New model2
Browse files- all_results.json +8 -8
- config.json +1 -1
- pytorch_model.bin +1 -1
- train_results.json +8 -8
- trainer_state.json +3765 -11
- training_args.bin +1 -1
all_results.json
CHANGED
|
@@ -1,15 +1,15 @@
|
|
| 1 |
{
|
| 2 |
-
"epoch":
|
| 3 |
-
"init_mem_cpu_alloc_delta":
|
| 4 |
"init_mem_cpu_peaked_delta": 18306,
|
| 5 |
-
"init_mem_gpu_alloc_delta":
|
| 6 |
"init_mem_gpu_peaked_delta": 0,
|
| 7 |
-
"total_flos":
|
| 8 |
"train_mem_cpu_alloc_delta": 0,
|
| 9 |
"train_mem_cpu_peaked_delta": 0,
|
| 10 |
-
"train_mem_gpu_alloc_delta":
|
| 11 |
"train_mem_gpu_peaked_delta": 0,
|
| 12 |
-
"train_runtime":
|
| 13 |
-
"train_samples":
|
| 14 |
-
"train_samples_per_second":
|
| 15 |
}
|
|
|
|
| 1 |
{
|
| 2 |
+
"epoch": 33.0,
|
| 3 |
+
"init_mem_cpu_alloc_delta": 65159,
|
| 4 |
"init_mem_cpu_peaked_delta": 18306,
|
| 5 |
+
"init_mem_gpu_alloc_delta": 1261972480,
|
| 6 |
"init_mem_gpu_peaked_delta": 0,
|
| 7 |
+
"total_flos": 4.638603148454345e+19,
|
| 8 |
"train_mem_cpu_alloc_delta": 0,
|
| 9 |
"train_mem_cpu_peaked_delta": 0,
|
| 10 |
+
"train_mem_gpu_alloc_delta": 6365176832,
|
| 11 |
"train_mem_gpu_peaked_delta": 0,
|
| 12 |
+
"train_runtime": 3655.2719,
|
| 13 |
+
"train_samples": 7434,
|
| 14 |
+
"train_samples_per_second": 2.095
|
| 15 |
}
|
config.json
CHANGED
|
@@ -1,5 +1,5 @@
|
|
| 1 |
{
|
| 2 |
-
"_name_or_path": "/workspace/data2/el/wav2vec2-large-xlsr-greek/checkpoint-
|
| 3 |
"activation_dropout": 0.0,
|
| 4 |
"apply_spec_augment": true,
|
| 5 |
"architectures": [
|
|
|
|
| 1 |
{
|
| 2 |
+
"_name_or_path": "/workspace/data2/el/wav2vec2-large-xlsr-greek/checkpoint-7656",
|
| 3 |
"activation_dropout": 0.0,
|
| 4 |
"apply_spec_augment": true,
|
| 5 |
"architectures": [
|
pytorch_model.bin
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 1262151127
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a303ff6f524b384a16a252f54365fe10e1a6f856be9b4e837b29419acc431958
|
| 3 |
size 1262151127
|
train_results.json
CHANGED
|
@@ -1,15 +1,15 @@
|
|
| 1 |
{
|
| 2 |
-
"epoch":
|
| 3 |
-
"init_mem_cpu_alloc_delta":
|
| 4 |
"init_mem_cpu_peaked_delta": 18306,
|
| 5 |
-
"init_mem_gpu_alloc_delta":
|
| 6 |
"init_mem_gpu_peaked_delta": 0,
|
| 7 |
-
"total_flos":
|
| 8 |
"train_mem_cpu_alloc_delta": 0,
|
| 9 |
"train_mem_cpu_peaked_delta": 0,
|
| 10 |
-
"train_mem_gpu_alloc_delta":
|
| 11 |
"train_mem_gpu_peaked_delta": 0,
|
| 12 |
-
"train_runtime":
|
| 13 |
-
"train_samples":
|
| 14 |
-
"train_samples_per_second":
|
| 15 |
}
|
|
|
|
| 1 |
{
|
| 2 |
+
"epoch": 33.0,
|
| 3 |
+
"init_mem_cpu_alloc_delta": 65159,
|
| 4 |
"init_mem_cpu_peaked_delta": 18306,
|
| 5 |
+
"init_mem_gpu_alloc_delta": 1261972480,
|
| 6 |
"init_mem_gpu_peaked_delta": 0,
|
| 7 |
+
"total_flos": 4.638603148454345e+19,
|
| 8 |
"train_mem_cpu_alloc_delta": 0,
|
| 9 |
"train_mem_cpu_peaked_delta": 0,
|
| 10 |
+
"train_mem_gpu_alloc_delta": 6365176832,
|
| 11 |
"train_mem_gpu_peaked_delta": 0,
|
| 12 |
+
"train_runtime": 3655.2719,
|
| 13 |
+
"train_samples": 7434,
|
| 14 |
+
"train_samples_per_second": 2.095
|
| 15 |
}
|
trainer_state.json
CHANGED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
-
"best_metric": 0.
|
| 3 |
-
"best_model_checkpoint": "/workspace/data2/el/wav2vec2-large-xlsr-greek/checkpoint-
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -1171,17 +1171,3771 @@
|
|
| 1171 |
"eval_wer": 0.47945896840617747,
|
| 1172 |
"step": 1740
|
| 1173 |
},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1174 |
{
|
| 1175 |
"epoch": 15.0,
|
| 1176 |
-
"
|
| 1177 |
-
"
|
| 1178 |
-
"
|
| 1179 |
-
"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1180 |
}
|
| 1181 |
],
|
| 1182 |
-
"max_steps":
|
| 1183 |
-
"num_train_epochs":
|
| 1184 |
-
"total_flos":
|
| 1185 |
"trial_name": null,
|
| 1186 |
"trial_params": null
|
| 1187 |
}
|
|
|
|
| 1 |
{
|
| 2 |
+
"best_metric": 0.42152013727667303,
|
| 3 |
+
"best_model_checkpoint": "/workspace/data2/el/wav2vec2-large-xlsr-greek/checkpoint-7656",
|
| 4 |
+
"epoch": 32.99784946236559,
|
| 5 |
+
"global_step": 7656,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 1171 |
"eval_wer": 0.47945896840617747,
|
| 1172 |
"step": 1740
|
| 1173 |
},
|
| 1174 |
+
{
|
| 1175 |
+
"epoch": 7.54,
|
| 1176 |
+
"learning_rate": 0.00018766233766233763,
|
| 1177 |
+
"loss": 0.7294,
|
| 1178 |
+
"step": 1750
|
| 1179 |
+
},
|
| 1180 |
+
{
|
| 1181 |
+
"epoch": 7.58,
|
| 1182 |
+
"learning_rate": 0.00018701298701298698,
|
| 1183 |
+
"loss": 0.521,
|
| 1184 |
+
"step": 1760
|
| 1185 |
+
},
|
| 1186 |
+
{
|
| 1187 |
+
"epoch": 7.63,
|
| 1188 |
+
"learning_rate": 0.00018636363636363634,
|
| 1189 |
+
"loss": 0.4495,
|
| 1190 |
+
"step": 1770
|
| 1191 |
+
},
|
| 1192 |
+
{
|
| 1193 |
+
"epoch": 7.67,
|
| 1194 |
+
"learning_rate": 0.00018571428571428572,
|
| 1195 |
+
"loss": 0.4751,
|
| 1196 |
+
"step": 1780
|
| 1197 |
+
},
|
| 1198 |
+
{
|
| 1199 |
+
"epoch": 7.71,
|
| 1200 |
+
"learning_rate": 0.00018506493506493507,
|
| 1201 |
+
"loss": 0.3635,
|
| 1202 |
+
"step": 1790
|
| 1203 |
+
},
|
| 1204 |
+
{
|
| 1205 |
+
"epoch": 7.76,
|
| 1206 |
+
"learning_rate": 0.0001844155844155844,
|
| 1207 |
+
"loss": 0.3542,
|
| 1208 |
+
"step": 1800
|
| 1209 |
+
},
|
| 1210 |
+
{
|
| 1211 |
+
"epoch": 7.8,
|
| 1212 |
+
"learning_rate": 0.00018376623376623375,
|
| 1213 |
+
"loss": 0.3538,
|
| 1214 |
+
"step": 1810
|
| 1215 |
+
},
|
| 1216 |
+
{
|
| 1217 |
+
"epoch": 7.84,
|
| 1218 |
+
"learning_rate": 0.0001831168831168831,
|
| 1219 |
+
"loss": 0.3165,
|
| 1220 |
+
"step": 1820
|
| 1221 |
+
},
|
| 1222 |
+
{
|
| 1223 |
+
"epoch": 7.89,
|
| 1224 |
+
"learning_rate": 0.00018246753246753245,
|
| 1225 |
+
"loss": 0.3112,
|
| 1226 |
+
"step": 1830
|
| 1227 |
+
},
|
| 1228 |
+
{
|
| 1229 |
+
"epoch": 7.93,
|
| 1230 |
+
"learning_rate": 0.0001818181818181818,
|
| 1231 |
+
"loss": 0.2908,
|
| 1232 |
+
"step": 1840
|
| 1233 |
+
},
|
| 1234 |
+
{
|
| 1235 |
+
"epoch": 7.97,
|
| 1236 |
+
"learning_rate": 0.00018116883116883116,
|
| 1237 |
+
"loss": 0.2654,
|
| 1238 |
+
"step": 1850
|
| 1239 |
+
},
|
| 1240 |
+
{
|
| 1241 |
+
"epoch": 8.0,
|
| 1242 |
+
"eval_loss": 0.5642479062080383,
|
| 1243 |
+
"eval_runtime": 162.5947,
|
| 1244 |
+
"eval_samples_per_second": 9.361,
|
| 1245 |
+
"eval_wer": 0.543151307156556,
|
| 1246 |
+
"step": 1856
|
| 1247 |
+
},
|
| 1248 |
+
{
|
| 1249 |
+
"epoch": 8.02,
|
| 1250 |
+
"learning_rate": 0.0001805194805194805,
|
| 1251 |
+
"loss": 0.2839,
|
| 1252 |
+
"step": 1860
|
| 1253 |
+
},
|
| 1254 |
+
{
|
| 1255 |
+
"epoch": 8.06,
|
| 1256 |
+
"learning_rate": 0.00017987012987012984,
|
| 1257 |
+
"loss": 0.2609,
|
| 1258 |
+
"step": 1870
|
| 1259 |
+
},
|
| 1260 |
+
{
|
| 1261 |
+
"epoch": 8.1,
|
| 1262 |
+
"learning_rate": 0.0001792207792207792,
|
| 1263 |
+
"loss": 0.2634,
|
| 1264 |
+
"step": 1880
|
| 1265 |
+
},
|
| 1266 |
+
{
|
| 1267 |
+
"epoch": 8.15,
|
| 1268 |
+
"learning_rate": 0.00017857142857142854,
|
| 1269 |
+
"loss": 0.2854,
|
| 1270 |
+
"step": 1890
|
| 1271 |
+
},
|
| 1272 |
+
{
|
| 1273 |
+
"epoch": 8.19,
|
| 1274 |
+
"learning_rate": 0.0001779220779220779,
|
| 1275 |
+
"loss": 0.2951,
|
| 1276 |
+
"step": 1900
|
| 1277 |
+
},
|
| 1278 |
+
{
|
| 1279 |
+
"epoch": 8.23,
|
| 1280 |
+
"learning_rate": 0.00017727272727272728,
|
| 1281 |
+
"loss": 0.319,
|
| 1282 |
+
"step": 1910
|
| 1283 |
+
},
|
| 1284 |
+
{
|
| 1285 |
+
"epoch": 8.28,
|
| 1286 |
+
"learning_rate": 0.00017662337662337663,
|
| 1287 |
+
"loss": 0.2618,
|
| 1288 |
+
"step": 1920
|
| 1289 |
+
},
|
| 1290 |
+
{
|
| 1291 |
+
"epoch": 8.32,
|
| 1292 |
+
"learning_rate": 0.00017597402597402596,
|
| 1293 |
+
"loss": 0.2858,
|
| 1294 |
+
"step": 1930
|
| 1295 |
+
},
|
| 1296 |
+
{
|
| 1297 |
+
"epoch": 8.36,
|
| 1298 |
+
"learning_rate": 0.0001753246753246753,
|
| 1299 |
+
"loss": 0.2259,
|
| 1300 |
+
"step": 1940
|
| 1301 |
+
},
|
| 1302 |
+
{
|
| 1303 |
+
"epoch": 8.4,
|
| 1304 |
+
"learning_rate": 0.00017467532467532466,
|
| 1305 |
+
"loss": 0.2892,
|
| 1306 |
+
"step": 1950
|
| 1307 |
+
},
|
| 1308 |
+
{
|
| 1309 |
+
"epoch": 8.45,
|
| 1310 |
+
"learning_rate": 0.00017402597402597401,
|
| 1311 |
+
"loss": 0.2668,
|
| 1312 |
+
"step": 1960
|
| 1313 |
+
},
|
| 1314 |
+
{
|
| 1315 |
+
"epoch": 8.49,
|
| 1316 |
+
"learning_rate": 0.00017337662337662337,
|
| 1317 |
+
"loss": 0.2553,
|
| 1318 |
+
"step": 1970
|
| 1319 |
+
},
|
| 1320 |
+
{
|
| 1321 |
+
"epoch": 8.53,
|
| 1322 |
+
"learning_rate": 0.00017272727272727272,
|
| 1323 |
+
"loss": 0.2421,
|
| 1324 |
+
"step": 1980
|
| 1325 |
+
},
|
| 1326 |
+
{
|
| 1327 |
+
"epoch": 8.58,
|
| 1328 |
+
"learning_rate": 0.00017207792207792205,
|
| 1329 |
+
"loss": 0.2892,
|
| 1330 |
+
"step": 1990
|
| 1331 |
+
},
|
| 1332 |
+
{
|
| 1333 |
+
"epoch": 8.62,
|
| 1334 |
+
"learning_rate": 0.0001714285714285714,
|
| 1335 |
+
"loss": 0.242,
|
| 1336 |
+
"step": 2000
|
| 1337 |
+
},
|
| 1338 |
+
{
|
| 1339 |
+
"epoch": 8.66,
|
| 1340 |
+
"learning_rate": 0.00017077922077922075,
|
| 1341 |
+
"loss": 0.2353,
|
| 1342 |
+
"step": 2010
|
| 1343 |
+
},
|
| 1344 |
+
{
|
| 1345 |
+
"epoch": 8.71,
|
| 1346 |
+
"learning_rate": 0.0001701298701298701,
|
| 1347 |
+
"loss": 0.2486,
|
| 1348 |
+
"step": 2020
|
| 1349 |
+
},
|
| 1350 |
+
{
|
| 1351 |
+
"epoch": 8.75,
|
| 1352 |
+
"learning_rate": 0.00016948051948051948,
|
| 1353 |
+
"loss": 0.2952,
|
| 1354 |
+
"step": 2030
|
| 1355 |
+
},
|
| 1356 |
+
{
|
| 1357 |
+
"epoch": 8.79,
|
| 1358 |
+
"learning_rate": 0.00016883116883116884,
|
| 1359 |
+
"loss": 0.2422,
|
| 1360 |
+
"step": 2040
|
| 1361 |
+
},
|
| 1362 |
+
{
|
| 1363 |
+
"epoch": 8.83,
|
| 1364 |
+
"learning_rate": 0.00016818181818181816,
|
| 1365 |
+
"loss": 0.2354,
|
| 1366 |
+
"step": 2050
|
| 1367 |
+
},
|
| 1368 |
+
{
|
| 1369 |
+
"epoch": 8.88,
|
| 1370 |
+
"learning_rate": 0.00016753246753246751,
|
| 1371 |
+
"loss": 0.2395,
|
| 1372 |
+
"step": 2060
|
| 1373 |
+
},
|
| 1374 |
+
{
|
| 1375 |
+
"epoch": 8.92,
|
| 1376 |
+
"learning_rate": 0.00016688311688311687,
|
| 1377 |
+
"loss": 0.2699,
|
| 1378 |
+
"step": 2070
|
| 1379 |
+
},
|
| 1380 |
+
{
|
| 1381 |
+
"epoch": 8.96,
|
| 1382 |
+
"learning_rate": 0.00016623376623376622,
|
| 1383 |
+
"loss": 0.2858,
|
| 1384 |
+
"step": 2080
|
| 1385 |
+
},
|
| 1386 |
+
{
|
| 1387 |
+
"epoch": 9.0,
|
| 1388 |
+
"eval_loss": 0.5172072052955627,
|
| 1389 |
+
"eval_runtime": 166.1449,
|
| 1390 |
+
"eval_samples_per_second": 9.161,
|
| 1391 |
+
"eval_wer": 0.5154940950842838,
|
| 1392 |
+
"step": 2088
|
| 1393 |
+
},
|
| 1394 |
+
{
|
| 1395 |
+
"epoch": 9.01,
|
| 1396 |
+
"learning_rate": 0.00016558441558441557,
|
| 1397 |
+
"loss": 0.268,
|
| 1398 |
+
"step": 2090
|
| 1399 |
+
},
|
| 1400 |
+
{
|
| 1401 |
+
"epoch": 9.05,
|
| 1402 |
+
"learning_rate": 0.00016493506493506493,
|
| 1403 |
+
"loss": 0.1977,
|
| 1404 |
+
"step": 2100
|
| 1405 |
+
},
|
| 1406 |
+
{
|
| 1407 |
+
"epoch": 9.09,
|
| 1408 |
+
"learning_rate": 0.00016428571428571428,
|
| 1409 |
+
"loss": 0.2355,
|
| 1410 |
+
"step": 2110
|
| 1411 |
+
},
|
| 1412 |
+
{
|
| 1413 |
+
"epoch": 9.14,
|
| 1414 |
+
"learning_rate": 0.0001636363636363636,
|
| 1415 |
+
"loss": 0.2398,
|
| 1416 |
+
"step": 2120
|
| 1417 |
+
},
|
| 1418 |
+
{
|
| 1419 |
+
"epoch": 9.18,
|
| 1420 |
+
"learning_rate": 0.00016298701298701296,
|
| 1421 |
+
"loss": 0.2471,
|
| 1422 |
+
"step": 2130
|
| 1423 |
+
},
|
| 1424 |
+
{
|
| 1425 |
+
"epoch": 9.22,
|
| 1426 |
+
"learning_rate": 0.0001623376623376623,
|
| 1427 |
+
"loss": 0.1975,
|
| 1428 |
+
"step": 2140
|
| 1429 |
+
},
|
| 1430 |
+
{
|
| 1431 |
+
"epoch": 9.27,
|
| 1432 |
+
"learning_rate": 0.0001616883116883117,
|
| 1433 |
+
"loss": 0.239,
|
| 1434 |
+
"step": 2150
|
| 1435 |
+
},
|
| 1436 |
+
{
|
| 1437 |
+
"epoch": 9.31,
|
| 1438 |
+
"learning_rate": 0.00016103896103896104,
|
| 1439 |
+
"loss": 0.2285,
|
| 1440 |
+
"step": 2160
|
| 1441 |
+
},
|
| 1442 |
+
{
|
| 1443 |
+
"epoch": 9.35,
|
| 1444 |
+
"learning_rate": 0.0001603896103896104,
|
| 1445 |
+
"loss": 0.2506,
|
| 1446 |
+
"step": 2170
|
| 1447 |
+
},
|
| 1448 |
+
{
|
| 1449 |
+
"epoch": 9.4,
|
| 1450 |
+
"learning_rate": 0.00015974025974025972,
|
| 1451 |
+
"loss": 0.2347,
|
| 1452 |
+
"step": 2180
|
| 1453 |
+
},
|
| 1454 |
+
{
|
| 1455 |
+
"epoch": 9.44,
|
| 1456 |
+
"learning_rate": 0.00015909090909090907,
|
| 1457 |
+
"loss": 0.2491,
|
| 1458 |
+
"step": 2190
|
| 1459 |
+
},
|
| 1460 |
+
{
|
| 1461 |
+
"epoch": 9.48,
|
| 1462 |
+
"learning_rate": 0.00015844155844155843,
|
| 1463 |
+
"loss": 0.1968,
|
| 1464 |
+
"step": 2200
|
| 1465 |
+
},
|
| 1466 |
+
{
|
| 1467 |
+
"epoch": 9.52,
|
| 1468 |
+
"learning_rate": 0.00015779220779220778,
|
| 1469 |
+
"loss": 0.2403,
|
| 1470 |
+
"step": 2210
|
| 1471 |
+
},
|
| 1472 |
+
{
|
| 1473 |
+
"epoch": 9.57,
|
| 1474 |
+
"learning_rate": 0.00015714285714285713,
|
| 1475 |
+
"loss": 0.2518,
|
| 1476 |
+
"step": 2220
|
| 1477 |
+
},
|
| 1478 |
+
{
|
| 1479 |
+
"epoch": 9.61,
|
| 1480 |
+
"learning_rate": 0.00015649350649350649,
|
| 1481 |
+
"loss": 0.2112,
|
| 1482 |
+
"step": 2230
|
| 1483 |
+
},
|
| 1484 |
+
{
|
| 1485 |
+
"epoch": 9.65,
|
| 1486 |
+
"learning_rate": 0.0001558441558441558,
|
| 1487 |
+
"loss": 0.2068,
|
| 1488 |
+
"step": 2240
|
| 1489 |
+
},
|
| 1490 |
+
{
|
| 1491 |
+
"epoch": 9.7,
|
| 1492 |
+
"learning_rate": 0.00015519480519480516,
|
| 1493 |
+
"loss": 0.2092,
|
| 1494 |
+
"step": 2250
|
| 1495 |
+
},
|
| 1496 |
+
{
|
| 1497 |
+
"epoch": 9.74,
|
| 1498 |
+
"learning_rate": 0.00015454545454545452,
|
| 1499 |
+
"loss": 0.1868,
|
| 1500 |
+
"step": 2260
|
| 1501 |
+
},
|
| 1502 |
+
{
|
| 1503 |
+
"epoch": 9.78,
|
| 1504 |
+
"learning_rate": 0.0001538961038961039,
|
| 1505 |
+
"loss": 0.2085,
|
| 1506 |
+
"step": 2270
|
| 1507 |
+
},
|
| 1508 |
+
{
|
| 1509 |
+
"epoch": 9.83,
|
| 1510 |
+
"learning_rate": 0.00015324675324675325,
|
| 1511 |
+
"loss": 0.2356,
|
| 1512 |
+
"step": 2280
|
| 1513 |
+
},
|
| 1514 |
+
{
|
| 1515 |
+
"epoch": 9.87,
|
| 1516 |
+
"learning_rate": 0.0001525974025974026,
|
| 1517 |
+
"loss": 0.2055,
|
| 1518 |
+
