"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-127934 → checkpoint-128555}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-127934 → checkpoint-128555}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-127934 → checkpoint-128555}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-127934 → checkpoint-128555}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-127934 → checkpoint-128555}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-127934 → checkpoint-128555}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-127934 → checkpoint-128555}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-127934 → checkpoint-128555}/trainer_state.json +798 -3
- model-bin/finetune/base/{checkpoint-127934 → checkpoint-128555}/training_args.bin +0 -0
- model-bin/finetune/base/log/1630154741.1674275/events.out.tfevents.1630154741.86bb0ddabf9b.4092.161 +3 -0
- model-bin/finetune/base/log/1630155136.7040803/events.out.tfevents.1630155136.86bb0ddabf9b.4092.163 +3 -0
- model-bin/finetune/base/log/1630155523.684827/events.out.tfevents.1630155523.86bb0ddabf9b.4092.165 +3 -0
- model-bin/finetune/base/log/1630155913.5081563/events.out.tfevents.1630155913.86bb0ddabf9b.4092.167 +3 -0
- model-bin/finetune/base/log/1630156305.7181728/events.out.tfevents.1630156305.86bb0ddabf9b.4092.169 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630154741.86bb0ddabf9b.4092.160 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630155136.86bb0ddabf9b.4092.162 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630155523.86bb0ddabf9b.4092.164 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630155913.86bb0ddabf9b.4092.166 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630156305.86bb0ddabf9b.4092.168 +3 -0
model-bin/finetune/base/{checkpoint-127934 → checkpoint-128555}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-127934 → checkpoint-128555}/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:c4872d2d605f0df9699326b1aeda83c25f476149c23a890b18875fa234bc0bd0
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-127934 → checkpoint-128555}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-127934 → checkpoint-128555}/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 sha256:3e5842c2e57c4a8f6df64403064767e0e93aa1658df0bcf1ad882e217674d101
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-127934 → checkpoint-128555}/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 sha256:91d532c78c8ed7d8fd84b67ddb497f628adc2269f28d9602d1e3e333c2f626c8
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-127934 → checkpoint-128555}/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 sha256:820eb3431598c473dfb26db745012fa47099e34b99ef93977d6575999336dec2
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-127934 → checkpoint-128555}/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 sha256:378ca1171fb69646f63afd641ae75e12c76d1cbcbef7c20565766d3e98564e6e
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-127934 → checkpoint-128555}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.17162025681719809,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-124947",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -262134,11 +262134,806 @@
|
|
| 262134 |
"eval_steps_per_second": 0.75,
|
| 262135 |
"eval_wer": 0.18386253268584235,
|
| 262136 |
"step": 127934
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 262137 |
}
|
| 262138 |
],
|
| 262139 |
"max_steps": 620000,
|
| 262140 |
"num_train_epochs": 5000,
|
| 262141 |
-
"total_flos": 3.
|
| 262142 |
"trial_name": null,
|
| 262143 |
"trial_params": null
|
| 262144 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.17162025681719809,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-124947",
|
| 4 |
+
"epoch": 1035.995983935743,
|
| 5 |
+
"global_step": 128555,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 262134 |
"eval_steps_per_second": 0.75,
|
