-06,
|
| 251683 |
+
"loss": 0.3329,
|
| 251684 |
+
"step": 119715
|
| 251685 |
+
},
|
| 251686 |
+
{
|
| 251687 |
+
"epoch": 957.06,
|
| 251688 |
+
"learning_rate": 8.098189102564104e-06,
|
| 251689 |
+
"loss": 0.2888,
|
| 251690 |
+
"step": 119720
|
| 251691 |
+
},
|
| 251692 |
+
{
|
| 251693 |
+
"epoch": 957.1,
|
| 251694 |
+
"learning_rate": 8.098108974358975e-06,
|
| 251695 |
+
"loss": 0.3254,
|
| 251696 |
+
"step": 119725
|
| 251697 |
+
},
|
| 251698 |
+
{
|
| 251699 |
+
"epoch": 957.14,
|
| 251700 |
+
"learning_rate": 8.098028846153847e-06,
|
| 251701 |
+
"loss": 0.3474,
|
| 251702 |
+
"step": 119730
|
| 251703 |
+
},
|
| 251704 |
+
{
|
| 251705 |
+
"epoch": 957.18,
|
| 251706 |
+
"learning_rate": 8.097948717948718e-06,
|
| 251707 |
+
"loss": 0.7834,
|
| 251708 |
+
"step": 119735
|
| 251709 |
+
},
|
| 251710 |
+
{
|
| 251711 |
+
"epoch": 957.22,
|
| 251712 |
+
"learning_rate": 8.097868589743591e-06,
|
| 251713 |
+
"loss": 0.9519,
|
| 251714 |
+
"step": 119740
|
| 251715 |
+
},
|
| 251716 |
+
{
|
| 251717 |
+
"epoch": 957.26,
|
| 251718 |
+
"learning_rate": 8.097788461538462e-06,
|
| 251719 |
+
"loss": 0.3019,
|
| 251720 |
+
"step": 119745
|
| 251721 |
+
},
|
| 251722 |
+
{
|
| 251723 |
+
"epoch": 957.3,
|
| 251724 |
+
"learning_rate": 8.097708333333334e-06,
|
| 251725 |
+
"loss": 0.2702,
|
| 251726 |
+
"step": 119750
|
| 251727 |
+
},
|
| 251728 |
+
{
|
| 251729 |
+
"epoch": 957.34,
|
| 251730 |
+
"learning_rate": 8.097628205128207e-06,
|
| 251731 |
+
"loss": 0.3897,
|
| 251732 |
+
"step": 119755
|
| 251733 |
+
},
|
| 251734 |
+
{
|
| 251735 |
+
"epoch": 957.38,
|
| 251736 |
+
"learning_rate": 8.097548076923078e-06,
|
| 251737 |
+
"loss": 0.8302,
|
| 251738 |
+
"step": 119760
|
| 251739 |
+
},
|
| 251740 |
+
{
|
| 251741 |
+
"epoch": 957.42,
|
| 251742 |
+
"learning_rate": 8.09746794871795e-06,
|
| 251743 |
+
"loss": 0.9148,
|
| 251744 |
+
"step": 119765
|
| 251745 |
+
},
|
| 251746 |
+
{
|
| 251747 |
+
"epoch": 957.46,
|
| 251748 |
+
"learning_rate": 8.097387820512821e-06,
|
| 251749 |
+
"loss": 0.2377,
|
| 251750 |
+
"step": 119770
|
| 251751 |
+
},
|
| 251752 |
+
{
|
| 251753 |
+
"epoch": 957.5,
|
| 251754 |
+
"learning_rate": 8.097307692307694e-06,
|
| 251755 |
+
"loss": 0.3443,
|
| 251756 |
+
"step": 119775
|
| 251757 |
+
},
|
| 251758 |
+
{
|
| 251759 |
+
"epoch": 957.54,
|
| 251760 |
+
"learning_rate": 8.097227564102564e-06,
|
| 251761 |
+
"loss": 0.3632,
|
| 251762 |
+
"step": 119780
|
| 251763 |
+
},
|
| 251764 |
+
{
|
| 251765 |
+
"epoch": 957.58,
|
| 251766 |
+
"learning_rate": 8.097147435897437e-06,
|
| 251767 |
+
"loss": 0.7574,
|
| 251768 |
+
"step": 119785
|
| 251769 |
+
},
|
| 251770 |
+
{
|
| 251771 |
+
"epoch": 957.62,
|
| 251772 |
+
"learning_rate": 8.097067307692308e-06,
|
| 251773 |
+
"loss": 0.8573,
|
| 251774 |
+
"step": 119790
|
| 251775 |
+
},
|
| 251776 |
+
{
|
| 251777 |
+
"epoch": 957.66,
|
| 251778 |
+
"learning_rate": 8.09698717948718e-06,
|
| 251779 |
+
"loss": 0.2621,
|
| 251780 |
+
"step": 119795
|
| 251781 |
+
},
|
| 251782 |
+
{
|
| 251783 |
+
"epoch": 957.7,
|
| 251784 |
+
"learning_rate": 8.096907051282052e-06,
|
| 251785 |
+
"loss": 0.319,
|
| 251786 |
+
"step": 119800
|
| 251787 |
+
},
|
| 251788 |
+
{
|
| 251789 |
+
"epoch": 957.74,
|
| 251790 |
+
"learning_rate": 8.096826923076924e-06,
|
| 251791 |
+
"loss": 0.359,
|
| 251792 |
+
"step": 119805
|
| 251793 |
+
},
|
| 251794 |
+
{
|
| 251795 |
+
"epoch": 957.78,
|
| 251796 |
+
"learning_rate": 8.096746794871795e-06,
|
| 251797 |
+
"loss": 0.6987,
|
| 251798 |
+
"step": 119810
|
| 251799 |
+
},
|
| 251800 |
+
{
|
| 251801 |
+
"epoch": 957.82,
|
| 251802 |
+
"learning_rate": 8.096666666666667e-06,
|
