"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-93337 → checkpoint-93960}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-93337 → checkpoint-93960}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-93337 → checkpoint-93960}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-93337 → checkpoint-93960}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-93337 → checkpoint-93960}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-93337 → checkpoint-93960}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-93337 → checkpoint-93960}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-93337 → checkpoint-93960}/trainer_state.json +798 -3
- model-bin/finetune/base/{checkpoint-93337 → checkpoint-93960}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629957682.4055254/events.out.tfevents.1629957682.8e89bd551565.924.81 +3 -0
- model-bin/finetune/base/log/1629958130.6461391/events.out.tfevents.1629958130.8e89bd551565.924.83 +3 -0
- model-bin/finetune/base/log/1629958565.4751377/events.out.tfevents.1629958565.8e89bd551565.924.85 +3 -0
- model-bin/finetune/base/log/1629959004.1539824/events.out.tfevents.1629959004.8e89bd551565.924.87 +3 -0
- model-bin/finetune/base/log/1629959449.6561632/events.out.tfevents.1629959452.8e89bd551565.924.89 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629957682.8e89bd551565.924.80 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629958130.8e89bd551565.924.82 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629958564.8e89bd551565.924.84 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629959004.8e89bd551565.924.86 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629959449.8e89bd551565.924.88 +3 -0
model-bin/finetune/base/{checkpoint-93337 → checkpoint-93960}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-93337 → checkpoint-93960}/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:d17ec4c2fabf26148ba9f60beaad001b3799a3176977e6859fa7e6018f2db434
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-93337 → checkpoint-93960}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-93337 → checkpoint-93960}/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 sha256:2ec6551a97db0f833d76aed5735e0b34957ecb5d651053551090b61d2c0caa7a
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-93337 → checkpoint-93960}/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:bb79bc39717b4fbbb27b011a7feffc6343508dc1c8a09652858314b9dda02de0
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-93337 → checkpoint-93960}/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:fa2098ffd50df5cf9c42261163807b82f181e741115f8e20fd104e5182b93533
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-93337 → checkpoint-93960}/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 sha256:8c63fed4c8116d16f4e14d3f5b432cbccea3336c281a082e52182f2419bb6712
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-93337 → checkpoint-93960}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.17637692697401752,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-92342",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -218118,11 +218118,806 @@
|
|
| 218118 |
"eval_steps_per_second": 0.694,
|
| 218119 |
"eval_wer": 0.1765940508069769,
|
| 218120 |
"step": 93337
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 218121 |
}
|
| 218122 |
],
|
| 218123 |
"max_steps": 620000,
|
| 218124 |
"num_train_epochs": 5000,
|
| 218125 |
-
"total_flos": 2.
|
| 218126 |
"trial_name": null,
|
| 218127 |
"trial_params": null
|
| 218128 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.17637692697401752,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-92342",
|
| 4 |
+
"epoch": 756.995983935743,
|
| 5 |
+
"global_step": 93960,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 218118 |
"eval_steps_per_second": 0.694,
|
| 218119 |
"eval_wer": 0.1765940508069769,
