"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-75162 → checkpoint-75787}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-75162 → checkpoint-75787}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-75162 → checkpoint-75787}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-75162 → checkpoint-75787}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-75162 → checkpoint-75787}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-75162 → checkpoint-75787}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-75162 → checkpoint-75787}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-75162 → checkpoint-75787}/trainer_state.json +798 -3
- model-bin/finetune/base/{checkpoint-75162 → checkpoint-75787}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629882392.1236827/events.out.tfevents.1629882392.7e498afd5545.905.93 +3 -0
- model-bin/finetune/base/log/1629883056.4519293/events.out.tfevents.1629883056.7e498afd5545.905.95 +3 -0
- model-bin/finetune/base/log/1629883697.070274/events.out.tfevents.1629883697.7e498afd5545.905.97 +3 -0
- model-bin/finetune/base/log/1629884340.6409886/events.out.tfevents.1629884340.7e498afd5545.905.99 +3 -0
- model-bin/finetune/base/log/1629885006.4519074/events.out.tfevents.1629885006.7e498afd5545.905.101 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629882392.7e498afd5545.905.92 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629883056.7e498afd5545.905.94 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629883697.7e498afd5545.905.96 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629884340.7e498afd5545.905.98 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629885006.7e498afd5545.905.100 +3 -0
model-bin/finetune/base/{checkpoint-75162 → checkpoint-75787}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-75162 → checkpoint-75787}/optimizer.pt
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 722165393
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:241cf9a0d25f6dfe6c4b09f0250d75c7b8556e0708ad5a758a4ae728e99ed97f
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-75162 → checkpoint-75787}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-75162 → checkpoint-75787}/pytorch_model.bin
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 377909911
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:07eddd7e3b2c639376f65debd5c6e71a40abb39191deddef03737b5becdd9cb4
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-75162 → checkpoint-75787}/rng_state.pth
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 14503
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ff20bb0fbeb061a0d474c60bb2f68154a4d0944df8867df4e634564652ff6a60
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-75162 → checkpoint-75787}/scaler.pt
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 559
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c6aaef4b9249c4bb80c33d2e747e6a2aee3a93d57119debf58fe887e8e98126a
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-75162 → checkpoint-75787}/scheduler.pt