| 262135 |
"eval_wer": 0.18386253268584235,
|
| 262136 |
"step": 127934
|
| 262137 |
+
},
|
| 262138 |
+
{
|
| 262139 |
+
"epoch": 1031.01,
|
| 262140 |
+
"learning_rate": 7.966586538461539e-06,
|
| 262141 |
+
"loss": 0.4199,
|
| 262142 |
+
"step": 127935
|
| 262143 |
+
},
|
| 262144 |
+
{
|
| 262145 |
+
"epoch": 1031.05,
|
| 262146 |
+
"learning_rate": 7.96650641025641e-06,
|
| 262147 |
+
"loss": 0.4543,
|
| 262148 |
+
"step": 127940
|
| 262149 |
+
},
|
| 262150 |
+
{
|
| 262151 |
+
"epoch": 1031.09,
|
| 262152 |
+
"learning_rate": 7.966426282051284e-06,
|
| 262153 |
+
"loss": 0.3229,
|
| 262154 |
+
"step": 127945
|
| 262155 |
+
},
|
| 262156 |
+
{
|
| 262157 |
+
"epoch": 1031.13,
|
| 262158 |
+
"learning_rate": 7.966346153846153e-06,
|
| 262159 |
+
"loss": 0.3324,
|
| 262160 |
+
"step": 127950
|
| 262161 |
+
},
|
| 262162 |
+
{
|
| 262163 |
+
"epoch": 1031.17,
|
| 262164 |
+
"learning_rate": 7.966266025641026e-06,
|
| 262165 |
+
"loss": 0.5669,
|
| 262166 |
+
"step": 127955
|
| 262167 |
+
},
|
| 262168 |
+
{
|
| 262169 |
+
"epoch": 1031.21,
|
| 262170 |
+
"learning_rate": 7.9661858974359e-06,
|
| 262171 |
+
"loss": 1.0836,
|
| 262172 |
+
"step": 127960
|
| 262173 |
+
},
|
| 262174 |
+
{
|
| 262175 |
+
"epoch": 1031.25,
|
| 262176 |
+
"learning_rate": 7.966105769230769e-06,
|
| 262177 |
+
"loss": 0.3034,
|
| 262178 |
+
"step": 127965
|
| 262179 |
+
},
|
| 262180 |
+
{
|
| 262181 |
+
"epoch": 1031.29,
|
| 262182 |
+
"learning_rate": 7.966025641025642e-06,
|
| 262183 |
+
"loss": 0.2907,
|
| 262184 |
+
"step": 127970
|
| 262185 |
+
},
|
| 262186 |
+
{
|
| 262187 |
+
"epoch": 1031.33,
|
| 262188 |
+
"learning_rate": 7.965945512820513e-06,
|
| 262189 |
+
"loss": 0.2937,
|
| 262190 |
+
"step": 127975
|
| 262191 |
+
},
|
| 262192 |
+
{
|
| 262193 |
+
"epoch": 1031.37,
|
| 262194 |
+
"learning_rate": 7.965865384615385e-06,
|
| 262195 |
+
"loss": 0.6083,
|
| 262196 |
+
"step": 127980
|
| 262197 |
+
},
|
| 262198 |
+
{
|
| 262199 |
+
"epoch": 1031.41,
|
| 262200 |
+
"learning_rate": 7.965785256410256e-06,
|
| 262201 |
+
"loss": 1.0449,
|
| 262202 |
+
"step": 127985
|
| 262203 |
+
},
|
| 262204 |
+
{
|
| 262205 |
+
"epoch": 1031.45,
|
| 262206 |
+
"learning_rate": 7.965705128205129e-06,
|
| 262207 |
+
"loss": 0.3381,
|
| 262208 |
+
"step": 127990
|
| 262209 |
+
},
|
| 262210 |
+
{
|
| 262211 |
+
"epoch": 1031.49,
|
| 262212 |
+
"learning_rate": 7.965625e-06,
|
| 262213 |
+
"loss": 0.3963,
|
| 262214 |
+
"step": 127995
|
| 262215 |
+
},
|
| 262216 |
+
{
|
| 262217 |
+
"epoch": 1031.53,
|
| 262218 |
+
"learning_rate": 7.965544871794872e-06,
|
| 262219 |
+
"loss": 0.4325,
|
| 262220 |
+
"step": 128000
|
| 262221 |
+
},
|
| 262222 |
+
{
|
| 262223 |
+
"epoch": 1031.57,
|
| 262224 |
+
"learning_rate": 7.965464743589745e-06,
|
| 262225 |
+
"loss": 0.5177,
|
| 262226 |
+
"step": 128005
|
| 262227 |
+
},
|
| 262228 |
+
{
|
| 262229 |
+
"epoch": 1031.61,
|
| 262230 |
+
"learning_rate": 7.965384615384616e-06,
|
| 262231 |
+
"loss": 1.0997,
|
| 262232 |
+
"step": 128010
|
| 262233 |
+
},
|
| 262234 |
+
{
|
| 262235 |
+
"epoch": 1031.65,
|
| 262236 |
+
"learning_rate": 7.965304487179488e-06,
|
| 262237 |
+
"loss": 0.336,
|
| 262238 |
+
"step": 128015
|
| 262239 |
+
},
|
| 262240 |
+
{
|
| 262241 |
+
"epoch": 1031.69,
|
| 262242 |
+
"learning_rate": 7.965224358974359e-06,
|
| 262243 |
+
"loss": 0.3077,
|
| 262244 |
+
"step": 128020
|
| 262245 |
+
},
|
| 262246 |
+
{
|
| 262247 |
+
"epoch": 1031.73,
|
| 262248 |
+
"learning_rate": 7.965144230769232e-06,
|
| 262249 |
+
"loss": 0.3324,
|
| 262250 |
+
"step": 128025
|
| 262251 |
+
},
|
| 262252 |
+
{
|
| 262253 |
+
"epoch": 1031.77,
|
| 262254 |
+
"learning_rate": 7.965064102564103e-06,
|
| 262255 |
+
"loss": 0.5266,
|
| 262256 |
+
"step": 128030
|
| 262257 |
+
},
|
| 262258 |
+
{
|
| 262259 |
+
"epoch": 1031.81,
|
| 262260 |
+
"learning_rate": 7.964983974358975e-06,
|
| 262261 |
+
"loss": 1.0535,