"step": 2290
|
| 1519 |
+
},
|
| 1520 |
+
{
|
| 1521 |
+
"epoch": 9.91,
|
| 1522 |
+
"learning_rate": 0.00015194805194805193,
|
| 1523 |
+
"loss": 0.18,
|
| 1524 |
+
"step": 2300
|
| 1525 |
+
},
|
| 1526 |
+
{
|
| 1527 |
+
"epoch": 9.95,
|
| 1528 |
+
"learning_rate": 0.00015129870129870128,
|
| 1529 |
+
"loss": 0.221,
|
| 1530 |
+
"step": 2310
|
| 1531 |
+
},
|
| 1532 |
+
{
|
| 1533 |
+
"epoch": 10.0,
|
| 1534 |
+
"learning_rate": 0.00015064935064935063,
|
| 1535 |
+
"loss": 0.203,
|
| 1536 |
+
"step": 2320
|
| 1537 |
+
},
|
| 1538 |
+
{
|
| 1539 |
+
"epoch": 10.0,
|
| 1540 |
+
"eval_loss": 0.5122302770614624,
|
| 1541 |
+
"eval_runtime": 165.9,
|
| 1542 |
+
"eval_samples_per_second": 9.174,
|
| 1543 |
+
"eval_wer": 0.4932875744423135,
|
| 1544 |
+
"step": 2320
|
| 1545 |
+
},
|
| 1546 |
+
{
|
| 1547 |
+
"epoch": 10.04,
|
| 1548 |
+
"learning_rate": 0.00015,
|
| 1549 |
+
"loss": 0.1914,
|
| 1550 |
+
"step": 2330
|
| 1551 |
+
},
|
| 1552 |
+
{
|
| 1553 |
+
"epoch": 10.09,
|
| 1554 |
+
"learning_rate": 0.00014935064935064934,
|
| 1555 |
+
"loss": 0.1825,
|
| 1556 |
+
"step": 2340
|
| 1557 |
+
},
|
| 1558 |
+
{
|
| 1559 |
+
"epoch": 10.13,
|
| 1560 |
+
"learning_rate": 0.0001487012987012987,
|
| 1561 |
+
"loss": 0.1756,
|
| 1562 |
+
"step": 2350
|
| 1563 |
+
},
|
| 1564 |
+
{
|
| 1565 |
+
"epoch": 10.17,
|
| 1566 |
+
"learning_rate": 0.00014805194805194805,
|
| 1567 |
+
"loss": 0.2172,
|
| 1568 |
+
"step": 2360
|
| 1569 |
+
},
|
| 1570 |
+
{
|
| 1571 |
+
"epoch": 10.22,
|
| 1572 |
+
"learning_rate": 0.0001474025974025974,
|
| 1573 |
+
"loss": 0.1853,
|
| 1574 |
+
"step": 2370
|
| 1575 |
+
},
|
| 1576 |
+
{
|
| 1577 |
+
"epoch": 10.26,
|
| 1578 |
+
"learning_rate": 0.00014675324675324672,
|
| 1579 |
+
"loss": 0.1908,
|
| 1580 |
+
"step": 2380
|
| 1581 |
+
},
|
| 1582 |
+
{
|
| 1583 |
+
"epoch": 10.3,
|
| 1584 |
+
"learning_rate": 0.00014610389610389608,
|
| 1585 |
+
"loss": 0.1766,
|
| 1586 |
+
"step": 2390
|
| 1587 |
+
},
|
| 1588 |
+
{
|
| 1589 |
+
"epoch": 10.34,
|
| 1590 |
+
"learning_rate": 0.00014545454545454546,
|
| 1591 |
+
"loss": 0.1714,
|
| 1592 |
+
"step": 2400
|
| 1593 |
+
},
|
| 1594 |
+
{
|
| 1595 |
+
"epoch": 10.39,
|
| 1596 |
+
"learning_rate": 0.00014480519480519478,
|
| 1597 |
+
"loss": 0.1723,
|
| 1598 |
+
"step": 2410
|
| 1599 |
+
},
|
| 1600 |
+
{
|
| 1601 |
+
"epoch": 10.43,
|
| 1602 |
+
"learning_rate": 0.00014415584415584414,
|
| 1603 |
+
"loss": 0.205,
|
| 1604 |
+
"step": 2420
|
| 1605 |
+
},
|
| 1606 |
+
{
|
| 1607 |
+
"epoch": 10.47,
|
| 1608 |
+
"learning_rate": 0.0001435064935064935,
|
| 1609 |
+
"loss": 0.1578,
|
| 1610 |
+
"step": 2430
|
| 1611 |
+
},
|
| 1612 |
+
{
|
| 1613 |
+
"epoch": 10.52,
|
| 1614 |
+
"learning_rate": 0.00014285714285714284,
|
| 1615 |
+
"loss": 0.1727,
|
| 1616 |
+
"step": 2440
|
| 1617 |
+
},
|
| 1618 |
+
{
|
| 1619 |
+
"epoch": 10.56,
|
| 1620 |
+
"learning_rate": 0.0001422077922077922,
|
| 1621 |
+
"loss": 0.1761,
|
| 1622 |
+
"step": 2450
|
| 1623 |
+
},
|
| 1624 |
+
{
|
| 1625 |
+
"epoch": 10.6,
|
| 1626 |
+
"learning_rate": 0.00014155844155844155,
|
| 1627 |
+
"loss": 0.1958,
|
| 1628 |
+
"step": 2460
|
| 1629 |
+
},
|
| 1630 |
+
{
|
| 1631 |
+
"epoch": 10.65,
|
| 1632 |
+
"learning_rate": 0.0001409090909090909,
|
| 1633 |
+
"loss": 0.1804,
|
| 1634 |
+
"step": 2470
|
| 1635 |
+
},
|
| 1636 |
+
{
|
| 1637 |
+
"epoch": 10.69,
|
| 1638 |
+
"learning_rate": 0.00014025974025974025,
|
| 1639 |
+
"loss": 0.2003,
|
| 1640 |
+
"step": 2480
|
| 1641 |
+
},
|
| 1642 |
+
{
|
| 1643 |
+
"epoch": 10.73,
|
| 1644 |
+
"learning_rate": 0.0001396103896103896,
|
| 1645 |
+
"loss": 0.2051,
|
| 1646 |
+
"step": 2490
|
| 1647 |
+
},
|
| 1648 |
+
{
|
| 1649 |
+
"epoch": 10.77,
|
| 1650 |
+
"learning_rate": 0.00013896103896103896,
|
| 1651 |
+
"loss": 0.1825,
|
| 1652 |
+
"step": 2500
|
| 1653 |
+
},
|
| 1654 |
+
{
|
| 1655 |
+
"epoch": 10.82,
|
| 1656 |
+
"learning_rate": 0.00013831168831168828,
|
| 1657 |
+
"loss": 0.192,
|
| 1658 |
+
"step": 2510
|
| 1659 |
+
},
|
| 1660 |
+
{
|
| 1661 |
+
"epoch": 10.86,
|
| 1662 |
+
"learning_rate": 0.00013766233766233766,
|
| 1663 |
+
"loss": 0.1578,
|
| 1664 |
+
"step": 2520
|
| 1665 |
+
},
|
| 1666 |
+
{
|
| 1667 |
+
"epoch": 10.9,
|
| 1668 |
+
"learning_rate": 0.00013701298701298702,
|
| 1669 |
+
"loss": 0.1688,
|
| 1670 |
+
"step": 2530
|
| 1671 |
+
},
|
| 1672 |
+
{
|
| 1673 |
+
"epoch": 10.95,
|
| 1674 |
+
"learning_rate": 0.00013636363636363634,
|
| 1675 |
+
"loss": 0.1489,
|
| 1676 |
+
"step": 2540
|
| 1677 |
+
},
|
| 1678 |
+
{
|
| 1679 |
+
"epoch": 10.99,
|
| 1680 |
+
"learning_rate": 0.0001357142857142857,
|
| 1681 |
+
"loss": 0.1646,
|
| 1682 |
+
"step": 2550
|
| 1683 |
+
},
|
| 1684 |
+
{
|
| 1685 |
+
"epoch": 11.0,
|
| 1686 |
+
"eval_loss": 0.5022615790367126,
|
| 1687 |
+
"eval_runtime": 166.6997,
|
| 1688 |
+
"eval_samples_per_second": 9.13,
|
| 1689 |
+
"eval_wer": 0.4882406379327748,
|
| 1690 |
+
"step": 2552
|
| 1691 |
+
},
|
| 1692 |
+
{
|
| 1693 |
+
"epoch": 11.03,
|
| 1694 |
+
"learning_rate": 0.00013506493506493505,
|
| 1695 |
+
"loss": 0.2222,
|
| 1696 |
+
"step": 2560
|
| 1697 |
+
},
|
| 1698 |
+
{
|
| 1699 |
+
"epoch": 11.08,
|
| 1700 |
+
"learning_rate": 0.0001344155844155844,
|
| 1701 |
+
"loss": 0.148,
|
| 1702 |
+
"step": 2570
|
| 1703 |
+
},
|
| 1704 |
+
{
|
| 1705 |
+
"epoch": 11.12,
|
| 1706 |
+
"learning_rate": 0.00013376623376623375,
|
| 1707 |
+
"loss": 0.1729,
|
| 1708 |
+
"step": 2580
|
| 1709 |
+
},
|
| 1710 |
+
{
|
| 1711 |
+
"epoch": 11.16,
|
| 1712 |
+
"learning_rate": 0.0001331168831168831,
|
| 1713 |
+
"loss": 0.1621,
|
| 1714 |
+
"step": 2590
|
| 1715 |
+
},
|
| 1716 |
+
{
|
| 1717 |
+
"epoch": 11.21,
|
| 1718 |
+
"learning_rate": 0.00013246753246753246,
|
| 1719 |
+
"loss": 0.1835,
|
| 1720 |
+
"step": 2600
|
| 1721 |
+
},
|
| 1722 |
+
{
|
| 1723 |
+
"epoch": 11.25,
|
| 1724 |
+
"learning_rate": 0.0001318181818181818,
|
| 1725 |
+
"loss": 0.1588,
|
| 1726 |
+
"step": 2610
|
| 1727 |
+
},
|
| 1728 |
+
{
|
| 1729 |
+
"epoch": 11.29,
|
| 1730 |
+
"learning_rate": 0.00013116883116883116,
|
| 1731 |
+
"loss": 0.167,
|
| 1732 |
+
"step": 2620
|
| 1733 |
+
},
|
| 1734 |
+
{
|
| 1735 |
+
"epoch": 11.34,
|
| 1736 |
+
"learning_rate": 0.0001305194805194805,
|
| 1737 |
+
"loss": 0.143,
|
| 1738 |
+
"step": 2630
|
| 1739 |
+
},
|
| 1740 |
+
{
|
| 1741 |
+
"epoch": 11.38,
|
| 1742 |
+
"learning_rate": 0.00012987012987012987,
|
| 1743 |
+
"loss": 0.1527,
|
| 1744 |
+
"step": 2640
|
| 1745 |
+
},
|
| 1746 |
+
{
|
| 1747 |
+
"epoch": 11.42,
|
| 1748 |
+
"learning_rate": 0.00012922077922077922,
|
| 1749 |
+
"loss": 0.1446,
|
| 1750 |
+
"step": 2650
|
| 1751 |
+
},
|
| 1752 |
+
{
|
| 1753 |
+
"epoch": 11.46,
|
| 1754 |
+
"learning_rate": 0.00012857142857142855,
|
| 1755 |
+
"loss": 0.1462,
|
| 1756 |
+
"step": 2660
|
| 1757 |
+
},
|
| 1758 |
+
{
|
| 1759 |
+
"epoch": 11.51,
|
| 1760 |
+
"learning_rate": 0.0001279220779220779,
|
| 1761 |
+
"loss": 0.1539,
|
| 1762 |
+
"step": 2670
|
| 1763 |
+
},
|
| 1764 |
+
{
|
| 1765 |
+
"epoch": 11.55,
|
| 1766 |
+
"learning_rate": 0.00012727272727272725,
|
| 1767 |
+
"loss": 0.1925,
|
| 1768 |
+
"step": 2680
|
| 1769 |
+
},
|
| 1770 |
+
{
|
| 1771 |
+
"epoch": 11.59,
|
| 1772 |
+
"learning_rate": 0.0001266233766233766,
|
| 1773 |
+
"loss": 0.1926,
|
| 1774 |
+
"step": 2690
|
| 1775 |
+
},
|
| 1776 |
+
{
|
| 1777 |
+
"epoch": 11.64,
|
| 1778 |
+
"learning_rate": 0.00012597402597402596,
|
| 1779 |
+
"loss": 0.1805,
|
| 1780 |
+
"step": 2700
|
| 1781 |
+
},
|
| 1782 |
+
{
|
| 1783 |
+
"epoch": 11.68,
|
| 1784 |
+
"learning_rate": 0.0001253246753246753,
|
| 1785 |
+
"loss": 0.1829,
|
| 1786 |
+
"step": 2710
|
| 1787 |
+
},
|
| 1788 |
+
{
|
| 1789 |
+
"epoch": 11.72,
|
| 1790 |
+
"learning_rate": 0.00012467532467532467,
|
| 1791 |
+
"loss": 0.1378,
|
| 1792 |
+
"step": 2720
|
| 1793 |
+
},
|
| 1794 |
+
{
|
| 1795 |
+
"epoch": 11.77,
|
| 1796 |
+
"learning_rate": 0.00012402597402597402,
|
| 1797 |
+
"loss": 0.1783,
|
| 1798 |
+
"step": 2730
|
| 1799 |
+
},
|
| 1800 |
+
{
|
| 1801 |
+
"epoch": 11.81,
|
| 1802 |
+
"learning_rate": 0.00012337662337662337,
|
| 1803 |
+
"loss": 0.1356,
|
| 1804 |
+
"step": 2740
|
| 1805 |
+
},
|
| 1806 |
+
{
|
| 1807 |
+
"epoch": 11.85,
|
| 1808 |
+
"learning_rate": 0.00012272727272727272,
|
| 1809 |
+
"loss": 0.1687,
|
| 1810 |
+
"step": 2750
|
| 1811 |
+
},
|
| 1812 |
+
{
|
| 1813 |
+
"epoch": 11.89,
|
| 1814 |
+
"learning_rate": 0.00012207792207792208,
|
| 1815 |
+
"loss": 0.1537,
|
| 1816 |
+
"step": 2760
|
| 1817 |
+
},
|
| 1818 |
+
{
|
| 1819 |
+
"epoch": 11.94,
|
| 1820 |
+
"learning_rate": 0.00012142857142857142,
|
| 1821 |
+
"loss": 0.1416,
|
| 1822 |
+
"step": 2770
|
| 1823 |
+
},
|
| 1824 |
+
{
|
| 1825 |
+
"epoch": 11.98,
|
| 1826 |
+
"learning_rate": 0.00012077922077922077,
|
| 1827 |
+
"loss": 0.1853,
|
| 1828 |
+
"step": 2780
|
| 1829 |
+
},
|
| 1830 |
+
{
|
| 1831 |
+
"epoch": 12.0,
|
| 1832 |
+
"eval_loss": 0.48231056332588196,
|
| 1833 |
+
"eval_runtime": 168.3536,
|
| 1834 |
+
"eval_samples_per_second": 9.04,
|
| 1835 |
+
"eval_wer": 0.4799636620571313,
|
| 1836 |
+
"step": 2784
|
| 1837 |
+
},
|
| 1838 |
+
{
|
| 1839 |
+
"epoch": 12.03,
|
| 1840 |
+
"learning_rate": 0.00012012987012987012,
|
| 1841 |
+
"loss": 0.1258,
|
| 1842 |
+
"step": 2790
|
| 1843 |
+
},
|
| 1844 |
+
{
|
| 1845 |
+
"epoch": 12.07,
|
| 1846 |
+
"learning_rate": 0.00011948051948051947,
|
| 1847 |
+
"loss": 0.1321,
|
| 1848 |
+
"step": 2800
|
| 1849 |
+
},
|
| 1850 |
+
{
|
| 1851 |
+
"epoch": 12.11,
|
| 1852 |
+
"learning_rate": 0.00011883116883116883,
|
| 1853 |
+
"loss": 0.1744,
|
| 1854 |
+
"step": 2810
|
| 1855 |
+
},
|
| 1856 |
+
{
|
| 1857 |
+
"epoch": 12.15,
|
| 1858 |
+
"learning_rate": 0.00011818181818181817,
|
| 1859 |
+
"loss": 0.1272,
|
| 1860 |
+
"step": 2820
|
| 1861 |
+
},
|
| 1862 |
+
{
|
| 1863 |
+
"epoch": 12.2,
|
| 1864 |
+
"learning_rate": 0.00011753246753246752,
|
| 1865 |
+
"loss": 0.1375,
|
| 1866 |
+
"step": 2830
|
| 1867 |
+
},
|
| 1868 |
+
{
|
| 1869 |
+
"epoch": 12.24,
|
| 1870 |
+
"learning_rate": 0.00011688311688311687,
|
| 1871 |
+
"loss": 0.1372,
|
| 1872 |
+
"step": 2840
|
| 1873 |
+
},
|
| 1874 |
+
{
|
| 1875 |
+
"epoch": 12.28,
|
| 1876 |
+
"learning_rate": 0.00011623376623376621,
|
| 1877 |
+
"loss": 0.1467,
|
| 1878 |
+
"step": 2850
|
| 1879 |
+
},
|
| 1880 |
+
{
|
| 1881 |
+
"epoch": 12.33,
|
| 1882 |
+
"learning_rate": 0.00011558441558441558,
|
| 1883 |
+
"loss": 0.1322,
|
| 1884 |
+
"step": 2860
|
| 1885 |
+
},
|
| 1886 |
+
{
|
| 1887 |
+
"epoch": 12.37,
|
| 1888 |
+
"learning_rate": 0.00011493506493506493,
|
| 1889 |
+
"loss": 0.1502,
|
| 1890 |
+
"step": 2870
|
| 1891 |
+
},
|
| 1892 |
+
{
|
| 1893 |
+
"epoch": 12.41,
|
| 1894 |
+
"learning_rate": 0.00011428571428571427,
|
| 1895 |
+
"loss": 0.1538,
|
| 1896 |
+
"step": 2880
|
| 1897 |
+
},
|
| 1898 |
+
{
|
| 1899 |
+
"epoch": 12.46,
|
| 1900 |
+
"learning_rate": 0.00011363636363636362,
|
| 1901 |
+
"loss": 0.1297,
|
| 1902 |
+
"step": 2890
|
| 1903 |
+
},
|
| 1904 |
+
{
|
| 1905 |
+
"epoch": 12.5,
|
| 1906 |
+
"learning_rate": 0.00011298701298701298,
|
| 1907 |
+
"loss": 0.1433,
|
| 1908 |
+
"step": 2900
|
| 1909 |
+
},
|
| 1910 |
+
{
|
| 1911 |
+
"epoch": 12.54,
|
| 1912 |
+
"learning_rate": 0.00011233766233766231,
|
| 1913 |
+
"loss": 0.2079,
|
| 1914 |
+
"step": 2910
|
| 1915 |
+
},
|
| 1916 |
+
{
|
| 1917 |
+
"epoch": 12.58,
|
| 1918 |
+
"learning_rate": 0.00011168831168831168,
|
| 1919 |
+
"loss": 0.1315,
|
| 1920 |
+
"step": 2920
|
| 1921 |
+
},
|
| 1922 |
+
{
|
| 1923 |
+
"epoch": 12.63,
|
| 1924 |
+
"learning_rate": 0.00011103896103896103,
|
| 1925 |
+
"loss": 0.134,
|
| 1926 |
+
"step": 2930
|
| 1927 |
+
},
|
| 1928 |
+
{
|
| 1929 |
+
"epoch": 12.67,
|
| 1930 |
+
"learning_rate": 0.00011038961038961037,
|
| 1931 |
+
"loss": 0.1684,
|
| 1932 |
+
"step": 2940
|
| 1933 |
+
},
|
| 1934 |
+
{
|
| 1935 |
+
"epoch": 12.71,
|
| 1936 |
+
"learning_rate": 0.00010974025974025973,
|
| 1937 |
+
"loss": 0.1354,
|
| 1938 |
+
"step": 2950
|
| 1939 |
+
},
|
| 1940 |
+
{
|
| 1941 |
+
"epoch": 12.76,
|
| 1942 |
+
"learning_rate": 0.00010909090909090908,
|
| 1943 |
+
"loss": 0.1471,
|
| 1944 |
+
"step": 2960
|
| 1945 |
+
},
|
| 1946 |
+
{
|
| 1947 |
+
"epoch": 12.8,
|
| 1948 |
+
"learning_rate": 0.00010844155844155842,
|
| 1949 |
+
"loss": 0.1138,
|
| 1950 |
+
"step": 2970
|
| 1951 |
+
},
|
| 1952 |
+
{
|
| 1953 |
+
"epoch": 12.84,
|
| 1954 |
+
"learning_rate": 0.00010779220779220778,
|
| 1955 |
+
"loss": 0.11,
|
| 1956 |
+
"step": 2980
|
| 1957 |
+
},
|
| 1958 |
+
{
|
| 1959 |
+
"epoch": 12.89,
|
| 1960 |
+
"learning_rate": 0.00010714285714285714,
|
| 1961 |
+
"loss": 0.1385,
|
| 1962 |
+
"step": 2990
|
| 1963 |
+
},
|
| 1964 |
+
{
|
| 1965 |
+
"epoch": 12.93,
|
| 1966 |
+
"learning_rate": 0.00010649350649350649,
|
| 1967 |
+
"loss": 0.1339,
|
| 1968 |
+
"step": 3000
|
| 1969 |
+
},
|
| 1970 |
+
{
|
| 1971 |
+
"epoch": 12.97,
|
| 1972 |
+
"learning_rate": 0.00010584415584415583,
|
| 1973 |
+
"loss": 0.142,
|
| 1974 |
+
"step": 3010
|
| 1975 |
+
},
|
| 1976 |
+
{
|
| 1977 |
+
"epoch": 13.0,
|
| 1978 |
+
"eval_loss": 0.49411994218826294,
|
| 1979 |
+
"eval_runtime": 170.3026,
|
| 1980 |
+
"eval_samples_per_second": 8.937,
|
| 1981 |
+
"eval_wer": 0.4742101544362572,
|
| 1982 |
+
"step": 3016
|
| 1983 |
+
},
|
| 1984 |
+
{
|
| 1985 |
+
"epoch": 13.02,
|
| 1986 |
+
"learning_rate": 0.00010519480519480518,
|
| 1987 |
+
"loss": 0.1411,
|
| 1988 |
+
"step": 3020
|
| 1989 |
+
},
|
| 1990 |
+
{
|
| 1991 |
+
"epoch": 13.06,
|
| 1992 |
+
"learning_rate": 0.00010454545454545455,
|
| 1993 |
+
"loss": 0.133,
|
| 1994 |
+
"step": 3030
|
| 1995 |
+
},
|
| 1996 |
+
{
|
| 1997 |
+
"epoch": 13.1,
|
| 1998 |
+
"learning_rate": 0.00010389610389610389,
|
| 1999 |
+
"loss": 0.1171,
|
| 2000 |
+
"step": 3040
|
| 2001 |
+
},
|
| 2002 |
+
{
|
| 2003 |
+
"epoch": 13.15,
|
| 2004 |
+
"learning_rate": 0.00010324675324675324,
|
| 2005 |
+
"loss": 0.1575,
|
| 2006 |
+
"step": 3050
|
| 2007 |
+
},
|