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 623
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8d36509b0709cb1c6a514d0b59c68cd5a3a2275d7fb28dc5813d880c7d4a4f0e
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-75162 → checkpoint-75787}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.18412114350410416,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-69565",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -194994,11 +194994,806 @@
|
|
| 194994 |
"eval_steps_per_second": 0.665,
|
| 194995 |
"eval_wer": 0.2013764427557531,
|
| 194996 |
"step": 75162
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 194997 |
}
|
| 194998 |
],
|
| 194999 |
"max_steps": 625000,
|
| 195000 |
"num_train_epochs": 5000,
|
| 195001 |
-
"total_flos": 2.
|
| 195002 |
"trial_name": null,
|
| 195003 |
"trial_params": null
|
| 195004 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.18412114350410416,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-69565",
|
| 4 |
+
"epoch": 605.9960159362549,
|
| 5 |
+
"global_step": 75787,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 194994 |
"eval_steps_per_second": 0.665,
|
| 194995 |
"eval_wer": 0.2013764427557531,
|
| 194996 |
"step": 75162
|
| 194997 |
+
},
|
| 194998 |
+
{
|
| 194999 |
+
"epoch": 601.02,
|
| 195000 |
+
"learning_rate": 8.811923076923078e-06,
|
| 195001 |
+
"loss": 0.3253,
|
| 195002 |
+
"step": 75165
|
| 195003 |
+
},
|
| 195004 |
+
{
|
| 195005 |
+
"epoch": 601.06,
|
| 195006 |
+
"learning_rate": 8.811842948717949e-06,
|
| 195007 |
+
"loss": 0.3342,
|
| 195008 |
+
"step": 75170
|
| 195009 |
+
},
|
| 195010 |
+
{
|
| 195011 |
+
"epoch": 601.1,
|
| 195012 |
+
"learning_rate": 8.811762820512822e-06,
|
| 195013 |
+
"loss": 0.383,
|
| 195014 |
+
"step": 75175
|
| 195015 |
+
},
|
| 195016 |
+
{
|
| 195017 |
+
"epoch": 601.14,
|
| 195018 |
+
"learning_rate": 8.811682692307693e-06,
|
| 195019 |
+
"loss": 0.4493,
|
| 195020 |
+
"step": 75180
|
| 195021 |
+
},
|
| 195022 |
+
{
|
| 195023 |
+
"epoch": 601.18,
|
| 195024 |
+
"learning_rate": 8.811602564102565e-06,
|
| 195025 |
+
"loss": 0.7843,
|
| 195026 |
+
"step": 75185
|
| 195027 |
+
},
|
| 195028 |
+
{
|
| 195029 |
+
"epoch": 601.22,
|
| 195030 |
+
"learning_rate": 8.811522435897438e-06,
|
| 195031 |
+
"loss": 0.8807,
|
| 195032 |
+
"step": 75190
|
| 195033 |
+
},
|
| 195034 |
+
{
|
| 195035 |
+
"epoch": 601.26,
|
| 195036 |
+
"learning_rate": 8.811442307692309e-06,
|
| 195037 |
+
"loss": 0.2823,
|
| 195038 |
+
"step": 75195
|
| 195039 |
+
},
|
| 195040 |
+
{
|
| 195041 |
+
"epoch": 601.3,
|
| 195042 |
+
"learning_rate": 8.81136217948718e-06,
|
| 195043 |
+
"loss": 0.387,
|
| 195044 |
+
"step": 75200
|
| 195045 |
+
},
|
| 195046 |
+
{
|
| 195047 |
+
"epoch": 601.34,
|
| 195048 |
+
"learning_rate": 8.811282051282052e-06,
|
| 195049 |
+
"loss": 0.3841,
|
| 195050 |
+
"step": 75205
|
| 195051 |
+
},
|
| 195052 |
+
{
|
| 195053 |
+
"epoch": 601.38,
|
| 195054 |
+
"learning_rate": 8.811201923076925e-06,
|
| 195055 |
+
"loss": 0.7324,
|
| 195056 |