|
| 262262 |
+
"step": 128035
|
| 262263 |
+
},
|
| 262264 |
+
{
|
| 262265 |
+
"epoch": 1031.85,
|
| 262266 |
+
"learning_rate": 7.964903846153846e-06,
|
| 262267 |
+
"loss": 0.3257,
|
| 262268 |
+
"step": 128040
|
| 262269 |
+
},
|
| 262270 |
+
{
|
| 262271 |
+
"epoch": 1031.89,
|
| 262272 |
+
"learning_rate": 7.96482371794872e-06,
|
| 262273 |
+
"loss": 0.2812,
|
| 262274 |
+
"step": 128045
|
| 262275 |
+
},
|
| 262276 |
+
{
|
| 262277 |
+
"epoch": 1031.93,
|
| 262278 |
+
"learning_rate": 7.96474358974359e-06,
|
| 262279 |
+
"loss": 0.3752,
|
| 262280 |
+
"step": 128050
|
| 262281 |
+
},
|
| 262282 |
+
{
|
| 262283 |
+
"epoch": 1031.97,
|
| 262284 |
+
"learning_rate": 7.964663461538462e-06,
|
| 262285 |
+
"loss": 0.5509,
|
| 262286 |
+
"step": 128055
|
| 262287 |
+
},
|
| 262288 |
+
{
|
| 262289 |
+
"epoch": 1032.0,
|
| 262290 |
+
"eval_loss": 0.3837217092514038,
|
| 262291 |
+
"eval_runtime": 36.4857,
|
| 262292 |
+
"eval_samples_per_second": 23.05,
|
| 262293 |
+
"eval_steps_per_second": 0.74,
|
| 262294 |
+
"eval_wer": 0.18022603098570467,
|
| 262295 |
+
"step": 128058
|
| 262296 |
+
},
|
| 262297 |
+
{
|
| 262298 |
+
"epoch": 1032.02,
|
| 262299 |
+
"learning_rate": 7.964583333333335e-06,
|
| 262300 |
+
"loss": 0.35,
|
| 262301 |
+
"step": 128060
|
| 262302 |
+
},
|
| 262303 |
+
{
|
| 262304 |
+
"epoch": 1032.06,
|
| 262305 |
+
"learning_rate": 7.964503205128206e-06,
|
| 262306 |
+
"loss": 0.3237,
|
| 262307 |
+
"step": 128065
|
| 262308 |
+
},
|
| 262309 |
+
{
|
| 262310 |
+
"epoch": 1032.1,
|
| 262311 |
+
"learning_rate": 7.964423076923078e-06,
|
| 262312 |
+
"loss": 0.3172,
|
| 262313 |
+
"step": 128070
|
| 262314 |
+
},
|
| 262315 |
+
{
|
| 262316 |
+
"epoch": 1032.14,
|
| 262317 |
+
"learning_rate": 7.964342948717949e-06,
|
| 262318 |
+
"loss": 0.3093,
|
| 262319 |
+
"step": 128075
|
| 262320 |
+
},
|
| 262321 |
+
{
|
| 262322 |
+
"epoch": 1032.18,
|
| 262323 |
+
"learning_rate": 7.964262820512822e-06,
|
| 262324 |
+
"loss": 0.5876,
|
| 262325 |
+
"step": 128080
|
| 262326 |
+
},
|
| 262327 |
+
{
|
| 262328 |
+
"epoch": 1032.22,
|
| 262329 |
+
"learning_rate": 7.964182692307693e-06,
|
| 262330 |
+
"loss": 1.0142,
|
| 262331 |
+
"step": 128085
|
| 262332 |
+
},
|
| 262333 |
+
{
|
| 262334 |
+
"epoch": 1032.26,
|
| 262335 |
+
"learning_rate": 7.964102564102565e-06,
|
| 262336 |
+
"loss": 0.9576,
|
| 262337 |
+
"step": 128090
|
| 262338 |
+
},
|
| 262339 |
+
{
|
| 262340 |
+
"epoch": 1032.3,
|
| 262341 |
+
"learning_rate": 7.964022435897436e-06,
|
| 262342 |
+
"loss": 0.2773,
|
| 262343 |
+
"step": 128095
|
| 262344 |
+
},
|
| 262345 |
+
{
|
| 262346 |
+
"epoch": 1032.34,
|
| 262347 |
+
"learning_rate": 7.96394230769231e-06,
|
| 262348 |
+
"loss": 0.3659,
|
| 262349 |
+
"step": 128100
|
| 262350 |
+
},
|
| 262351 |
+
{
|
| 262352 |
+
"epoch": 1032.38,
|
| 262353 |
+
"learning_rate": 7.96386217948718e-06,
|
| 262354 |
+
"loss": 0.5929,
|
| 262355 |
+
"step": 128105
|
| 262356 |
+
},
|
| 262357 |
+
{
|
| 262358 |
+
"epoch": 1032.42,
|
| 262359 |
+
"learning_rate": 7.963782051282052e-06,
|
| 262360 |
+
"loss": 0.8956,
|
| 262361 |
+
"step": 128110
|
| 262362 |
+
},
|
| 262363 |
+
{
|
| 262364 |
+
"epoch": 1032.46,
|
| 262365 |
+
"learning_rate": 7.963701923076925e-06,
|
| 262366 |
+
"loss": 0.2758,
|
| 262367 |
+
"step": 128115
|
| 262368 |
+
},
|
| 262369 |
+
{
|
| 262370 |
+
"epoch": 1032.5,
|
| 262371 |
+
"learning_rate": 7.963621794871795e-06,
|
| 262372 |
+
"loss": 0.2807,
|
| 262373 |
+
"step": 128120
|
| 262374 |
+
},
|
| 262375 |
+
{
|
| 262376 |
+
"epoch": 1032.54,
|
| 262377 |
+
"learning_rate": 7.963541666666668e-06,
|
| 262378 |
+
"loss": 0.3192,
|
| 262379 |
+
"step": 128125
|
| 262380 |
+
},
|
| 262381 |
+
{
|
| 262382 |
+
"epoch": 1032.58,
|
| 262383 |
+
"learning_rate": 7.963461538461539e-06,
|
| 262384 |
+
"loss": 0.5896,
|
| 262385 |
+
"step": 128130
|
| 262386 |
+
},
|
| 262387 |
+
{
|
