"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-108271 → checkpoint-108894}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-108271 → checkpoint-108894}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-108271 → checkpoint-108894}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-108271 → checkpoint-108894}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-108271 → checkpoint-108894}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-108271 → checkpoint-108894}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-108271 → checkpoint-108894}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-108271 → checkpoint-108894}/trainer_state.json +792 -3
- model-bin/finetune/base/{checkpoint-108271 → checkpoint-108894}/training_args.bin +0 -0
- model-bin/finetune/base/log/1630036151.3539824/events.out.tfevents.1630036151.52f5c7e305a3.886.31 +3 -0
- model-bin/finetune/base/log/1630036627.3019376/events.out.tfevents.1630036627.52f5c7e305a3.886.33 +3 -0
- model-bin/finetune/base/log/1630037085.2077386/events.out.tfevents.1630037085.52f5c7e305a3.886.35 +3 -0
- model-bin/finetune/base/log/1630037540.4465652/events.out.tfevents.1630037540.52f5c7e305a3.886.37 +3 -0
- model-bin/finetune/base/log/1630038001.097344/events.out.tfevents.1630038001.52f5c7e305a3.886.39 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630036151.52f5c7e305a3.886.30 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630036627.52f5c7e305a3.886.32 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630037085.52f5c7e305a3.886.34 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630037540.52f5c7e305a3.886.36 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630038001.52f5c7e305a3.886.38 +3 -0
model-bin/finetune/base/{checkpoint-108271 → checkpoint-108894}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-108271 → checkpoint-108894}/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 sha256:5ed84f90fce277309fcc079c636a580b0a9951877365de9d22e0a45b325f4959
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-108271 → checkpoint-108894}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-108271 → checkpoint-108894}/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 sha256:39714b9078cb72ab6d02184b330358f312f7c0beda11d9d911841507ed9f4072
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-108271 → checkpoint-108894}/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 sha256:11325d828073bc28dba0e547a02b539378ed66d4d97bcd76ac0fa374d53c78a0
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-108271 → checkpoint-108894}/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:b566898de17294f298a5b112ea4a529d2e3e5586226b104d900242be9b719cae
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-108271 → checkpoint-108894}/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:aa4b134d7e3400d11766c99ee8aa673798295021b7bb19c2a048b99ee2bff2af
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-108271 → checkpoint-108894}/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,
|
|
@@ -237120,11 +237120,800 @@
|
|
| 237120 |
"eval_steps_per_second": 0.636,
|
| 237121 |
"eval_wer": 0.1901564344746163,
|
| 237122 |
"step": 108271
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 237123 |
}
|
| 237124 |
],
|
| 237125 |
"max_steps": 620000,
|
| 237126 |
"num_train_epochs": 5000,
|
| 237127 |
-
"total_flos": 3.
|
| 237128 |
"trial_name": null,
|
| 237129 |
"trial_params": null
|
| 237130 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1743826049391605,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-101551",
|
| 4 |
+
"epoch": 878.0,
|
| 5 |
+
"global_step": 108894,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 237120 |
"eval_steps_per_second": 0.636,
|
| 237121 |
"eval_wer": 0.1901564344746163,
|
| 237122 |
"step": 108271
|
| 237123 |
+
},
|
| 237124 |
+
{