+
"step": 75210
|
| 195057 |
+
},
|
| 195058 |
+
{
|
| 195059 |
+
"epoch": 601.42,
|
| 195060 |
+
"learning_rate": 8.811121794871795e-06,
|
| 195061 |
+
"loss": 0.8773,
|
| 195062 |
+
"step": 75215
|
| 195063 |
+
},
|
| 195064 |
+
{
|
| 195065 |
+
"epoch": 601.46,
|
| 195066 |
+
"learning_rate": 8.811041666666668e-06,
|
| 195067 |
+
"loss": 0.2424,
|
| 195068 |
+
"step": 75220
|
| 195069 |
+
},
|
| 195070 |
+
{
|
| 195071 |
+
"epoch": 601.5,
|
| 195072 |
+
"learning_rate": 8.810961538461539e-06,
|
| 195073 |
+
"loss": 0.366,
|
| 195074 |
+
"step": 75225
|
| 195075 |
+
},
|
| 195076 |
+
{
|
| 195077 |
+
"epoch": 601.54,
|
| 195078 |
+
"learning_rate": 8.81088141025641e-06,
|
| 195079 |
+
"loss": 0.5086,
|
| 195080 |
+
"step": 75230
|
| 195081 |
+
},
|
| 195082 |
+
{
|
| 195083 |
+
"epoch": 601.58,
|
| 195084 |
+
"learning_rate": 8.810801282051282e-06,
|
| 195085 |
+
"loss": 0.8468,
|
| 195086 |
+
"step": 75235
|
| 195087 |
+
},
|
| 195088 |
+
{
|
| 195089 |
+
"epoch": 601.62,
|
| 195090 |
+
"learning_rate": 8.810721153846155e-06,
|
| 195091 |
+
"loss": 1.0401,
|
| 195092 |
+
"step": 75240
|
| 195093 |
+
},
|
| 195094 |
+
{
|
| 195095 |
+
"epoch": 601.66,
|
| 195096 |
+
"learning_rate": 8.810641025641026e-06,
|
| 195097 |
+
"loss": 0.3034,
|
| 195098 |
+
"step": 75245
|
| 195099 |
+
},
|
| 195100 |
+
{
|
| 195101 |
+
"epoch": 601.7,
|
| 195102 |
+
"learning_rate": 8.810560897435897e-06,
|
| 195103 |
+
"loss": 0.3393,
|
| 195104 |
+
"step": 75250
|
| 195105 |
+
},
|
| 195106 |
+
{
|
| 195107 |
+
"epoch": 601.74,
|
| 195108 |
+
"learning_rate": 8.81048076923077e-06,
|
| 195109 |
+
"loss": 0.3378,
|
| 195110 |
+
"step": 75255
|
| 195111 |
+
},
|
| 195112 |
+
{
|
| 195113 |
+
"epoch": 601.78,
|
| 195114 |
+
"learning_rate": 8.810400641025642e-06,
|
| 195115 |
+
"loss": 0.7611,
|
| 195116 |
+
"step": 75260
|
| 195117 |
+
},
|
| 195118 |
+
{
|
| 195119 |
+
"epoch": 601.82,
|
| 195120 |
+
"learning_rate": 8.810320512820513e-06,
|
| 195121 |
+
"loss": 1.0786,
|
| 195122 |
+
"step": 75265
|
| 195123 |
+
},
|
| 195124 |
+
{
|
| 195125 |
+
"epoch": 601.86,
|
| 195126 |
+
"learning_rate": 8.810240384615385e-06,
|
| 195127 |
+
"loss": 0.3302,
|
| 195128 |
+
"step": 75270
|
| 195129 |
+
},
|
| 195130 |
+
{
|
| 195131 |
+
"epoch": 601.9,
|
| 195132 |
+
"learning_rate": 8.810160256410258e-06,
|
| 195133 |
+
"loss": 0.2704,
|
| 195134 |
+
"step": 75275
|
| 195135 |
+
},
|
| 195136 |
+
{
|
| 195137 |
+
"epoch": 601.94,
|
| 195138 |
+
"learning_rate": 8.810080128205129e-06,
|
| 195139 |
+
"loss": 0.4022,
|
| 195140 |
+
"step": 75280
|
| 195141 |
+
},
|
| 195142 |
+
{
|
| 195143 |
+
"epoch": 601.98,
|
| 195144 |
+
"learning_rate": 8.81e-06,
|
| 195145 |
+
"loss": 0.8502,
|
| 195146 |
+
"step": 75285
|
| 195147 |
+
},
|
| 195148 |
+
{
|
| 195149 |
+
"epoch": 602.0,
|
| 195150 |
+
"eval_loss": 0.3613823354244232,
|
| 195151 |
+
"eval_runtime": 40.6158,
|
| 195152 |
+
"eval_samples_per_second": 20.706,
|
| 195153 |
+
"eval_steps_per_second": 0.665,
|
| 195154 |
+
"eval_wer": 0.1880054390610463,