|
| 218120 |
"step": 93337
|
| 218121 |
+
},
|
| 218122 |
+
{
|
| 218123 |
+
"epoch": 746.02,
|
| 218124 |
+
"learning_rate": 8.520785256410257e-06,
|
| 218125 |
+
"loss": 0.2982,
|
| 218126 |
+
"step": 93340
|
| 218127 |
+
},
|
| 218128 |
+
{
|
| 218129 |
+
"epoch": 746.06,
|
| 218130 |
+
"learning_rate": 8.52070512820513e-06,
|
| 218131 |
+
"loss": 0.2931,
|
| 218132 |
+
"step": 93345
|
| 218133 |
+
},
|
| 218134 |
+
{
|
| 218135 |
+
"epoch": 746.1,
|
| 218136 |
+
"learning_rate": 8.520625e-06,
|
| 218137 |
+
"loss": 0.2853,
|
| 218138 |
+
"step": 93350
|
| 218139 |
+
},
|
| 218140 |
+
{
|
| 218141 |
+
"epoch": 746.14,
|
| 218142 |
+
"learning_rate": 8.520544871794872e-06,
|
| 218143 |
+
"loss": 0.3717,
|
| 218144 |
+
"step": 93355
|
| 218145 |
+
},
|
| 218146 |
+
{
|
| 218147 |
+
"epoch": 746.18,
|
| 218148 |
+
"learning_rate": 8.520464743589746e-06,
|
| 218149 |
+
"loss": 0.8425,
|
| 218150 |
+
"step": 93360
|
| 218151 |
+
},
|
| 218152 |
+
{
|
| 218153 |
+
"epoch": 746.22,
|
| 218154 |
+
"learning_rate": 8.520384615384615e-06,
|
| 218155 |
+
"loss": 0.8917,
|
| 218156 |
+
"step": 93365
|
| 218157 |
+
},
|
| 218158 |
+
{
|
| 218159 |
+
"epoch": 746.26,
|
| 218160 |
+
"learning_rate": 8.520304487179488e-06,
|
| 218161 |
+
"loss": 0.2906,
|
| 218162 |
+
"step": 93370
|
| 218163 |
+
},
|
| 218164 |
+
{
|
| 218165 |
+
"epoch": 746.3,
|
| 218166 |
+
"learning_rate": 8.52022435897436e-06,
|
| 218167 |
+
"loss": 0.3077,
|
| 218168 |
+
"step": 93375
|
| 218169 |
+
},
|
| 218170 |
+
{
|
| 218171 |
+
"epoch": 746.34,
|
| 218172 |
+
"learning_rate": 8.520144230769231e-06,
|
| 218173 |
+
"loss": 0.3911,
|
| 218174 |
+
"step": 93380
|
| 218175 |
+
},
|
| 218176 |
+
{
|
| 218177 |
+
"epoch": 746.38,
|
| 218178 |
+
"learning_rate": 8.520064102564102e-06,
|
| 218179 |
+
"loss": 0.7335,
|
| 218180 |
+
"step": 93385
|
| 218181 |
+
},
|
| 218182 |
+
{
|
| 218183 |
+
"epoch": 746.42,
|
| 218184 |
+
"learning_rate": 8.519983974358975e-06,
|
| 218185 |
+
"loss": 0.8044,
|
| 218186 |
+
"step": 93390
|
| 218187 |
+
},
|
| 218188 |
+
{
|
| 218189 |
+
"epoch": 746.46,
|
| 218190 |
+
"learning_rate": 8.519903846153847e-06,
|
| 218191 |
+
"loss": 0.3038,
|
| 218192 |
+
"step": 93395
|
| 218193 |
+
},
|
| 218194 |
+
{
|
| 218195 |
+
"epoch": 746.5,
|
| 218196 |
+
"learning_rate": 8.519823717948718e-06,
|
| 218197 |
+
"loss": 0.2585,
|
| 218198 |
+
"step": 93400
|
| 218199 |
+
},
|
| 218200 |
+
{
|
| 218201 |
+
"epoch": 746.54,
|
| 218202 |
+
"learning_rate": 8.519743589743591e-06,
|
| 218203 |
+
"loss": 0.4126,
|
| 218204 |
+
"step": 93405
|
| 218205 |
+
},
|
| 218206 |
+
{
|
| 218207 |
+
"epoch": 746.58,
|
| 218208 |
+
"learning_rate": 8.519663461538462e-06,
|
| 218209 |
+
"loss": 0.7608,
|
| 218210 |
+
"step": 93410
|
| 218211 |
+
},
|
| 218212 |
+
{
|
| 218213 |
+
"epoch": 746.62,
|
| 218214 |
+
"learning_rate": 8.519583333333334e-06,
|
| 218215 |
+
"loss": 1.017,
|
| 218216 |
+
"step": 93415
|
| 218217 |
+
},
|
| 218218 |
+
{
|
| 218219 |
+
"epoch": 746.66,
|
| 218220 |
+
"learning_rate": 8.519503205128205e-06,
|
| 218221 |
+
"loss": 0.3271,
|
| 218222 |
+
"step": 93420
|
| 218223 |
+
},
|
| 218224 |
+
{
|
| 218225 |
+
"epoch": 746.7,
|
| 218226 |
+
"learning_rate": 8.519423076923078e-06,
|
| 218227 |
+
"loss": 0.4955,
|
| 218228 |
+
"step": 93425
|
| 218229 |
+
},
|
| 218230 |
+
{
|
| 218231 |
+
"epoch": 746.74,
|
| 218232 |
+
"learning_rate": 8.51934294871795e-06,
|
| 218233 |
+
"loss": 0.4092,
|
| 218234 |
+
"step": 93430
|
| 218235 |
+
},
|
| 218236 |
+
{
|
| 218237 |
+
"epoch": 746.78,
|
| 218238 |
+
"learning_rate": 8.519262820512821e-06,
|
| 218239 |
+
"loss": 0.723,
|
| 218240 |
+
"step": 93435
|
| 218241 |
+
},
|
| 218242 |
+
{
|
| 218243 |
+
"epoch": 746.82,
|
| 218244 |
+
"learning_rate": 8.519182692307692e-06,
|
| 218245 |
+
"loss": 0.8715,
|
| 218246 |
+
"step": 93440