| 2008 |
+
{
|
| 2009 |
+
"epoch": 13.19,
|
| 2010 |
+
"learning_rate": 0.0001025974025974026,
|
| 2011 |
+
"loss": 0.1194,
|
| 2012 |
+
"step": 3060
|
| 2013 |
+
},
|
| 2014 |
+
{
|
| 2015 |
+
"epoch": 13.23,
|
| 2016 |
+
"learning_rate": 0.00010194805194805193,
|
| 2017 |
+
"loss": 0.1312,
|
| 2018 |
+
"step": 3070
|
| 2019 |
+
},
|
| 2020 |
+
{
|
| 2021 |
+
"epoch": 13.28,
|
| 2022 |
+
"learning_rate": 0.00010129870129870129,
|
| 2023 |
+
"loss": 0.1214,
|
| 2024 |
+
"step": 3080
|
| 2025 |
+
},
|
| 2026 |
+
{
|
| 2027 |
+
"epoch": 13.32,
|
| 2028 |
+
"learning_rate": 0.00010064935064935065,
|
| 2029 |
+
"loss": 0.1174,
|
| 2030 |
+
"step": 3090
|
| 2031 |
+
},
|
| 2032 |
+
{
|
| 2033 |
+
"epoch": 13.36,
|
| 2034 |
+
"learning_rate": 9.999999999999999e-05,
|
| 2035 |
+
"loss": 0.142,
|
| 2036 |
+
"step": 3100
|
| 2037 |
+
},
|
| 2038 |
+
{
|
| 2039 |
+
"epoch": 13.4,
|
| 2040 |
+
"learning_rate": 9.935064935064934e-05,
|
| 2041 |
+
"loss": 0.1143,
|
| 2042 |
+
"step": 3110
|
| 2043 |
+
},
|
| 2044 |
+
{
|
| 2045 |
+
"epoch": 13.45,
|
| 2046 |
+
"learning_rate": 9.87012987012987e-05,
|
| 2047 |
+
"loss": 0.1363,
|
| 2048 |
+
"step": 3120
|
| 2049 |
+
},
|
| 2050 |
+
{
|
| 2051 |
+
"epoch": 13.49,
|
| 2052 |
+
"learning_rate": 9.805194805194804e-05,
|
| 2053 |
+
"loss": 0.1297,
|
| 2054 |
+
"step": 3130
|
| 2055 |
+
},
|
| 2056 |
+
{
|
| 2057 |
+
"epoch": 13.53,
|
| 2058 |
+
"learning_rate": 9.740259740259739e-05,
|
| 2059 |
+
"loss": 0.1373,
|
| 2060 |
+
"step": 3140
|
| 2061 |
+
},
|
| 2062 |
+
{
|
| 2063 |
+
"epoch": 13.58,
|
| 2064 |
+
"learning_rate": 9.675324675324676e-05,
|
| 2065 |
+
"loss": 0.092,
|
| 2066 |
+
"step": 3150
|
| 2067 |
+
},
|
| 2068 |
+
{
|
| 2069 |
+
"epoch": 13.62,
|
| 2070 |
+
"learning_rate": 9.61038961038961e-05,
|
| 2071 |
+
"loss": 0.1177,
|
| 2072 |
+
"step": 3160
|
| 2073 |
+
},
|
| 2074 |
+
{
|
| 2075 |
+
"epoch": 13.66,
|
| 2076 |
+
"learning_rate": 9.545454545454545e-05,
|
| 2077 |
+
"loss": 0.1292,
|
| 2078 |
+
"step": 3170
|
| 2079 |
+
},
|
| 2080 |
+
{
|
| 2081 |
+
"epoch": 13.71,
|
| 2082 |
+
"learning_rate": 9.48051948051948e-05,
|
| 2083 |
+
"loss": 0.098,
|
| 2084 |
+
"step": 3180
|
| 2085 |
+
},
|
| 2086 |
+
{
|
| 2087 |
+
"epoch": 13.75,
|
| 2088 |
+
"learning_rate": 9.415584415584414e-05,
|
| 2089 |
+
"loss": 0.114,
|
| 2090 |
+
"step": 3190
|
| 2091 |
+
},
|
| 2092 |
+
{
|
| 2093 |
+
"epoch": 13.79,
|
| 2094 |
+
"learning_rate": 9.350649350649349e-05,
|
| 2095 |
+
"loss": 0.101,
|
| 2096 |
+
"step": 3200
|
| 2097 |
+
},
|
| 2098 |
+
{
|
| 2099 |
+
"epoch": 13.83,
|
| 2100 |
+
"learning_rate": 9.285714285714286e-05,
|
| 2101 |
+
"loss": 0.1173,
|
| 2102 |
+
"step": 3210
|
| 2103 |
+
},
|
| 2104 |
+
{
|
| 2105 |
+
"epoch": 13.88,
|
| 2106 |
+
"learning_rate": 9.22077922077922e-05,
|
| 2107 |
+
"loss": 0.0998,
|
| 2108 |
+
"step": 3220
|
| 2109 |
+
},
|
| 2110 |
+
{
|
| 2111 |
+
"epoch": 13.92,
|
| 2112 |
+
"learning_rate": 9.155844155844155e-05,
|
| 2113 |
+
"loss": 0.1186,
|
| 2114 |
+
"step": 3230
|
| 2115 |
+
},
|
| 2116 |
+
{
|
| 2117 |
+
"epoch": 13.96,
|
| 2118 |
+
"learning_rate": 9.09090909090909e-05,
|
| 2119 |
+
"loss": 0.1155,
|
| 2120 |
+
"step": 3240
|
| 2121 |
+
},
|
| 2122 |
+
{
|
| 2123 |
+
"epoch": 14.0,
|
| 2124 |
+
"eval_loss": 0.4869328737258911,
|
| 2125 |
+
"eval_runtime": 173.2877,
|
| 2126 |
+
"eval_samples_per_second": 8.783,
|
| 2127 |
+
"eval_wer": 0.46189562935298273,
|
| 2128 |
+
"step": 3248
|
| 2129 |
+
},
|
| 2130 |
+
{
|
| 2131 |
+
"epoch": 14.01,
|
| 2132 |
+
"learning_rate": 9.025974025974026e-05,
|
| 2133 |
+
"loss": 0.1039,
|
| 2134 |
+
"step": 3250
|
| 2135 |
+
},
|
| 2136 |
+
{
|
| 2137 |
+
"epoch": 14.05,
|
| 2138 |
+
"learning_rate": 8.96103896103896e-05,
|
| 2139 |
+
"loss": 0.1142,
|
| 2140 |
+
"step": 3260
|
| 2141 |
+
},
|
| 2142 |
+
{
|
| 2143 |
+
"epoch": 14.09,
|
| 2144 |
+
"learning_rate": 8.896103896103895e-05,
|
| 2145 |
+
"loss": 0.106,
|
| 2146 |
+
"step": 3270
|
| 2147 |
+
},
|
| 2148 |
+
{
|
| 2149 |
+
"epoch": 14.14,
|
| 2150 |
+
"learning_rate": 8.831168831168831e-05,
|
| 2151 |
+
"loss": 0.1123,
|
| 2152 |
+
"step": 3280
|
| 2153 |
+
},
|
| 2154 |
+
{
|
| 2155 |
+
"epoch": 14.18,
|
| 2156 |
+
"learning_rate": 8.766233766233765e-05,
|
| 2157 |
+
"loss": 0.121,
|
| 2158 |
+
"step": 3290
|
| 2159 |
+
},
|
| 2160 |
+
{
|
| 2161 |
+
"epoch": 14.22,
|
| 2162 |
+
"learning_rate": 8.701298701298701e-05,
|
| 2163 |
+
"loss": 0.1108,
|
| 2164 |
+
"step": 3300
|
| 2165 |
+
},
|
| 2166 |
+
{
|
| 2167 |
+
"epoch": 14.27,
|
| 2168 |
+
"learning_rate": 8.636363636363636e-05,
|
| 2169 |
+
"loss": 0.1195,
|
| 2170 |
+
"step": 3310
|
| 2171 |
+
},
|
| 2172 |
+
{
|
| 2173 |
+
"epoch": 14.31,
|
| 2174 |
+
"learning_rate": 8.57142857142857e-05,
|
| 2175 |
+
"loss": 0.1116,
|
| 2176 |
+
"step": 3320
|
| 2177 |
+
},
|
| 2178 |
+
{
|
| 2179 |
+
"epoch": 14.35,
|
| 2180 |
+
"learning_rate": 8.506493506493505e-05,
|
| 2181 |
+
"loss": 0.1008,
|
| 2182 |
+
"step": 3330
|
| 2183 |
+
},
|
| 2184 |
+
{
|
| 2185 |
+
"epoch": 14.4,
|
| 2186 |
+
"learning_rate": 8.441558441558442e-05,
|
| 2187 |
+
"loss": 0.1013,
|
| 2188 |
+
"step": 3340
|
| 2189 |
+
},
|
| 2190 |
+
{
|
| 2191 |
+
"epoch": 14.44,
|
| 2192 |
+
"learning_rate": 8.376623376623376e-05,
|
| 2193 |
+
"loss": 0.1282,
|
| 2194 |
+
"step": 3350
|
| 2195 |
+
},
|
| 2196 |
+
{
|
| 2197 |
+
"epoch": 14.48,
|
| 2198 |
+
"learning_rate": 8.311688311688311e-05,
|
| 2199 |
+
"loss": 0.1175,
|
| 2200 |
+
"step": 3360
|
| 2201 |
+
},
|
| 2202 |
+
{
|
| 2203 |
+
"epoch": 14.52,
|
| 2204 |
+
"learning_rate": 8.246753246753246e-05,
|
| 2205 |
+
"loss": 0.1254,
|
| 2206 |
+
"step": 3370
|
| 2207 |
+
},
|
| 2208 |
+
{
|
| 2209 |
+
"epoch": 14.57,
|
| 2210 |
+
"learning_rate": 8.18181818181818e-05,
|
| 2211 |
+
"loss": 0.1169,
|
| 2212 |
+
"step": 3380
|
| 2213 |
+
},
|
| 2214 |
+
{
|
| 2215 |
+
"epoch": 14.61,
|
| 2216 |
+
"learning_rate": 8.116883116883116e-05,
|
| 2217 |
+
"loss": 0.101,
|
| 2218 |
+
"step": 3390
|
| 2219 |
+
},
|
| 2220 |
+
{
|
| 2221 |
+
"epoch": 14.65,
|
| 2222 |
+
"learning_rate": 8.051948051948052e-05,
|
| 2223 |
+
"loss": 0.134,
|
| 2224 |
+
"step": 3400
|
| 2225 |
+
},
|
| 2226 |
+
{
|
| 2227 |
+
"epoch": 14.7,
|
| 2228 |
+
"learning_rate": 7.987012987012986e-05,
|
| 2229 |
+
"loss": 0.1124,
|
| 2230 |
+
"step": 3410
|
| 2231 |
+
},
|
| 2232 |
+
{
|
| 2233 |
+
"epoch": 14.74,
|
| 2234 |
+
"learning_rate": 7.922077922077921e-05,
|
| 2235 |
+
"loss": 0.0964,
|
| 2236 |
+
"step": 3420
|
| 2237 |
+
},
|
| 2238 |
+
{
|
| 2239 |
+
"epoch": 14.78,
|
| 2240 |
+
"learning_rate": 7.857142857142857e-05,
|
| 2241 |
+
"loss": 0.0959,
|
| 2242 |
+
"step": 3430
|
| 2243 |
+
},
|
| 2244 |
+
{
|
| 2245 |
+
"epoch": 14.83,
|
| 2246 |
+
"learning_rate": 7.79220779220779e-05,
|
| 2247 |
+
"loss": 0.0944,
|
| 2248 |
+
"step": 3440
|
| 2249 |
+
},
|
| 2250 |
+
{
|
| 2251 |
+
"epoch": 14.87,
|
| 2252 |
+
"learning_rate": 7.727272727272726e-05,
|
| 2253 |
+
"loss": 0.1302,
|
| 2254 |
+
"step": 3450
|
| 2255 |
+
},
|
| 2256 |
+
{
|
| 2257 |
+
"epoch": 14.91,
|
| 2258 |
+
"learning_rate": 7.662337662337662e-05,
|
| 2259 |
+
"loss": 0.1181,
|
| 2260 |
+
"step": 3460
|
| 2261 |
+
},
|
| 2262 |
+
{
|
| 2263 |
+
"epoch": 14.95,
|
| 2264 |
+
"learning_rate": 7.597402597402596e-05,
|
| 2265 |
+
"loss": 0.0713,
|
| 2266 |
+
"step": 3470
|
| 2267 |
+
},
|
| 2268 |
+
{
|
| 2269 |
+
"epoch": 15.0,
|
| 2270 |
+
"learning_rate": 7.532467532467532e-05,
|
| 2271 |
+
"loss": 0.0849,
|
| 2272 |
+
"step": 3480
|
| 2273 |
+
},
|
| 2274 |
{
|
| 2275 |
"epoch": 15.0,
|
| 2276 |
+
"eval_loss": 0.4888065457344055,
|
| 2277 |
+
"eval_runtime": 168.5146,
|
| 2278 |
+
"eval_samples_per_second": 9.032,
|
| 2279 |
+
"eval_wer": 0.4614918744322196,
|
| 2280 |
+
"step": 3480
|
| 2281 |
+
},
|
| 2282 |
+
{
|
| 2283 |
+
"epoch": 15.04,
|
| 2284 |
+
"learning_rate": 7.467532467532467e-05,
|
| 2285 |
+
"loss": 0.1545,
|
| 2286 |
+
"step": 3490
|
| 2287 |
+
},
|
| 2288 |
+
{
|
| 2289 |
+
"epoch": 15.09,
|
| 2290 |
+
"learning_rate": 7.402597402597402e-05,
|
| 2291 |
+
"loss": 0.0952,
|
| 2292 |
+
"step": 3500
|
| 2293 |
+
},
|
| 2294 |
+
{
|
| 2295 |
+
"epoch": 15.13,
|
| 2296 |
+
"learning_rate": 7.337662337662336e-05,
|
| 2297 |
+
"loss": 0.1229,
|
| 2298 |
+
"step": 3510
|
| 2299 |
+
},
|
| 2300 |
+
{
|
| 2301 |
+
"epoch": 15.17,
|
| 2302 |
+
"learning_rate": 7.272727272727273e-05,
|
| 2303 |
+
"loss": 0.086,
|
| 2304 |
+
"step": 3520
|
| 2305 |
+
},
|
| 2306 |
+
{
|
| 2307 |
+
"epoch": 15.22,
|
| 2308 |
+
"learning_rate": 7.207792207792207e-05,
|
| 2309 |
+
"loss": 0.093,
|
| 2310 |
+
"step": 3530
|
| 2311 |
+
},
|
| 2312 |
+
{
|
| 2313 |
+
"epoch": 15.26,
|
| 2314 |
+
"learning_rate": 7.142857142857142e-05,
|
| 2315 |
+
"loss": 0.1197,
|
| 2316 |
+
"step": 3540
|
| 2317 |
+
},
|
| 2318 |
+
{
|
| 2319 |
+
"epoch": 15.3,
|
| 2320 |
+
"learning_rate": 7.077922077922077e-05,
|
| 2321 |
+
"loss": 0.1104,
|
| 2322 |
+
"step": 3550
|
| 2323 |
+
},
|
| 2324 |
+
{
|
| 2325 |
+
"epoch": 15.34,
|
| 2326 |
+
"learning_rate": 7.012987012987013e-05,
|
| 2327 |
+
"loss": 0.1101,
|
| 2328 |
+
"step": 3560
|
| 2329 |
+
},
|
| 2330 |
+
{
|
| 2331 |
+
"epoch": 15.39,
|
| 2332 |
+
"learning_rate": 6.948051948051948e-05,
|
| 2333 |
+
"loss": 0.1005,
|
| 2334 |
+
"step": 3570
|
| 2335 |
+
},
|
| 2336 |
+
{
|
| 2337 |
+
"epoch": 15.43,
|
| 2338 |
+
"learning_rate": 6.883116883116883e-05,
|
| 2339 |
+
"loss": 0.0979,
|
| 2340 |
+
"step": 3580
|
| 2341 |
+
},
|
| 2342 |
+
{
|
| 2343 |
+
"epoch": 15.47,
|
| 2344 |
+
"learning_rate": 6.818181818181817e-05,
|
| 2345 |
+
"loss": 0.0979,
|
| 2346 |
+
"step": 3590
|
| 2347 |
+
},
|
| 2348 |
+
{
|
| 2349 |
+
"epoch": 15.52,
|
| 2350 |
+
"learning_rate": 6.753246753246752e-05,
|
| 2351 |
+
"loss": 0.0762,
|
| 2352 |
+
"step": 3600
|
| 2353 |
+
},
|
| 2354 |
+
{
|
| 2355 |
+
"epoch": 15.56,
|
| 2356 |
+
"learning_rate": 6.688311688311688e-05,
|
| 2357 |
+
"loss": 0.1136,
|
| 2358 |
+
"step": 3610
|
| 2359 |
+
},
|
| 2360 |
+
{
|
| 2361 |
+
"epoch": 15.6,
|
| 2362 |
+
"learning_rate": 6.623376623376623e-05,
|
| 2363 |
+
"loss": 0.1039,
|
| 2364 |
+
"step": 3620
|
| 2365 |
+
},
|
| 2366 |
+
{
|
| 2367 |
+
"epoch": 15.65,
|
| 2368 |
+
"learning_rate": 6.558441558441558e-05,
|
| 2369 |
+
"loss": 0.0897,
|
| 2370 |
+
"step": 3630
|
| 2371 |
+
},
|
| 2372 |
+
{
|
| 2373 |
+
"epoch": 15.69,
|
| 2374 |
+
"learning_rate": 6.493506493506494e-05,
|
| 2375 |
+
"loss": 0.0883,
|
| 2376 |
+
"step": 3640
|
| 2377 |
+
},
|
| 2378 |
+
{
|
| 2379 |
+
"epoch": 15.73,
|
| 2380 |
+
"learning_rate": 6.428571428571427e-05,
|
| 2381 |
+
"loss": 0.0931,
|
| 2382 |
+
"step": 3650
|
| 2383 |
+
},
|
| 2384 |
+
{
|
| 2385 |
+
"epoch": 15.77,
|
| 2386 |
+
"learning_rate": 6.363636363636363e-05,
|
| 2387 |
+
"loss": 0.0881,
|
| 2388 |
+
"step": 3660
|
| 2389 |
+
},
|
| 2390 |
+
{
|
| 2391 |
+
"epoch": 15.82,
|
| 2392 |
+
"learning_rate": 6.298701298701298e-05,
|
| 2393 |
+
"loss": 0.1156,
|
| 2394 |
+
"step": 3670
|
| 2395 |
+
},
|
| 2396 |
+
{
|
| 2397 |
+
"epoch": 15.86,
|
| 2398 |
+
"learning_rate": 6.233766233766233e-05,
|
| 2399 |
+
"loss": 0.0939,
|
| 2400 |
+
"step": 3680
|
| 2401 |
+
},
|
| 2402 |
+
{
|
| 2403 |
+
"epoch": 15.9,
|
| 2404 |
+
"learning_rate": 6.168831168831169e-05,
|
| 2405 |
+
"loss": 0.1109,
|
| 2406 |
+
"step": 3690
|
| 2407 |
+
},
|
| 2408 |
+
{
|
| 2409 |
+
"epoch": 15.95,
|
| 2410 |
+
"learning_rate": 6.103896103896104e-05,
|
| 2411 |
+
"loss": 0.0996,
|
| 2412 |
+
"step": 3700
|
| 2413 |
+
},
|
| 2414 |
+
{
|
| 2415 |
+
"epoch": 15.99,
|
| 2416 |
+
"learning_rate": 6.0389610389610384e-05,
|
| 2417 |
+
"loss": 0.0927,
|
| 2418 |
+
"step": 3710
|
| 2419 |
+
},
|
| 2420 |
+
{
|
| 2421 |
+
"epoch": 16.0,
|
| 2422 |
+
"eval_loss": 0.47719869017601013,
|
| 2423 |
+
"eval_runtime": 172.1905,
|
| 2424 |
+
"eval_samples_per_second": 8.839,
|
| 2425 |
+
"eval_wer": 0.448369839507419,
|
| 2426 |
+
"step": 3712
|
| 2427 |
+
},
|
| 2428 |
+
{
|
| 2429 |
+
"epoch": 16.03,
|
| 2430 |
+
"learning_rate": 5.974025974025974e-05,
|
| 2431 |
+
"loss": 0.096,
|
| 2432 |
+
"step": 3720
|
| 2433 |
+
},
|
| 2434 |
+
{
|
| 2435 |
+
"epoch": 16.08,
|
| 2436 |
+
"learning_rate": 5.909090909090908e-05,
|
| 2437 |
+
"loss": 0.1205,
|
| 2438 |
+
"step": 3730
|
| 2439 |
+
},
|
| 2440 |
+
{
|
| 2441 |
+
"epoch": 16.12,
|
| 2442 |
+
"learning_rate": 5.8441558441558436e-05,
|
| 2443 |
+
"loss": 0.0864,
|
| 2444 |
+
"step": 3740
|
| 2445 |
+
},
|
| 2446 |
+
{
|
| 2447 |
+
"epoch": 16.16,
|
| 2448 |
+
"learning_rate": 5.779220779220779e-05,
|
| 2449 |
+
"loss": 0.1027,
|
| 2450 |
+
"step": 3750
|
| 2451 |
+
},
|
| 2452 |
+
{
|
| 2453 |
+
"epoch": 16.21,
|
| 2454 |
+
"learning_rate": 5.7142857142857135e-05,
|
| 2455 |
+
"loss": 0.0853,
|
| 2456 |
+
"step": 3760
|
| 2457 |
+
},
|
| 2458 |
+
{
|
| 2459 |
+
"epoch": 16.25,
|
| 2460 |
+
"learning_rate": 5.649350649350649e-05,
|
| 2461 |
+
"loss": 0.0866,
|
| 2462 |
+
"step": 3770
|
| 2463 |
+
},
|
| 2464 |
+
{
|
| 2465 |
+
"epoch": 16.29,
|
| 2466 |
+
"learning_rate": 5.584415584415584e-05,
|
| 2467 |
+
"loss": 0.0815,
|
| 2468 |
+
"step": 3780
|
| 2469 |
+
},
|
| 2470 |
+
{
|
| 2471 |
+
"epoch": 16.34,
|
| 2472 |
+
"learning_rate": 5.519480519480519e-05,
|
| 2473 |
+
"loss": 0.0971,
|
| 2474 |
+
"step": 3790
|
| 2475 |
+
},
|
| 2476 |
+
{
|
| 2477 |
+
"epoch": 16.38,
|
| 2478 |
+
"learning_rate": 5.454545454545454e-05,
|
| 2479 |
+
"loss": 0.0737,
|
| 2480 |
+
"step": 3800
|
| 2481 |
+
},
|
| 2482 |
+
{
|
| 2483 |
+
"epoch": 16.42,
|
| 2484 |
+
"learning_rate": 5.389610389610389e-05,
|
| 2485 |
+
"loss": 0.1072,
|
| 2486 |
+
"step": 3810
|
| 2487 |
+
},
|
| 2488 |
+
{
|
| 2489 |
+
"epoch": 16.46,
|
| 2490 |
+
"learning_rate": 5.3246753246753245e-05,
|
| 2491 |
+
"loss": 0.1073,
|
| 2492 |
+
"step": 3820
|
| 2493 |
+
},
|
| 2494 |
+
{
|
| 2495 |
+
"epoch": 16.51,
|
| 2496 |
+
"learning_rate": 5.259740259740259e-05,
|
| 2497 |
+
"loss": 0.0915,
|
| 2498 |
+
"step": 3830
|
| 2499 |
+
},
|