| 262388 |
+
"epoch": 1032.62,
|
| 262389 |
+
"learning_rate": 7.96338141025641e-06,
|
| 262390 |
+
"loss": 0.9102,
|
| 262391 |
+
"step": 128135
|
| 262392 |
+
},
|
| 262393 |
+
{
|
| 262394 |
+
"epoch": 1032.66,
|
| 262395 |
+
"learning_rate": 7.963301282051282e-06,
|
| 262396 |
+
"loss": 0.2534,
|
| 262397 |
+
"step": 128140
|
| 262398 |
+
},
|
| 262399 |
+
{
|
| 262400 |
+
"epoch": 1032.7,
|
| 262401 |
+
"learning_rate": 7.963221153846155e-06,
|
| 262402 |
+
"loss": 0.2776,
|
| 262403 |
+
"step": 128145
|
| 262404 |
+
},
|
| 262405 |
+
{
|
| 262406 |
+
"epoch": 1032.74,
|
| 262407 |
+
"learning_rate": 7.963141025641026e-06,
|
| 262408 |
+
"loss": 0.3249,
|
| 262409 |
+
"step": 128150
|
| 262410 |
+
},
|
| 262411 |
+
{
|
| 262412 |
+
"epoch": 1032.78,
|
| 262413 |
+
"learning_rate": 7.963060897435898e-06,
|
| 262414 |
+
"loss": 0.5992,
|
| 262415 |
+
"step": 128155
|
| 262416 |
+
},
|
| 262417 |
+
{
|
| 262418 |
+
"epoch": 1032.82,
|
| 262419 |
+
"learning_rate": 7.96298076923077e-06,
|
| 262420 |
+
"loss": 0.9844,
|
| 262421 |
+
"step": 128160
|
| 262422 |
+
},
|
| 262423 |
+
{
|
| 262424 |
+
"epoch": 1032.86,
|
| 262425 |
+
"learning_rate": 7.962900641025642e-06,
|
| 262426 |
+
"loss": 0.2506,
|
| 262427 |
+
"step": 128165
|
| 262428 |
+
},
|
| 262429 |
+
{
|
| 262430 |
+
"epoch": 1032.9,
|
| 262431 |
+
"learning_rate": 7.962820512820513e-06,
|
| 262432 |
+
"loss": 0.338,
|
| 262433 |
+
"step": 128170
|
| 262434 |
+
},
|
| 262435 |
+
{
|
| 262436 |
+
"epoch": 1032.94,
|
| 262437 |
+
"learning_rate": 7.962740384615385e-06,
|
| 262438 |
+
"loss": 0.3421,
|
| 262439 |
+
"step": 128175
|
| 262440 |
+
},
|
| 262441 |
+
{
|
| 262442 |
+
"epoch": 1032.98,
|
| 262443 |
+
"learning_rate": 7.962660256410258e-06,
|
| 262444 |
+
"loss": 0.7898,
|
| 262445 |
+
"step": 128180
|
| 262446 |
+
},
|
| 262447 |
+
{
|
| 262448 |
+
"epoch": 1033.0,
|
| 262449 |
+
"eval_loss": 0.30437222123146057,
|
| 262450 |
+
"eval_runtime": 35.9032,
|
| 262451 |
+
"eval_samples_per_second": 23.424,
|
| 262452 |
+
"eval_steps_per_second": 0.752,
|
| 262453 |
+
"eval_wer": 0.17522637238256933,
|
| 262454 |
+
"step": 128182
|
| 262455 |
+
},
|
| 262456 |
+
{
|
| 262457 |
+
"epoch": 1033.02,
|
| 262458 |
+
"learning_rate": 7.962580128205129e-06,
|
| 262459 |
+
"loss": 0.3256,
|
| 262460 |
+
"step": 128185
|
| 262461 |
+
},
|
| 262462 |
+
{
|
| 262463 |
+
"epoch": 1033.06,
|
| 262464 |
+
"learning_rate": 7.9625e-06,
|
| 262465 |
+
"loss": 0.2859,
|
| 262466 |
+
"step": 128190
|
| 262467 |
+
},
|
| 262468 |
+
{
|
| 262469 |
+
"epoch": 1033.1,
|
| 262470 |
+
"learning_rate": 7.962419871794872e-06,
|
| 262471 |
+
"loss": 0.2552,
|
| 262472 |
+
"step": 128195
|
| 262473 |
+
},
|
| 262474 |
+
{
|
| 262475 |
+
"epoch": 1033.14,
|
| 262476 |
+
"learning_rate": 7.962339743589745e-06,
|
| 262477 |
+
"loss": 0.3579,
|
| 262478 |
+
"step": 128200
|
| 262479 |
+
},
|
| 262480 |
+
{
|
| 262481 |
+
"epoch": 1033.18,
|
| 262482 |
+
"learning_rate": 7.962259615384616e-06,
|
| 262483 |
+
"loss": 0.788,
|
| 262484 |
+
"step": 128205
|
| 262485 |
+
},
|
| 262486 |
+
{
|
| 262487 |
+
"epoch": 1033.22,
|
| 262488 |
+
"learning_rate": 7.962179487179488e-06,
|
| 262489 |
+
"loss": 0.761,
|
| 262490 |
+
"step": 128210
|
| 262491 |
+
},
|
| 262492 |
+
{
|
| 262493 |
+
"epoch": 1033.27,
|
| 262494 |
+
"learning_rate": 7.96209935897436e-06,
|
| 262495 |
+
"loss": 0.2531,
|
| 262496 |
+
"step": 128215
|
| 262497 |
+
},
|
| 262498 |
+
{
|
| 262499 |
+
"epoch": 1033.31,
|
| 262500 |
+
"learning_rate": 7.962019230769232e-06,
|
| 262501 |
+
"loss": 0.2436,
|
| 262502 |
+
"step": 128220
|
| 262503 |
+
},
|
| 262504 |
+
{
|
| 262505 |
+
"epoch": 1033.35,
|
| 262506 |
+
"learning_rate": 7.961939102564103e-06,
|
| 262507 |
+
"loss": 0.3021,
|
| 262508 |
+
"step": 128225
|
| 262509 |
+
},
|
| 262510 |
+
{
|
| 262511 |
+
"epoch": 1033.39,
|
| 262512 |
+
"learning_rate": 7.961858974358975e-06,