|
| 237125 |
+
"epoch": 866.03,
|
| 237126 |
+
"learning_rate": 8.267657512116318e-06,
|
| 237127 |
+
"loss": 0.3466,
|
| 237128 |
+
"step": 108275
|
| 237129 |
+
},
|
| 237130 |
+
{
|
| 237131 |
+
"epoch": 866.07,
|
| 237132 |
+
"learning_rate": 8.267576736672053e-06,
|
| 237133 |
+
"loss": 0.307,
|
| 237134 |
+
"step": 108280
|
| 237135 |
+
},
|
| 237136 |
+
{
|
| 237137 |
+
"epoch": 866.11,
|
| 237138 |
+
"learning_rate": 8.267495961227788e-06,
|
| 237139 |
+
"loss": 0.2886,
|
| 237140 |
+
"step": 108285
|
| 237141 |
+
},
|
| 237142 |
+
{
|
| 237143 |
+
"epoch": 866.15,
|
| 237144 |
+
"learning_rate": 8.267415185783523e-06,
|
| 237145 |
+
"loss": 0.3951,
|
| 237146 |
+
"step": 108290
|
| 237147 |
+
},
|
| 237148 |
+
{
|
| 237149 |
+
"epoch": 866.19,
|
| 237150 |
+
"learning_rate": 8.267334410339258e-06,
|
| 237151 |
+
"loss": 0.865,
|
| 237152 |
+
"step": 108295
|
| 237153 |
+
},
|
| 237154 |
+
{
|
| 237155 |
+
"epoch": 866.23,
|
| 237156 |
+
"learning_rate": 8.267253634894993e-06,
|
| 237157 |
+
"loss": 0.6746,
|
| 237158 |
+
"step": 108300
|
| 237159 |
+
},
|
| 237160 |
+
{
|
| 237161 |
+
"epoch": 866.27,
|
| 237162 |
+
"learning_rate": 8.267172859450728e-06,
|
| 237163 |
+
"loss": 0.2599,
|
| 237164 |
+
"step": 108305
|
| 237165 |
+
},
|
| 237166 |
+
{
|
| 237167 |
+
"epoch": 866.31,
|
| 237168 |
+
"learning_rate": 8.267092084006463e-06,
|
| 237169 |
+
"loss": 0.2727,
|
| 237170 |
+
"step": 108310
|
| 237171 |
+
},
|
| 237172 |
+
{
|
| 237173 |
+
"epoch": 866.35,
|
| 237174 |
+
"learning_rate": 8.267011308562198e-06,
|
| 237175 |
+
"loss": 0.4177,
|
| 237176 |
+
"step": 108315
|
| 237177 |
+
},
|
| 237178 |
+
{
|
| 237179 |
+
"epoch": 866.39,
|
| 237180 |
+
"learning_rate": 8.266930533117933e-06,
|
| 237181 |
+
"loss": 0.9478,
|
| 237182 |
+
"step": 108320
|
| 237183 |
+
},
|
| 237184 |
+
{
|
| 237185 |
+
"epoch": 866.43,
|
| 237186 |
+
"learning_rate": 8.26684975767367e-06,
|
| 237187 |
+
"loss": 0.5904,
|
| 237188 |
+
"step": 108325
|
| 237189 |
+
},
|
| 237190 |
+
{
|
| 237191 |
+
"epoch": 866.47,
|
| 237192 |
+
"learning_rate": 8.266768982229403e-06,
|
| 237193 |
+
"loss": 0.2861,
|
| 237194 |
+
"step": 108330
|
| 237195 |
+
},
|
| 237196 |
+
{
|
| 237197 |
+
"epoch": 866.51,
|
| 237198 |
+
"learning_rate": 8.26668820678514e-06,
|
| 237199 |
+
"loss": 0.2866,
|
| 237200 |
+
"step": 108335
|
| 237201 |
+
},
|
| 237202 |
+
{
|
| 237203 |
+
"epoch": 866.55,
|
| 237204 |
+
"learning_rate": 8.266607431340873e-06,
|
| 237205 |
+
"loss": 0.363,
|
| 237206 |
+
"step": 108340
|
| 237207 |
+
},
|
| 237208 |
+
{
|
| 237209 |
+
"epoch": 866.59,
|
| 237210 |
+
"learning_rate": 8.266526655896609e-06,
|
| 237211 |
+
"loss": 0.8137,
|
| 237212 |
+
"step": 108345
|
| 237213 |
+
},
|
| 237214 |
+
{
|
| 237215 |
+
"epoch": 866.63,
|
| 237216 |
+
"learning_rate": 8.266445880452343e-06,
|
| 237217 |
+
"loss": 0.6726,
|
| 237218 |
+
"step": 108350
|
| 237219 |
+
},
|
| 237220 |
+
{
|
| 237221 |
+
"epoch": 866.67,
|
| 237222 |
+
"learning_rate": 8.266365105008079e-06,
|
| 237223 |
+
"loss": 0.2582,
|
| 237224 |
+
"step": 108355
|
| 237225 |
+
},
|
| 237226 |
+
{
|
| 237227 |
+
"epoch": 866.71,
|
| 237228 |
+
"learning_rate": 8.266284329563813e-06,
|
| 237229 |
+
"loss": 0.3101,
|
| 237230 |
+
"step": 108360
|
| 237231 |
+
},
|
| 237232 |
+
{
|
| 237233 |
+
"epoch": 866.75,
|
| 237234 |
+
"learning_rate": 8.266203554119549e-06,
|
| 237235 |
+
"loss": 0.4569,
|
| 237236 |
+
"step": 108365
|
| 237237 |
+
},
|
| 237238 |
+
{
|
| 237239 |
+
"epoch": 866.79,
|
| 237240 |
+
"learning_rate": 8.266122778675283e-06,
|
| 237241 |
+
"loss": 0.9037,
|
| 237242 |
+
"step": 108370
|
| 237243 |
+
},
|
| 237244 |
+
{
|
| 237245 |
+
"epoch": 866.83,
|
| 237246 |
+
"learning_rate": 8.266042003231019e-06,
|
| 237247 |
+
"loss": 0.5905,
|
| 237248 |
+
"step": 108375
|
| 237249 |
+
},
|