| 2500 |
+
{
|
| 2501 |
+
"epoch": 16.55,
|
| 2502 |
+
"learning_rate": 5.1948051948051944e-05,
|
| 2503 |
+
"loss": 0.1018,
|
| 2504 |
+
"step": 3840
|
| 2505 |
+
},
|
| 2506 |
+
{
|
| 2507 |
+
"epoch": 16.59,
|
| 2508 |
+
"learning_rate": 5.12987012987013e-05,
|
| 2509 |
+
"loss": 0.0941,
|
| 2510 |
+
"step": 3850
|
| 2511 |
+
},
|
| 2512 |
+
{
|
| 2513 |
+
"epoch": 16.64,
|
| 2514 |
+
"learning_rate": 5.064935064935064e-05,
|
| 2515 |
+
"loss": 0.0725,
|
| 2516 |
+
"step": 3860
|
| 2517 |
+
},
|
| 2518 |
+
{
|
| 2519 |
+
"epoch": 16.68,
|
| 2520 |
+
"learning_rate": 4.9999999999999996e-05,
|
| 2521 |
+
"loss": 0.0905,
|
| 2522 |
+
"step": 3870
|
| 2523 |
+
},
|
| 2524 |
+
{
|
| 2525 |
+
"epoch": 16.72,
|
| 2526 |
+
"learning_rate": 4.935064935064935e-05,
|
| 2527 |
+
"loss": 0.095,
|
| 2528 |
+
"step": 3880
|
| 2529 |
+
},
|
| 2530 |
+
{
|
| 2531 |
+
"epoch": 16.77,
|
| 2532 |
+
"learning_rate": 4.8701298701298694e-05,
|
| 2533 |
+
"loss": 0.0855,
|
| 2534 |
+
"step": 3890
|
| 2535 |
+
},
|
| 2536 |
+
{
|
| 2537 |
+
"epoch": 16.81,
|
| 2538 |
+
"learning_rate": 4.805194805194805e-05,
|
| 2539 |
+
"loss": 0.0754,
|
| 2540 |
+
"step": 3900
|
| 2541 |
+
},
|
| 2542 |
+
{
|
| 2543 |
+
"epoch": 16.85,
|
| 2544 |
+
"learning_rate": 4.74025974025974e-05,
|
| 2545 |
+
"loss": 0.0687,
|
| 2546 |
+
"step": 3910
|
| 2547 |
+
},
|
| 2548 |
+
{
|
| 2549 |
+
"epoch": 16.89,
|
| 2550 |
+
"learning_rate": 4.6753246753246746e-05,
|
| 2551 |
+
"loss": 0.0884,
|
| 2552 |
+
"step": 3920
|
| 2553 |
+
},
|
| 2554 |
+
{
|
| 2555 |
+
"epoch": 16.94,
|
| 2556 |
+
"learning_rate": 4.61038961038961e-05,
|
| 2557 |
+
"loss": 0.0887,
|
| 2558 |
+
"step": 3930
|
| 2559 |
+
},
|
| 2560 |
+
{
|
| 2561 |
+
"epoch": 16.98,
|
| 2562 |
+
"learning_rate": 4.545454545454545e-05,
|
| 2563 |
+
"loss": 0.1013,
|
| 2564 |
+
"step": 3940
|
| 2565 |
+
},
|
| 2566 |
+
{
|
| 2567 |
+
"epoch": 17.0,
|
| 2568 |
+
"eval_loss": 0.48812878131866455,
|
| 2569 |
+
"eval_runtime": 172.0812,
|
| 2570 |
+
"eval_samples_per_second": 8.845,
|
| 2571 |
+
"eval_wer": 0.44079943474311095,
|
| 2572 |
+
"step": 3944
|
| 2573 |
+
},
|
| 2574 |
+
{
|
| 2575 |
+
"epoch": 17.03,
|
| 2576 |
+
"learning_rate": 4.48051948051948e-05,
|
| 2577 |
+
"loss": 0.1091,
|
| 2578 |
+
"step": 3950
|
| 2579 |
+
},
|
| 2580 |
+
{
|
| 2581 |
+
"epoch": 17.07,
|
| 2582 |
+
"learning_rate": 4.415584415584416e-05,
|
| 2583 |
+
"loss": 0.0671,
|
| 2584 |
+
"step": 3960
|
| 2585 |
+
},
|
| 2586 |
+
{
|
| 2587 |
+
"epoch": 17.11,
|
| 2588 |
+
"learning_rate": 4.3506493506493503e-05,
|
| 2589 |
+
"loss": 0.1076,
|
| 2590 |
+
"step": 3970
|
| 2591 |
+
},
|
| 2592 |
+
{
|
| 2593 |
+
"epoch": 17.15,
|
| 2594 |
+
"learning_rate": 4.285714285714285e-05,
|
| 2595 |
+
"loss": 0.0881,
|
| 2596 |
+
"step": 3980
|
| 2597 |
+
},
|
| 2598 |
+
{
|
| 2599 |
+
"epoch": 17.2,
|
| 2600 |
+
"learning_rate": 4.220779220779221e-05,
|
| 2601 |
+
"loss": 0.0764,
|
| 2602 |
+
"step": 3990
|
| 2603 |
+
},
|
| 2604 |
+
{
|
| 2605 |
+
"epoch": 17.24,
|
| 2606 |
+
"learning_rate": 4.1558441558441555e-05,
|
| 2607 |
+
"loss": 0.0892,
|
| 2608 |
+
"step": 4000
|
| 2609 |
+
},
|
| 2610 |
+
{
|
| 2611 |
+
"epoch": 17.28,
|
| 2612 |
+
"learning_rate": 4.09090909090909e-05,
|
| 2613 |
+
"loss": 0.1031,
|
| 2614 |
+
"step": 4010
|
| 2615 |
+
},
|
| 2616 |
+
{
|
| 2617 |
+
"epoch": 17.33,
|
| 2618 |
+
"learning_rate": 4.025974025974026e-05,
|
| 2619 |
+
"loss": 0.102,
|
| 2620 |
+
"step": 4020
|
| 2621 |
+
},
|
| 2622 |
+
{
|
| 2623 |
+
"epoch": 17.37,
|
| 2624 |
+
"learning_rate": 3.961038961038961e-05,
|
| 2625 |
+
"loss": 0.0653,
|
| 2626 |
+
"step": 4030
|
| 2627 |
+
},
|
| 2628 |
+
{
|
| 2629 |
+
"epoch": 17.41,
|
| 2630 |
+
"learning_rate": 3.896103896103895e-05,
|
| 2631 |
+
"loss": 0.0822,
|
| 2632 |
+
"step": 4040
|
| 2633 |
+
},
|
| 2634 |
+
{
|
| 2635 |
+
"epoch": 17.46,
|
| 2636 |
+
"learning_rate": 3.831168831168831e-05,
|
| 2637 |
+
"loss": 0.0911,
|
| 2638 |
+
"step": 4050
|
| 2639 |
+
},
|
| 2640 |
+
{
|
| 2641 |
+
"epoch": 17.5,
|
| 2642 |
+
"learning_rate": 3.766233766233766e-05,
|
| 2643 |
+
"loss": 0.0743,
|
| 2644 |
+
"step": 4060
|
| 2645 |
+
},
|
| 2646 |
+
{
|
| 2647 |
+
"epoch": 17.54,
|
| 2648 |
+
"learning_rate": 3.701298701298701e-05,
|
| 2649 |
+
"loss": 0.0768,
|
| 2650 |
+
"step": 4070
|
| 2651 |
+
},
|
| 2652 |
+
{
|
| 2653 |
+
"epoch": 17.58,
|
| 2654 |
+
"learning_rate": 3.6363636363636364e-05,
|
| 2655 |
+
"loss": 0.0888,
|
| 2656 |
+
"step": 4080
|
| 2657 |
+
},
|
| 2658 |
+
{
|
| 2659 |
+
"epoch": 17.63,
|
| 2660 |
+
"learning_rate": 3.571428571428571e-05,
|
| 2661 |
+
"loss": 0.0753,
|
| 2662 |
+
"step": 4090
|
| 2663 |
+
},
|
| 2664 |
+
{
|
| 2665 |
+
"epoch": 17.67,
|
| 2666 |
+
"learning_rate": 3.506493506493506e-05,
|
| 2667 |
+
"loss": 0.0513,
|
| 2668 |
+
"step": 4100
|
| 2669 |
+
},
|
| 2670 |
+
{
|
| 2671 |
+
"epoch": 17.71,
|
| 2672 |
+
"learning_rate": 3.4415584415584416e-05,
|
| 2673 |
+
"loss": 0.0764,
|
| 2674 |
+
"step": 4110
|
| 2675 |
+
},
|
| 2676 |
+
{
|
| 2677 |
+
"epoch": 17.76,
|
| 2678 |
+
"learning_rate": 3.376623376623376e-05,
|
| 2679 |
+
"loss": 0.0788,
|
| 2680 |
+
"step": 4120
|
| 2681 |
+
},
|
| 2682 |
+
{
|
| 2683 |
+
"epoch": 17.8,
|
| 2684 |
+
"learning_rate": 3.3116883116883115e-05,
|
| 2685 |
+
"loss": 0.0754,
|
| 2686 |
+
"step": 4130
|
| 2687 |
+
},
|
| 2688 |
+
{
|
| 2689 |
+
"epoch": 17.84,
|
| 2690 |
+
"learning_rate": 3.246753246753247e-05,
|
| 2691 |
+
"loss": 0.1319,
|
| 2692 |
+
"step": 4140
|
| 2693 |
+
},
|
| 2694 |
+
{
|
| 2695 |
+
"epoch": 17.89,
|
| 2696 |
+
"learning_rate": 3.1818181818181814e-05,
|
| 2697 |
+
"loss": 0.0941,
|
| 2698 |
+
"step": 4150
|
| 2699 |
+
},
|
| 2700 |
+
{
|
| 2701 |
+
"epoch": 17.93,
|
| 2702 |
+
"learning_rate": 3.1168831168831166e-05,
|
| 2703 |
+
"loss": 0.0838,
|
| 2704 |
+
"step": 4160
|
| 2705 |
+
},
|
| 2706 |
+
{
|
| 2707 |
+
"epoch": 17.97,
|
| 2708 |
+
"learning_rate": 3.051948051948052e-05,
|
| 2709 |
+
"loss": 0.0669,
|
| 2710 |
+
"step": 4170
|
| 2711 |
+
},
|
| 2712 |
+
{
|
| 2713 |
+
"epoch": 18.0,
|
| 2714 |
+
"eval_loss": 0.47913026809692383,
|
| 2715 |
+
"eval_runtime": 175.3638,
|
| 2716 |
+
"eval_samples_per_second": 8.679,
|
| 2717 |
+
"eval_wer": 0.4378722115675785,
|
| 2718 |
+
"step": 4176
|
| 2719 |
+
},
|
| 2720 |
+
{
|
| 2721 |
+
"epoch": 18.02,
|
| 2722 |
+
"learning_rate": 2.987012987012987e-05,
|
| 2723 |
+
"loss": 0.0772,
|
| 2724 |
+
"step": 4180
|
| 2725 |
+
},
|
| 2726 |
+
{
|
| 2727 |
+
"epoch": 18.06,
|
| 2728 |
+
"learning_rate": 2.9220779220779218e-05,
|
| 2729 |
+
"loss": 0.0737,
|
| 2730 |
+
"step": 4190
|
| 2731 |
+
},
|
| 2732 |
+
{
|
| 2733 |
+
"epoch": 18.1,
|
| 2734 |
+
"learning_rate": 2.8571428571428567e-05,
|
| 2735 |
+
"loss": 0.098,
|
| 2736 |
+
"step": 4200
|
| 2737 |
+
},
|
| 2738 |
+
{
|
| 2739 |
+
"epoch": 18.15,
|
| 2740 |
+
"learning_rate": 2.792207792207792e-05,
|
| 2741 |
+
"loss": 0.0967,
|
| 2742 |
+
"step": 4210
|
| 2743 |
+
},
|
| 2744 |
+
{
|
| 2745 |
+
"epoch": 18.19,
|
| 2746 |
+
"learning_rate": 2.727272727272727e-05,
|
| 2747 |
+
"loss": 0.0766,
|
| 2748 |
+
"step": 4220
|
| 2749 |
+
},
|
| 2750 |
+
{
|
| 2751 |
+
"epoch": 18.23,
|
| 2752 |
+
"learning_rate": 2.6623376623376623e-05,
|
| 2753 |
+
"loss": 0.0758,
|
| 2754 |
+
"step": 4230
|
| 2755 |
+
},
|
| 2756 |
+
{
|
| 2757 |
+
"epoch": 18.28,
|
| 2758 |
+
"learning_rate": 2.5974025974025972e-05,
|
| 2759 |
+
"loss": 0.0734,
|
| 2760 |
+
"step": 4240
|
| 2761 |
+
},
|
| 2762 |
+
{
|
| 2763 |
+
"epoch": 18.32,
|
| 2764 |
+
"learning_rate": 2.532467532467532e-05,
|
| 2765 |
+
"loss": 0.0573,
|
| 2766 |
+
"step": 4250
|
| 2767 |
+
},
|
| 2768 |
+
{
|
| 2769 |
+
"epoch": 18.36,
|
| 2770 |
+
"learning_rate": 2.4675324675324674e-05,
|
| 2771 |
+
"loss": 0.0711,
|
| 2772 |
+
"step": 4260
|
| 2773 |
+
},
|
| 2774 |
+
{
|
| 2775 |
+
"epoch": 18.4,
|
| 2776 |
+
"learning_rate": 2.4025974025974024e-05,
|
| 2777 |
+
"loss": 0.0686,
|
| 2778 |
+
"step": 4270
|
| 2779 |
+
},
|
| 2780 |
+
{
|
| 2781 |
+
"epoch": 18.45,
|
| 2782 |
+
"learning_rate": 2.3376623376623373e-05,
|
| 2783 |
+
"loss": 0.0851,
|
| 2784 |
+
"step": 4280
|
| 2785 |
+
},
|
| 2786 |
+
{
|
| 2787 |
+
"epoch": 18.49,
|
| 2788 |
+
"learning_rate": 2.2727272727272726e-05,
|
| 2789 |
+
"loss": 0.075,
|
| 2790 |
+
"step": 4290
|
| 2791 |
+
},
|
| 2792 |
+
{
|
| 2793 |
+
"epoch": 18.53,
|
| 2794 |
+
"learning_rate": 2.207792207792208e-05,
|
| 2795 |
+
"loss": 0.0806,
|
| 2796 |
+
"step": 4300
|
| 2797 |
+
},
|
| 2798 |
+
{
|
| 2799 |
+
"epoch": 18.58,
|
| 2800 |
+
"learning_rate": 2.1428571428571425e-05,
|
| 2801 |
+
"loss": 0.0751,
|
| 2802 |
+
"step": 4310
|
| 2803 |
+
},
|
| 2804 |
+
{
|
| 2805 |
+
"epoch": 18.62,
|
| 2806 |
+
"learning_rate": 2.0779220779220778e-05,
|
| 2807 |
+
"loss": 0.0741,
|
| 2808 |
+
"step": 4320
|
| 2809 |
+
},
|
| 2810 |
+
{
|
| 2811 |
+
"epoch": 18.66,
|
| 2812 |
+
"learning_rate": 2.012987012987013e-05,
|
| 2813 |
+
"loss": 0.0683,
|
| 2814 |
+
"step": 4330
|
| 2815 |
+
},
|
| 2816 |
+
{
|
| 2817 |
+
"epoch": 18.71,
|
| 2818 |
+
"learning_rate": 1.9480519480519476e-05,
|
| 2819 |
+
"loss": 0.0853,
|
| 2820 |
+
"step": 4340
|
| 2821 |
+
},
|
| 2822 |
+
{
|
| 2823 |
+
"epoch": 18.75,
|
| 2824 |
+
"learning_rate": 1.883116883116883e-05,
|
| 2825 |
+
"loss": 0.0759,
|
| 2826 |
+
"step": 4350
|
| 2827 |
+
},
|
| 2828 |
+
{
|
| 2829 |
+
"epoch": 18.79,
|
| 2830 |
+
"learning_rate": 1.8181818181818182e-05,
|
| 2831 |
+
"loss": 0.0749,
|
| 2832 |
+
"step": 4360
|
| 2833 |
+
},
|
| 2834 |
+
{
|
| 2835 |
+
"epoch": 18.83,
|
| 2836 |
+
"learning_rate": 1.753246753246753e-05,
|
| 2837 |
+
"loss": 0.0884,
|
| 2838 |
+
"step": 4370
|
| 2839 |
+
},
|
| 2840 |
+
{
|
| 2841 |
+
"epoch": 18.88,
|
| 2842 |
+
"learning_rate": 1.688311688311688e-05,
|
| 2843 |
+
"loss": 0.0657,
|
| 2844 |
+
"step": 4380
|
| 2845 |
+
},
|
| 2846 |
+
{
|
| 2847 |
+
"epoch": 18.92,
|
| 2848 |
+
"learning_rate": 1.6233766233766234e-05,
|
| 2849 |
+
"loss": 0.0698,
|
| 2850 |
+
"step": 4390
|
| 2851 |
+
},
|
| 2852 |
+
{
|
| 2853 |
+
"epoch": 18.96,
|
| 2854 |
+
"learning_rate": 1.5584415584415583e-05,
|
| 2855 |
+
"loss": 0.0723,
|
| 2856 |
+
"step": 4400
|
| 2857 |
+
},
|
| 2858 |
+
{
|
| 2859 |
+
"epoch": 19.0,
|
| 2860 |
+
"eval_loss": 0.4765560030937195,
|
| 2861 |
+
"eval_runtime": 176.1902,
|
| 2862 |
+
"eval_samples_per_second": 8.638,
|
| 2863 |
+
"eval_wer": 0.43716564045624307,
|
| 2864 |
+
"step": 4408
|
| 2865 |
+
},
|
| 2866 |
+
{
|
| 2867 |
+
"epoch": 19.01,
|
| 2868 |
+
"learning_rate": 1.4935064935064934e-05,
|
| 2869 |
+
"loss": 0.0774,
|
| 2870 |
+
"step": 4410
|
| 2871 |
+
},
|
| 2872 |
+
{
|
| 2873 |
+
"epoch": 19.05,
|
| 2874 |
+
"learning_rate": 1.4285714285714284e-05,
|
| 2875 |
+
"loss": 0.0721,
|
| 2876 |
+
"step": 4420
|
| 2877 |
+
},
|
| 2878 |
+
{
|
| 2879 |
+
"epoch": 19.09,
|
| 2880 |
+
"learning_rate": 1.3636363636363635e-05,
|
| 2881 |
+
"loss": 0.0752,
|
| 2882 |
+
"step": 4430
|
| 2883 |
+
},
|
| 2884 |
+
{
|
| 2885 |
+
"epoch": 19.14,
|
| 2886 |
+
"learning_rate": 1.2987012987012986e-05,
|
| 2887 |
+
"loss": 0.0556,
|
| 2888 |
+
"step": 4440
|
| 2889 |
+
},
|
| 2890 |
+
{
|
| 2891 |
+
"epoch": 19.18,
|
| 2892 |
+
"learning_rate": 1.2337662337662337e-05,
|
| 2893 |
+
"loss": 0.0749,
|
| 2894 |
+
"step": 4450
|
| 2895 |
+
},
|
| 2896 |
+
{
|
| 2897 |
+
"epoch": 19.22,
|
| 2898 |
+
"learning_rate": 1.1688311688311687e-05,
|
| 2899 |
+
"loss": 0.0645,
|
| 2900 |
+
"step": 4460
|
| 2901 |
+
},
|
| 2902 |
+
{
|
| 2903 |
+
"epoch": 19.27,
|
| 2904 |
+
"learning_rate": 1.103896103896104e-05,
|
| 2905 |
+
"loss": 0.0757,
|
| 2906 |
+
"step": 4470
|
| 2907 |
+
},
|
| 2908 |
+
{
|
| 2909 |
+
"epoch": 19.31,
|
| 2910 |
+
"learning_rate": 1.0389610389610389e-05,
|
| 2911 |
+
"loss": 0.0726,
|
| 2912 |
+
"step": 4480
|
| 2913 |
+
},
|
| 2914 |
+
{
|
| 2915 |
+
"epoch": 19.35,
|
| 2916 |
+
"learning_rate": 9.740259740259738e-06,
|
| 2917 |
+
"loss": 0.073,
|
| 2918 |
+
"step": 4490
|
| 2919 |
+
},
|
| 2920 |
+
{
|
| 2921 |
+
"epoch": 19.4,
|
| 2922 |
+
"learning_rate": 9.090909090909091e-06,
|
| 2923 |
+
"loss": 0.0711,
|
| 2924 |
+
"step": 4500
|
| 2925 |
+
},
|
| 2926 |
+
{
|
| 2927 |
+
"epoch": 19.44,
|
| 2928 |
+
"learning_rate": 8.44155844155844e-06,
|
| 2929 |
+
"loss": 0.0813,
|
| 2930 |
+
"step": 4510
|
| 2931 |
+
},
|
| 2932 |
+
{
|
| 2933 |
+
"epoch": 19.48,
|
| 2934 |
+
"learning_rate": 7.792207792207792e-06,
|
| 2935 |
+
"loss": 0.0847,
|
| 2936 |
+
"step": 4520
|
| 2937 |
+
},
|
| 2938 |
+
{
|
| 2939 |
+
"epoch": 19.52,
|
| 2940 |
+
"learning_rate": 7.142857142857142e-06,
|
| 2941 |
+
"loss": 0.1081,
|
| 2942 |
+
"step": 4530
|
| 2943 |
+
},
|
| 2944 |
+
{
|
| 2945 |
+
"epoch": 19.57,
|
| 2946 |
+
"learning_rate": 6.493506493506493e-06,
|
| 2947 |
+
"loss": 0.0694,
|
| 2948 |
+
"step": 4540
|
| 2949 |
+
},
|
| 2950 |
+
{
|
| 2951 |
+
"epoch": 19.61,
|
| 2952 |
+
"learning_rate": 5.844155844155843e-06,
|
| 2953 |
+
"loss": 0.0687,
|
| 2954 |
+
"step": 4550
|
| 2955 |
+
},
|
| 2956 |
+
{
|
| 2957 |
+
"epoch": 19.65,
|
| 2958 |
+
"learning_rate": 5.194805194805194e-06,
|
| 2959 |
+
"loss": 0.1179,
|
| 2960 |
+
"step": 4560
|
| 2961 |
+
},
|
| 2962 |
+
{
|
| 2963 |
+
"epoch": 19.7,
|
| 2964 |
+
"learning_rate": 4.5454545454545455e-06,
|
| 2965 |
+
"loss": 0.0648,
|
| 2966 |
+
"step": 4570
|
| 2967 |
+
},
|
| 2968 |
+
{
|
| 2969 |
+
"epoch": 19.74,
|
| 2970 |
+
"learning_rate": 3.896103896103896e-06,
|
| 2971 |
+
"loss": 0.0782,
|
| 2972 |
+
"step": 4580
|
| 2973 |
+
},
|
| 2974 |
+
{
|
| 2975 |
+
"epoch": 19.78,
|
| 2976 |
+
"learning_rate": 3.2467532467532465e-06,
|
| 2977 |
+
"loss": 0.063,
|
| 2978 |
+
"step": 4590
|
| 2979 |
+
},
|
| 2980 |
+
{
|
| 2981 |
+
"epoch": 19.83,
|
| 2982 |
+
"learning_rate": 2.597402597402597e-06,
|
| 2983 |
+
"loss": 0.0658,
|
| 2984 |
+
"step": 4600
|
| 2985 |
+
},
|
| 2986 |
+
{
|
| 2987 |
+
"epoch": 19.87,
|
| 2988 |
+
"learning_rate": 1.948051948051948e-06,
|
| 2989 |
+
"loss": 0.0613,