|
| 262513 |
+
"loss": 0.6576,
|
| 262514 |
+
"step": 128230
|
| 262515 |
+
},
|
| 262516 |
+
{
|
| 262517 |
+
"epoch": 1033.43,
|
| 262518 |
+
"learning_rate": 7.961778846153848e-06,
|
| 262519 |
+
"loss": 0.8004,
|
| 262520 |
+
"step": 128235
|
| 262521 |
+
},
|
| 262522 |
+
{
|
| 262523 |
+
"epoch": 1033.47,
|
| 262524 |
+
"learning_rate": 7.961698717948717e-06,
|
| 262525 |
+
"loss": 0.2714,
|
| 262526 |
+
"step": 128240
|
| 262527 |
+
},
|
| 262528 |
+
{
|
| 262529 |
+
"epoch": 1033.51,
|
| 262530 |
+
"learning_rate": 7.96161858974359e-06,
|
| 262531 |
+
"loss": 0.3474,
|
| 262532 |
+
"step": 128245
|
| 262533 |
+
},
|
| 262534 |
+
{
|
| 262535 |
+
"epoch": 1033.55,
|
| 262536 |
+
"learning_rate": 7.961538461538462e-06,
|
| 262537 |
+
"loss": 0.3564,
|
| 262538 |
+
"step": 128250
|
| 262539 |
+
},
|
| 262540 |
+
{
|
| 262541 |
+
"epoch": 1033.59,
|
| 262542 |
+
"learning_rate": 7.961458333333333e-06,
|
| 262543 |
+
"loss": 0.6588,
|
| 262544 |
+
"step": 128255
|
| 262545 |
+
},
|
| 262546 |
+
{
|
| 262547 |
+
"epoch": 1033.63,
|
| 262548 |
+
"learning_rate": 7.961378205128206e-06,
|
| 262549 |
+
"loss": 0.7798,
|
| 262550 |
+
"step": 128260
|
| 262551 |
+
},
|
| 262552 |
+
{
|
| 262553 |
+
"epoch": 1033.67,
|
| 262554 |
+
"learning_rate": 7.961298076923078e-06,
|
| 262555 |
+
"loss": 0.2631,
|
| 262556 |
+
"step": 128265
|
| 262557 |
+
},
|
| 262558 |
+
{
|
| 262559 |
+
"epoch": 1033.71,
|
| 262560 |
+
"learning_rate": 7.961217948717949e-06,
|
| 262561 |
+
"loss": 0.2682,
|
| 262562 |
+
"step": 128270
|
| 262563 |
+
},
|
| 262564 |
+
{
|
| 262565 |
+
"epoch": 1033.75,
|
| 262566 |
+
"learning_rate": 7.96113782051282e-06,
|
| 262567 |
+
"loss": 0.3668,
|
| 262568 |
+
"step": 128275
|
| 262569 |
+
},
|
| 262570 |
+
{
|
| 262571 |
+
"epoch": 1033.79,
|
| 262572 |
+
"learning_rate": 7.961057692307693e-06,
|
| 262573 |
+
"loss": 0.7907,
|
| 262574 |
+
"step": 128280
|
| 262575 |
+
},
|
| 262576 |
+
{
|
| 262577 |
+
"epoch": 1033.83,
|
| 262578 |
+
"learning_rate": 7.960977564102565e-06,
|
| 262579 |
+
"loss": 0.7451,
|
| 262580 |
+
"step": 128285
|
| 262581 |
+
},
|
| 262582 |
+
{
|
| 262583 |
+
"epoch": 1033.87,
|
| 262584 |
+
"learning_rate": 7.960897435897436e-06,
|
| 262585 |
+
"loss": 0.2511,
|
| 262586 |
+
"step": 128290
|
| 262587 |
+
},
|
| 262588 |
+
{
|
| 262589 |
+
"epoch": 1033.91,
|
| 262590 |
+
"learning_rate": 7.960817307692307e-06,
|
| 262591 |
+
"loss": 0.2528,
|
| 262592 |
+
"step": 128295
|
| 262593 |
+
},
|
| 262594 |
+
{
|
| 262595 |
+
"epoch": 1033.95,
|
| 262596 |
+
"learning_rate": 7.96073717948718e-06,
|
| 262597 |
+
"loss": 0.4353,
|
| 262598 |
+
"step": 128300
|
| 262599 |
+
},
|
| 262600 |
+
{
|
| 262601 |
+
"epoch": 1033.99,
|
| 262602 |
+
"learning_rate": 7.960657051282052e-06,
|
| 262603 |
+
"loss": 0.7537,
|
| 262604 |
+
"step": 128305
|
| 262605 |
+
},
|
| 262606 |
+
{
|
| 262607 |
+
"epoch": 1034.0,
|
| 262608 |
+
"eval_loss": 0.4501224160194397,
|
| 262609 |
+
"eval_runtime": 35.5986,
|
| 262610 |
+
"eval_samples_per_second": 23.625,
|
| 262611 |
+
"eval_steps_per_second": 0.758,
|
| 262612 |
+
"eval_wer": 0.18477849299507762,
|
| 262613 |
+
"step": 128306
|
| 262614 |
+
},
|
| 262615 |
+
{
|
| 262616 |
+
"epoch": 1026.03,
|
| 262617 |
+
"learning_rate": 7.960576923076923e-06,
|
| 262618 |
+
"loss": 0.3182,
|
| 262619 |
+
"step": 128310
|
| 262620 |
+
},
|
| 262621 |
+
{
|
| 262622 |
+
"epoch": 1026.07,
|
| 262623 |
+
"learning_rate": 7.960496794871796e-06,
|
| 262624 |
+
"loss": 0.2978,
|
| 262625 |
+
"step": 128315
|
| 262626 |
+
},
|
| 262627 |
+
{
|
| 262628 |
+
"epoch": 1026.11,
|
| 262629 |
+
"learning_rate": 7.960416666666668e-06,
|
| 262630 |
+
"loss": 0.2738,
|
| 262631 |
+
"step": 128320
|
| 262632 |
+
},
|
| 262633 |
+
{
|
| 262634 |
+
"epoch": 1026.15,
|
| 262635 |
+
"learning_rate": 7.960336538461539e-06,
|
| 262636 |
+
"loss": 0.2883,
|
| 262637 |
+
"step": 128325