| 237250 |
+
{
|
| 237251 |
+
"epoch": 866.87,
|
| 237252 |
+
"learning_rate": 8.265961227786753e-06,
|
| 237253 |
+
"loss": 0.2725,
|
| 237254 |
+
"step": 108380
|
| 237255 |
+
},
|
| 237256 |
+
{
|
| 237257 |
+
"epoch": 866.91,
|
| 237258 |
+
"learning_rate": 8.265880452342489e-06,
|
| 237259 |
+
"loss": 0.304,
|
| 237260 |
+
"step": 108385
|
| 237261 |
+
},
|
| 237262 |
+
{
|
| 237263 |
+
"epoch": 866.95,
|
| 237264 |
+
"learning_rate": 8.265799676898225e-06,
|
| 237265 |
+
"loss": 0.4479,
|
| 237266 |
+
"step": 108390
|
| 237267 |
+
},
|
| 237268 |
+
{
|
| 237269 |
+
"epoch": 866.99,
|
| 237270 |
+
"learning_rate": 8.265718901453959e-06,
|
| 237271 |
+
"loss": 0.8521,
|
| 237272 |
+
"step": 108395
|
| 237273 |
+
},
|
| 237274 |
+
{
|
| 237275 |
+
"epoch": 867.0,
|
| 237276 |
+
"eval_loss": 0.38071098923683167,
|
| 237277 |
+
"eval_runtime": 42.7441,
|
| 237278 |
+
"eval_samples_per_second": 19.605,
|
| 237279 |
+
"eval_steps_per_second": 0.632,
|
| 237280 |
+
"eval_wer": 0.17914788425185402,
|
| 237281 |
+
"step": 108396
|
| 237282 |
+
},
|
| 237283 |
+
{
|
| 237284 |
+
"epoch": 867.03,
|
| 237285 |
+
"learning_rate": 8.265638126009695e-06,
|
| 237286 |
+
"loss": 0.3092,
|
| 237287 |
+
"step": 108400
|
| 237288 |
+
},
|
| 237289 |
+
{
|
| 237290 |
+
"epoch": 867.07,
|
| 237291 |
+
"learning_rate": 8.265557350565429e-06,
|
| 237292 |
+
"loss": 0.2916,
|
| 237293 |
+
"step": 108405
|
| 237294 |
+
},
|
| 237295 |
+
{
|
| 237296 |
+
"epoch": 867.11,
|
| 237297 |
+
"learning_rate": 8.265476575121165e-06,
|
| 237298 |
+
"loss": 0.2953,
|
| 237299 |
+
"step": 108410
|
| 237300 |
+
},
|
| 237301 |
+
{
|
| 237302 |
+
"epoch": 867.15,
|
| 237303 |
+
"learning_rate": 8.265395799676899e-06,
|
| 237304 |
+
"loss": 0.4351,
|
| 237305 |
+
"step": 108415
|
| 237306 |
+
},
|
| 237307 |
+
{
|
| 237308 |
+
"epoch": 867.19,
|
| 237309 |
+
"learning_rate": 8.265315024232635e-06,
|
| 237310 |
+
"loss": 1.1041,
|
| 237311 |
+
"step": 108420
|
| 237312 |
+
},
|
| 237313 |
+
{
|
| 237314 |
+
"epoch": 867.23,
|
| 237315 |
+
"learning_rate": 8.265234248788369e-06,
|
| 237316 |
+
"loss": 0.6721,
|
| 237317 |
+
"step": 108425
|
| 237318 |
+
},
|
| 237319 |
+
{
|
| 237320 |
+
"epoch": 867.27,
|
| 237321 |
+
"learning_rate": 8.265153473344105e-06,
|
| 237322 |
+
"loss": 0.3183,
|
| 237323 |
+
"step": 108430
|
| 237324 |
+
},
|
| 237325 |
+
{
|
| 237326 |
+
"epoch": 867.31,
|
| 237327 |
+
"learning_rate": 8.265072697899839e-06,
|
| 237328 |
+
"loss": 0.3486,
|
| 237329 |
+
"step": 108435
|
| 237330 |
+
},
|
| 237331 |
+
{
|
| 237332 |
+
"epoch": 867.35,
|
| 237333 |
+
"learning_rate": 8.264991922455575e-06,
|
| 237334 |
+
"loss": 0.3618,
|
| 237335 |
+
"step": 108440
|
| 237336 |
+
},
|
| 237337 |
+
{
|
| 237338 |
+
"epoch": 867.39,
|
| 237339 |
+
"learning_rate": 8.264911147011309e-06,
|
| 237340 |
+
"loss": 0.9431,
|
| 237341 |
+
"step": 108445
|
| 237342 |
+
},
|
| 237343 |
+
{
|
| 237344 |
+
"epoch": 867.43,
|
| 237345 |
+
"learning_rate": 8.264830371567045e-06,
|
| 237346 |
+
"loss": 0.7192,
|
| 237347 |
+
"step": 108450
|
| 237348 |
+
},
|
| 237349 |
+
{
|
| 237350 |
+
"epoch": 867.47,
|
| 237351 |
+
"learning_rate": 8.26474959612278e-06,
|
| 237352 |
+
"loss": 0.2688,
|
| 237353 |
+
"step": 108455
|
| 237354 |
+
},
|
| 237355 |
+
{
|
| 237356 |
+
"epoch": 867.51,
|
| 237357 |
+
"learning_rate": 8.264668820678515e-06,
|
| 237358 |
+
"loss": 0.2996,
|
| 237359 |
+
"step": 108460
|
| 237360 |
+
},
|
| 237361 |
+
{
|
| 237362 |
+
"epoch": 867.55,
|
| 237363 |
+
"learning_rate": 8.26458804523425e-06,
|
| 237364 |
+
"loss": 0.4245,
|
| 237365 |
+
"step": 108465
|
| 237366 |
+
},
|
| 237367 |
+
{
|
| 237368 |
+
"epoch": 867.59,
|
| 237369 |
+
"learning_rate": 8.264507269789985e-06,
|
| 237370 |
+
"loss": 0.9526,
|
| 237371 |
+
"step": 108470
|
| 237372 |
+
},
|
| 237373 |
+
{
|
| 237374 |
+