|
| 2990 |
+
"step": 4610
|
| 2991 |
+
},
|
| 2992 |
+
{
|
| 2993 |
+
"epoch": 19.91,
|
| 2994 |
+
"learning_rate": 1.2987012987012986e-06,
|
| 2995 |
+
"loss": 0.0873,
|
| 2996 |
+
"step": 4620
|
| 2997 |
+
},
|
| 2998 |
+
{
|
| 2999 |
+
"epoch": 19.95,
|
| 3000 |
+
"learning_rate": 6.493506493506493e-07,
|
| 3001 |
+
"loss": 0.1031,
|
| 3002 |
+
"step": 4630
|
| 3003 |
+
},
|
| 3004 |
+
{
|
| 3005 |
+
"epoch": 20.0,
|
| 3006 |
+
"learning_rate": 0.0,
|
| 3007 |
+
"loss": 0.0675,
|
| 3008 |
+
"step": 4640
|
| 3009 |
+
},
|
| 3010 |
+
{
|
| 3011 |
+
"epoch": 20.0,
|
| 3012 |
+
"eval_loss": 0.4751843214035034,
|
| 3013 |
+
"eval_runtime": 177.6725,
|
| 3014 |
+
"eval_samples_per_second": 8.566,
|
| 3015 |
+
"eval_wer": 0.4374684566468154,
|
| 3016 |
+
"step": 4640
|
| 3017 |
+
},
|
| 3018 |
+
{
|
| 3019 |
+
"epoch": 20.04,
|
| 3020 |
+
"learning_rate": 5.9688581314878885e-05,
|
| 3021 |
+
"loss": 0.0605,
|
| 3022 |
+
"step": 4650
|
| 3023 |
+
},
|
| 3024 |
+
{
|
| 3025 |
+
"epoch": 20.09,
|
| 3026 |
+
"learning_rate": 5.9169550173010376e-05,
|
| 3027 |
+
"loss": 0.0568,
|
| 3028 |
+
"step": 4660
|
| 3029 |
+
},
|
| 3030 |
+
{
|
| 3031 |
+
"epoch": 20.13,
|
| 3032 |
+
"learning_rate": 5.865051903114186e-05,
|
| 3033 |
+
"loss": 0.0844,
|
| 3034 |
+
"step": 4670
|
| 3035 |
+
},
|
| 3036 |
+
{
|
| 3037 |
+
"epoch": 20.17,
|
| 3038 |
+
"learning_rate": 5.813148788927335e-05,
|
| 3039 |
+
"loss": 0.0892,
|
| 3040 |
+
"step": 4680
|
| 3041 |
+
},
|
| 3042 |
+
{
|
| 3043 |
+
"epoch": 20.22,
|
| 3044 |
+
"learning_rate": 5.761245674740484e-05,
|
| 3045 |
+
"loss": 0.0686,
|
| 3046 |
+
"step": 4690
|
| 3047 |
+
},
|
| 3048 |
+
{
|
| 3049 |
+
"epoch": 20.26,
|
| 3050 |
+
"learning_rate": 5.7093425605536325e-05,
|
| 3051 |
+
"loss": 0.0704,
|
| 3052 |
+
"step": 4700
|
| 3053 |
+
},
|
| 3054 |
+
{
|
| 3055 |
+
"epoch": 20.3,
|
| 3056 |
+
"learning_rate": 5.6574394463667816e-05,
|
| 3057 |
+
"loss": 0.0687,
|
| 3058 |
+
"step": 4710
|
| 3059 |
+
},
|
| 3060 |
+
{
|
| 3061 |
+
"epoch": 20.34,
|
| 3062 |
+
"learning_rate": 5.605536332179931e-05,
|
| 3063 |
+
"loss": 0.1011,
|
| 3064 |
+
"step": 4720
|
| 3065 |
+
},
|
| 3066 |
+
{
|
| 3067 |
+
"epoch": 20.39,
|
| 3068 |
+
"learning_rate": 5.553633217993079e-05,
|
| 3069 |
+
"loss": 0.0844,
|
| 3070 |
+
"step": 4730
|
| 3071 |
+
},
|
| 3072 |
+
{
|
| 3073 |
+
"epoch": 20.43,
|
| 3074 |
+
"learning_rate": 5.501730103806228e-05,
|
| 3075 |
+
"loss": 0.0722,
|
| 3076 |
+
"step": 4740
|
| 3077 |
+
},
|
| 3078 |
+
{
|
| 3079 |
+
"epoch": 20.47,
|
| 3080 |
+
"learning_rate": 5.449826989619377e-05,
|
| 3081 |
+
"loss": 0.057,
|
| 3082 |
+
"step": 4750
|
| 3083 |
+
},
|
| 3084 |
+
{
|
| 3085 |
+
"epoch": 20.52,
|
| 3086 |
+
"learning_rate": 5.3979238754325256e-05,
|
| 3087 |
+
"loss": 0.0596,
|
| 3088 |
+
"step": 4760
|
| 3089 |
+
},
|
| 3090 |
+
{
|
| 3091 |
+
"epoch": 20.56,
|
| 3092 |
+
"learning_rate": 5.346020761245675e-05,
|
| 3093 |
+
"loss": 0.0786,
|
| 3094 |
+
"step": 4770
|
| 3095 |
+
},
|
| 3096 |
+
{
|
| 3097 |
+
"epoch": 20.6,
|
| 3098 |
+
"learning_rate": 5.294117647058824e-05,
|
| 3099 |
+
"loss": 0.0835,
|
| 3100 |
+
"step": 4780
|
| 3101 |
+
},
|
| 3102 |
+
{
|
| 3103 |
+
"epoch": 20.65,
|
| 3104 |
+
"learning_rate": 5.242214532871972e-05,
|
| 3105 |
+
"loss": 0.066,
|
| 3106 |
+
"step": 4790
|
| 3107 |
+
},
|
| 3108 |
+
{
|
| 3109 |
+
"epoch": 20.69,
|
| 3110 |
+
"learning_rate": 5.1903114186851205e-05,
|
| 3111 |
+
"loss": 0.0721,
|
| 3112 |
+
"step": 4800
|
| 3113 |
+
},
|
| 3114 |
+
{
|
| 3115 |
+
"epoch": 20.73,
|
| 3116 |
+
"learning_rate": 5.138408304498269e-05,
|
| 3117 |
+
"loss": 0.0978,
|
| 3118 |
+
"step": 4810
|
| 3119 |
+
},
|
| 3120 |
+
{
|
| 3121 |
+
"epoch": 20.77,
|
| 3122 |
+
"learning_rate": 5.086505190311418e-05,
|
| 3123 |
+
"loss": 0.0893,
|
| 3124 |
+
"step": 4820
|
| 3125 |
+
},
|
| 3126 |
+
{
|
| 3127 |
+
"epoch": 20.82,
|
| 3128 |
+
"learning_rate": 5.0346020761245664e-05,
|
| 3129 |
+
"loss": 0.0723,
|
| 3130 |
+
"step": 4830
|
| 3131 |
+
},
|
| 3132 |
+
{
|
| 3133 |
+
"epoch": 20.86,
|
| 3134 |
+
"learning_rate": 4.9826989619377155e-05,
|
| 3135 |
+
"loss": 0.0697,
|
| 3136 |
+
"step": 4840
|
| 3137 |
+
},
|
| 3138 |
+
{
|
| 3139 |
+
"epoch": 20.9,
|
| 3140 |
+
"learning_rate": 4.9307958477508645e-05,
|
| 3141 |
+
"loss": 0.0928,
|
| 3142 |
+
"step": 4850
|
| 3143 |
+
},
|
| 3144 |
+
{
|
| 3145 |
+
"epoch": 20.95,
|
| 3146 |
+
"learning_rate": 4.878892733564013e-05,
|
| 3147 |
+
"loss": 0.0894,
|
| 3148 |
+
"step": 4860
|
| 3149 |
+
},
|
| 3150 |
+
{
|
| 3151 |
+
"epoch": 20.99,
|
| 3152 |
+
"learning_rate": 4.826989619377162e-05,
|
| 3153 |
+
"loss": 0.0782,
|
| 3154 |
+
"step": 4870
|
| 3155 |
+
},
|
| 3156 |
+
{
|
| 3157 |
+
"epoch": 21.0,
|
| 3158 |
+
"eval_loss": 0.46749043464660645,
|
| 3159 |
+
"eval_runtime": 156.5519,
|
| 3160 |
+
"eval_samples_per_second": 9.722,
|
| 3161 |
+
"eval_wer": 0.4334309074391844,
|
| 3162 |
+
"step": 4872
|
| 3163 |
+
},
|
| 3164 |
+
{
|
| 3165 |
+
"epoch": 21.03,
|
| 3166 |
+
"learning_rate": 4.775086505190311e-05,
|
| 3167 |
+
"loss": 0.1052,
|
| 3168 |
+
"step": 4880
|
| 3169 |
+
},
|
| 3170 |
+
{
|
| 3171 |
+
"epoch": 21.08,
|
| 3172 |
+
"learning_rate": 4.7231833910034595e-05,
|
| 3173 |
+
"loss": 0.0648,
|
| 3174 |
+
"step": 4890
|
| 3175 |
+
},
|
| 3176 |
+
{
|
| 3177 |
+
"epoch": 21.12,
|
| 3178 |
+
"learning_rate": 4.6712802768166085e-05,
|
| 3179 |
+
"loss": 0.0676,
|
| 3180 |
+
"step": 4900
|
| 3181 |
+
},
|
| 3182 |
+
{
|
| 3183 |
+
"epoch": 21.16,
|
| 3184 |
+
"learning_rate": 4.6193771626297576e-05,
|
| 3185 |
+
"loss": 0.0642,
|
| 3186 |
+
"step": 4910
|
| 3187 |
+
},
|
| 3188 |
+
{
|
| 3189 |
+
"epoch": 21.21,
|
| 3190 |
+
"learning_rate": 4.567474048442906e-05,
|
| 3191 |
+
"loss": 0.0926,
|
| 3192 |
+
"step": 4920
|
| 3193 |
+
},
|
| 3194 |
+
{
|
| 3195 |
+
"epoch": 21.25,
|
| 3196 |
+
"learning_rate": 4.515570934256055e-05,
|
| 3197 |
+
"loss": 0.0781,
|
| 3198 |
+
"step": 4930
|
| 3199 |
+
},
|
| 3200 |
+
{
|
| 3201 |
+
"epoch": 21.29,
|
| 3202 |
+
"learning_rate": 4.463667820069204e-05,
|
| 3203 |
+
"loss": 0.0668,
|
| 3204 |
+
"step": 4940
|
| 3205 |
+
},
|
| 3206 |
+
{
|
| 3207 |
+
"epoch": 21.34,
|
| 3208 |
+
"learning_rate": 4.4117647058823526e-05,
|
| 3209 |
+
"loss": 0.0725,
|
| 3210 |
+
"step": 4950
|
| 3211 |
+
},
|
| 3212 |
+
{
|
| 3213 |
+
"epoch": 21.38,
|
| 3214 |
+
"learning_rate": 4.3598615916955016e-05,
|
| 3215 |
+
"loss": 0.0785,
|
| 3216 |
+
"step": 4960
|
| 3217 |
+
},
|
| 3218 |
+
{
|
| 3219 |
+
"epoch": 21.42,
|
| 3220 |
+
"learning_rate": 4.307958477508651e-05,
|
| 3221 |
+
"loss": 0.057,
|
| 3222 |
+
"step": 4970
|
| 3223 |
+
},
|
| 3224 |
+
{
|
| 3225 |
+
"epoch": 21.46,
|
| 3226 |
+
"learning_rate": 4.256055363321799e-05,
|
| 3227 |
+
"loss": 0.0659,
|
| 3228 |
+
"step": 4980
|
| 3229 |
+
},
|
| 3230 |
+
{
|
| 3231 |
+
"epoch": 21.51,
|
| 3232 |
+
"learning_rate": 4.204152249134948e-05,
|
| 3233 |
+
"loss": 0.0864,
|
| 3234 |
+
"step": 4990
|
| 3235 |
+
},
|
| 3236 |
+
{
|
| 3237 |
+
"epoch": 21.55,
|
| 3238 |
+
"learning_rate": 4.1522491349480966e-05,
|
| 3239 |
+
"loss": 0.0731,
|
| 3240 |
+
"step": 5000
|
| 3241 |
+
},
|
| 3242 |
+
{
|
| 3243 |
+
"epoch": 21.59,
|
| 3244 |
+
"learning_rate": 4.1003460207612456e-05,
|
| 3245 |
+
"loss": 0.0744,
|
| 3246 |
+
"step": 5010
|
| 3247 |
+
},
|
| 3248 |
+
{
|
| 3249 |
+
"epoch": 21.64,
|
| 3250 |
+
"learning_rate": 4.048442906574395e-05,
|
| 3251 |
+
"loss": 0.0769,
|
| 3252 |
+
"step": 5020
|
| 3253 |
+
},
|
| 3254 |
+
{
|
| 3255 |
+
"epoch": 21.68,
|
| 3256 |
+
"learning_rate": 3.9965397923875424e-05,
|
| 3257 |
+
"loss": 0.079,
|
| 3258 |
+
"step": 5030
|
| 3259 |
+
},
|
| 3260 |
+
{
|
| 3261 |
+
"epoch": 21.72,
|
| 3262 |
+
"learning_rate": 3.9446366782006915e-05,
|
| 3263 |
+
"loss": 0.0875,
|
| 3264 |
+
"step": 5040
|
| 3265 |
+
},
|
| 3266 |
+
{
|
| 3267 |
+
"epoch": 21.77,
|
| 3268 |
+
"learning_rate": 3.89273356401384e-05,
|
| 3269 |
+
"loss": 0.0705,
|
| 3270 |
+
"step": 5050
|
| 3271 |
+
},
|
| 3272 |
+
{
|
| 3273 |
+
"epoch": 21.81,
|
| 3274 |
+
"learning_rate": 3.840830449826989e-05,
|
| 3275 |
+
"loss": 0.0723,
|
| 3276 |
+
"step": 5060
|
| 3277 |
+
},
|
| 3278 |
+
{
|
| 3279 |
+
"epoch": 21.85,
|
| 3280 |
+
"learning_rate": 3.788927335640138e-05,
|
| 3281 |
+
"loss": 0.0833,
|
| 3282 |
+
"step": 5070
|
| 3283 |
+
},
|
| 3284 |
+
{
|
| 3285 |
+
"epoch": 21.89,
|
| 3286 |
+
"learning_rate": 3.7370242214532864e-05,
|
| 3287 |
+
"loss": 0.0914,
|
| 3288 |
+
"step": 5080
|
| 3289 |
+
},
|
| 3290 |
+
{
|
| 3291 |
+
"epoch": 21.94,
|
| 3292 |
+
"learning_rate": 3.6851211072664355e-05,
|
| 3293 |
+
"loss": 0.0754,
|
| 3294 |
+
"step": 5090
|
| 3295 |
+
},
|
| 3296 |
+
{
|
| 3297 |
+
"epoch": 21.98,
|
| 3298 |
+
"learning_rate": 3.6332179930795846e-05,
|
| 3299 |
+
"loss": 0.0848,
|
| 3300 |
+
"step": 5100
|
| 3301 |
+
},
|
| 3302 |
+
{
|
| 3303 |
+
"epoch": 22.0,
|
| 3304 |
+
"eval_loss": 0.4894248843193054,
|
| 3305 |
+
"eval_runtime": 156.3927,
|
| 3306 |
+
"eval_samples_per_second": 9.732,
|
| 3307 |
+
"eval_wer": 0.4378722115675785,
|
| 3308 |
+
"step": 5104
|
| 3309 |
+
},
|
| 3310 |
+
{
|
| 3311 |
+
"epoch": 22.03,
|
| 3312 |
+
"learning_rate": 3.581314878892733e-05,
|
| 3313 |
+
"loss": 0.0783,
|
| 3314 |
+
"step": 5110
|
| 3315 |
+
},
|
| 3316 |
+
{
|
| 3317 |
+
"epoch": 22.07,
|
| 3318 |
+
"learning_rate": 3.529411764705882e-05,
|
| 3319 |
+
"loss": 0.0698,
|
| 3320 |
+
"step": 5120
|
| 3321 |
+
},
|
| 3322 |
+
{
|
| 3323 |
+
"epoch": 22.11,
|
| 3324 |
+
"learning_rate": 3.477508650519031e-05,
|
| 3325 |
+
"loss": 0.0885,
|
| 3326 |
+
"step": 5130
|
| 3327 |
+
},
|
| 3328 |
+
{
|
| 3329 |
+
"epoch": 22.15,
|
| 3330 |
+
"learning_rate": 3.4256055363321795e-05,
|
| 3331 |
+
"loss": 0.0613,
|
| 3332 |
+
"step": 5140
|
| 3333 |
+
},
|
| 3334 |
+
{
|
| 3335 |
+
"epoch": 22.2,
|
| 3336 |
+
"learning_rate": 3.3737024221453286e-05,
|
| 3337 |
+
"loss": 0.0567,
|
| 3338 |
+
"step": 5150
|
| 3339 |
+
},
|
| 3340 |
+
{
|
| 3341 |
+
"epoch": 22.24,
|
| 3342 |
+
"learning_rate": 3.3217993079584777e-05,
|
| 3343 |
+
"loss": 0.0905,
|
| 3344 |
+
"step": 5160
|
| 3345 |
+
},
|
| 3346 |
+
{
|
| 3347 |
+
"epoch": 22.28,
|
| 3348 |
+
"learning_rate": 3.269896193771626e-05,
|
| 3349 |
+
"loss": 0.0865,
|
| 3350 |
+
"step": 5170
|
| 3351 |
+
},
|
| 3352 |
+
{
|
| 3353 |
+
"epoch": 22.33,
|
| 3354 |
+
"learning_rate": 3.217993079584775e-05,
|
| 3355 |
+
"loss": 0.0881,
|
| 3356 |
+
"step": 5180
|
| 3357 |
+
},
|
| 3358 |
+
{
|
| 3359 |
+
"epoch": 22.37,
|
| 3360 |
+
"learning_rate": 3.1660899653979235e-05,
|
| 3361 |
+
"loss": 0.0702,
|
| 3362 |
+
"step": 5190
|
| 3363 |
+
},
|
| 3364 |
+
{
|
| 3365 |
+
"epoch": 22.41,
|
| 3366 |
+
"learning_rate": 3.114186851211072e-05,
|
| 3367 |
+
"loss": 0.0688,
|
| 3368 |
+
"step": 5200
|
| 3369 |
+
},
|
| 3370 |
+
{
|
| 3371 |
+
"epoch": 22.46,
|
| 3372 |
+
"learning_rate": 3.062283737024221e-05,
|
| 3373 |
+
"loss": 0.0713,
|
| 3374 |
+
"step": 5210
|
| 3375 |
+
},
|
| 3376 |
+
{
|
| 3377 |
+
"epoch": 22.5,
|
| 3378 |
+
"learning_rate": 3.0103806228373697e-05,
|
| 3379 |
+
"loss": 0.0615,
|
| 3380 |
+
"step": 5220
|
| 3381 |
+
},
|
| 3382 |
+
{
|
| 3383 |
+
"epoch": 22.54,
|
| 3384 |
+
"learning_rate": 2.9584775086505188e-05,
|
| 3385 |
+
"loss": 0.0761,
|
| 3386 |
+
"step": 5230
|
| 3387 |
+
},
|
| 3388 |
+
{
|
| 3389 |
+
"epoch": 22.58,
|
| 3390 |
+
"learning_rate": 2.9065743944636675e-05,
|
| 3391 |
+
"loss": 0.0751,
|
| 3392 |
+
"step": 5240
|
| 3393 |
+
},
|
| 3394 |
+
{
|
| 3395 |
+
"epoch": 22.63,
|
| 3396 |
+
"learning_rate": 2.8546712802768163e-05,
|
| 3397 |
+
"loss": 0.0767,
|
| 3398 |
+
"step": 5250
|
| 3399 |
+
},
|
| 3400 |
+
{
|
| 3401 |
+
"epoch": 22.67,
|
| 3402 |
+
"learning_rate": 2.8027681660899653e-05,
|
| 3403 |
+
"loss": 0.0765,
|
| 3404 |
+
"step": 5260
|
| 3405 |
+
},
|
| 3406 |
+
{
|
| 3407 |
+
"epoch": 22.71,
|
| 3408 |
+
"learning_rate": 2.750865051903114e-05,
|
| 3409 |
+
"loss": 0.0652,
|
| 3410 |
+
"step": 5270
|
| 3411 |
+
},
|
| 3412 |
+
{
|
| 3413 |
+
"epoch": 22.76,
|
| 3414 |
+
"learning_rate": 2.6989619377162628e-05,
|
| 3415 |
+
"loss": 0.077,
|
| 3416 |
+
"step": 5280
|
| 3417 |
+
},
|
| 3418 |
+
{
|
| 3419 |
+
"epoch": 22.8,
|
| 3420 |
+
"learning_rate": 2.647058823529412e-05,
|
| 3421 |
+
"loss": 0.0643,
|
| 3422 |
+
"step": 5290
|
| 3423 |
+
},
|
| 3424 |
+
{
|
| 3425 |
+
"epoch": 22.84,
|
| 3426 |
+
"learning_rate": 2.5951557093425603e-05,
|
| 3427 |
+
"loss": 0.0646,
|
| 3428 |
+
"step": 5300
|
| 3429 |
+
},
|
| 3430 |
+
{
|
| 3431 |
+
"epoch": 22.89,
|
| 3432 |
+
"learning_rate": 2.543252595155709e-05,
|
| 3433 |
+
"loss": 0.084,
|
| 3434 |
+
"step": 5310
|
| 3435 |
+
},
|
| 3436 |
+
{
|
| 3437 |
+
"epoch": 22.93,
|
| 3438 |
+
"learning_rate": 2.4913494809688577e-05,
|
| 3439 |
+
"loss": 0.0615,
|
| 3440 |
+
"step": 5320
|
| 3441 |
+
},
|
| 3442 |
+
{
|
| 3443 |
+
"epoch": 22.97,
|
| 3444 |
+
"learning_rate": 2.4394463667820065e-05,
|
| 3445 |
+
"loss": 0.0859,
|
| 3446 |
+
"step": 5330
|
| 3447 |
+
},
|
| 3448 |
+
{
|
| 3449 |
+
"epoch": 23.0,
|
| 3450 |
+
"eval_loss": 0.4741663634777069,
|
| 3451 |
+
"eval_runtime": 158.2043,
|
| 3452 |
+
"eval_samples_per_second": 9.62,
|
| 3453 |
+
"eval_wer": 0.43252245886746743,
|
| 3454 |
+
"step": 5336
|
| 3455 |
+
},
|
| 3456 |
+
{
|
| 3457 |
+
"epoch": 23.02,
|
| 3458 |
+
"learning_rate": 2.3875432525951555e-05,
|
| 3459 |
+
"loss": 0.085,
|
| 3460 |
+
"step": 5340
|
| 3461 |
+
},
|
| 3462 |
+
{
|
| 3463 |
+
"epoch": 23.06,
|
| 3464 |
+
"learning_rate": 2.3356401384083043e-05,
|
| 3465 |
+
"loss": 0.0685,
|
| 3466 |
+
"step": 5350
|
| 3467 |
+
},
|
| 3468 |
+
{
|
| 3469 |
+
"epoch": 23.1,
|
| 3470 |
+
"learning_rate": 2.283737024221453e-05,
|
| 3471 |
+
"loss": 0.0572,
|
| 3472 |
+
"step": 5360
|
| 3473 |
+
},
|
| 3474 |
+
{
|
| 3475 |
+
"epoch": 23.15,
|
| 3476 |
+
"learning_rate": 2.231833910034602e-05,
|
| 3477 |
+
"loss": 0.0752,
|
| 3478 |
+
"step": 5370
|
| 3479 |
+
},