|
| 195155 |
+
"step": 75287
|
| 195156 |
+
},
|
| 195157 |
+
{
|
| 195158 |
+
"epoch": 602.02,
|
| 195159 |
+
"learning_rate": 8.809919871794873e-06,
|
| 195160 |
+
"loss": 0.3835,
|
| 195161 |
+
"step": 75290
|
| 195162 |
+
},
|
| 195163 |
+
{
|
| 195164 |
+
"epoch": 602.06,
|
| 195165 |
+
"learning_rate": 8.809839743589745e-06,
|
| 195166 |
+
"loss": 0.2873,
|
| 195167 |
+
"step": 75295
|
| 195168 |
+
},
|
| 195169 |
+
{
|
| 195170 |
+
"epoch": 602.1,
|
| 195171 |
+
"learning_rate": 8.809759615384616e-06,
|
| 195172 |
+
"loss": 0.2944,
|
| 195173 |
+
"step": 75300
|
| 195174 |
+
},
|
| 195175 |
+
{
|
| 195176 |
+
"epoch": 602.14,
|
| 195177 |
+
"learning_rate": 8.809679487179487e-06,
|
| 195178 |
+
"loss": 0.3795,
|
| 195179 |
+
"step": 75305
|
| 195180 |
+
},
|
| 195181 |
+
{
|
| 195182 |
+
"epoch": 602.18,
|
| 195183 |
+
"learning_rate": 8.80959935897436e-06,
|
| 195184 |
+
"loss": 0.8076,
|
| 195185 |
+
"step": 75310
|
| 195186 |
+
},
|
| 195187 |
+
{
|
| 195188 |
+
"epoch": 602.22,
|
| 195189 |
+
"learning_rate": 8.809519230769232e-06,
|
| 195190 |
+
"loss": 0.8927,
|
| 195191 |
+
"step": 75315
|
| 195192 |
+
},
|
| 195193 |
+
{
|
| 195194 |
+
"epoch": 602.26,
|
| 195195 |
+
"learning_rate": 8.809439102564103e-06,
|
| 195196 |
+
"loss": 0.3271,
|
| 195197 |
+
"step": 75320
|
| 195198 |
+
},
|
| 195199 |
+
{
|
| 195200 |
+
"epoch": 602.3,
|
| 195201 |
+
"learning_rate": 8.809358974358975e-06,
|
| 195202 |
+
"loss": 0.3014,
|
| 195203 |
+
"step": 75325
|
| 195204 |
+
},
|
| 195205 |
+
{
|
| 195206 |
+
"epoch": 602.34,
|
| 195207 |
+
"learning_rate": 8.809278846153848e-06,
|
| 195208 |
+
"loss": 0.449,
|
| 195209 |
+
"step": 75330
|
| 195210 |
+
},
|
| 195211 |
+
{
|
| 195212 |
+
"epoch": 602.38,
|
| 195213 |
+
"learning_rate": 8.809198717948717e-06,
|
| 195214 |
+
"loss": 0.9804,
|
| 195215 |
+
"step": 75335
|
| 195216 |
+
},
|
| 195217 |
+
{
|
| 195218 |
+
"epoch": 602.42,
|
| 195219 |
+
"learning_rate": 8.80911858974359e-06,
|
| 195220 |
+
"loss": 0.8483,
|
| 195221 |
+
"step": 75340
|
| 195222 |
+
},
|
| 195223 |
+
{
|
| 195224 |
+
"epoch": 602.46,
|
| 195225 |
+
"learning_rate": 8.809038461538463e-06,
|
| 195226 |
+
"loss": 0.307,
|
| 195227 |
+
"step": 75345
|
| 195228 |
+
},
|
| 195229 |
+
{
|
| 195230 |
+
"epoch": 602.5,
|
| 195231 |
+
"learning_rate": 8.808958333333335e-06,
|
| 195232 |
+
"loss": 0.3112,
|
| 195233 |
+
"step": 75350
|
| 195234 |
+
},
|
| 195235 |
+
{
|
| 195236 |
+
"epoch": 602.54,
|
| 195237 |
+
"learning_rate": 8.808878205128206e-06,
|
| 195238 |
+
"loss": 0.4685,
|
| 195239 |
+
"step": 75355
|
| 195240 |
+
},
|
| 195241 |
+
{
|
| 195242 |
+
"epoch": 602.58,
|
| 195243 |
+
"learning_rate": 8.808798076923078e-06,
|
| 195244 |
+
"loss": 0.8196,
|
| 195245 |
+
"step": 75360
|
| 195246 |
+
},
|
| 195247 |
+
{
|
| 195248 |
+
"epoch": 602.62,
|
| 195249 |
+
"learning_rate": 8.80871794871795e-06,
|
| 195250 |
+
"loss": 0.7176,
|
| 195251 |
+
"step": 75365
|
| 195252 |
+
},
|
| 195253 |
+
{
|
| 195254 |
+
"epoch": 602.66,
|