|
| 262638 |
+
},
|
| 262639 |
+
{
|
| 262640 |
+
"epoch": 1026.19,
|
| 262641 |
+
"learning_rate": 7.96025641025641e-06,
|
| 262642 |
+
"loss": 0.8658,
|
| 262643 |
+
"step": 128330
|
| 262644 |
+
},
|
| 262645 |
+
{
|
| 262646 |
+
"epoch": 1026.23,
|
| 262647 |
+
"learning_rate": 7.960176282051283e-06,
|
| 262648 |
+
"loss": 0.6159,
|
| 262649 |
+
"step": 128335
|
| 262650 |
+
},
|
| 262651 |
+
{
|
| 262652 |
+
"epoch": 1026.27,
|
| 262653 |
+
"learning_rate": 7.960096153846155e-06,
|
| 262654 |
+
"loss": 0.3185,
|
| 262655 |
+
"step": 128340
|
| 262656 |
+
},
|
| 262657 |
+
{
|
| 262658 |
+
"epoch": 1026.31,
|
| 262659 |
+
"learning_rate": 7.960016025641026e-06,
|
| 262660 |
+
"loss": 0.317,
|
| 262661 |
+
"step": 128345
|
| 262662 |
+
},
|
| 262663 |
+
{
|
| 262664 |
+
"epoch": 1026.35,
|
| 262665 |
+
"learning_rate": 7.959935897435897e-06,
|
| 262666 |
+
"loss": 0.4537,
|
| 262667 |
+
"step": 128350
|
| 262668 |
+
},
|
| 262669 |
+
{
|
| 262670 |
+
"epoch": 1026.39,
|
| 262671 |
+
"learning_rate": 7.95985576923077e-06,
|
| 262672 |
+
"loss": 0.8634,
|
| 262673 |
+
"step": 128355
|
| 262674 |
+
},
|
| 262675 |
+
{
|
| 262676 |
+
"epoch": 1026.43,
|
| 262677 |
+
"learning_rate": 7.959775641025642e-06,
|
| 262678 |
+
"loss": 0.6575,
|
| 262679 |
+
"step": 128360
|
| 262680 |
+
},
|
| 262681 |
+
{
|
| 262682 |
+
"epoch": 1026.47,
|
| 262683 |
+
"learning_rate": 7.959695512820513e-06,
|
| 262684 |
+
"loss": 0.2603,
|
| 262685 |
+
"step": 128365
|
| 262686 |
+
},
|
| 262687 |
+
{
|
| 262688 |
+
"epoch": 1026.51,
|
| 262689 |
+
"learning_rate": 7.959615384615386e-06,
|
| 262690 |
+
"loss": 0.3384,
|
| 262691 |
+
"step": 128370
|
| 262692 |
+
},
|
| 262693 |
+
{
|
| 262694 |
+
"epoch": 1026.55,
|
| 262695 |
+
"learning_rate": 7.959535256410258e-06,
|
| 262696 |
+
"loss": 0.3691,
|
| 262697 |
+
"step": 128375
|
| 262698 |
+
},
|
| 262699 |
+
{
|
| 262700 |
+
"epoch": 1026.59,
|
| 262701 |
+
"learning_rate": 7.959455128205129e-06,
|
| 262702 |
+
"loss": 1.2005,
|
| 262703 |
+
"step": 128380
|
| 262704 |
+
},
|
| 262705 |
+
{
|
| 262706 |
+
"epoch": 1026.63,
|
| 262707 |
+
"learning_rate": 7.959375e-06,
|
| 262708 |
+
"loss": 0.5745,
|
| 262709 |
+
"step": 128385
|
| 262710 |
+
},
|
| 262711 |
+
{
|
| 262712 |
+
"epoch": 1026.67,
|
| 262713 |
+
"learning_rate": 7.959294871794873e-06,
|
| 262714 |
+
"loss": 0.2822,
|
| 262715 |
+
"step": 128390
|
| 262716 |
+
},
|
| 262717 |
+
{
|
| 262718 |
+
"epoch": 1026.71,
|
| 262719 |
+
"learning_rate": 7.959214743589743e-06,
|
| 262720 |
+
"loss": 0.2759,
|
| 262721 |
+
"step": 128395
|
| 262722 |
+
},
|
| 262723 |
+
{
|
| 262724 |
+
"epoch": 1026.75,
|
| 262725 |
+
"learning_rate": 7.959134615384616e-06,
|
| 262726 |
+
"loss": 0.3691,
|
| 262727 |
+
"step": 128400
|
| 262728 |
+
},
|
| 262729 |
+
{
|
| 262730 |
+
"epoch": 1026.79,
|
| 262731 |
+
"learning_rate": 7.95905448717949e-06,
|
| 262732 |
+
"loss": 1.295,
|
| 262733 |
+
"step": 128405
|
| 262734 |
+
},
|
| 262735 |
+
{
|
| 262736 |
+
"epoch": 1026.83,
|
| 262737 |
+
"learning_rate": 7.958974358974359e-06,
|
| 262738 |
+
"loss": 0.665,
|
| 262739 |
+
"step": 128410
|
| 262740 |
+
},
|
| 262741 |
+
{
|
| 262742 |
+
"epoch": 1026.87,
|
| 262743 |
+
"learning_rate": 7.958894230769232e-06,
|
| 262744 |
+
"loss": 0.2374,
|
| 262745 |
+
"step": 128415
|
| 262746 |
+
},
|
| 262747 |
+
{
|
| 262748 |
+
"epoch": 1026.91,
|
| 262749 |
+
"learning_rate": 7.958814102564103e-06,
|
| 262750 |
+
"loss": 0.3356,
|
| 262751 |
+
"step": 128420
|
| 262752 |
+
},
|
| 262753 |
+
{
|
| 262754 |
+
"epoch": 1026.95,
|
| 262755 |
+
"learning_rate": 7.958733974358975e-06,
|
| 262756 |
+
"loss": 0.349,
|
| 262757 |
+
"step": 128425
|
| 262758 |
+
},
|
| 262759 |
+
{
|
| 262760 |
+
"epoch": 1026.99,
|
| 262761 |
+
"learning_rate": 7.958653846153846e-06,
|
| 262762 |
+
"loss": 0.913,
|
| 262763 |
+
"step": 128430
|
| 262764 |
+
},
|
| 262765 |
+
{
|