|
| 218247 |
+
},
|
| 218248 |
+
{
|
| 218249 |
+
"epoch": 746.86,
|
| 218250 |
+
"learning_rate": 8.519102564102565e-06,
|
| 218251 |
+
"loss": 0.2983,
|
| 218252 |
+
"step": 93445
|
| 218253 |
+
},
|
| 218254 |
+
{
|
| 218255 |
+
"epoch": 746.9,
|
| 218256 |
+
"learning_rate": 8.519022435897437e-06,
|
| 218257 |
+
"loss": 0.2913,
|
| 218258 |
+
"step": 93450
|
| 218259 |
+
},
|
| 218260 |
+
{
|
| 218261 |
+
"epoch": 746.94,
|
| 218262 |
+
"learning_rate": 8.518942307692308e-06,
|
| 218263 |
+
"loss": 0.384,
|
| 218264 |
+
"step": 93455
|
| 218265 |
+
},
|
| 218266 |
+
{
|
| 218267 |
+
"epoch": 746.98,
|
| 218268 |
+
"learning_rate": 8.518862179487181e-06,
|
| 218269 |
+
"loss": 0.7787,
|
| 218270 |
+
"step": 93460
|
| 218271 |
+
},
|
| 218272 |
+
{
|
| 218273 |
+
"epoch": 747.0,
|
| 218274 |
+
"eval_loss": 0.4655894935131073,
|
| 218275 |
+
"eval_runtime": 39.6595,
|
| 218276 |
+
"eval_samples_per_second": 21.13,
|
| 218277 |
+
"eval_steps_per_second": 0.681,
|
| 218278 |
+
"eval_wer": 0.19279159265837773,
|
| 218279 |
+
"step": 93462
|
| 218280 |
+
},
|
| 218281 |
+
{
|
| 218282 |
+
"epoch": 753.02,
|
| 218283 |
+
"learning_rate": 8.518782051282053e-06,
|
| 218284 |
+
"loss": 0.3159,
|
| 218285 |
+
"step": 93465
|
| 218286 |
+
},
|
| 218287 |
+
{
|
| 218288 |
+
"epoch": 753.06,
|
| 218289 |
+
"learning_rate": 8.518701923076924e-06,
|
| 218290 |
+
"loss": 0.2954,
|
| 218291 |
+
"step": 93470
|
| 218292 |
+
},
|
| 218293 |
+
{
|
| 218294 |
+
"epoch": 753.1,
|
| 218295 |
+
"learning_rate": 8.518621794871795e-06,
|
| 218296 |
+
"loss": 0.32,
|
| 218297 |
+
"step": 93475
|
| 218298 |
+
},
|
| 218299 |
+
{
|
| 218300 |
+
"epoch": 753.14,
|
| 218301 |
+
"learning_rate": 8.518541666666668e-06,
|
| 218302 |
+
"loss": 0.4195,
|
| 218303 |
+
"step": 93480
|
| 218304 |
+
},
|
| 218305 |
+
{
|
| 218306 |
+
"epoch": 753.18,
|
| 218307 |
+
"learning_rate": 8.518461538461538e-06,
|
| 218308 |
+
"loss": 0.709,
|
| 218309 |
+
"step": 93485
|
| 218310 |
+
},
|
| 218311 |
+
{
|
| 218312 |
+
"epoch": 753.22,
|
| 218313 |
+
"learning_rate": 8.518381410256411e-06,
|
| 218314 |
+
"loss": 0.8771,
|
| 218315 |
+
"step": 93490
|
| 218316 |
+
},
|
| 218317 |
+
{
|
| 218318 |
+
"epoch": 753.27,
|
| 218319 |
+
"learning_rate": 8.518301282051282e-06,
|
| 218320 |
+
"loss": 0.2633,
|
| 218321 |
+
"step": 93495
|
| 218322 |
+
},
|
| 218323 |
+
{
|
| 218324 |
+
"epoch": 753.31,
|
| 218325 |
+
"learning_rate": 8.518221153846154e-06,
|
| 218326 |
+
"loss": 0.2711,
|
| 218327 |
+
"step": 93500
|
| 218328 |
+
},
|
| 218329 |
+
{
|
| 218330 |
+
"epoch": 753.35,
|
| 218331 |
+
"learning_rate": 8.518141025641027e-06,
|
| 218332 |
+
"loss": 0.4551,
|
| 218333 |
+
"step": 93505
|
| 218334 |
+
},
|
| 218335 |
+
{
|
| 218336 |
+
"epoch": 753.39,
|
| 218337 |
+
"learning_rate": 8.518060897435898e-06,
|
| 218338 |
+
"loss": 0.7144,
|
| 218339 |
+
"step": 93510
|
| 218340 |
+
},
|
| 218341 |
+
{
|
| 218342 |
+
"epoch": 753.43,
|
| 218343 |
+
"learning_rate": 8.51798076923077e-06,
|
| 218344 |
+
"loss": 0.8337,
|
| 218345 |
+
"step": 93515
|
| 218346 |
+
},
|
| 218347 |
+
{
|
| 218348 |
+
"epoch": 753.47,
|
| 218349 |
+
"learning_rate": 8.517900641025641e-06,
|
| 218350 |
+
"loss": 0.2779,
|
| 218351 |
+
"step": 93520
|
| 218352 |
+
},
|
| 218353 |
+
{
|
| 218354 |
+
"epoch": 753.51,
|
| 218355 |
+
"learning_rate": 8.517820512820514e-06,
|
| 218356 |
+
"loss": 0.4137,
|
| 218357 |
+
"step": 93525
|
| 218358 |
+
},
|
| 218359 |
+
{
|
| 218360 |
+
"epoch": 753.55,
|
| 218361 |
+
"learning_rate": 8.517740384615385e-06,
|
| 218362 |
+
"loss": 0.3697,
|
| 218363 |
+
"step": 93530
|
| 218364 |
+
},
|
| 218365 |
+
{
|
| 218366 |
+
"epoch": 753.59,
|
| 218367 |
+
"learning_rate": 8.517660256410257e-06,
|
| 218368 |
+
"loss": 0.7943,
|
| 218369 |
+
"step": 93535
|
| 218370 |
+
},
|
| 218371 |
+
{
|
| 218372 |
+
"epoch": 753.63,