| 251803 |
+
"loss": 0.8702,
|
| 251804 |
+
"step": 119815
|
| 251805 |
+
},
|
| 251806 |
+
{
|
| 251807 |
+
"epoch": 957.86,
|
| 251808 |
+
"learning_rate": 8.09658653846154e-06,
|
| 251809 |
+
"loss": 0.286,
|
| 251810 |
+
"step": 119820
|
| 251811 |
+
},
|
| 251812 |
+
{
|
| 251813 |
+
"epoch": 957.9,
|
| 251814 |
+
"learning_rate": 8.096506410256411e-06,
|
| 251815 |
+
"loss": 0.3204,
|
| 251816 |
+
"step": 119825
|
| 251817 |
+
},
|
| 251818 |
+
{
|
| 251819 |
+
"epoch": 957.94,
|
| 251820 |
+
"learning_rate": 8.096426282051282e-06,
|
| 251821 |
+
"loss": 0.3385,
|
| 251822 |
+
"step": 119830
|
| 251823 |
+
},
|
| 251824 |
+
{
|
| 251825 |
+
"epoch": 957.98,
|
| 251826 |
+
"learning_rate": 8.096346153846154e-06,
|
| 251827 |
+
"loss": 0.7305,
|
| 251828 |
+
"step": 119835
|
| 251829 |
+
},
|
| 251830 |
+
{
|
| 251831 |
+
"epoch": 958.0,
|
| 251832 |
+
"eval_loss": 0.3832470774650574,
|
| 251833 |
+
"eval_runtime": 35.8919,
|
| 251834 |
+
"eval_samples_per_second": 23.459,
|
| 251835 |
+
"eval_steps_per_second": 0.752,
|
| 251836 |
+
"eval_wer": 0.18405627198124266,
|
| 251837 |
+
"step": 119837
|
| 251838 |
}
|
| 251839 |
],
|
| 251840 |
+
"max_steps": 625000,
|
| 251841 |
"num_train_epochs": 5000,
|
| 251842 |
+
"total_flos": 3.3724359344258666e+20,
|
| 251843 |
"trial_name": null,
|
| 251844 |
"trial_params": null
|
| 251845 |
}
|
model-bin/finetune/base/{checkpoint-119214 → checkpoint-119837}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1630126542.1800706/events.out.tfevents.1630126542.86bb0ddabf9b.4092.21
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:71100b3015524c1965a509c46fdea4a639439b2de0747befab5132d8f800c65c
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630126929.6951962/events.out.tfevents.1630126929.86bb0ddabf9b.4092.23
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:4d022c13d5feb07a03a5b2f64e5569613c54087c5750e168deb0e1500951855f
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630127435.6478999/events.out.tfevents.1630127435.86bb0ddabf9b.4092.25
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6b0f058ce4e6c01995a8e318e4ff1dc2896fdc28d66911f25bcb5de460d51c83
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630127823.8657374/events.out.tfevents.1630127823.86bb0ddabf9b.4092.27
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1cd697448d0b8f2e04100dc0afb38e1fad46b8f9134eedf6cf554363e1c1fd24
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630128219.1986852/events.out.tfevents.1630128219.86bb0ddabf9b.4092.29
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9be0965e61d0531927651184a75847eb43cb0a3e9696d69a8d6b47def617382f
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1630126542.86bb0ddabf9b.4092.20
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c953c11b616283054438e1511b46182ba4fd7b044a53773d0f8f77c8f99db2ce
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630126929.86bb0ddabf9b.4092.22
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:29e0f519e51883526f6f8f3b8aa29c4ee84ef460c908a27744c156bc9788ca2e
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630127435.86bb0ddabf9b.4092.24
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a4ebd9caa59a057a99231525f9b6bd9db136d8db05927f91b8859d3c3bc7f480
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630127823.86bb0ddabf9b.4092.26
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:070a8876bd4ac5bec8e33e2345df637e8e70aeb40b92cbdd2a4ec945bfe0bf93
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630128219.86bb0ddabf9b.4092.28
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:efb42f9468425e80997fc57776cfba0f6fbdc74d6d17ba1c4cb1bd13a9f5496f
|
| 3 |
+
size 8622
|