"epoch": 867.63,
|
| 237375 |
+
"learning_rate": 8.26442649434572e-06,
|
| 237376 |
+
"loss": 0.6272,
|
| 237377 |
+
"step": 108475
|
| 237378 |
+
},
|
| 237379 |
+
{
|
| 237380 |
+
"epoch": 867.67,
|
| 237381 |
+
"learning_rate": 8.264345718901455e-06,
|
| 237382 |
+
"loss": 0.2743,
|
| 237383 |
+
"step": 108480
|
| 237384 |
+
},
|
| 237385 |
+
{
|
| 237386 |
+
"epoch": 867.71,
|
| 237387 |
+
"learning_rate": 8.26426494345719e-06,
|
| 237388 |
+
"loss": 0.2969,
|
| 237389 |
+
"step": 108485
|
| 237390 |
+
},
|
| 237391 |
+
{
|
| 237392 |
+
"epoch": 867.75,
|
| 237393 |
+
"learning_rate": 8.264184168012925e-06,
|
| 237394 |
+
"loss": 0.3703,
|
| 237395 |
+
"step": 108490
|
| 237396 |
+
},
|
| 237397 |
+
{
|
| 237398 |
+
"epoch": 867.79,
|
| 237399 |
+
"learning_rate": 8.26410339256866e-06,
|
| 237400 |
+
"loss": 0.9945,
|
| 237401 |
+
"step": 108495
|
| 237402 |
+
},
|
| 237403 |
+
{
|
| 237404 |
+
"epoch": 867.83,
|
| 237405 |
+
"learning_rate": 8.264022617124395e-06,
|
| 237406 |
+
"loss": 0.6777,
|
| 237407 |
+
"step": 108500
|
| 237408 |
+
},
|
| 237409 |
+
{
|
| 237410 |
+
"epoch": 867.87,
|
| 237411 |
+
"learning_rate": 8.26394184168013e-06,
|
| 237412 |
+
"loss": 0.2981,
|
| 237413 |
+
"step": 108505
|
| 237414 |
+
},
|
| 237415 |
+
{
|
| 237416 |
+
"epoch": 867.91,
|
| 237417 |
+
"learning_rate": 8.263861066235865e-06,
|
| 237418 |
+
"loss": 0.3543,
|
| 237419 |
+
"step": 108510
|
| 237420 |
+
},
|
| 237421 |
+
{
|
| 237422 |
+
"epoch": 867.95,
|
| 237423 |
+
"learning_rate": 8.2637802907916e-06,
|
| 237424 |
+
"loss": 0.3779,
|
| 237425 |
+
"step": 108515
|
| 237426 |
+
},
|
| 237427 |
+
{
|
| 237428 |
+
"epoch": 867.99,
|
| 237429 |
+
"learning_rate": 8.263699515347335e-06,
|
| 237430 |
+
"loss": 0.8175,
|
| 237431 |
+
"step": 108520
|
| 237432 |
+
},
|
| 237433 |
+
{
|
| 237434 |
+
"epoch": 868.0,
|
| 237435 |
+
"eval_loss": 0.4252279996871948,
|
| 237436 |
+
"eval_runtime": 42.7199,
|
| 237437 |
+
"eval_samples_per_second": 19.616,
|
| 237438 |
+
"eval_steps_per_second": 0.632,
|
| 237439 |
+
"eval_wer": 0.1883531803659599,
|
| 237440 |
+
"step": 108521
|
| 237441 |
+
},
|
| 237442 |
+
{
|
| 237443 |
+
"epoch": 875.03,
|
| 237444 |
+
"learning_rate": 8.26361873990307e-06,
|
| 237445 |
+
"loss": 0.3295,
|
| 237446 |
+
"step": 108525
|
| 237447 |
+
},
|
| 237448 |
+
{
|
| 237449 |
+
"epoch": 875.07,
|
| 237450 |
+
"learning_rate": 8.263537964458806e-06,
|
| 237451 |
+
"loss": 0.3735,
|
| 237452 |
+
"step": 108530
|
| 237453 |
+
},
|
| 237454 |
+
{
|
| 237455 |
+
"epoch": 875.11,
|
| 237456 |
+
"learning_rate": 8.26345718901454e-06,
|
| 237457 |
+
"loss": 0.2969,
|
| 237458 |
+
"step": 108535
|
| 237459 |
+
},
|
| 237460 |
+
{
|
| 237461 |
+
"epoch": 875.15,
|
| 237462 |
+
"learning_rate": 8.263376413570276e-06,
|
| 237463 |
+
"loss": 0.3783,
|
| 237464 |
+
"step": 108540
|
| 237465 |
+
},
|
| 237466 |
+
{
|
| 237467 |
+
"epoch": 875.19,
|
| 237468 |
+
"learning_rate": 8.26329563812601e-06,
|
| 237469 |
+
"loss": 0.9893,
|
| 237470 |
+
"step": 108545
|
| 237471 |
+
},
|
| 237472 |
+
{
|
| 237473 |
+
"epoch": 875.23,
|
| 237474 |
+
"learning_rate": 8.263214862681746e-06,
|
| 237475 |
+
"loss": 0.6453,
|
| 237476 |
+
"step": 108550
|
| 237477 |
+
},
|
| 237478 |
+
{
|
| 237479 |
+
"epoch": 875.27,
|
| 237480 |
+
"learning_rate": 8.26313408723748e-06,
|
| 237481 |
+
"loss": 0.2648,
|
| 237482 |
+
"step": 108555
|
| 237483 |
+
},
|
| 237484 |
+
{
|
| 237485 |
+
"epoch": 875.31,
|
| 237486 |
+
"learning_rate": 8.263053311793216e-06,
|
| 237487 |
+
"loss": 0.332,
|
| 237488 |
+
"step": 108560
|
| 237489 |
+
},
|
| 237490 |
+
{
|
| 237491 |
+
"epoch": 875.35,
|
| 237492 |
+
"learning_rate": 8.26297253634895e-06,
|
| 237493 |
+
"loss": 0.4552,
|
| 237494 |
+
"step": 108565
|
| 237495 |
+
},
|
| 237496 |
+
{
|
| 237497 |
+
"epoch": 875.39,
|
| 237498 |
+