|
| 218373 |
+
"learning_rate": 8.517580128205128e-06,
|
| 218374 |
+
"loss": 0.8019,
|
| 218375 |
+
"step": 93540
|
| 218376 |
+
},
|
| 218377 |
+
{
|
| 218378 |
+
"epoch": 753.67,
|
| 218379 |
+
"learning_rate": 8.517500000000001e-06,
|
| 218380 |
+
"loss": 0.3453,
|
| 218381 |
+
"step": 93545
|
| 218382 |
+
},
|
| 218383 |
+
{
|
| 218384 |
+
"epoch": 753.71,
|
| 218385 |
+
"learning_rate": 8.517419871794872e-06,
|
| 218386 |
+
"loss": 0.3126,
|
| 218387 |
+
"step": 93550
|
| 218388 |
+
},
|
| 218389 |
+
{
|
| 218390 |
+
"epoch": 753.75,
|
| 218391 |
+
"learning_rate": 8.517339743589744e-06,
|
| 218392 |
+
"loss": 0.3679,
|
| 218393 |
+
"step": 93555
|
| 218394 |
+
},
|
| 218395 |
+
{
|
| 218396 |
+
"epoch": 753.79,
|
| 218397 |
+
"learning_rate": 8.517259615384617e-06,
|
| 218398 |
+
"loss": 0.7963,
|
| 218399 |
+
"step": 93560
|
| 218400 |
+
},
|
| 218401 |
+
{
|
| 218402 |
+
"epoch": 753.83,
|
| 218403 |
+
"learning_rate": 8.517179487179488e-06,
|
| 218404 |
+
"loss": 0.8713,
|
| 218405 |
+
"step": 93565
|
| 218406 |
+
},
|
| 218407 |
+
{
|
| 218408 |
+
"epoch": 753.87,
|
| 218409 |
+
"learning_rate": 8.51709935897436e-06,
|
| 218410 |
+
"loss": 0.2662,
|
| 218411 |
+
"step": 93570
|
| 218412 |
+
},
|
| 218413 |
+
{
|
| 218414 |
+
"epoch": 753.91,
|
| 218415 |
+
"learning_rate": 8.517019230769231e-06,
|
| 218416 |
+
"loss": 0.306,
|
| 218417 |
+
"step": 93575
|
| 218418 |
+
},
|
| 218419 |
+
{
|
| 218420 |
+
"epoch": 753.95,
|
| 218421 |
+
"learning_rate": 8.516939102564104e-06,
|
| 218422 |
+
"loss": 0.3898,
|
| 218423 |
+
"step": 93580
|
| 218424 |
+
},
|
| 218425 |
+
{
|
| 218426 |
+
"epoch": 753.99,
|
| 218427 |
+
"learning_rate": 8.516858974358975e-06,
|
| 218428 |
+
"loss": 0.7886,
|
| 218429 |
+
"step": 93585
|
| 218430 |
+
},
|
| 218431 |
+
{
|
| 218432 |
+
"epoch": 754.0,
|
| 218433 |
+
"eval_loss": 0.33248934149742126,
|
| 218434 |
+
"eval_runtime": 40.6199,
|
| 218435 |
+
"eval_samples_per_second": 20.655,
|
| 218436 |
+
"eval_steps_per_second": 0.665,
|
| 218437 |
+
"eval_wer": 0.19042131796903133,
|
| 218438 |
+
"step": 93586
|
| 218439 |
+
},
|
| 218440 |
+
{
|
| 218441 |
+
"epoch": 748.03,
|
| 218442 |
+
"learning_rate": 8.516778846153847e-06,
|
| 218443 |
+
"loss": 0.3288,
|
| 218444 |
+
"step": 93590
|
| 218445 |
+
},
|
| 218446 |
+
{
|
| 218447 |
+
"epoch": 748.07,
|
| 218448 |
+
"learning_rate": 8.516698717948718e-06,
|
| 218449 |
+
"loss": 0.2822,
|
| 218450 |
+
"step": 93595
|
| 218451 |
+
},
|
| 218452 |
+
{
|
| 218453 |
+
"epoch": 748.11,
|
| 218454 |
+
"learning_rate": 8.516618589743591e-06,
|
| 218455 |
+
"loss": 0.3252,
|
| 218456 |
+
"step": 93600
|
| 218457 |
+
},
|
| 218458 |
+
{
|
| 218459 |
+
"epoch": 748.15,
|
| 218460 |
+
"learning_rate": 8.516538461538462e-06,
|
| 218461 |
+
"loss": 0.3961,
|
| 218462 |
+
"step": 93605
|
| 218463 |
+
},
|
| 218464 |
+
{
|
| 218465 |
+
"epoch": 748.19,
|
| 218466 |
+
"learning_rate": 8.516458333333334e-06,
|
| 218467 |
+
"loss": 0.9088,
|
| 218468 |
+
"step": 93610
|
| 218469 |
+
},
|
| 218470 |
+
{
|
| 218471 |
+
"epoch": 748.23,
|
| 218472 |
+
"learning_rate": 8.516378205128207e-06,
|
| 218473 |
+
"loss": 0.7068,
|
| 218474 |
+
"step": 93615
|
| 218475 |
+
},
|
| 218476 |
+
{
|
| 218477 |
+
"epoch": 748.27,
|
| 218478 |
+
"learning_rate": 8.516298076923078e-06,
|
| 218479 |
+
"loss": 0.2668,
|
| 218480 |
+
"step": 93620
|
| 218481 |
+
},
|
| 218482 |
+
{
|
| 218483 |
+
"epoch": 748.31,
|
| 218484 |
+
"learning_rate": 8.51621794871795e-06,
|
| 218485 |
+
"loss": 0.298,
|
| 218486 |
+
"step": 93625
|
| 218487 |
+
},
|
| 218488 |
+
{
|
| 218489 |
+
"epoch": 748.35,
|
| 218490 |
+
"learning_rate": 8.516137820512821e-06,
|
| 218491 |
+
"loss": 0.415,
|
| 218492 |
+
"step": 93630
|
| 218493 |
+
},
|
| 218494 |
+
{
|
| 218495 |
+
"epoch": 748.39,
|
| 218496 |
+
"learning_rate": 8.516057692307694e-06,
|