| 262766 |
+
"epoch": 1027.0,
|
| 262767 |
+
"eval_loss": 0.39013320207595825,
|
| 262768 |
+
"eval_runtime": 36.0878,
|
| 262769 |
+
"eval_samples_per_second": 23.304,
|
| 262770 |
+
"eval_steps_per_second": 0.748,
|
| 262771 |
+
"eval_wer": 0.18398896791987226,
|
| 262772 |
+
"step": 128431
|
| 262773 |
+
},
|
| 262774 |
+
{
|
| 262775 |
+
"epoch": 1035.03,
|
| 262776 |
+
"learning_rate": 7.958573717948719e-06,
|
| 262777 |
+
"loss": 0.3237,
|
| 262778 |
+
"step": 128435
|
| 262779 |
+
},
|
| 262780 |
+
{
|
| 262781 |
+
"epoch": 1035.07,
|
| 262782 |
+
"learning_rate": 7.95849358974359e-06,
|
| 262783 |
+
"loss": 0.277,
|
| 262784 |
+
"step": 128440
|
| 262785 |
+
},
|
| 262786 |
+
{
|
| 262787 |
+
"epoch": 1035.11,
|
| 262788 |
+
"learning_rate": 7.958413461538462e-06,
|
| 262789 |
+
"loss": 0.3544,
|
| 262790 |
+
"step": 128445
|
| 262791 |
+
},
|
| 262792 |
+
{
|
| 262793 |
+
"epoch": 1035.15,
|
| 262794 |
+
"learning_rate": 7.958333333333333e-06,
|
| 262795 |
+
"loss": 0.3829,
|
| 262796 |
+
"step": 128450
|
| 262797 |
+
},
|
| 262798 |
+
{
|
| 262799 |
+
"epoch": 1035.19,
|
| 262800 |
+
"learning_rate": 7.958253205128206e-06,
|
| 262801 |
+
"loss": 0.919,
|
| 262802 |
+
"step": 128455
|
| 262803 |
+
},
|
| 262804 |
+
{
|
| 262805 |
+
"epoch": 1035.23,
|
| 262806 |
+
"learning_rate": 7.958173076923077e-06,
|
| 262807 |
+
"loss": 0.6866,
|
| 262808 |
+
"step": 128460
|
| 262809 |
+
},
|
| 262810 |
+
{
|
| 262811 |
+
"epoch": 1035.27,
|
| 262812 |
+
"learning_rate": 7.958092948717949e-06,
|
| 262813 |
+
"loss": 0.2282,
|
| 262814 |
+
"step": 128465
|
| 262815 |
+
},
|
| 262816 |
+
{
|
| 262817 |
+
"epoch": 1035.31,
|
| 262818 |
+
"learning_rate": 7.958012820512822e-06,
|
| 262819 |
+
"loss": 0.3411,
|
| 262820 |
+
"step": 128470
|
| 262821 |
+
},
|
| 262822 |
+
{
|
| 262823 |
+
"epoch": 1035.35,
|
| 262824 |
+
"learning_rate": 7.957932692307693e-06,
|
| 262825 |
+
"loss": 0.4395,
|
| 262826 |
+
"step": 128475
|
| 262827 |
+
},
|
| 262828 |
+
{
|
| 262829 |
+
"epoch": 1035.39,
|
| 262830 |
+
"learning_rate": 7.957852564102565e-06,
|
| 262831 |
+
"loss": 0.9639,
|
| 262832 |
+
"step": 128480
|
| 262833 |
+
},
|
| 262834 |
+
{
|
| 262835 |
+
"epoch": 1035.43,
|
| 262836 |
+
"learning_rate": 7.957772435897436e-06,
|
| 262837 |
+
"loss": 0.6705,
|
| 262838 |
+
"step": 128485
|
| 262839 |
+
},
|
| 262840 |
+
{
|
| 262841 |
+
"epoch": 1035.47,
|
| 262842 |
+
"learning_rate": 7.957692307692309e-06,
|
| 262843 |
+
"loss": 0.3297,
|
| 262844 |
+
"step": 128490
|
| 262845 |
+
},
|
| 262846 |
+
{
|
| 262847 |
+
"epoch": 1035.51,
|
| 262848 |
+
"learning_rate": 7.95761217948718e-06,
|
| 262849 |
+
"loss": 0.3632,
|
| 262850 |
+
"step": 128495
|
| 262851 |
+
},
|
| 262852 |
+
{
|
| 262853 |
+
"epoch": 1035.55,
|
| 262854 |
+
"learning_rate": 7.957532051282052e-06,
|
| 262855 |
+
"loss": 0.4183,
|
| 262856 |
+
"step": 128500
|
| 262857 |
+
},
|
| 262858 |
+
{
|
| 262859 |
+
"epoch": 1035.59,
|
| 262860 |
+
"learning_rate": 7.957451923076925e-06,
|
| 262861 |
+
"loss": 0.9173,
|
| 262862 |
+
"step": 128505
|
| 262863 |
+
},
|
| 262864 |
+
{
|
| 262865 |
+
"epoch": 1035.63,
|
| 262866 |
+
"learning_rate": 7.957371794871796e-06,
|
| 262867 |
+
"loss": 0.6212,
|
| 262868 |
+
"step": 128510
|
| 262869 |
+
},
|
| 262870 |
+
{
|
| 262871 |
+
"epoch": 1035.67,
|
| 262872 |
+
"learning_rate": 7.957291666666668e-06,
|
| 262873 |
+
"loss": 0.2969,
|
| 262874 |
+
"step": 128515
|
| 262875 |
+
},
|
| 262876 |
+
{
|
| 262877 |
+
"epoch": 1035.71,
|
| 262878 |
+
"learning_rate": 7.957211538461539e-06,
|
| 262879 |
+
"loss": 0.344,
|
| 262880 |
+
"step": 128520
|
| 262881 |
+
},
|
| 262882 |
+
{
|
| 262883 |
+
"epoch": 1035.76,
|
| 262884 |
+
"learning_rate": 7.957131410256412e-06,
|
| 262885 |
+
"loss": 0.3409,
|
| 262886 |
+
"step": 128525
|
| 262887 |
+
},
|
| 262888 |
+
{
|
| 262889 |
+
"epoch": 1035.8,
|
| 262890 |
+
"learning_rate": 7.957051282051282e-06,