|
| 3480 |
+
{
|
| 3481 |
+
"epoch": 23.19,
|
| 3482 |
+
"learning_rate": 2.1799307958477508e-05,
|
| 3483 |
+
"loss": 0.084,
|
| 3484 |
+
"step": 5380
|
| 3485 |
+
},
|
| 3486 |
+
{
|
| 3487 |
+
"epoch": 23.23,
|
| 3488 |
+
"learning_rate": 2.1280276816608995e-05,
|
| 3489 |
+
"loss": 0.0603,
|
| 3490 |
+
"step": 5390
|
| 3491 |
+
},
|
| 3492 |
+
{
|
| 3493 |
+
"epoch": 23.28,
|
| 3494 |
+
"learning_rate": 2.0761245674740483e-05,
|
| 3495 |
+
"loss": 0.0702,
|
| 3496 |
+
"step": 5400
|
| 3497 |
+
},
|
| 3498 |
+
{
|
| 3499 |
+
"epoch": 23.32,
|
| 3500 |
+
"learning_rate": 2.0242214532871974e-05,
|
| 3501 |
+
"loss": 0.063,
|
| 3502 |
+
"step": 5410
|
| 3503 |
+
},
|
| 3504 |
+
{
|
| 3505 |
+
"epoch": 23.36,
|
| 3506 |
+
"learning_rate": 1.9723183391003457e-05,
|
| 3507 |
+
"loss": 0.0558,
|
| 3508 |
+
"step": 5420
|
| 3509 |
+
},
|
| 3510 |
+
{
|
| 3511 |
+
"epoch": 23.4,
|
| 3512 |
+
"learning_rate": 1.9204152249134945e-05,
|
| 3513 |
+
"loss": 0.0848,
|
| 3514 |
+
"step": 5430
|
| 3515 |
+
},
|
| 3516 |
+
{
|
| 3517 |
+
"epoch": 23.45,
|
| 3518 |
+
"learning_rate": 1.8685121107266432e-05,
|
| 3519 |
+
"loss": 0.0613,
|
| 3520 |
+
"step": 5440
|
| 3521 |
+
},
|
| 3522 |
+
{
|
| 3523 |
+
"epoch": 23.49,
|
| 3524 |
+
"learning_rate": 1.8166089965397923e-05,
|
| 3525 |
+
"loss": 0.0639,
|
| 3526 |
+
"step": 5450
|
| 3527 |
+
},
|
| 3528 |
+
{
|
| 3529 |
+
"epoch": 23.53,
|
| 3530 |
+
"learning_rate": 1.764705882352941e-05,
|
| 3531 |
+
"loss": 0.0778,
|
| 3532 |
+
"step": 5460
|
| 3533 |
+
},
|
| 3534 |
+
{
|
| 3535 |
+
"epoch": 23.58,
|
| 3536 |
+
"learning_rate": 1.7128027681660898e-05,
|
| 3537 |
+
"loss": 0.0645,
|
| 3538 |
+
"step": 5470
|
| 3539 |
+
},
|
| 3540 |
+
{
|
| 3541 |
+
"epoch": 23.62,
|
| 3542 |
+
"learning_rate": 1.6608996539792388e-05,
|
| 3543 |
+
"loss": 0.0766,
|
| 3544 |
+
"step": 5480
|
| 3545 |
+
},
|
| 3546 |
+
{
|
| 3547 |
+
"epoch": 23.66,
|
| 3548 |
+
"learning_rate": 1.6089965397923876e-05,
|
| 3549 |
+
"loss": 0.054,
|
| 3550 |
+
"step": 5490
|
| 3551 |
+
},
|
| 3552 |
+
{
|
| 3553 |
+
"epoch": 23.71,
|
| 3554 |
+
"learning_rate": 1.557093425605536e-05,
|
| 3555 |
+
"loss": 0.0695,
|
| 3556 |
+
"step": 5500
|
| 3557 |
+
},
|
| 3558 |
+
{
|
| 3559 |
+
"epoch": 23.75,
|
| 3560 |
+
"learning_rate": 1.5051903114186849e-05,
|
| 3561 |
+
"loss": 0.0682,
|
| 3562 |
+
"step": 5510
|
| 3563 |
+
},
|
| 3564 |
+
{
|
| 3565 |
+
"epoch": 23.79,
|
| 3566 |
+
"learning_rate": 1.4532871972318338e-05,
|
| 3567 |
+
"loss": 0.0762,
|
| 3568 |
+
"step": 5520
|
| 3569 |
+
},
|
| 3570 |
+
{
|
| 3571 |
+
"epoch": 23.83,
|
| 3572 |
+
"learning_rate": 1.4013840830449827e-05,
|
| 3573 |
+
"loss": 0.0729,
|
| 3574 |
+
"step": 5530
|
| 3575 |
+
},
|
| 3576 |
+
{
|
| 3577 |
+
"epoch": 23.88,
|
| 3578 |
+
"learning_rate": 1.3494809688581314e-05,
|
| 3579 |
+
"loss": 0.0704,
|
| 3580 |
+
"step": 5540
|
| 3581 |
+
},
|
| 3582 |
+
{
|
| 3583 |
+
"epoch": 23.92,
|
| 3584 |
+
"learning_rate": 1.2975778546712801e-05,
|
| 3585 |
+
"loss": 0.0629,
|
| 3586 |
+
"step": 5550
|
| 3587 |
+
},
|
| 3588 |
+
{
|
| 3589 |
+
"epoch": 23.96,
|
| 3590 |
+
"learning_rate": 1.2456747404844289e-05,
|
| 3591 |
+
"loss": 0.0739,
|
| 3592 |
+
"step": 5560
|
| 3593 |
+
},
|
| 3594 |
+
{
|
| 3595 |
+
"epoch": 24.0,
|
| 3596 |
+
"eval_loss": 0.48862332105636597,
|
| 3597 |
+
"eval_runtime": 159.9632,
|
| 3598 |
+
"eval_samples_per_second": 9.515,
|
| 3599 |
+
"eval_wer": 0.4284849096598365,
|
| 3600 |
+
"step": 5568
|
| 3601 |
+
},
|
| 3602 |
+
{
|
| 3603 |
+
"epoch": 24.01,
|
| 3604 |
+
"learning_rate": 1.1937716262975778e-05,
|
| 3605 |
+
"loss": 0.075,
|
| 3606 |
+
"step": 5570
|
| 3607 |
+
},
|
| 3608 |
+
{
|
| 3609 |
+
"epoch": 24.05,
|
| 3610 |
+
"learning_rate": 1.1418685121107265e-05,
|
| 3611 |
+
"loss": 0.0693,
|
| 3612 |
+
"step": 5580
|
| 3613 |
+
},
|
| 3614 |
+
{
|
| 3615 |
+
"epoch": 24.09,
|
| 3616 |
+
"learning_rate": 1.0899653979238754e-05,
|
| 3617 |
+
"loss": 0.0599,
|
| 3618 |
+
"step": 5590
|
| 3619 |
+
},
|
| 3620 |
+
{
|
| 3621 |
+
"epoch": 24.14,
|
| 3622 |
+
"learning_rate": 1.0380622837370241e-05,
|
| 3623 |
+
"loss": 0.0598,
|
| 3624 |
+
"step": 5600
|
| 3625 |
+
},
|
| 3626 |
+
{
|
| 3627 |
+
"epoch": 24.18,
|
| 3628 |
+
"learning_rate": 9.861591695501729e-06,
|
| 3629 |
+
"loss": 0.069,
|
| 3630 |
+
"step": 5610
|
| 3631 |
+
},
|
| 3632 |
+
{
|
| 3633 |
+
"epoch": 24.22,
|
| 3634 |
+
"learning_rate": 9.342560553633216e-06,
|
| 3635 |
+
"loss": 0.0602,
|
| 3636 |
+
"step": 5620
|
| 3637 |
+
},
|
| 3638 |
+
{
|
| 3639 |
+
"epoch": 24.27,
|
| 3640 |
+
"learning_rate": 8.823529411764705e-06,
|
| 3641 |
+
"loss": 0.0616,
|
| 3642 |
+
"step": 5630
|
| 3643 |
+
},
|
| 3644 |
+
{
|
| 3645 |
+
"epoch": 24.31,
|
| 3646 |
+
"learning_rate": 8.304498269896194e-06,
|
| 3647 |
+
"loss": 0.072,
|
| 3648 |
+
"step": 5640
|
| 3649 |
+
},
|
| 3650 |
+
{
|
| 3651 |
+
"epoch": 24.35,
|
| 3652 |
+
"learning_rate": 7.78546712802768e-06,
|
| 3653 |
+
"loss": 0.0703,
|
| 3654 |
+
"step": 5650
|
| 3655 |
+
},
|
| 3656 |
+
{
|
| 3657 |
+
"epoch": 24.4,
|
| 3658 |
+
"learning_rate": 7.266435986159169e-06,
|
| 3659 |
+
"loss": 0.0583,
|
| 3660 |
+
"step": 5660
|
| 3661 |
+
},
|
| 3662 |
+
{
|
| 3663 |
+
"epoch": 24.44,
|
| 3664 |
+
"learning_rate": 6.747404844290657e-06,
|
| 3665 |
+
"loss": 0.0663,
|
| 3666 |
+
"step": 5670
|
| 3667 |
+
},
|
| 3668 |
+
{
|
| 3669 |
+
"epoch": 24.48,
|
| 3670 |
+
"learning_rate": 6.228373702422144e-06,
|
| 3671 |
+
"loss": 0.0681,
|
| 3672 |
+
"step": 5680
|
| 3673 |
+
},
|
| 3674 |
+
{
|
| 3675 |
+
"epoch": 24.52,
|
| 3676 |
+
"learning_rate": 5.7093425605536325e-06,
|
| 3677 |
+
"loss": 0.0584,
|
| 3678 |
+
"step": 5690
|
| 3679 |
+
},
|
| 3680 |
+
{
|
| 3681 |
+
"epoch": 24.57,
|
| 3682 |
+
"learning_rate": 5.190311418685121e-06,
|
| 3683 |
+
"loss": 0.0581,
|
| 3684 |
+
"step": 5700
|
| 3685 |
+
},
|
| 3686 |
+
{
|
| 3687 |
+
"epoch": 24.61,
|
| 3688 |
+
"learning_rate": 4.671280276816608e-06,
|
| 3689 |
+
"loss": 0.0757,
|
| 3690 |
+
"step": 5710
|
| 3691 |
+
},
|
| 3692 |
+
{
|
| 3693 |
+
"epoch": 24.65,
|
| 3694 |
+
"learning_rate": 4.152249134948097e-06,
|
| 3695 |
+
"loss": 0.0675,
|
| 3696 |
+
"step": 5720
|
| 3697 |
+
},
|
| 3698 |
+
{
|
| 3699 |
+
"epoch": 24.7,
|
| 3700 |
+
"learning_rate": 3.6332179930795844e-06,
|
| 3701 |
+
"loss": 0.0588,
|
| 3702 |
+
"step": 5730
|
| 3703 |
+
},
|
| 3704 |
+
{
|
| 3705 |
+
"epoch": 24.74,
|
| 3706 |
+
"learning_rate": 3.114186851211072e-06,
|
| 3707 |
+
"loss": 0.0774,
|
| 3708 |
+
"step": 5740
|
| 3709 |
+
},
|
| 3710 |
+
{
|
| 3711 |
+
"epoch": 24.78,
|
| 3712 |
+
"learning_rate": 2.5951557093425604e-06,
|
| 3713 |
+
"loss": 0.0684,
|
| 3714 |
+
"step": 5750
|
| 3715 |
+
},
|
| 3716 |
+
{
|
| 3717 |
+
"epoch": 24.83,
|
| 3718 |
+
"learning_rate": 2.0761245674740485e-06,
|
| 3719 |
+
"loss": 0.0591,
|
| 3720 |
+
"step": 5760
|
| 3721 |
+
},
|
| 3722 |
+
{
|
| 3723 |
+
"epoch": 24.87,
|
| 3724 |
+
"learning_rate": 1.557093425605536e-06,
|
| 3725 |
+
"loss": 0.0611,
|
| 3726 |
+
"step": 5770
|
| 3727 |
+
},
|
| 3728 |
+
{
|
| 3729 |
+
"epoch": 24.91,
|
| 3730 |
+
"learning_rate": 1.0380622837370243e-06,
|
| 3731 |
+
"loss": 0.058,
|
| 3732 |
+
"step": 5780
|
| 3733 |
+
},
|
| 3734 |
+
{
|
| 3735 |
+
"epoch": 24.95,
|
| 3736 |
+
"learning_rate": 5.190311418685121e-07,
|
| 3737 |
+
"loss": 0.086,
|
| 3738 |
+
"step": 5790
|
| 3739 |
+
},
|
| 3740 |
+
{
|
| 3741 |
+
"epoch": 25.0,
|
| 3742 |
+
"learning_rate": 0.0,
|
| 3743 |
+
"loss": 0.0505,
|
| 3744 |
+
"step": 5800
|
| 3745 |
+
},
|
| 3746 |
+
{
|
| 3747 |
+
"epoch": 25.0,
|
| 3748 |
+
"eval_loss": 0.48566901683807373,
|
| 3749 |
+
"eval_runtime": 159.4769,
|
| 3750 |
+
"eval_samples_per_second": 9.544,
|
| 3751 |
+
"eval_wer": 0.42787927727869185,
|
| 3752 |
+
"step": 5800
|
| 3753 |
+
},
|
| 3754 |
+
{
|
| 3755 |
+
"epoch": 25.04,
|
| 3756 |
+
"learning_rate": 4.971181556195965e-05,
|
| 3757 |
+
"loss": 0.0735,
|
| 3758 |
+
"step": 5810
|
| 3759 |
+
},
|
| 3760 |
+
{
|
| 3761 |
+
"epoch": 25.09,
|
| 3762 |
+
"learning_rate": 4.9279538904899127e-05,
|
| 3763 |
+
"loss": 0.0769,
|
| 3764 |
+
"step": 5820
|
| 3765 |
+
},
|
| 3766 |
+
{
|
| 3767 |
+
"epoch": 25.13,
|
| 3768 |
+
"learning_rate": 4.884726224783861e-05,
|
| 3769 |
+
"loss": 0.0613,
|
| 3770 |
+
"step": 5830
|
| 3771 |
+
},
|
| 3772 |
+
{
|
| 3773 |
+
"epoch": 25.17,
|
| 3774 |
+
"learning_rate": 4.841498559077809e-05,
|
| 3775 |
+
"loss": 0.0602,
|
| 3776 |
+
"step": 5840
|
| 3777 |
+
},
|
| 3778 |
+
{
|
| 3779 |
+
"epoch": 25.22,
|
| 3780 |
+
"learning_rate": 4.798270893371757e-05,
|
| 3781 |
+
"loss": 0.0509,
|
| 3782 |
+
"step": 5850
|
| 3783 |
+
},
|
| 3784 |
+
{
|
| 3785 |
+
"epoch": 25.26,
|
| 3786 |
+
"learning_rate": 4.755043227665706e-05,
|
| 3787 |
+
"loss": 0.0785,
|
| 3788 |
+
"step": 5860
|
| 3789 |
+
},
|
| 3790 |
+
{
|
| 3791 |
+
"epoch": 25.3,
|
| 3792 |
+
"learning_rate": 4.711815561959654e-05,
|
| 3793 |
+
"loss": 0.0735,
|
| 3794 |
+
"step": 5870
|
| 3795 |
+
},
|
| 3796 |
+
{
|
| 3797 |
+
"epoch": 25.34,
|
| 3798 |
+
"learning_rate": 4.668587896253602e-05,
|
| 3799 |
+
"loss": 0.0764,
|
| 3800 |
+
"step": 5880
|
| 3801 |
+
},
|
| 3802 |
+
{
|
| 3803 |
+
"epoch": 25.39,
|
| 3804 |
+
"learning_rate": 4.6253602305475506e-05,
|
| 3805 |
+
"loss": 0.0737,
|
| 3806 |
+
"step": 5890
|
| 3807 |
+
},
|
| 3808 |
+
{
|
| 3809 |
+
"epoch": 25.43,
|
| 3810 |
+
"learning_rate": 4.582132564841498e-05,
|
| 3811 |
+
"loss": 0.0706,
|
| 3812 |
+
"step": 5900
|
| 3813 |
+
},
|
| 3814 |
+
{
|
| 3815 |
+
"epoch": 25.47,
|
| 3816 |
+
"learning_rate": 4.538904899135446e-05,
|
| 3817 |
+
"loss": 0.079,
|
| 3818 |
+
"step": 5910
|
| 3819 |
+
},
|
| 3820 |
+
{
|
| 3821 |
+
"epoch": 25.52,
|
| 3822 |
+
"learning_rate": 4.495677233429394e-05,
|
| 3823 |
+
"loss": 0.0556,
|
| 3824 |
+
"step": 5920
|
| 3825 |
+
},
|
| 3826 |
+
{
|
| 3827 |
+
"epoch": 25.56,
|
| 3828 |
+
"learning_rate": 4.4524495677233426e-05,
|
| 3829 |
+
"loss": 0.0683,
|
| 3830 |
+
"step": 5930
|
| 3831 |
+
},
|
| 3832 |
+
{
|
| 3833 |
+
"epoch": 25.6,
|
| 3834 |
+
"learning_rate": 4.4092219020172906e-05,
|
| 3835 |
+
"loss": 0.0973,
|
| 3836 |
+
"step": 5940
|
| 3837 |
+
},
|
| 3838 |
+
{
|
| 3839 |
+
"epoch": 25.65,
|
| 3840 |
+
"learning_rate": 4.3659942363112386e-05,
|
| 3841 |
+
"loss": 0.0622,
|
| 3842 |
+
"step": 5950
|
| 3843 |
+
},
|
| 3844 |
+
{
|
| 3845 |
+
"epoch": 25.69,
|
| 3846 |
+
"learning_rate": 4.322766570605187e-05,
|
| 3847 |
+
"loss": 0.0762,
|
| 3848 |
+
"step": 5960
|
| 3849 |
+
},
|
| 3850 |
+
{
|
| 3851 |
+
"epoch": 25.73,
|
| 3852 |
+
"learning_rate": 4.279538904899135e-05,
|
| 3853 |
+
"loss": 0.0716,
|
| 3854 |
+
"step": 5970
|
| 3855 |
+
},
|
| 3856 |
+
{
|
| 3857 |
+
"epoch": 25.77,
|
| 3858 |
+
"learning_rate": 4.236311239193083e-05,
|
| 3859 |
+
"loss": 0.0459,
|
| 3860 |
+
"step": 5980
|
| 3861 |
+
},
|
| 3862 |
+
{
|
| 3863 |
+
"epoch": 25.82,
|
| 3864 |
+
"learning_rate": 4.193083573487032e-05,
|
| 3865 |
+
"loss": 0.0619,
|
| 3866 |
+
"step": 5990
|
| 3867 |
+
},
|
| 3868 |
+
{
|
| 3869 |
+
"epoch": 25.86,
|
| 3870 |
+
"learning_rate": 4.14985590778098e-05,
|
| 3871 |
+
"loss": 0.0568,
|
| 3872 |
+
"step": 6000
|
| 3873 |
+
},
|
| 3874 |
+
{
|
| 3875 |
+
"epoch": 25.9,
|
| 3876 |
+
"learning_rate": 4.106628242074927e-05,
|
| 3877 |
+
"loss": 0.087,
|
| 3878 |
+
"step": 6010
|
| 3879 |
+
},
|
| 3880 |
+
{
|
| 3881 |
+
"epoch": 25.95,
|
| 3882 |
+
"learning_rate": 4.063400576368875e-05,
|
| 3883 |
+
"loss": 0.0805,
|
| 3884 |
+
"step": 6020
|
| 3885 |
+
},
|
| 3886 |
+
{
|
| 3887 |
+
"epoch": 25.99,
|
| 3888 |
+
"learning_rate": 4.020172910662824e-05,
|
| 3889 |
+
"loss": 0.0513,
|
| 3890 |
+
"step": 6030
|
| 3891 |
+
},
|
| 3892 |
+
{
|
| 3893 |
+
"epoch": 26.0,
|
| 3894 |
+
"eval_loss": 0.4976740777492523,
|
| 3895 |
+
"eval_runtime": 157.8423,
|
| 3896 |
+
"eval_samples_per_second": 9.643,
|
| 3897 |
+
"eval_wer": 0.429595235691935,
|
| 3898 |
+
"step": 6032
|
| 3899 |
+
},
|
| 3900 |
+
{
|
| 3901 |
+
"epoch": 26.03,
|
| 3902 |
+
"learning_rate": 3.976945244956772e-05,
|
| 3903 |
+
"loss": 0.0881,
|
| 3904 |
+
"step": 6040
|
| 3905 |
+
},
|
| 3906 |
+
{
|
| 3907 |
+
"epoch": 26.08,
|
| 3908 |
+
"learning_rate": 3.93371757925072e-05,
|
| 3909 |
+
"loss": 0.0585,
|
| 3910 |
+
"step": 6050
|
| 3911 |
+
},
|
| 3912 |
+
{
|
| 3913 |
+
"epoch": 26.12,
|
| 3914 |
+
"learning_rate": 3.8904899135446685e-05,
|
| 3915 |
+
"loss": 0.0816,
|
| 3916 |
+
"step": 6060
|
| 3917 |
+
},
|
| 3918 |
+
{
|
| 3919 |
+
"epoch": 26.16,
|
| 3920 |
+
"learning_rate": 3.8472622478386165e-05,
|
| 3921 |
+
"loss": 0.068,
|
| 3922 |
+
"step": 6070
|
| 3923 |
+
},
|
| 3924 |
+
{
|
| 3925 |
+
"epoch": 26.21,
|
| 3926 |
+
"learning_rate": 3.8040345821325645e-05,
|
| 3927 |
+
"loss": 0.0543,
|
| 3928 |
+
"step": 6080
|
| 3929 |
+
},
|
| 3930 |
+
{
|
| 3931 |
+
"epoch": 26.25,
|
| 3932 |
+
"learning_rate": 3.7608069164265125e-05,
|
| 3933 |
+
"loss": 0.0794,
|
| 3934 |
+
"step": 6090
|
| 3935 |
+
},
|
| 3936 |
+
{
|
| 3937 |
+
"epoch": 26.29,
|
| 3938 |
+
"learning_rate": 3.7175792507204605e-05,
|
| 3939 |
+
"loss": 0.0541,
|
| 3940 |
+
"step": 6100
|
| 3941 |
+
},
|
| 3942 |
+
{
|
| 3943 |
+
"epoch": 26.34,
|
| 3944 |
+
"learning_rate": 3.6743515850144085e-05,
|
| 3945 |
+
"loss": 0.0509,
|
| 3946 |
+
"step": 6110
|
| 3947 |
+
},
|
| 3948 |
+
{
|
| 3949 |
+
"epoch": 26.38,
|
| 3950 |
+
"learning_rate": 3.631123919308357e-05,
|
| 3951 |
+
"loss": 0.0729,
|
| 3952 |
+
"step": 6120
|
| 3953 |
+
},
|
| 3954 |
+
{
|
| 3955 |
+
"epoch": 26.42,
|
| 3956 |
+
"learning_rate": 3.587896253602305e-05,
|
| 3957 |
+
"loss": 0.057,
|
| 3958 |
+
"step": 6130
|
| 3959 |
+
},
|
| 3960 |
+
{
|
| 3961 |
+
"epoch": 26.46,
|
| 3962 |
+
"learning_rate": 3.544668587896253e-05,
|
| 3963 |
+
"loss": 0.0469,
|
| 3964 |
+
"step": 6140
|
| 3965 |
+
},
|
| 3966 |
+
{
|
| 3967 |
+
"epoch": 26.51,
|
| 3968 |
+
"learning_rate": 3.501440922190201e-05,
|
| 3969 |
+
"loss": 0.0648,
|
| 3970 |
+