| 218497 |
+
"loss": 0.9226,
|
| 218498 |
+
"step": 93635
|
| 218499 |
+
},
|
| 218500 |
+
{
|
| 218501 |
+
"epoch": 748.43,
|
| 218502 |
+
"learning_rate": 8.515977564102564e-06,
|
| 218503 |
+
"loss": 0.69,
|
| 218504 |
+
"step": 93640
|
| 218505 |
+
},
|
| 218506 |
+
{
|
| 218507 |
+
"epoch": 748.47,
|
| 218508 |
+
"learning_rate": 8.515897435897437e-06,
|
| 218509 |
+
"loss": 0.2476,
|
| 218510 |
+
"step": 93645
|
| 218511 |
+
},
|
| 218512 |
+
{
|
| 218513 |
+
"epoch": 748.51,
|
| 218514 |
+
"learning_rate": 8.515817307692308e-06,
|
| 218515 |
+
"loss": 0.325,
|
| 218516 |
+
"step": 93650
|
| 218517 |
+
},
|
| 218518 |
+
{
|
| 218519 |
+
"epoch": 748.55,
|
| 218520 |
+
"learning_rate": 8.51573717948718e-06,
|
| 218521 |
+
"loss": 0.4583,
|
| 218522 |
+
"step": 93655
|
| 218523 |
+
},
|
| 218524 |
+
{
|
| 218525 |
+
"epoch": 748.59,
|
| 218526 |
+
"learning_rate": 8.515657051282052e-06,
|
| 218527 |
+
"loss": 0.8614,
|
| 218528 |
+
"step": 93660
|
| 218529 |
+
},
|
| 218530 |
+
{
|
| 218531 |
+
"epoch": 748.63,
|
| 218532 |
+
"learning_rate": 8.515576923076924e-06,
|
| 218533 |
+
"loss": 0.6535,
|
| 218534 |
+
"step": 93665
|
| 218535 |
+
},
|
| 218536 |
+
{
|
| 218537 |
+
"epoch": 748.67,
|
| 218538 |
+
"learning_rate": 8.515496794871795e-06,
|
| 218539 |
+
"loss": 0.3172,
|
| 218540 |
+
"step": 93670
|
| 218541 |
+
},
|
| 218542 |
+
{
|
| 218543 |
+
"epoch": 748.71,
|
| 218544 |
+
"learning_rate": 8.515416666666667e-06,
|
| 218545 |
+
"loss": 0.314,
|
| 218546 |
+
"step": 93675
|
| 218547 |
+
},
|
| 218548 |
+
{
|
| 218549 |
+
"epoch": 748.75,
|
| 218550 |
+
"learning_rate": 8.51533653846154e-06,
|
| 218551 |
+
"loss": 0.4656,
|
| 218552 |
+
"step": 93680
|
| 218553 |
+
},
|
| 218554 |
+
{
|
| 218555 |
+
"epoch": 748.79,
|
| 218556 |
+
"learning_rate": 8.515256410256411e-06,
|
| 218557 |
+
"loss": 0.9987,
|
| 218558 |
+
"step": 93685
|
| 218559 |
+
},
|
| 218560 |
+
{
|
| 218561 |
+
"epoch": 748.83,
|
| 218562 |
+
"learning_rate": 8.515176282051282e-06,
|
| 218563 |
+
"loss": 0.6576,
|
| 218564 |
+
"step": 93690
|
| 218565 |
+
},
|
| 218566 |
+
{
|
| 218567 |
+
"epoch": 748.87,
|
| 218568 |
+
"learning_rate": 8.515096153846154e-06,
|
| 218569 |
+
"loss": 0.3404,
|
| 218570 |
+
"step": 93695
|
| 218571 |
+
},
|
| 218572 |
+
{
|
| 218573 |
+
"epoch": 748.91,
|
| 218574 |
+
"learning_rate": 8.515016025641027e-06,
|
| 218575 |
+
"loss": 0.2821,
|
| 218576 |
+
"step": 93700
|
| 218577 |
+
},
|
| 218578 |
+
{
|
| 218579 |
+
"epoch": 748.95,
|
| 218580 |
+
"learning_rate": 8.514935897435898e-06,
|
| 218581 |
+
"loss": 0.3785,
|
| 218582 |
+
"step": 93705
|
| 218583 |
+
},
|
| 218584 |
+
{
|
| 218585 |
+
"epoch": 748.99,
|
| 218586 |
+
"learning_rate": 8.51485576923077e-06,
|
| 218587 |
+
"loss": 0.966,
|
| 218588 |
+
"step": 93710
|
| 218589 |
+
},
|
| 218590 |
+
{
|
| 218591 |
+
"epoch": 749.0,
|
| 218592 |
+
"eval_loss": 0.4866238534450531,
|
| 218593 |
+
"eval_runtime": 38.5067,
|
| 218594 |
+
"eval_samples_per_second": 21.788,
|
| 218595 |
+
"eval_steps_per_second": 0.701,
|
| 218596 |
+
"eval_wer": 0.19120631341600902,
|
| 218597 |
+
"step": 93711
|
| 218598 |
+
},
|
| 218599 |
+
{
|
| 218600 |
+
"epoch": 749.03,
|
| 218601 |
+
"learning_rate": 8.514775641025642e-06,
|
| 218602 |
+
"loss": 0.369,
|
| 218603 |
+
"step": 93715
|
| 218604 |
+
},
|
| 218605 |
+
{
|
| 218606 |
+
"epoch": 749.07,
|
| 218607 |
+
"learning_rate": 8.514695512820514e-06,
|
| 218608 |
+
"loss": 0.3605,
|
| 218609 |
+
"step": 93720
|
| 218610 |
+
},
|
| 218611 |
+
{
|
| 218612 |
+
"epoch": 749.11,
|
| 218613 |
+
"learning_rate": 8.514615384615385e-06,
|
| 218614 |
+
"loss": 0.2692,
|
| 218615 |
+
"step": 93725
|
| 218616 |
+
},
|
| 218617 |
+
{
|
| 218618 |
+
"epoch": 749.15,
|
| 218619 |
+
"learning_rate": 8.514535256410257e-06,
|
| 218620 |
+
"loss": 0.4142,
|
| 218621 |
+
"step": 93730
|
| 218622 |