| 195255 |
+
"learning_rate": 8.80863782051282e-06,
|
| 195256 |
+
"loss": 0.3299,
|
| 195257 |
+
"step": 75370
|
| 195258 |
+
},
|
| 195259 |
+
{
|
| 195260 |
+
"epoch": 602.7,
|
| 195261 |
+
"learning_rate": 8.808557692307693e-06,
|
| 195262 |
+
"loss": 0.3395,
|
| 195263 |
+
"step": 75375
|
| 195264 |
+
},
|
| 195265 |
+
{
|
| 195266 |
+
"epoch": 602.74,
|
| 195267 |
+
"learning_rate": 8.808477564102565e-06,
|
| 195268 |
+
"loss": 0.482,
|
| 195269 |
+
"step": 75380
|
| 195270 |
+
},
|
| 195271 |
+
{
|
| 195272 |
+
"epoch": 602.78,
|
| 195273 |
+
"learning_rate": 8.808397435897436e-06,
|
| 195274 |
+
"loss": 0.7871,
|
| 195275 |
+
"step": 75385
|
| 195276 |
+
},
|
| 195277 |
+
{
|
| 195278 |
+
"epoch": 602.82,
|
| 195279 |
+
"learning_rate": 8.808317307692307e-06,
|
| 195280 |
+
"loss": 1.0059,
|
| 195281 |
+
"step": 75390
|
| 195282 |
+
},
|
| 195283 |
+
{
|
| 195284 |
+
"epoch": 602.86,
|
| 195285 |
+
"learning_rate": 8.80823717948718e-06,
|
| 195286 |
+
"loss": 0.289,
|
| 195287 |
+
"step": 75395
|
| 195288 |
+
},
|
| 195289 |
+
{
|
| 195290 |
+
"epoch": 602.9,
|
| 195291 |
+
"learning_rate": 8.808157051282052e-06,
|
| 195292 |
+
"loss": 0.3244,
|
| 195293 |
+
"step": 75400
|
| 195294 |
+
},
|
| 195295 |
+
{
|
| 195296 |
+
"epoch": 602.94,
|
| 195297 |
+
"learning_rate": 8.808076923076923e-06,
|
| 195298 |
+
"loss": 0.4663,
|
| 195299 |
+
"step": 75405
|
| 195300 |
+
},
|
| 195301 |
+
{
|
| 195302 |
+
"epoch": 602.98,
|
| 195303 |
+
"learning_rate": 8.807996794871796e-06,
|
| 195304 |
+
"loss": 0.7288,
|
| 195305 |
+
"step": 75410
|
| 195306 |
+
},
|
| 195307 |
+
{
|
| 195308 |
+
"epoch": 603.0,
|
| 195309 |
+
"eval_loss": 0.4494644105434418,
|
| 195310 |
+
"eval_runtime": 40.3969,
|
| 195311 |
+
"eval_samples_per_second": 20.818,
|
| 195312 |
+
"eval_steps_per_second": 0.668,
|
| 195313 |
+
"eval_wer": 0.18498835177635412,
|
| 195314 |
+
"step": 75412
|
| 195315 |
+
},
|
| 195316 |
+
{
|
| 195317 |
+
"epoch": 603.02,
|
| 195318 |
+
"learning_rate": 8.807916666666668e-06,
|
| 195319 |
+
"loss": 0.4818,
|
| 195320 |
+
"step": 75415
|
| 195321 |
+
},
|
| 195322 |
+
{
|
| 195323 |
+
"epoch": 603.06,
|
| 195324 |
+
"learning_rate": 8.807836538461539e-06,
|
| 195325 |
+
"loss": 0.2858,
|
| 195326 |
+
"step": 75420
|
| 195327 |
+
},
|
| 195328 |
+
{
|
| 195329 |
+
"epoch": 603.1,
|
| 195330 |
+
"learning_rate": 8.80775641025641e-06,
|
| 195331 |
+
"loss": 0.3195,
|
| 195332 |
+
"step": 75425
|
| 195333 |
+
},
|
| 195334 |
+
{
|
| 195335 |
+
"epoch": 603.14,
|
| 195336 |
+
"learning_rate": 8.807676282051283e-06,
|
| 195337 |
+
"loss": 0.3992,
|
| 195338 |
+
"step": 75430
|
| 195339 |
+
},
|
| 195340 |
+
{
|
| 195341 |
+
"epoch": 603.18,
|
| 195342 |
+
"learning_rate": 8.807596153846155e-06,
|
| 195343 |
+
"loss": 0.8451,
|
| 195344 |
+
"step": 75435
|
| 195345 |
+
},
|
| 195346 |
+
{
|
| 195347 |
+
"epoch": 603.22,
|
| 195348 |
+
"learning_rate": 8.807516025641026e-06,
|
| 195349 |
+
"loss": 0.8592,
|
| 195350 |
+
"step": 75440
|
| 195351 |
+
},
|
| 195352 |
+
{
|