"learning_rate": 8.262891760904686e-06,
|
| 237499 |
+
"loss": 0.8424,
|
| 237500 |
+
"step": 108570
|
| 237501 |
+
},
|
| 237502 |
+
{
|
| 237503 |
+
"epoch": 875.43,
|
| 237504 |
+
"learning_rate": 8.26281098546042e-06,
|
| 237505 |
+
"loss": 0.7068,
|
| 237506 |
+
"step": 108575
|
| 237507 |
+
},
|
| 237508 |
+
{
|
| 237509 |
+
"epoch": 875.47,
|
| 237510 |
+
"learning_rate": 8.262730210016156e-06,
|
| 237511 |
+
"loss": 0.283,
|
| 237512 |
+
"step": 108580
|
| 237513 |
+
},
|
| 237514 |
+
{
|
| 237515 |
+
"epoch": 875.51,
|
| 237516 |
+
"learning_rate": 8.26264943457189e-06,
|
| 237517 |
+
"loss": 0.3432,
|
| 237518 |
+
"step": 108585
|
| 237519 |
+
},
|
| 237520 |
+
{
|
| 237521 |
+
"epoch": 875.55,
|
| 237522 |
+
"learning_rate": 8.262568659127626e-06,
|
| 237523 |
+
"loss": 0.4533,
|
| 237524 |
+
"step": 108590
|
| 237525 |
+
},
|
| 237526 |
+
{
|
| 237527 |
+
"epoch": 875.59,
|
| 237528 |
+
"learning_rate": 8.262487883683362e-06,
|
| 237529 |
+
"loss": 1.0342,
|
| 237530 |
+
"step": 108595
|
| 237531 |
+
},
|
| 237532 |
+
{
|
| 237533 |
+
"epoch": 875.63,
|
| 237534 |
+
"learning_rate": 8.262407108239096e-06,
|
| 237535 |
+
"loss": 0.5523,
|
| 237536 |
+
"step": 108600
|
| 237537 |
+
},
|
| 237538 |
+
{
|
| 237539 |
+
"epoch": 875.67,
|
| 237540 |
+
"learning_rate": 8.262326332794832e-06,
|
| 237541 |
+
"loss": 0.2626,
|
| 237542 |
+
"step": 108605
|
| 237543 |
+
},
|
| 237544 |
+
{
|
| 237545 |
+
"epoch": 875.71,
|
| 237546 |
+
"learning_rate": 8.262245557350566e-06,
|
| 237547 |
+
"loss": 0.4075,
|
| 237548 |
+
"step": 108610
|
| 237549 |
+
},
|
| 237550 |
+
{
|
| 237551 |
+
"epoch": 875.76,
|
| 237552 |
+
"learning_rate": 8.262164781906302e-06,
|
| 237553 |
+
"loss": 0.3639,
|
| 237554 |
+
"step": 108615
|
| 237555 |
+
},
|
| 237556 |
+
{
|
| 237557 |
+
"epoch": 875.8,
|
| 237558 |
+
"learning_rate": 8.262084006462036e-06,
|
| 237559 |
+
"loss": 0.869,
|
| 237560 |
+
"step": 108620
|
| 237561 |
+
},
|
| 237562 |
+
{
|
| 237563 |
+
"epoch": 875.84,
|
| 237564 |
+
"learning_rate": 8.262003231017772e-06,
|
| 237565 |
+
"loss": 0.7347,
|
| 237566 |
+
"step": 108625
|
| 237567 |
+
},
|
| 237568 |
+
{
|
| 237569 |
+
"epoch": 875.88,
|
| 237570 |
+
"learning_rate": 8.261922455573506e-06,
|
| 237571 |
+
"loss": 0.2967,
|
| 237572 |
+
"step": 108630
|
| 237573 |
+
},
|
| 237574 |
+
{
|
| 237575 |
+
"epoch": 875.92,
|
| 237576 |
+
"learning_rate": 8.261841680129242e-06,
|
| 237577 |
+
"loss": 0.3193,
|
| 237578 |
+
"step": 108635
|
| 237579 |
+
},
|
| 237580 |
+
{
|
| 237581 |
+
"epoch": 875.96,
|
| 237582 |
+
"learning_rate": 8.261760904684976e-06,
|
| 237583 |
+
"loss": 0.3686,
|
| 237584 |
+
"step": 108640
|
| 237585 |
+
},
|
| 237586 |
+
{
|
| 237587 |
+
"epoch": 876.0,
|
| 237588 |
+
"learning_rate": 8.261680129240712e-06,
|
| 237589 |
+
"loss": 1.0945,
|
| 237590 |
+
"step": 108645
|
| 237591 |
+
},
|
| 237592 |
+
{
|
| 237593 |
+
"epoch": 876.0,
|
| 237594 |
+
"eval_loss": 0.3863566219806671,
|
| 237595 |
+
"eval_runtime": 42.4014,
|
| 237596 |
+
"eval_samples_per_second": 19.764,
|
| 237597 |
+
"eval_steps_per_second": 0.637,
|
| 237598 |
+
"eval_wer": 0.18718179095537585,
|
| 237599 |
+
"step": 108645
|
| 237600 |
+
},
|
| 237601 |
+
{
|
| 237602 |
+
"epoch": 869.04,
|
| 237603 |
+
"learning_rate": 8.261599353796446e-06,
|
| 237604 |
+
"loss": 0.3436,
|
| 237605 |
+
"step": 108650
|
| 237606 |
+
},
|
| 237607 |
+
{
|
| 237608 |
+
"epoch": 869.08,
|
| 237609 |
+
"learning_rate": 8.261518578352182e-06,
|
| 237610 |
+
"loss": 0.3277,
|
| 237611 |
+
"step": 108655
|
| 237612 |
+
},
|
| 237613 |
+
{
|
| 237614 |
+
"epoch": 869.12,
|
| 237615 |
+
"learning_rate": 8.261437802907918e-06,
|
| 237616 |
+
"loss": 0.2893,
|
| 237617 |
+
"step": 108660
|
| 237618 |
+
},
|
| 237619 |
+
{
|
| 237620 |
+
"epoch": 869.16,
|
| 237621 |
+
"learning_rate": 8.261357027463652e-06,