+
},
|
| 218623 |
+
{
|
| 218624 |
+
"epoch": 749.19,
|
| 218625 |
+
"learning_rate": 8.51445512820513e-06,
|
| 218626 |
+
"loss": 0.9357,
|
| 218627 |
+
"step": 93735
|
| 218628 |
+
},
|
| 218629 |
+
{
|
| 218630 |
+
"epoch": 749.23,
|
| 218631 |
+
"learning_rate": 8.514375000000001e-06,
|
| 218632 |
+
"loss": 0.6897,
|
| 218633 |
+
"step": 93740
|
| 218634 |
+
},
|
| 218635 |
+
{
|
| 218636 |
+
"epoch": 749.27,
|
| 218637 |
+
"learning_rate": 8.514294871794872e-06,
|
| 218638 |
+
"loss": 0.3877,
|
| 218639 |
+
"step": 93745
|
| 218640 |
+
},
|
| 218641 |
+
{
|
| 218642 |
+
"epoch": 749.31,
|
| 218643 |
+
"learning_rate": 8.514214743589744e-06,
|
| 218644 |
+
"loss": 0.3677,
|
| 218645 |
+
"step": 93750
|
| 218646 |
+
},
|
| 218647 |
+
{
|
| 218648 |
+
"epoch": 749.35,
|
| 218649 |
+
"learning_rate": 8.514134615384617e-06,
|
| 218650 |
+
"loss": 0.4088,
|
| 218651 |
+
"step": 93755
|
| 218652 |
+
},
|
| 218653 |
+
{
|
| 218654 |
+
"epoch": 749.39,
|
| 218655 |
+
"learning_rate": 8.514054487179488e-06,
|
| 218656 |
+
"loss": 0.8744,
|
| 218657 |
+
"step": 93760
|
| 218658 |
+
},
|
| 218659 |
+
{
|
| 218660 |
+
"epoch": 749.43,
|
| 218661 |
+
"learning_rate": 8.51397435897436e-06,
|
| 218662 |
+
"loss": 0.651,
|
| 218663 |
+
"step": 93765
|
| 218664 |
+
},
|
| 218665 |
+
{
|
| 218666 |
+
"epoch": 749.47,
|
| 218667 |
+
"learning_rate": 8.513894230769232e-06,
|
| 218668 |
+
"loss": 0.2776,
|
| 218669 |
+
"step": 93770
|
| 218670 |
+
},
|
| 218671 |
+
{
|
| 218672 |
+
"epoch": 749.51,
|
| 218673 |
+
"learning_rate": 8.513814102564104e-06,
|
| 218674 |
+
"loss": 0.4864,
|
| 218675 |
+
"step": 93775
|
| 218676 |
+
},
|
| 218677 |
+
{
|
| 218678 |
+
"epoch": 749.55,
|
| 218679 |
+
"learning_rate": 8.513733974358975e-06,
|
| 218680 |
+
"loss": 0.4179,
|
| 218681 |
+
"step": 93780
|
| 218682 |
+
},
|
| 218683 |
+
{
|
| 218684 |
+
"epoch": 749.59,
|
| 218685 |
+
"learning_rate": 8.513653846153847e-06,
|
| 218686 |
+
"loss": 0.9494,
|
| 218687 |
+
"step": 93785
|
| 218688 |
+
},
|
| 218689 |
+
{
|
| 218690 |
+
"epoch": 749.63,
|
| 218691 |
+
"learning_rate": 8.51357371794872e-06,
|
| 218692 |
+
"loss": 0.6507,
|
| 218693 |
+
"step": 93790
|
| 218694 |
+
},
|
| 218695 |
+
{
|
| 218696 |
+
"epoch": 749.67,
|
| 218697 |
+
"learning_rate": 8.51349358974359e-06,
|
| 218698 |
+
"loss": 0.3145,
|
| 218699 |
+
"step": 93795
|
| 218700 |
+
},
|
| 218701 |
+
{
|
| 218702 |
+
"epoch": 749.71,
|
| 218703 |
+
"learning_rate": 8.513413461538462e-06,
|
| 218704 |
+
"loss": 0.3237,
|
| 218705 |
+
"step": 93800
|
| 218706 |
+
},
|
| 218707 |
+
{
|
| 218708 |
+
"epoch": 749.75,
|
| 218709 |
+
"learning_rate": 8.513333333333335e-06,
|
| 218710 |
+
"loss": 0.397,
|
| 218711 |
+
"step": 93805
|
| 218712 |
+
},
|
| 218713 |
+
{
|
| 218714 |
+
"epoch": 749.79,
|
| 218715 |
+
"learning_rate": 8.513253205128205e-06,
|
| 218716 |
+
"loss": 0.9229,
|
| 218717 |
+
"step": 93810
|
| 218718 |
+
},
|
| 218719 |
+
{
|
| 218720 |
+
"epoch": 749.83,
|
| 218721 |
+
"learning_rate": 8.513173076923078e-06,
|
| 218722 |
+
"loss": 0.6785,
|
| 218723 |
+
"step": 93815
|
| 218724 |
+
},
|
| 218725 |
+
{
|
| 218726 |
+
"epoch": 749.87,
|
| 218727 |
+
"learning_rate": 8.51309294871795e-06,
|
| 218728 |
+
"loss": 0.3045,
|
| 218729 |
+
"step": 93820
|
| 218730 |
+
},
|
| 218731 |
+
{
|
| 218732 |
+
"epoch": 749.91,
|
| 218733 |
+
"learning_rate": 8.51301282051282e-06,
|
| 218734 |
+
"loss": 0.362,
|
| 218735 |
+
"step": 93825
|
| 218736 |
+
},
|
| 218737 |
+
{
|
| 218738 |
+
"epoch": 749.95,
|
| 218739 |
+
"learning_rate": 8.512932692307692e-06,
|
| 218740 |
+
"loss": 0.4119,
|
| 218741 |
+
"step": 93830
|
| 218742 |
+
},
|
| 218743 |
+
{
|
| 218744 |
+
"epoch": 749.99,
|
| 218745 |
+
"learning_rate": 8.512852564102565e-06,
|
| 218746 |
+
"loss": 1.0564,
|
| 218747 |
+
"step": 93835
|
| 218748 |
+
},
|
| 218749 |
+
{
|