"step": 6150
|
| 3971 |
+
},
|
| 3972 |
+
{
|
| 3973 |
+
"epoch": 26.55,
|
| 3974 |
+
"learning_rate": 3.458213256484149e-05,
|
| 3975 |
+
"loss": 0.08,
|
| 3976 |
+
"step": 6160
|
| 3977 |
+
},
|
| 3978 |
+
{
|
| 3979 |
+
"epoch": 26.59,
|
| 3980 |
+
"learning_rate": 3.414985590778098e-05,
|
| 3981 |
+
"loss": 0.083,
|
| 3982 |
+
"step": 6170
|
| 3983 |
+
},
|
| 3984 |
+
{
|
| 3985 |
+
"epoch": 26.64,
|
| 3986 |
+
"learning_rate": 3.371757925072046e-05,
|
| 3987 |
+
"loss": 0.0671,
|
| 3988 |
+
"step": 6180
|
| 3989 |
+
},
|
| 3990 |
+
{
|
| 3991 |
+
"epoch": 26.68,
|
| 3992 |
+
"learning_rate": 3.328530259365994e-05,
|
| 3993 |
+
"loss": 0.0795,
|
| 3994 |
+
"step": 6190
|
| 3995 |
+
},
|
| 3996 |
+
{
|
| 3997 |
+
"epoch": 26.72,
|
| 3998 |
+
"learning_rate": 3.2853025936599424e-05,
|
| 3999 |
+
"loss": 0.0695,
|
| 4000 |
+
"step": 6200
|
| 4001 |
+
},
|
| 4002 |
+
{
|
| 4003 |
+
"epoch": 26.77,
|
| 4004 |
+
"learning_rate": 3.24207492795389e-05,
|
| 4005 |
+
"loss": 0.0525,
|
| 4006 |
+
"step": 6210
|
| 4007 |
+
},
|
| 4008 |
+
{
|
| 4009 |
+
"epoch": 26.81,
|
| 4010 |
+
"learning_rate": 3.1988472622478384e-05,
|
| 4011 |
+
"loss": 0.0615,
|
| 4012 |
+
"step": 6220
|
| 4013 |
+
},
|
| 4014 |
+
{
|
| 4015 |
+
"epoch": 26.85,
|
| 4016 |
+
"learning_rate": 3.1556195965417864e-05,
|
| 4017 |
+
"loss": 0.0768,
|
| 4018 |
+
"step": 6230
|
| 4019 |
+
},
|
| 4020 |
+
{
|
| 4021 |
+
"epoch": 26.89,
|
| 4022 |
+
"learning_rate": 3.1123919308357344e-05,
|
| 4023 |
+
"loss": 0.0719,
|
| 4024 |
+
"step": 6240
|
| 4025 |
+
},
|
| 4026 |
+
{
|
| 4027 |
+
"epoch": 26.94,
|
| 4028 |
+
"learning_rate": 3.069164265129683e-05,
|
| 4029 |
+
"loss": 0.0514,
|
| 4030 |
+
"step": 6250
|
| 4031 |
+
},
|
| 4032 |
+
{
|
| 4033 |
+
"epoch": 26.98,
|
| 4034 |
+
"learning_rate": 3.025936599423631e-05,
|
| 4035 |
+
"loss": 0.0673,
|
| 4036 |
+
"step": 6260
|
| 4037 |
+
},
|
| 4038 |
+
{
|
| 4039 |
+
"epoch": 27.0,
|
| 4040 |
+
"eval_loss": 0.487693190574646,
|
| 4041 |
+
"eval_runtime": 160.5521,
|
| 4042 |
+
"eval_samples_per_second": 9.48,
|
| 4043 |
+
"eval_wer": 0.4280811547390734,
|
| 4044 |
+
"step": 6264
|
| 4045 |
+
},
|
| 4046 |
+
{
|
| 4047 |
+
"epoch": 27.03,
|
| 4048 |
+
"learning_rate": 2.9827089337175787e-05,
|
| 4049 |
+
"loss": 0.0669,
|
| 4050 |
+
"step": 6270
|
| 4051 |
+
},
|
| 4052 |
+
{
|
| 4053 |
+
"epoch": 27.07,
|
| 4054 |
+
"learning_rate": 2.939481268011527e-05,
|
| 4055 |
+
"loss": 0.0484,
|
| 4056 |
+
"step": 6280
|
| 4057 |
+
},
|
| 4058 |
+
{
|
| 4059 |
+
"epoch": 27.11,
|
| 4060 |
+
"learning_rate": 2.8962536023054754e-05,
|
| 4061 |
+
"loss": 0.0671,
|
| 4062 |
+
"step": 6290
|
| 4063 |
+
},
|
| 4064 |
+
{
|
| 4065 |
+
"epoch": 27.15,
|
| 4066 |
+
"learning_rate": 2.8530259365994234e-05,
|
| 4067 |
+
"loss": 0.0603,
|
| 4068 |
+
"step": 6300
|
| 4069 |
+
},
|
| 4070 |
+
{
|
| 4071 |
+
"epoch": 27.2,
|
| 4072 |
+
"learning_rate": 2.8097982708933717e-05,
|
| 4073 |
+
"loss": 0.0593,
|
| 4074 |
+
"step": 6310
|
| 4075 |
+
},
|
| 4076 |
+
{
|
| 4077 |
+
"epoch": 27.24,
|
| 4078 |
+
"learning_rate": 2.7665706051873194e-05,
|
| 4079 |
+
"loss": 0.0494,
|
| 4080 |
+
"step": 6320
|
| 4081 |
+
},
|
| 4082 |
+
{
|
| 4083 |
+
"epoch": 27.28,
|
| 4084 |
+
"learning_rate": 2.7233429394812677e-05,
|
| 4085 |
+
"loss": 0.0747,
|
| 4086 |
+
"step": 6330
|
| 4087 |
+
},
|
| 4088 |
+
{
|
| 4089 |
+
"epoch": 27.33,
|
| 4090 |
+
"learning_rate": 2.680115273775216e-05,
|
| 4091 |
+
"loss": 0.0635,
|
| 4092 |
+
"step": 6340
|
| 4093 |
+
},
|
| 4094 |
+
{
|
| 4095 |
+
"epoch": 27.37,
|
| 4096 |
+
"learning_rate": 2.636887608069164e-05,
|
| 4097 |
+
"loss": 0.09,
|
| 4098 |
+
"step": 6350
|
| 4099 |
+
},
|
| 4100 |
+
{
|
| 4101 |
+
"epoch": 27.41,
|
| 4102 |
+
"learning_rate": 2.5936599423631124e-05,
|
| 4103 |
+
"loss": 0.0716,
|
| 4104 |
+
"step": 6360
|
| 4105 |
+
},
|
| 4106 |
+
{
|
| 4107 |
+
"epoch": 27.46,
|
| 4108 |
+
"learning_rate": 2.5504322766570603e-05,
|
| 4109 |
+
"loss": 0.0515,
|
| 4110 |
+
"step": 6370
|
| 4111 |
+
},
|
| 4112 |
+
{
|
| 4113 |
+
"epoch": 27.5,
|
| 4114 |
+
"learning_rate": 2.5072046109510083e-05,
|
| 4115 |
+
"loss": 0.061,
|
| 4116 |
+
"step": 6380
|
| 4117 |
+
},
|
| 4118 |
+
{
|
| 4119 |
+
"epoch": 27.54,
|
| 4120 |
+
"learning_rate": 2.4639769452449563e-05,
|
| 4121 |
+
"loss": 0.0673,
|
| 4122 |
+
"step": 6390
|
| 4123 |
+
},
|
| 4124 |
+
{
|
| 4125 |
+
"epoch": 27.58,
|
| 4126 |
+
"learning_rate": 2.4207492795389047e-05,
|
| 4127 |
+
"loss": 0.0603,
|
| 4128 |
+
"step": 6400
|
| 4129 |
+
},
|
| 4130 |
+
{
|
| 4131 |
+
"epoch": 27.63,
|
| 4132 |
+
"learning_rate": 2.377521613832853e-05,
|
| 4133 |
+
"loss": 0.0693,
|
| 4134 |
+
"step": 6410
|
| 4135 |
+
},
|
| 4136 |
+
{
|
| 4137 |
+
"epoch": 27.67,
|
| 4138 |
+
"learning_rate": 2.334293948126801e-05,
|
| 4139 |
+
"loss": 0.0606,
|
| 4140 |
+
"step": 6420
|
| 4141 |
+
},
|
| 4142 |
+
{
|
| 4143 |
+
"epoch": 27.71,
|
| 4144 |
+
"learning_rate": 2.291066282420749e-05,
|
| 4145 |
+
"loss": 0.0683,
|
| 4146 |
+
"step": 6430
|
| 4147 |
+
},
|
| 4148 |
+
{
|
| 4149 |
+
"epoch": 27.76,
|
| 4150 |
+
"learning_rate": 2.247838616714697e-05,
|
| 4151 |
+
"loss": 0.0758,
|
| 4152 |
+
"step": 6440
|
| 4153 |
+
},
|
| 4154 |
+
{
|
| 4155 |
+
"epoch": 27.8,
|
| 4156 |
+
"learning_rate": 2.2046109510086453e-05,
|
| 4157 |
+
"loss": 0.0551,
|
| 4158 |
+
"step": 6450
|
| 4159 |
+
},
|
| 4160 |
+
{
|
| 4161 |
+
"epoch": 27.84,
|
| 4162 |
+
"learning_rate": 2.1613832853025936e-05,
|
| 4163 |
+
"loss": 0.0528,
|
| 4164 |
+
"step": 6460
|
| 4165 |
+
},
|
| 4166 |
+
{
|
| 4167 |
+
"epoch": 27.89,
|
| 4168 |
+
"learning_rate": 2.1181556195965416e-05,
|
| 4169 |
+
"loss": 0.0696,
|
| 4170 |
+
"step": 6470
|
| 4171 |
+
},
|
| 4172 |
+
{
|
| 4173 |
+
"epoch": 27.93,
|
| 4174 |
+
"learning_rate": 2.07492795389049e-05,
|
| 4175 |
+
"loss": 0.0681,
|
| 4176 |
+
"step": 6480
|
| 4177 |
+
},
|
| 4178 |
+
{
|
| 4179 |
+
"epoch": 27.97,
|
| 4180 |
+
"learning_rate": 2.0317002881844376e-05,
|
| 4181 |
+
"loss": 0.0622,
|
| 4182 |
+
"step": 6490
|
| 4183 |
+
},
|
| 4184 |
+
{
|
| 4185 |
+
"epoch": 28.0,
|
| 4186 |
+
"eval_loss": 0.491367906332016,
|
| 4187 |
+
"eval_runtime": 161.8629,
|
| 4188 |
+
"eval_samples_per_second": 9.403,
|
| 4189 |
+
"eval_wer": 0.42626425759563946,
|
| 4190 |
+
"step": 6496
|
| 4191 |
+
},
|
| 4192 |
+
{
|
| 4193 |
+
"epoch": 28.02,
|
| 4194 |
+
"learning_rate": 1.988472622478386e-05,
|
| 4195 |
+
"loss": 0.0681,
|
| 4196 |
+
"step": 6500
|
| 4197 |
+
},
|
| 4198 |
+
{
|
| 4199 |
+
"epoch": 28.06,
|
| 4200 |
+
"learning_rate": 1.9452449567723343e-05,
|
| 4201 |
+
"loss": 0.0502,
|
| 4202 |
+
"step": 6510
|
| 4203 |
+
},
|
| 4204 |
+
{
|
| 4205 |
+
"epoch": 28.1,
|
| 4206 |
+
"learning_rate": 1.9020172910662823e-05,
|
| 4207 |
+
"loss": 0.0921,
|
| 4208 |
+
"step": 6520
|
| 4209 |
+
},
|
| 4210 |
+
{
|
| 4211 |
+
"epoch": 28.15,
|
| 4212 |
+
"learning_rate": 1.8587896253602302e-05,
|
| 4213 |
+
"loss": 0.0705,
|
| 4214 |
+
"step": 6530
|
| 4215 |
+
},
|
| 4216 |
+
{
|
| 4217 |
+
"epoch": 28.19,
|
| 4218 |
+
"learning_rate": 1.8155619596541786e-05,
|
| 4219 |
+
"loss": 0.0657,
|
| 4220 |
+
"step": 6540
|
| 4221 |
+
},
|
| 4222 |
+
{
|
| 4223 |
+
"epoch": 28.23,
|
| 4224 |
+
"learning_rate": 1.7723342939481266e-05,
|
| 4225 |
+
"loss": 0.0713,
|
| 4226 |
+
"step": 6550
|
| 4227 |
+
},
|
| 4228 |
+
{
|
| 4229 |
+
"epoch": 28.28,
|
| 4230 |
+
"learning_rate": 1.7291066282420746e-05,
|
| 4231 |
+
"loss": 0.0914,
|
| 4232 |
+
"step": 6560
|
| 4233 |
+
},
|
| 4234 |
+
{
|
| 4235 |
+
"epoch": 28.32,
|
| 4236 |
+
"learning_rate": 1.685878962536023e-05,
|
| 4237 |
+
"loss": 0.0531,
|
| 4238 |
+
"step": 6570
|
| 4239 |
+
},
|
| 4240 |
+
{
|
| 4241 |
+
"epoch": 28.36,
|
| 4242 |
+
"learning_rate": 1.6426512968299712e-05,
|
| 4243 |
+
"loss": 0.0578,
|
| 4244 |
+
"step": 6580
|
| 4245 |
+
},
|
| 4246 |
+
{
|
| 4247 |
+
"epoch": 28.4,
|
| 4248 |
+
"learning_rate": 1.5994236311239192e-05,
|
| 4249 |
+
"loss": 0.0659,
|
| 4250 |
+
"step": 6590
|
| 4251 |
+
},
|
| 4252 |
+
{
|
| 4253 |
+
"epoch": 28.45,
|
| 4254 |
+
"learning_rate": 1.5561959654178672e-05,
|
| 4255 |
+
"loss": 0.0522,
|
| 4256 |
+
"step": 6600
|
| 4257 |
+
},
|
| 4258 |
+
{
|
| 4259 |
+
"epoch": 28.49,
|
| 4260 |
+
"learning_rate": 1.5129682997118155e-05,
|
| 4261 |
+
"loss": 0.0738,
|
| 4262 |
+
"step": 6610
|
| 4263 |
+
},
|
| 4264 |
+
{
|
| 4265 |
+
"epoch": 28.53,
|
| 4266 |
+
"learning_rate": 1.4697406340057635e-05,
|
| 4267 |
+
"loss": 0.0731,
|
| 4268 |
+
"step": 6620
|
| 4269 |
+
},
|
| 4270 |
+
{
|
| 4271 |
+
"epoch": 28.58,
|
| 4272 |
+
"learning_rate": 1.4265129682997117e-05,
|
| 4273 |
+
"loss": 0.0408,
|
| 4274 |
+
"step": 6630
|
| 4275 |
+
},
|
| 4276 |
+
{
|
| 4277 |
+
"epoch": 28.62,
|
| 4278 |
+
"learning_rate": 1.3832853025936597e-05,
|
| 4279 |
+
"loss": 0.0617,
|
| 4280 |
+
"step": 6640
|
| 4281 |
+
},
|
| 4282 |
+
{
|
| 4283 |
+
"epoch": 28.66,
|
| 4284 |
+
"learning_rate": 1.340057636887608e-05,
|
| 4285 |
+
"loss": 0.0476,
|
| 4286 |
+
"step": 6650
|
| 4287 |
+
},
|
| 4288 |
+
{
|
| 4289 |
+
"epoch": 28.71,
|
| 4290 |
+
"learning_rate": 1.2968299711815562e-05,
|
| 4291 |
+
"loss": 0.0728,
|
| 4292 |
+
"step": 6660
|
| 4293 |
+
},
|
| 4294 |
+
{
|
| 4295 |
+
"epoch": 28.75,
|
| 4296 |
+
"learning_rate": 1.2536023054755042e-05,
|
| 4297 |
+
"loss": 0.0487,
|
| 4298 |
+
"step": 6670
|
| 4299 |
+
},
|
| 4300 |
+
{
|
| 4301 |
+
"epoch": 28.79,
|
| 4302 |
+
"learning_rate": 1.2103746397694523e-05,
|
| 4303 |
+
"loss": 0.0579,
|
| 4304 |
+
"step": 6680
|
| 4305 |
+
},
|
| 4306 |
+
{
|
| 4307 |
+
"epoch": 28.83,
|
| 4308 |
+
"learning_rate": 1.1671469740634005e-05,
|
| 4309 |
+
"loss": 0.0545,
|
| 4310 |
+
"step": 6690
|
| 4311 |
+
},
|
| 4312 |
+
{
|
| 4313 |
+
"epoch": 28.88,
|
| 4314 |
+
"learning_rate": 1.1239193083573485e-05,
|
| 4315 |
+
"loss": 0.0668,
|
| 4316 |
+
"step": 6700
|
| 4317 |
+
},
|
| 4318 |
+
{
|
| 4319 |
+
"epoch": 28.92,
|
| 4320 |
+
"learning_rate": 1.0806916426512968e-05,
|
| 4321 |
+
"loss": 0.0557,
|
| 4322 |
+
"step": 6710
|
| 4323 |
+
},
|
| 4324 |
+
{
|
| 4325 |
+
"epoch": 28.96,
|
| 4326 |
+
"learning_rate": 1.037463976945245e-05,
|
| 4327 |
+
"loss": 0.0538,
|
| 4328 |
+
"step": 6720
|
| 4329 |
+
},
|
| 4330 |
+
{
|
| 4331 |
+
"epoch": 29.0,
|
| 4332 |
+
"eval_loss": 0.4901193082332611,
|
| 4333 |
+
"eval_runtime": 162.4755,
|
| 4334 |
+
"eval_samples_per_second": 9.368,
|
| 4335 |
+
"eval_wer": 0.42353891188048853,
|
| 4336 |
+
"step": 6728
|
| 4337 |
+
},
|
| 4338 |
+
{
|
| 4339 |
+
"epoch": 29.01,
|
| 4340 |
+
"learning_rate": 9.94236311239193e-06,
|
| 4341 |
+
"loss": 0.0581,
|
| 4342 |
+
"step": 6730
|
| 4343 |
+
},
|
| 4344 |
+
{
|
| 4345 |
+
"epoch": 29.05,
|
| 4346 |
+
"learning_rate": 9.510086455331411e-06,
|
| 4347 |
+
"loss": 0.0577,
|
| 4348 |
+
"step": 6740
|
| 4349 |
+
},
|
| 4350 |
+
{
|
| 4351 |
+
"epoch": 29.09,
|
| 4352 |
+
"learning_rate": 9.077809798270893e-06,
|
| 4353 |
+
"loss": 0.0459,
|
| 4354 |
+
"step": 6750
|
| 4355 |
+
},
|
| 4356 |
+
{
|
| 4357 |
+
"epoch": 29.14,
|
| 4358 |
+
"learning_rate": 8.645533141210373e-06,
|
| 4359 |
+
"loss": 0.0537,
|
| 4360 |
+
"step": 6760
|
| 4361 |
+
},
|
| 4362 |
+
{
|
| 4363 |
+
"epoch": 29.18,
|
| 4364 |
+
"learning_rate": 8.213256484149856e-06,
|
| 4365 |
+
"loss": 0.0763,
|
| 4366 |
+
"step": 6770
|
| 4367 |
+
},
|
| 4368 |
+
{
|
| 4369 |
+
"epoch": 29.22,
|
| 4370 |
+
"learning_rate": 7.780979827089336e-06,
|
| 4371 |
+
"loss": 0.0517,
|
| 4372 |
+
"step": 6780
|
| 4373 |
+
},
|
| 4374 |
+
{
|
| 4375 |
+
"epoch": 29.27,
|
| 4376 |
+
"learning_rate": 7.348703170028818e-06,
|
| 4377 |
+
"loss": 0.0763,
|
| 4378 |
+
"step": 6790
|
| 4379 |
+
},
|
| 4380 |
+
{
|
| 4381 |
+
"epoch": 29.31,
|
| 4382 |
+
"learning_rate": 6.916426512968298e-06,
|
| 4383 |
+
"loss": 0.0468,
|
| 4384 |
+
"step": 6800
|
| 4385 |
+
},
|
| 4386 |
+
{
|
| 4387 |
+
"epoch": 29.35,
|
| 4388 |
+
"learning_rate": 6.484149855907781e-06,
|
| 4389 |
+
"loss": 0.0573,
|
| 4390 |
+
"step": 6810
|
| 4391 |
+
},
|
| 4392 |
+
{
|
| 4393 |
+
"epoch": 29.4,
|
| 4394 |
+
"learning_rate": 6.051873198847262e-06,
|
| 4395 |
+
"loss": 0.0535,
|
| 4396 |
+
"step": 6820
|
| 4397 |
+
},
|
| 4398 |
+
{
|
| 4399 |
+
"epoch": 29.44,
|
| 4400 |
+
"learning_rate": 5.619596541786742e-06,
|
| 4401 |
+
"loss": 0.0577,
|
| 4402 |
+
"step": 6830
|
| 4403 |
+
},
|
| 4404 |
+
{
|
| 4405 |
+
"epoch": 29.48,
|
| 4406 |
+
"learning_rate": 5.187319884726225e-06,
|
| 4407 |
+
"loss": 0.0643,
|
| 4408 |
+
"step": 6840
|
| 4409 |
+
},
|
| 4410 |
+
{
|
| 4411 |
+
"epoch": 29.52,
|
| 4412 |
+
"learning_rate": 4.755043227665706e-06,
|
| 4413 |
+
"loss": 0.0459,
|
| 4414 |
+
"step": 6850
|
| 4415 |
+
},
|
| 4416 |
+
{
|
| 4417 |
+
"epoch": 29.57,
|
| 4418 |
+
"learning_rate": 4.322766570605186e-06,
|
| 4419 |
+
"loss": 0.0596,
|
| 4420 |
+
"step": 6860
|
| 4421 |
+
},
|
| 4422 |
+
{
|
| 4423 |
+
"epoch": 29.61,
|
| 4424 |
+
"learning_rate": 3.890489913544668e-06,
|
| 4425 |
+
"loss": 0.0369,
|
| 4426 |
+
"step": 6870
|
| 4427 |
+
},
|
| 4428 |
+
{
|
| 4429 |
+
"epoch": 29.65,
|
| 4430 |
+
"learning_rate": 3.458213256484149e-06,
|
| 4431 |
+
"loss": 0.0723,
|
| 4432 |
+
"step": 6880
|
| 4433 |
+
},
|
| 4434 |
+
{
|
| 4435 |
+
"epoch": 29.7,
|
| 4436 |
+
"learning_rate": 3.025936599423631e-06,
|
| 4437 |
+
"loss": 0.0666,
|
| 4438 |
+
"step": 6890
|
| 4439 |
+
},
|
| 4440 |
+
{
|
| 4441 |
+
"epoch": 29.74,
|
| 4442 |
+
"learning_rate": 2.5936599423631124e-06,
|
| 4443 |
+
"loss": 0.046,
|
| 4444 |
+
"step": 6900
|
| 4445 |
+
},
|
| 4446 |
+
{
|
| 4447 |
+
"epoch": 29.78,
|
| 4448 |
+
"learning_rate": 2.161383285302593e-06,
|
| 4449 |
+
"loss": 0.05,
|
| 4450 |
+
"step": 6910
|
| 4451 |
+
},
|
| 4452 |
+
{
|
| 4453 |
+
"epoch": 29.83,
|
| 4454 |
+
"learning_rate": 1.7291066282420746e-06,
|
| 4455 |
+
"loss": 0.0543,
|
| 4456 |
+
"step": 6920
|
| 4457 |
+
},
|
| 4458 |
+
{
|
| 4459 |
+
"epoch": 29.87,
|
| 4460 |
+
"learning_rate": 1.2968299711815562e-06,