|
| 237622 |
+
"loss": 0.4526,
|
| 237623 |
+
"step": 108665
|
| 237624 |
+
},
|
| 237625 |
+
{
|
| 237626 |
+
"epoch": 869.2,
|
| 237627 |
+
"learning_rate": 8.261276252019388e-06,
|
| 237628 |
+
"loss": 1.1031,
|
| 237629 |
+
"step": 108670
|
| 237630 |
+
},
|
| 237631 |
+
{
|
| 237632 |
+
"epoch": 869.24,
|
| 237633 |
+
"learning_rate": 8.261195476575122e-06,
|
| 237634 |
+
"loss": 0.3105,
|
| 237635 |
+
"step": 108675
|
| 237636 |
+
},
|
| 237637 |
+
{
|
| 237638 |
+
"epoch": 869.28,
|
| 237639 |
+
"learning_rate": 8.261114701130858e-06,
|
| 237640 |
+
"loss": 0.3029,
|
| 237641 |
+
"step": 108680
|
| 237642 |
+
},
|
| 237643 |
+
{
|
| 237644 |
+
"epoch": 869.32,
|
| 237645 |
+
"learning_rate": 8.261033925686592e-06,
|
| 237646 |
+
"loss": 0.2574,
|
| 237647 |
+
"step": 108685
|
| 237648 |
+
},
|
| 237649 |
+
{
|
| 237650 |
+
"epoch": 869.36,
|
| 237651 |
+
"learning_rate": 8.260953150242328e-06,
|
| 237652 |
+
"loss": 0.4242,
|
| 237653 |
+
"step": 108690
|
| 237654 |
+
},
|
| 237655 |
+
{
|
| 237656 |
+
"epoch": 869.4,
|
| 237657 |
+
"learning_rate": 8.260872374798062e-06,
|
| 237658 |
+
"loss": 1.089,
|
| 237659 |
+
"step": 108695
|
| 237660 |
+
},
|
| 237661 |
+
{
|
| 237662 |
+
"epoch": 869.44,
|
| 237663 |
+
"learning_rate": 8.260791599353797e-06,
|
| 237664 |
+
"loss": 0.2905,
|
| 237665 |
+
"step": 108700
|
| 237666 |
+
},
|
| 237667 |
+
{
|
| 237668 |
+
"epoch": 869.48,
|
| 237669 |
+
"learning_rate": 8.260710823909532e-06,
|
| 237670 |
+
"loss": 0.3501,
|
| 237671 |
+
"step": 108705
|
| 237672 |
+
},
|
| 237673 |
+
{
|
| 237674 |
+
"epoch": 869.52,
|
| 237675 |
+
"learning_rate": 8.260630048465267e-06,
|
| 237676 |
+
"loss": 0.3259,
|
| 237677 |
+
"step": 108710
|
| 237678 |
+
},
|
| 237679 |
+
{
|
| 237680 |
+
"epoch": 869.56,
|
| 237681 |
+
"learning_rate": 8.260549273021002e-06,
|
| 237682 |
+
"loss": 0.4261,
|
| 237683 |
+
"step": 108715
|
| 237684 |
+
},
|
| 237685 |
+
{
|
| 237686 |
+
"epoch": 869.6,
|
| 237687 |
+
"learning_rate": 8.260468497576737e-06,
|
| 237688 |
+
"loss": 1.0349,
|
| 237689 |
+
"step": 108720
|
| 237690 |
+
},
|
| 237691 |
+
{
|
| 237692 |
+
"epoch": 869.64,
|
| 237693 |
+
"learning_rate": 8.260387722132472e-06,
|
| 237694 |
+
"loss": 0.2813,
|
| 237695 |
+
"step": 108725
|
| 237696 |
+
},
|
| 237697 |
+
{
|
| 237698 |
+
"epoch": 869.68,
|
| 237699 |
+
"learning_rate": 8.260306946688207e-06,
|
| 237700 |
+
"loss": 0.3317,
|
| 237701 |
+
"step": 108730
|
| 237702 |
+
},
|
| 237703 |
+
{
|
| 237704 |
+
"epoch": 869.72,
|
| 237705 |
+
"learning_rate": 8.260226171243943e-06,
|
| 237706 |
+
"loss": 0.3327,
|
| 237707 |
+
"step": 108735
|
| 237708 |
+
},
|
| 237709 |
+
{
|
| 237710 |
+
"epoch": 869.76,
|
| 237711 |
+
"learning_rate": 8.260145395799677e-06,
|
| 237712 |
+
"loss": 0.4383,
|
| 237713 |
+
"step": 108740
|
| 237714 |
+
},
|
| 237715 |
+
{
|
| 237716 |
+
"epoch": 869.8,
|
| 237717 |
+
"learning_rate": 8.260064620355413e-06,
|
| 237718 |
+
"loss": 1.236,
|
| 237719 |
+
"step": 108745
|
| 237720 |
+
},
|
| 237721 |
+
{
|
| 237722 |
+
"epoch": 869.84,
|
| 237723 |
+
"learning_rate": 8.259983844911147e-06,
|
| 237724 |
+
"loss": 0.2982,
|
| 237725 |
+
"step": 108750
|
| 237726 |
+
},
|
| 237727 |
+
{
|
| 237728 |
+
"epoch": 869.88,
|
| 237729 |
+
"learning_rate": 8.259903069466883e-06,
|
| 237730 |
+
"loss": 0.3329,
|
| 237731 |
+
"step": 108755
|
| 237732 |
+
},
|
| 237733 |
+
{
|
| 237734 |
+
"epoch": 869.92,
|
| 237735 |
+
"learning_rate": 8.259822294022617e-06,
|
| 237736 |
+
"loss": 0.363,
|
| 237737 |
+
"step": 108760
|
| 237738 |
+
},
|
| 237739 |
+
{
|
| 237740 |
+
"epoch": 869.96,
|
| 237741 |
+
"learning_rate": 8.259741518578353e-06,
|
| 237742 |
+
"loss": 0.3857,
|
| 237743 |
+
"step": 108765
|
| 237744 |
+
},
|
| 237745 |
+
{
|
| 237746 |
+
"epoch": 870.0,
|
| 237747 |
+
"learning_rate": 8.259660743134087e-06,