|
| 262891 |
+
"loss": 0.7752,
|
| 262892 |
+
"step": 128530
|
| 262893 |
+
},
|
| 262894 |
+
{
|
| 262895 |
+
"epoch": 1035.84,
|
| 262896 |
+
"learning_rate": 7.956971153846155e-06,
|
| 262897 |
+
"loss": 0.5721,
|
| 262898 |
+
"step": 128535
|
| 262899 |
+
},
|
| 262900 |
+
{
|
| 262901 |
+
"epoch": 1035.88,
|
| 262902 |
+
"learning_rate": 7.956891025641026e-06,
|
| 262903 |
+
"loss": 0.2648,
|
| 262904 |
+
"step": 128540
|
| 262905 |
+
},
|
| 262906 |
+
{
|
| 262907 |
+
"epoch": 1035.92,
|
| 262908 |
+
"learning_rate": 7.956810897435897e-06,
|
| 262909 |
+
"loss": 0.2546,
|
| 262910 |
+
"step": 128545
|
| 262911 |
+
},
|
| 262912 |
+
{
|
| 262913 |
+
"epoch": 1035.96,
|
| 262914 |
+
"learning_rate": 7.956730769230769e-06,
|
| 262915 |
+
"loss": 0.4116,
|
| 262916 |
+
"step": 128550
|
| 262917 |
+
},
|
| 262918 |
+
{
|
| 262919 |
+
"epoch": 1036.0,
|
| 262920 |
+
"learning_rate": 7.956650641025642e-06,
|
| 262921 |
+
"loss": 1.0995,
|
| 262922 |
+
"step": 128555
|
| 262923 |
+
},
|
| 262924 |
+
{
|
| 262925 |
+
"epoch": 1036.0,
|
| 262926 |
+
"eval_loss": 0.38994449377059937,
|
| 262927 |
+
"eval_runtime": 36.1195,
|
| 262928 |
+
"eval_samples_per_second": 23.284,
|
| 262929 |
+
"eval_steps_per_second": 0.748,
|
| 262930 |
+
"eval_wer": 0.17821493624772314,
|
| 262931 |
+
"step": 128555
|
| 262932 |
}
|
| 262933 |
],
|
| 262934 |
"max_steps": 620000,
|
| 262935 |
"num_train_epochs": 5000,
|
| 262936 |
+
"total_flos": 3.6175982925308166e+20,
|
| 262937 |
"trial_name": null,
|
| 262938 |
"trial_params": null
|
| 262939 |
}
|
model-bin/finetune/base/{checkpoint-127934 → checkpoint-128555}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1630154741.1674275/events.out.tfevents.1630154741.86bb0ddabf9b.4092.161
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:661e3daa2fc836b008b57bb92a5788644d8a30e3e09e155adb080655b760dab1
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630155136.7040803/events.out.tfevents.1630155136.86bb0ddabf9b.4092.163
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:03f781653158771bcb94e23be35d9b984f08f9011e961927b55c7ac16338f4e6
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630155523.684827/events.out.tfevents.1630155523.86bb0ddabf9b.4092.165
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c8126b6f05db9404a2b2cf59a0df4a36fc25cbc0775e78f07981d9292da0cb75
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630155913.5081563/events.out.tfevents.1630155913.86bb0ddabf9b.4092.167
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:bb5b6b16e4244a1fdf3278d955371f4c8e8c7ac98dbb0300151ada50cd8b93f7
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630156305.7181728/events.out.tfevents.1630156305.86bb0ddabf9b.4092.169
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e7cd7e45415f7b305ec0fb5d89e9fcdb9207439e325d5332292fe56388dd0828
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1630154741.86bb0ddabf9b.4092.160
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:4ad6684b494eda0155994ffa2c5f363983ace70e1771b5dd7e4b1ec9a5c51819
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630155136.86bb0ddabf9b.4092.162
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:563ff2a1378c58d9cf770f162ff08f8442a46ed33aa809771458c880cfb81c6e
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630155523.86bb0ddabf9b.4092.164
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f164e4a3d71dcaf98fa4449520013d91669b622d917080135801329a4593a33b
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630155913.86bb0ddabf9b.4092.166
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:49ea6309158cbac808ea94eaee73da384a5188daf2b2ff45600a228823cd3917
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630156305.86bb0ddabf9b.4092.168
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f436f3703fc0f994efd708dac5cd4663aa34f0d92f7d73f762f24a5e0505e3da
|
| 3 |
+
size 8622
|