|
| 4461 |
+
"loss": 0.0571,
|
| 4462 |
+
"step": 6930
|
| 4463 |
+
},
|
| 4464 |
+
{
|
| 4465 |
+
"epoch": 29.91,
|
| 4466 |
+
"learning_rate": 8.645533141210373e-07,
|
| 4467 |
+
"loss": 0.078,
|
| 4468 |
+
"step": 6940
|
| 4469 |
+
},
|
| 4470 |
+
{
|
| 4471 |
+
"epoch": 29.95,
|
| 4472 |
+
"learning_rate": 4.3227665706051865e-07,
|
| 4473 |
+
"loss": 0.0628,
|
| 4474 |
+
"step": 6950
|
| 4475 |
+
},
|
| 4476 |
+
{
|
| 4477 |
+
"epoch": 30.0,
|
| 4478 |
+
"learning_rate": 0.0,
|
| 4479 |
+
"loss": 0.0481,
|
| 4480 |
+
"step": 6960
|
| 4481 |
+
},
|
| 4482 |
+
{
|
| 4483 |
+
"epoch": 30.0,
|
| 4484 |
+
"eval_loss": 0.49076756834983826,
|
| 4485 |
+
"eval_runtime": 163.7648,
|
| 4486 |
+
"eval_samples_per_second": 9.294,
|
| 4487 |
+
"eval_wer": 0.4218229534672454,
|
| 4488 |
+
"step": 6960
|
| 4489 |
+
},
|
| 4490 |
+
{
|
| 4491 |
+
"epoch": 30.04,
|
| 4492 |
+
"learning_rate": 2.695128339444735e-05,
|
| 4493 |
+
"loss": 0.0523,
|
| 4494 |
+
"step": 6970
|
| 4495 |
+
},
|
| 4496 |
+
{
|
| 4497 |
+
"epoch": 30.09,
|
| 4498 |
+
"learning_rate": 2.655840754321634e-05,
|
| 4499 |
+
"loss": 0.048,
|
| 4500 |
+
"step": 6980
|
| 4501 |
+
},
|
| 4502 |
+
{
|
| 4503 |
+
"epoch": 30.13,
|
| 4504 |
+
"learning_rate": 2.6165531691985332e-05,
|
| 4505 |
+
"loss": 0.0873,
|
| 4506 |
+
"step": 6990
|
| 4507 |
+
},
|
| 4508 |
+
{
|
| 4509 |
+
"epoch": 30.17,
|
| 4510 |
+
"learning_rate": 2.577265584075432e-05,
|
| 4511 |
+
"loss": 0.0555,
|
| 4512 |
+
"step": 7000
|
| 4513 |
+
},
|
| 4514 |
+
{
|
| 4515 |
+
"epoch": 30.22,
|
| 4516 |
+
"learning_rate": 2.537977998952331e-05,
|
| 4517 |
+
"loss": 0.058,
|
| 4518 |
+
"step": 7010
|
| 4519 |
+
},
|
| 4520 |
+
{
|
| 4521 |
+
"epoch": 30.26,
|
| 4522 |
+
"learning_rate": 2.49869041382923e-05,
|
| 4523 |
+
"loss": 0.0639,
|
| 4524 |
+
"step": 7020
|
| 4525 |
+
},
|
| 4526 |
+
{
|
| 4527 |
+
"epoch": 30.3,
|
| 4528 |
+
"learning_rate": 2.4594028287061287e-05,
|
| 4529 |
+
"loss": 0.0592,
|
| 4530 |
+
"step": 7030
|
| 4531 |
+
},
|
| 4532 |
+
{
|
| 4533 |
+
"epoch": 30.34,
|
| 4534 |
+
"learning_rate": 2.4201152435830276e-05,
|
| 4535 |
+
"loss": 0.053,
|
| 4536 |
+
"step": 7040
|
| 4537 |
+
},
|
| 4538 |
+
{
|
| 4539 |
+
"epoch": 30.39,
|
| 4540 |
+
"learning_rate": 2.380827658459926e-05,
|
| 4541 |
+
"loss": 0.0677,
|
| 4542 |
+
"step": 7050
|
| 4543 |
+
},
|
| 4544 |
+
{
|
| 4545 |
+
"epoch": 30.43,
|
| 4546 |
+
"learning_rate": 2.341540073336825e-05,
|
| 4547 |
+
"loss": 0.0879,
|
| 4548 |
+
"step": 7060
|
| 4549 |
+
},
|
| 4550 |
+
{
|
| 4551 |
+
"epoch": 30.47,
|
| 4552 |
+
"learning_rate": 2.3022524882137243e-05,
|
| 4553 |
+
"loss": 0.0505,
|
| 4554 |
+
"step": 7070
|
| 4555 |
+
},
|
| 4556 |
+
{
|
| 4557 |
+
"epoch": 30.52,
|
| 4558 |
+
"learning_rate": 2.262964903090623e-05,
|
| 4559 |
+
"loss": 0.0496,
|
| 4560 |
+
"step": 7080
|
| 4561 |
+
},
|
| 4562 |
+
{
|
| 4563 |
+
"epoch": 30.56,
|
| 4564 |
+
"learning_rate": 2.223677317967522e-05,
|
| 4565 |
+
"loss": 0.0481,
|
| 4566 |
+
"step": 7090
|
| 4567 |
+
},
|
| 4568 |
+
{
|
| 4569 |
+
"epoch": 30.6,
|
| 4570 |
+
"learning_rate": 2.184389732844421e-05,
|
| 4571 |
+
"loss": 0.0528,
|
| 4572 |
+
"step": 7100
|
| 4573 |
+
},
|
| 4574 |
+
{
|
| 4575 |
+
"epoch": 30.65,
|
| 4576 |
+
"learning_rate": 2.1451021477213198e-05,
|
| 4577 |
+
"loss": 0.0675,
|
| 4578 |
+
"step": 7110
|
| 4579 |
+
},
|
| 4580 |
+
{
|
| 4581 |
+
"epoch": 30.69,
|
| 4582 |
+
"learning_rate": 2.1058145625982187e-05,
|
| 4583 |
+
"loss": 0.077,
|
| 4584 |
+
"step": 7120
|
| 4585 |
+
},
|
| 4586 |
+
{
|
| 4587 |
+
"epoch": 30.73,
|
| 4588 |
+
"learning_rate": 2.0665269774751176e-05,
|
| 4589 |
+
"loss": 0.1024,
|
| 4590 |
+
"step": 7130
|
| 4591 |
+
},
|
| 4592 |
+
{
|
| 4593 |
+
"epoch": 30.77,
|
| 4594 |
+
"learning_rate": 2.0272393923520165e-05,
|
| 4595 |
+
"loss": 0.0617,
|
| 4596 |
+
"step": 7140
|
| 4597 |
+
},
|
| 4598 |
+
{
|
| 4599 |
+
"epoch": 30.82,
|
| 4600 |
+
"learning_rate": 1.9879518072289157e-05,
|
| 4601 |
+
"loss": 0.0541,
|
| 4602 |
+
"step": 7150
|
| 4603 |
+
},
|
| 4604 |
+
{
|
| 4605 |
+
"epoch": 30.86,
|
| 4606 |
+
"learning_rate": 1.9486642221058146e-05,
|
| 4607 |
+
"loss": 0.0581,
|
| 4608 |
+
"step": 7160
|
| 4609 |
+
},
|
| 4610 |
+
{
|
| 4611 |
+
"epoch": 30.9,
|
| 4612 |
+
"learning_rate": 1.9093766369827135e-05,
|
| 4613 |
+
"loss": 0.0542,
|
| 4614 |
+
"step": 7170
|
| 4615 |
+
},
|
| 4616 |
+
{
|
| 4617 |
+
"epoch": 30.95,
|
| 4618 |
+
"learning_rate": 1.870089051859612e-05,
|
| 4619 |
+
"loss": 0.0494,
|
| 4620 |
+
"step": 7180
|
| 4621 |
+
},
|
| 4622 |
+
{
|
| 4623 |
+
"epoch": 30.99,
|
| 4624 |
+
"learning_rate": 1.830801466736511e-05,
|
| 4625 |
+
"loss": 0.0637,
|
| 4626 |
+
"step": 7190
|
| 4627 |
+
},
|
| 4628 |
+
{
|
| 4629 |
+
"epoch": 31.0,
|
| 4630 |
+
"eval_loss": 0.5002039074897766,
|
| 4631 |
+
"eval_runtime": 155.4344,
|
| 4632 |
+
"eval_samples_per_second": 9.792,
|
| 4633 |
+
"eval_wer": 0.42343797315029774,
|
| 4634 |
+
"step": 7192
|
| 4635 |
+
},
|
| 4636 |
+
{
|
| 4637 |
+
"epoch": 31.03,
|
| 4638 |
+
"learning_rate": 1.79151388161341e-05,
|
| 4639 |
+
"loss": 0.0592,
|
| 4640 |
+
"step": 7200
|
| 4641 |
+
},
|
| 4642 |
+
{
|
| 4643 |
+
"epoch": 31.08,
|
| 4644 |
+
"learning_rate": 1.752226296490309e-05,
|
| 4645 |
+
"loss": 0.083,
|
| 4646 |
+
"step": 7210
|
| 4647 |
+
},
|
| 4648 |
+
{
|
| 4649 |
+
"epoch": 31.12,
|
| 4650 |
+
"learning_rate": 1.712938711367208e-05,
|
| 4651 |
+
"loss": 0.0548,
|
| 4652 |
+
"step": 7220
|
| 4653 |
+
},
|
| 4654 |
+
{
|
| 4655 |
+
"epoch": 31.16,
|
| 4656 |
+
"learning_rate": 1.6736511262441068e-05,
|
| 4657 |
+
"loss": 0.0679,
|
| 4658 |
+
"step": 7230
|
| 4659 |
+
},
|
| 4660 |
+
{
|
| 4661 |
+
"epoch": 31.21,
|
| 4662 |
+
"learning_rate": 1.6343635411210057e-05,
|
| 4663 |
+
"loss": 0.0572,
|
| 4664 |
+
"step": 7240
|
| 4665 |
+
},
|
| 4666 |
+
{
|
| 4667 |
+
"epoch": 31.25,
|
| 4668 |
+
"learning_rate": 1.5950759559979046e-05,
|
| 4669 |
+
"loss": 0.0431,
|
| 4670 |
+
"step": 7250
|
| 4671 |
+
},
|
| 4672 |
+
{
|
| 4673 |
+
"epoch": 31.29,
|
| 4674 |
+
"learning_rate": 1.5557883708748035e-05,
|
| 4675 |
+
"loss": 0.0493,
|
| 4676 |
+
"step": 7260
|
| 4677 |
+
},
|
| 4678 |
+
{
|
| 4679 |
+
"epoch": 31.34,
|
| 4680 |
+
"learning_rate": 1.5165007857517022e-05,
|
| 4681 |
+
"loss": 0.0553,
|
| 4682 |
+
"step": 7270
|
| 4683 |
+
},
|
| 4684 |
+
{
|
| 4685 |
+
"epoch": 31.38,
|
| 4686 |
+
"learning_rate": 1.4772132006286012e-05,
|
| 4687 |
+
"loss": 0.0437,
|
| 4688 |
+
"step": 7280
|
| 4689 |
+
},
|
| 4690 |
+
{
|
| 4691 |
+
"epoch": 31.42,
|
| 4692 |
+
"learning_rate": 1.4379256155055001e-05,
|
| 4693 |
+
"loss": 0.0407,
|
| 4694 |
+
"step": 7290
|
| 4695 |
+
},
|
| 4696 |
+
{
|
| 4697 |
+
"epoch": 31.46,
|
| 4698 |
+
"learning_rate": 1.398638030382399e-05,
|
| 4699 |
+
"loss": 0.0591,
|
| 4700 |
+
"step": 7300
|
| 4701 |
+
},
|
| 4702 |
+
{
|
| 4703 |
+
"epoch": 31.51,
|
| 4704 |
+
"learning_rate": 1.3593504452592979e-05,
|
| 4705 |
+
"loss": 0.062,
|
| 4706 |
+
"step": 7310
|
| 4707 |
+
},
|
| 4708 |
+
{
|
| 4709 |
+
"epoch": 31.55,
|
| 4710 |
+
"learning_rate": 1.320062860136197e-05,
|
| 4711 |
+
"loss": 0.0405,
|
| 4712 |
+
"step": 7320
|
| 4713 |
+
},
|
| 4714 |
+
{
|
| 4715 |
+
"epoch": 31.59,
|
| 4716 |
+
"learning_rate": 1.2807752750130958e-05,
|
| 4717 |
+
"loss": 0.0548,
|
| 4718 |
+
"step": 7330
|
| 4719 |
+
},
|
| 4720 |
+
{
|
| 4721 |
+
"epoch": 31.64,
|
| 4722 |
+
"learning_rate": 1.2414876898899946e-05,
|
| 4723 |
+
"loss": 0.0573,
|
| 4724 |
+
"step": 7340
|
| 4725 |
+
},
|
| 4726 |
+
{
|
| 4727 |
+
"epoch": 31.68,
|
| 4728 |
+
"learning_rate": 1.2022001047668934e-05,
|
| 4729 |
+
"loss": 0.0693,
|
| 4730 |
+
"step": 7350
|
| 4731 |
+
},
|
| 4732 |
+
{
|
| 4733 |
+
"epoch": 31.72,
|
| 4734 |
+
"learning_rate": 1.1629125196437925e-05,
|
| 4735 |
+
"loss": 0.1054,
|
| 4736 |
+
"step": 7360
|
| 4737 |
+
},
|
| 4738 |
+
{
|
| 4739 |
+
"epoch": 31.77,
|
| 4740 |
+
"learning_rate": 1.1236249345206914e-05,
|
| 4741 |
+
"loss": 0.0425,
|
| 4742 |
+
"step": 7370
|
| 4743 |
+
},
|
| 4744 |
+
{
|
| 4745 |
+
"epoch": 31.81,
|
| 4746 |
+
"learning_rate": 1.0843373493975903e-05,
|
| 4747 |
+
"loss": 0.0618,
|
| 4748 |
+
"step": 7380
|
| 4749 |
+
},
|
| 4750 |
+
{
|
| 4751 |
+
"epoch": 31.85,
|
| 4752 |
+
"learning_rate": 1.045049764274489e-05,
|
| 4753 |
+
"loss": 0.0534,
|
| 4754 |
+
"step": 7390
|
| 4755 |
+
},
|
| 4756 |
+
{
|
| 4757 |
+
"epoch": 31.89,
|
| 4758 |
+
"learning_rate": 1.005762179151388e-05,
|
| 4759 |
+
"loss": 0.0515,
|
| 4760 |
+
"step": 7400
|
| 4761 |
+
},
|
| 4762 |
+
{
|
| 4763 |
+
"epoch": 31.94,
|
| 4764 |
+
"learning_rate": 9.66474594028287e-06,
|
| 4765 |
+
"loss": 0.0827,
|
| 4766 |
+
"step": 7410
|
| 4767 |
+
},
|
| 4768 |
+
{
|
| 4769 |
+
"epoch": 31.98,
|
| 4770 |
+
"learning_rate": 9.27187008905186e-06,
|
| 4771 |
+
"loss": 0.0635,
|
| 4772 |
+
"step": 7420
|
| 4773 |
+
},
|
| 4774 |
+
{
|
| 4775 |
+
"epoch": 32.0,
|
| 4776 |
+
"eval_loss": 0.5024468898773193,
|
| 4777 |
+
"eval_runtime": 158.8322,
|
| 4778 |
+
"eval_samples_per_second": 9.582,
|
| 4779 |
+
"eval_wer": 0.4231351569597254,
|
| 4780 |
+
"step": 7424
|
| 4781 |
+
},
|
| 4782 |
+
{
|
| 4783 |
+
"epoch": 32.03,
|
| 4784 |
+
"learning_rate": 8.878994237820847e-06,
|
| 4785 |
+
"loss": 0.0644,
|
| 4786 |
+
"step": 7430
|
| 4787 |
+
},
|
| 4788 |
+
{
|
| 4789 |
+
"epoch": 32.07,
|
| 4790 |
+
"learning_rate": 8.486118386589838e-06,
|
| 4791 |
+
"loss": 0.0466,
|
| 4792 |
+
"step": 7440
|
| 4793 |
+
},
|
| 4794 |
+
{
|
| 4795 |
+
"epoch": 32.11,
|
| 4796 |
+
"learning_rate": 8.093242535358827e-06,
|
| 4797 |
+
"loss": 0.064,
|
| 4798 |
+
"step": 7450
|
| 4799 |
+
},
|
| 4800 |
+
{
|
| 4801 |
+
"epoch": 32.15,
|
| 4802 |
+
"learning_rate": 7.700366684127815e-06,
|
| 4803 |
+
"loss": 0.0596,
|
| 4804 |
+
"step": 7460
|
| 4805 |
+
},
|
| 4806 |
+
{
|
| 4807 |
+
"epoch": 32.2,
|
| 4808 |
+
"learning_rate": 7.3074908328968034e-06,
|
| 4809 |
+
"loss": 0.0634,
|
| 4810 |
+
"step": 7470
|
| 4811 |
+
},
|
| 4812 |
+
{
|
| 4813 |
+
"epoch": 32.24,
|
| 4814 |
+
"learning_rate": 6.914614981665793e-06,
|
| 4815 |
+
"loss": 0.0476,
|
| 4816 |
+
"step": 7480
|
| 4817 |
+
},
|
| 4818 |
+
{
|
| 4819 |
+
"epoch": 32.28,
|
| 4820 |
+
"learning_rate": 6.521739130434782e-06,
|
| 4821 |
+
"loss": 0.0367,
|
| 4822 |
+
"step": 7490
|
| 4823 |
+
},
|
| 4824 |
+
{
|
| 4825 |
+
"epoch": 32.33,
|
| 4826 |
+
"learning_rate": 6.128863279203772e-06,
|
| 4827 |
+
"loss": 0.0649,
|
| 4828 |
+
"step": 7500
|
| 4829 |
+
},
|
| 4830 |
+
{
|
| 4831 |
+
"epoch": 32.37,
|
| 4832 |
+
"learning_rate": 5.73598742797276e-06,
|
| 4833 |
+
"loss": 0.0461,
|
| 4834 |
+
"step": 7510
|
| 4835 |
+
},
|
| 4836 |
+
{
|
| 4837 |
+
"epoch": 32.41,
|
| 4838 |
+
"learning_rate": 5.3431115767417495e-06,
|
| 4839 |
+
"loss": 0.0726,
|
| 4840 |
+
"step": 7520
|
| 4841 |
+
},
|
| 4842 |
+
{
|
| 4843 |
+
"epoch": 32.46,
|
| 4844 |
+
"learning_rate": 4.9502357255107375e-06,
|
| 4845 |
+
"loss": 0.062,
|
| 4846 |
+
"step": 7530
|
| 4847 |
+
},
|
| 4848 |
+
{
|
| 4849 |
+
"epoch": 32.5,
|
| 4850 |
+
"learning_rate": 4.557359874279727e-06,
|
| 4851 |
+
"loss": 0.0567,
|
| 4852 |
+
"step": 7540
|
| 4853 |
+
},
|
| 4854 |
+
{
|
| 4855 |
+
"epoch": 32.54,
|
| 4856 |
+
"learning_rate": 4.164484023048716e-06,
|
| 4857 |
+
"loss": 0.0706,
|
| 4858 |
+
"step": 7550
|
| 4859 |
+
},
|
| 4860 |
+
{
|
| 4861 |
+
"epoch": 32.58,
|
| 4862 |
+
"learning_rate": 3.7716081718177054e-06,
|
| 4863 |
+
"loss": 0.0568,
|
| 4864 |
+
"step": 7560
|
| 4865 |
+
},
|
| 4866 |
+
{
|
| 4867 |
+
"epoch": 32.63,
|
| 4868 |
+
"learning_rate": 3.3787323205866943e-06,
|
| 4869 |
+
"loss": 0.0491,
|
| 4870 |
+
"step": 7570
|
| 4871 |
+
},
|
| 4872 |
+
{
|
| 4873 |
+
"epoch": 32.67,
|
| 4874 |
+
"learning_rate": 2.9858564693556836e-06,
|
| 4875 |
+
"loss": 0.0654,
|
| 4876 |
+
"step": 7580
|
| 4877 |
+
},
|
| 4878 |
+
{
|
| 4879 |
+
"epoch": 32.71,
|
| 4880 |
+
"learning_rate": 2.5929806181246724e-06,
|
| 4881 |
+
"loss": 0.0648,
|
| 4882 |
+
"step": 7590
|
| 4883 |
+
},
|
| 4884 |
+
{
|
| 4885 |
+
"epoch": 32.76,
|
| 4886 |
+
"learning_rate": 2.2001047668936617e-06,
|
| 4887 |
+
"loss": 0.0631,
|
| 4888 |
+
"step": 7600
|
| 4889 |
+
},
|
| 4890 |
+
{
|
| 4891 |
+
"epoch": 32.8,
|
| 4892 |
+
"learning_rate": 1.8072289156626506e-06,
|
| 4893 |
+
"loss": 0.0488,
|
| 4894 |
+
"step": 7610
|
| 4895 |
+
},
|
| 4896 |
+
{
|
| 4897 |
+
"epoch": 32.84,
|
| 4898 |
+
"learning_rate": 1.4143530644316395e-06,
|
| 4899 |
+
"loss": 0.0547,
|
| 4900 |
+
"step": 7620
|
| 4901 |
+
},
|
| 4902 |
+
{
|
| 4903 |
+
"epoch": 32.89,
|
| 4904 |
+
"learning_rate": 1.0214772132006286e-06,
|
| 4905 |
+
"loss": 0.061,
|
| 4906 |
+
"step": 7630
|
| 4907 |
+
},
|
| 4908 |
+
{
|
| 4909 |
+
"epoch": 32.93,
|
| 4910 |
+
"learning_rate": 6.286013619696175e-07,
|
| 4911 |
+
"loss": 0.0495,
|
| 4912 |
+
"step": 7640
|
| 4913 |
+
},
|
| 4914 |
+
{
|
| 4915 |
+
"epoch": 32.97,
|
| 4916 |
+
"learning_rate": 2.357255107386066e-07,
|
| 4917 |
+
"loss": 0.0501,
|
| 4918 |
+
"step": 7650
|
| 4919 |
+
},
|
| 4920 |
+
{
|
| 4921 |
+
"epoch": 33.0,
|
| 4922 |
+
"eval_loss": 0.49998876452445984,
|
| 4923 |
+
"eval_runtime": 161.3568,
|
| 4924 |
+
"eval_samples_per_second": 9.433,
|
| 4925 |
+
"eval_wer": 0.42152013727667303,
|
| 4926 |
+
"step": 7656
|
| 4927 |
+
},
|
| 4928 |
+
{
|
| 4929 |
+
"epoch": 33.0,
|
| 4930 |
+
"step": 7656,
|
| 4931 |
+
"total_flos": 4.638603148454345e+19,
|
| 4932 |
+
"train_runtime": 3655.2719,
|
| 4933 |
+
"train_samples_per_second": 2.095
|
| 4934 |
}
|
| 4935 |
],
|
| 4936 |
+
"max_steps": 7656,
|
| 4937 |
+
"num_train_epochs": 33,
|
| 4938 |
+
"total_flos": 4.638603148454345e+19,
|
| 4939 |
"trial_name": null,
|
| 4940 |
"trial_params": null
|
| 4941 |
}
|
training_args.bin
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 2351
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:cb557f92a18c44747d9bda5647e3c4c26918f83d098cf2176b9933f5b2902c2d
|
| 3 |
size 2351
|