| 218750 |
+
"epoch": 750.0,
|
| 218751 |
+
"eval_loss": 0.3657403886318207,
|
| 218752 |
+
"eval_runtime": 40.3739,
|
| 218753 |
+
"eval_samples_per_second": 20.756,
|
| 218754 |
+
"eval_steps_per_second": 0.669,
|
| 218755 |
+
"eval_wer": 0.18745001090671126,
|
| 218756 |
+
"step": 93836
|
| 218757 |
+
},
|
| 218758 |
+
{
|
| 218759 |
+
"epoch": 756.03,
|
| 218760 |
+
"learning_rate": 8.512772435897437e-06,
|
| 218761 |
+
"loss": 0.2971,
|
| 218762 |
+
"step": 93840
|
| 218763 |
+
},
|
| 218764 |
+
{
|
| 218765 |
+
"epoch": 756.07,
|
| 218766 |
+
"learning_rate": 8.512692307692308e-06,
|
| 218767 |
+
"loss": 0.3112,
|
| 218768 |
+
"step": 93845
|
| 218769 |
+
},
|
| 218770 |
+
{
|
| 218771 |
+
"epoch": 756.11,
|
| 218772 |
+
"learning_rate": 8.51261217948718e-06,
|
| 218773 |
+
"loss": 0.323,
|
| 218774 |
+
"step": 93850
|
| 218775 |
+
},
|
| 218776 |
+
{
|
| 218777 |
+
"epoch": 756.15,
|
| 218778 |
+
"learning_rate": 8.512532051282052e-06,
|
| 218779 |
+
"loss": 0.3942,
|
| 218780 |
+
"step": 93855
|
| 218781 |
+
},
|
| 218782 |
+
{
|
| 218783 |
+
"epoch": 756.19,
|
| 218784 |
+
"learning_rate": 8.512451923076924e-06,
|
| 218785 |
+
"loss": 0.9728,
|
| 218786 |
+
"step": 93860
|
| 218787 |
+
},
|
| 218788 |
+
{
|
| 218789 |
+
"epoch": 756.23,
|
| 218790 |
+
"learning_rate": 8.512371794871795e-06,
|
| 218791 |
+
"loss": 0.6976,
|
| 218792 |
+
"step": 93865
|
| 218793 |
+
},
|
| 218794 |
+
{
|
| 218795 |
+
"epoch": 756.27,
|
| 218796 |
+
"learning_rate": 8.512291666666668e-06,
|
| 218797 |
+
"loss": 0.3086,
|
| 218798 |
+
"step": 93870
|
| 218799 |
+
},
|
| 218800 |
+
{
|
| 218801 |
+
"epoch": 756.31,
|
| 218802 |
+
"learning_rate": 8.51221153846154e-06,
|
| 218803 |
+
"loss": 0.2799,
|
| 218804 |
+
"step": 93875
|
| 218805 |
+
},
|
| 218806 |
+
{
|
| 218807 |
+
"epoch": 756.35,
|
| 218808 |
+
"learning_rate": 8.512131410256411e-06,
|
| 218809 |
+
"loss": 0.3701,
|
| 218810 |
+
"step": 93880
|
| 218811 |
+
},
|
| 218812 |
+
{
|
| 218813 |
+
"epoch": 756.39,
|
| 218814 |
+
"learning_rate": 8.512051282051282e-06,
|
| 218815 |
+
"loss": 0.9532,
|
| 218816 |
+
"step": 93885
|
| 218817 |
+
},
|
| 218818 |
+
{
|
| 218819 |
+
"epoch": 756.43,
|
| 218820 |
+
"learning_rate": 8.511971153846155e-06,
|
| 218821 |
+
"loss": 0.7129,
|
| 218822 |
+
"step": 93890
|
| 218823 |
+
},
|
| 218824 |
+
{
|
| 218825 |
+
"epoch": 756.47,
|
| 218826 |
+
"learning_rate": 8.511891025641027e-06,
|
| 218827 |
+
"loss": 0.308,
|
| 218828 |
+
"step": 93895
|
| 218829 |
+
},
|
| 218830 |
+
{
|
| 218831 |
+
"epoch": 756.51,
|
| 218832 |
+
"learning_rate": 8.511810897435898e-06,
|
| 218833 |
+
"loss": 0.322,
|
| 218834 |
+
"step": 93900
|
| 218835 |
+
},
|
| 218836 |
+
{
|
| 218837 |
+
"epoch": 756.55,
|
| 218838 |
+
"learning_rate": 8.511730769230771e-06,
|
| 218839 |
+
"loss": 0.4047,
|
| 218840 |
+
"step": 93905
|
| 218841 |
+
},
|
| 218842 |
+
{
|
| 218843 |
+
"epoch": 756.59,
|
| 218844 |
+
"learning_rate": 8.511650641025642e-06,
|
| 218845 |
+
"loss": 0.8819,
|
| 218846 |
+
"step": 93910
|
| 218847 |
+
},
|
| 218848 |
+
{
|
| 218849 |
+
"epoch": 756.63,
|
| 218850 |
+
"learning_rate": 8.511570512820514e-06,
|
| 218851 |
+
"loss": 0.7582,
|
| 218852 |
+
"step": 93915
|
| 218853 |
+
},
|
| 218854 |
+
{
|
| 218855 |
+
"epoch": 756.67,
|
| 218856 |
+
"learning_rate": 8.511490384615385e-06,
|
| 218857 |
+
"loss": 0.286,
|
| 218858 |
+
"step": 93920
|
| 218859 |
+
},
|
| 218860 |
+
{
|
| 218861 |
+
"epoch": 756.71,
|
| 218862 |
+
"learning_rate": 8.511410256410258e-06,
|
| 218863 |
+
"loss": 0.326,
|
| 218864 |
+
"step": 93925
|
| 218865 |
+
},
|
| 218866 |
+
{
|
| 218867 |
+
"epoch": 756.76,
|
| 218868 |
+
"learning_rate": 8.511330128205128e-06,
|
| 218869 |
+
"loss": 0.4234,
|
| 218870 |
+
"step": 93930
|
| 218871 |
+
},
|
| 218872 |
+
{
|
| 218873 |
+
"epoch": 756.8,
|
| 218874 |
+
"learning_rate": 8.511250000000001e-06,