|
| 237748 |
+
"loss": 1.1321,
|
| 237749 |
+
"step": 108770
|
| 237750 |
+
},
|
| 237751 |
+
{
|
| 237752 |
+
"epoch": 870.0,
|
| 237753 |
+
"eval_loss": 0.38697919249534607,
|
| 237754 |
+
"eval_runtime": 40.4796,
|
| 237755 |
+
"eval_samples_per_second": 20.677,
|
| 237756 |
+
"eval_steps_per_second": 0.667,
|
| 237757 |
+
"eval_wer": 0.19202253855278767,
|
| 237758 |
+
"step": 108770
|
| 237759 |
+
},
|
| 237760 |
+
{
|
| 237761 |
+
"epoch": 877.04,
|
| 237762 |
+
"learning_rate": 8.259579967689823e-06,
|
| 237763 |
+
"loss": 0.3274,
|
| 237764 |
+
"step": 108775
|
| 237765 |
+
},
|
| 237766 |
+
{
|
| 237767 |
+
"epoch": 877.08,
|
| 237768 |
+
"learning_rate": 8.259499192245557e-06,
|
| 237769 |
+
"loss": 0.3668,
|
| 237770 |
+
"step": 108780
|
| 237771 |
+
},
|
| 237772 |
+
{
|
| 237773 |
+
"epoch": 877.12,
|
| 237774 |
+
"learning_rate": 8.259418416801293e-06,
|
| 237775 |
+
"loss": 0.3338,
|
| 237776 |
+
"step": 108785
|
| 237777 |
+
},
|
| 237778 |
+
{
|
| 237779 |
+
"epoch": 877.16,
|
| 237780 |
+
"learning_rate": 8.259337641357027e-06,
|
| 237781 |
+
"loss": 0.4149,
|
| 237782 |
+
"step": 108790
|
| 237783 |
+
},
|
| 237784 |
+
{
|
| 237785 |
+
"epoch": 877.2,
|
| 237786 |
+
"learning_rate": 8.259256865912763e-06,
|
| 237787 |
+
"loss": 1.1563,
|
| 237788 |
+
"step": 108795
|
| 237789 |
+
},
|
| 237790 |
+
{
|
| 237791 |
+
"epoch": 877.24,
|
| 237792 |
+
"learning_rate": 8.259176090468499e-06,
|
| 237793 |
+
"loss": 0.3301,
|
| 237794 |
+
"step": 108800
|
| 237795 |
+
},
|
| 237796 |
+
{
|
| 237797 |
+
"epoch": 877.28,
|
| 237798 |
+
"learning_rate": 8.259095315024233e-06,
|
| 237799 |
+
"loss": 0.2894,
|
| 237800 |
+
"step": 108805
|
| 237801 |
+
},
|
| 237802 |
+
{
|
| 237803 |
+
"epoch": 877.32,
|
| 237804 |
+
"learning_rate": 8.259014539579969e-06,
|
| 237805 |
+
"loss": 0.2696,
|
| 237806 |
+
"step": 108810
|
| 237807 |
+
},
|
| 237808 |
+
{
|
| 237809 |
+
"epoch": 877.36,
|
| 237810 |
+
"learning_rate": 8.258933764135703e-06,
|
| 237811 |
+
"loss": 0.4193,
|
| 237812 |
+
"step": 108815
|
| 237813 |
+
},
|
| 237814 |
+
{
|
| 237815 |
+
"epoch": 877.4,
|
| 237816 |
+
"learning_rate": 8.258852988691439e-06,
|
| 237817 |
+
"loss": 1.1781,
|
| 237818 |
+
"step": 108820
|
| 237819 |
+
},
|
| 237820 |
+
{
|
| 237821 |
+
"epoch": 877.44,
|
| 237822 |
+
"learning_rate": 8.258772213247173e-06,
|
| 237823 |
+
"loss": 0.3014,
|
| 237824 |
+
"step": 108825
|
| 237825 |
+
},
|
| 237826 |
+
{
|
| 237827 |
+
"epoch": 877.48,
|
| 237828 |
+
"learning_rate": 8.258691437802909e-06,
|
| 237829 |
+
"loss": 0.3459,
|
| 237830 |
+
"step": 108830
|
| 237831 |
+
},
|
| 237832 |
+
{
|
| 237833 |
+
"epoch": 877.52,
|
| 237834 |
+
"learning_rate": 8.258610662358643e-06,
|
| 237835 |
+
"loss": 0.2947,
|
| 237836 |
+
"step": 108835
|
| 237837 |
+
},
|
| 237838 |
+
{
|
| 237839 |
+
"epoch": 877.56,
|
| 237840 |
+
"learning_rate": 8.258529886914379e-06,
|
| 237841 |
+
"loss": 0.6715,
|
| 237842 |
+
"step": 108840
|
| 237843 |
+
},
|
| 237844 |
+
{
|
| 237845 |
+
"epoch": 877.6,
|
| 237846 |
+
"learning_rate": 8.258449111470113e-06,
|
| 237847 |
+
"loss": 1.1457,
|
| 237848 |
+
"step": 108845
|
| 237849 |
+
},
|
| 237850 |
+
{
|
| 237851 |
+
"epoch": 877.65,
|
| 237852 |
+
"learning_rate": 8.258368336025849e-06,
|
| 237853 |
+
"loss": 0.2831,
|
| 237854 |
+
"step": 108850
|
| 237855 |
+
},
|
| 237856 |
+
{
|
| 237857 |
+
"epoch": 877.69,
|
| 237858 |
+
"learning_rate": 8.258287560581583e-06,
|
| 237859 |
+
"loss": 0.2946,
|
| 237860 |
+
"step": 108855
|
| 237861 |
+
},
|
| 237862 |
+
{
|
| 237863 |
+
"epoch": 877.73,
|
| 237864 |
+
"learning_rate": 8.258206785137319e-06,
|
| 237865 |
+
"loss": 0.3625,
|
| 237866 |
+
"step": 108860
|
| 237867 |
+
},
|
| 237868 |
+
{
|
| 237869 |
+
"epoch": 877.77,
|
| 237870 |
+
"learning_rate": 8.258126009693055e-06,
|
| 237871 |