|
| 218875 |
+
"loss": 0.8456,
|
| 218876 |
+
"step": 93935
|
| 218877 |
+
},
|
| 218878 |
+
{
|
| 218879 |
+
"epoch": 756.84,
|
| 218880 |
+
"learning_rate": 8.511169871794872e-06,
|
| 218881 |
+
"loss": 0.6125,
|
| 218882 |
+
"step": 93940
|
| 218883 |
+
},
|
| 218884 |
+
{
|
| 218885 |
+
"epoch": 756.88,
|
| 218886 |
+
"learning_rate": 8.511089743589744e-06,
|
| 218887 |
+
"loss": 0.2815,
|
| 218888 |
+
"step": 93945
|
| 218889 |
+
},
|
| 218890 |
+
{
|
| 218891 |
+
"epoch": 756.92,
|
| 218892 |
+
"learning_rate": 8.511009615384615e-06,
|
| 218893 |
+
"loss": 0.3681,
|
| 218894 |
+
"step": 93950
|
| 218895 |
+
},
|
| 218896 |
+
{
|
| 218897 |
+
"epoch": 756.96,
|
| 218898 |
+
"learning_rate": 8.510929487179488e-06,
|
| 218899 |
+
"loss": 0.4144,
|
| 218900 |
+
"step": 93955
|
| 218901 |
+
},
|
| 218902 |
+
{
|
| 218903 |
+
"epoch": 757.0,
|
| 218904 |
+
"learning_rate": 8.51084935897436e-06,
|
| 218905 |
+
"loss": 1.381,
|
| 218906 |
+
"step": 93960
|
| 218907 |
+
},
|
| 218908 |
+
{
|
| 218909 |
+
"epoch": 757.0,
|
| 218910 |
+
"eval_loss": 0.3931798040866852,
|
| 218911 |
+
"eval_runtime": 39.9111,
|
| 218912 |
+
"eval_samples_per_second": 20.997,
|
| 218913 |
+
"eval_steps_per_second": 0.677,
|
| 218914 |
+
"eval_wer": 0.18971618401097712,
|
| 218915 |
+
"step": 93960
|
| 218916 |
}
|
| 218917 |
],
|
| 218918 |
"max_steps": 620000,
|
| 218919 |
"num_train_epochs": 5000,
|
| 218920 |
+
"total_flos": 2.6441517960043604e+20,
|
| 218921 |
"trial_name": null,
|
| 218922 |
"trial_params": null
|
| 218923 |
}
|
model-bin/finetune/base/{checkpoint-93337 → checkpoint-93960}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629957682.4055254/events.out.tfevents.1629957682.8e89bd551565.924.81
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:08d105a78974f4235280dc7827683e6e3a02694328e40455e1f73a5dc7409ca9
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629958130.6461391/events.out.tfevents.1629958130.8e89bd551565.924.83
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b2444dcaf2d538fbf02c4eaaf5596949914d14932ae1c55a8297d6eafb1a16d3
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629958565.4751377/events.out.tfevents.1629958565.8e89bd551565.924.85
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1a20a88ca95d6bc9a281c7be2358a950b88090ce81a72127f6fa436dc3ac5ece
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629959004.1539824/events.out.tfevents.1629959004.8e89bd551565.924.87
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1332b148c9309be6a77dc7699c5c92d81dfaa642851a89ce97d1b0f76df3ca8c
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629959449.6561632/events.out.tfevents.1629959452.8e89bd551565.924.89
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:230ecdd4f4f83e9fede175e00b6bd646cba700243f95f1fe38ee3279124b1444
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629957682.8e89bd551565.924.80
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:62dc008537e2ba3f1407197e57f85e60c50b420a27db552981753562a26523c8
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629958130.8e89bd551565.924.82
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:96c60b3ff922ed4778261d4c34a02d7b87b681251ab2e3f0e56228f0dd414727
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629958564.8e89bd551565.924.84
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:5511f356e623b81a41a778588a50378720fdce16512852166215e84262114767
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629959004.8e89bd551565.924.86
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b46e42f14ae69899e0202eccc573e1d0e0f284b7cdc43ed9271ebe7616bb09c7
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629959449.8e89bd551565.924.88
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:76ecb0389ff04935739f34be9a4d9d2b1ebd5043d47b450d6bf3730518d4f1bc
|
| 3 |
+
size 8622
|