+
"loss": 0.4454,
|
| 237872 |
+
"step": 108865
|
| 237873 |
+
},
|
| 237874 |
+
{
|
| 237875 |
+
"epoch": 877.81,
|
| 237876 |
+
"learning_rate": 8.258045234248789e-06,
|
| 237877 |
+
"loss": 1.1977,
|
| 237878 |
+
"step": 108870
|
| 237879 |
+
},
|
| 237880 |
+
{
|
| 237881 |
+
"epoch": 877.85,
|
| 237882 |
+
"learning_rate": 8.257964458804525e-06,
|
| 237883 |
+
"loss": 0.3407,
|
| 237884 |
+
"step": 108875
|
| 237885 |
+
},
|
| 237886 |
+
{
|
| 237887 |
+
"epoch": 877.89,
|
| 237888 |
+
"learning_rate": 8.257883683360259e-06,
|
| 237889 |
+
"loss": 0.2812,
|
| 237890 |
+
"step": 108880
|
| 237891 |
+
},
|
| 237892 |
+
{
|
| 237893 |
+
"epoch": 877.93,
|
| 237894 |
+
"learning_rate": 8.257802907915995e-06,
|
| 237895 |
+
"loss": 0.3153,
|
| 237896 |
+
"step": 108885
|
| 237897 |
+
},
|
| 237898 |
+
{
|
| 237899 |
+
"epoch": 877.97,
|
| 237900 |
+
"learning_rate": 8.257722132471729e-06,
|
| 237901 |
+
"loss": 0.4498,
|
| 237902 |
+
"step": 108890
|
| 237903 |
+
},
|
| 237904 |
+
{
|
| 237905 |
+
"epoch": 878.0,
|
| 237906 |
+
"eval_loss": 0.38034501671791077,
|
| 237907 |
+
"eval_runtime": 41.1105,
|
| 237908 |
+
"eval_samples_per_second": 20.36,
|
| 237909 |
+
"eval_steps_per_second": 0.657,
|
| 237910 |
+
"eval_wer": 0.17886361930496134,
|
| 237911 |
+
"step": 108894
|
| 237912 |
}
|
| 237913 |
],
|
| 237914 |
"max_steps": 620000,
|
| 237915 |
"num_train_epochs": 5000,
|
| 237916 |
+
"total_flos": 3.064483777918432e+20,
|
| 237917 |
"trial_name": null,
|
| 237918 |
"trial_params": null
|
| 237919 |
}
|
model-bin/finetune/base/{checkpoint-108271 → checkpoint-108894}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1630036151.3539824/events.out.tfevents.1630036151.52f5c7e305a3.886.31
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d8979ad7190107b053f5dc3753d1f3e3cd99a18dd874e753edf6cf6163c56a0a
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630036627.3019376/events.out.tfevents.1630036627.52f5c7e305a3.886.33
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:099b34f0762b8a8fbce9ec702092040f086c3dd03937a2b15801a8b8780bc3fa
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630037085.2077386/events.out.tfevents.1630037085.52f5c7e305a3.886.35
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2a61b70b0cc1eb0d73f1c6c306b4e7f59db13fda5ab379ad35c4062860ef42d9
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630037540.4465652/events.out.tfevents.1630037540.52f5c7e305a3.886.37
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2d47fa4a624ff2a49371cf4c70fff9dc5303467a8cd9f50cb0f31a7a7bc4660e
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630038001.097344/events.out.tfevents.1630038001.52f5c7e305a3.886.39
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:484559fcf494a8b6da0187528b4b65a964988ac6aace915ba4151f76b263c5f6
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1630036151.52f5c7e305a3.886.30
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:adc661535a47900cb7aba9330dcc9ce4d63b486c30e296f3691ef0d17428f11f
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630036627.52f5c7e305a3.886.32
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:90a06474bcf4675383f2112d6b78846b6e1177aeec0b1cddac6d88db28655438
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630037085.52f5c7e305a3.886.34
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b0bd75cfbde5a1a29ceeb94393c79178ffddc24843b8b2567a5ee06f6ef86672
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630037540.52f5c7e305a3.886.36
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:504982e47e775fa6f7dcd298faf3ebf5912a7c4ca45a43996e923afbcc019760
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630038001.52f5c7e305a3.886.38
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b32e93b5a44af40b43a90d9d69bbb7cf412f13384ef68a8d152be85397ea722b
|
| 3 |
+
size 8462
|