| 195353 |
+
"epoch": 603.26,
|
| 195354 |
+
"learning_rate": 8.807435897435899e-06,
|
| 195355 |
+
"loss": 0.3101,
|
| 195356 |
+
"step": 75445
|
| 195357 |
+
},
|
| 195358 |
+
{
|
| 195359 |
+
"epoch": 603.3,
|
| 195360 |
+
"learning_rate": 8.80735576923077e-06,
|
| 195361 |
+
"loss": 0.3108,
|
| 195362 |
+
"step": 75450
|
| 195363 |
+
},
|
| 195364 |
+
{
|
| 195365 |
+
"epoch": 603.34,
|
| 195366 |
+
"learning_rate": 8.807275641025642e-06,
|
| 195367 |
+
"loss": 0.3803,
|
| 195368 |
+
"step": 75455
|
| 195369 |
+
},
|
| 195370 |
+
{
|
| 195371 |
+
"epoch": 603.38,
|
| 195372 |
+
"learning_rate": 8.807195512820513e-06,
|
| 195373 |
+
"loss": 0.8631,
|
| 195374 |
+
"step": 75460
|
| 195375 |
+
},
|
| 195376 |
+
{
|
| 195377 |
+
"epoch": 603.42,
|
| 195378 |
+
"learning_rate": 8.807115384615386e-06,
|
| 195379 |
+
"loss": 0.9733,
|
| 195380 |
+
"step": 75465
|
| 195381 |
+
},
|
| 195382 |
+
{
|
| 195383 |
+
"epoch": 603.46,
|
| 195384 |
+
"learning_rate": 8.807035256410258e-06,
|
| 195385 |
+
"loss": 0.3768,
|
| 195386 |
+
"step": 75470
|
| 195387 |
+
},
|
| 195388 |
+
{
|
| 195389 |
+
"epoch": 603.5,
|
| 195390 |
+
"learning_rate": 8.806955128205129e-06,
|
| 195391 |
+
"loss": 0.3623,
|
| 195392 |
+
"step": 75475
|
| 195393 |
+
},
|
| 195394 |
+
{
|
| 195395 |
+
"epoch": 603.54,
|
| 195396 |
+
"learning_rate": 8.806875e-06,
|
| 195397 |
+
"loss": 0.3788,
|
| 195398 |
+
"step": 75480
|
| 195399 |
+
},
|
| 195400 |
+
{
|
| 195401 |
+
"epoch": 603.58,
|
| 195402 |
+
"learning_rate": 8.806794871794873e-06,
|
| 195403 |
+
"loss": 0.7772,
|
| 195404 |
+
"step": 75485
|
| 195405 |
+
},
|
| 195406 |
+
{
|
| 195407 |
+
"epoch": 603.62,
|
| 195408 |
+
"learning_rate": 8.806714743589743e-06,
|
| 195409 |
+
"loss": 0.9261,
|
| 195410 |
+
"step": 75490
|
| 195411 |
+
},
|
| 195412 |
+
{
|
| 195413 |
+
"epoch": 603.66,
|
| 195414 |
+
"learning_rate": 8.806634615384616e-06,
|
| 195415 |
+
"loss": 0.352,
|
| 195416 |
+
"step": 75495
|
| 195417 |
+
},
|
| 195418 |
+
{
|
| 195419 |
+
"epoch": 603.7,
|
| 195420 |
+
"learning_rate": 8.806554487179489e-06,
|
| 195421 |
+
"loss": 0.3595,
|
| 195422 |
+
"step": 75500
|
| 195423 |
+
},
|
| 195424 |
+
{
|
| 195425 |
+
"epoch": 603.74,
|
| 195426 |
+
"learning_rate": 8.806474358974359e-06,
|
| 195427 |
+
"loss": 0.3123,
|
| 195428 |
+
"step": 75505
|
| 195429 |
+
},
|
| 195430 |
+
{
|
| 195431 |
+
"epoch": 603.78,
|
| 195432 |
+
"learning_rate": 8.806394230769232e-06,
|
| 195433 |
+
"loss": 0.7048,
|
| 195434 |
+
"step": 75510
|
| 195435 |
+
},
|
| 195436 |
+
{
|
| 195437 |
+
"epoch": 603.82,
|
| 195438 |
+
"learning_rate": 8.806314102564103e-06,
|
| 195439 |
+
"loss": 0.9546,
|
| 195440 |
+
"step": 75515
|
| 195441 |
+
},
|
| 195442 |
+
{
|
| 195443 |
+
"epoch": 603.86,
|
| 195444 |
+
"learning_rate": 8.806233974358975e-06,
|
| 195445 |
+
"loss": 0.29,
|
| 195446 |
+
"step": 75520
|
| 195447 |
+
},
|
| 195448 |
+
{
|
| 195449 |
+
"epoch": 603.9,
|
| 195450 |
+
"learning_rate": 8.806153846153846e-06,
|
| 195451 |
+
"loss": 0.3054,
|
| 195452 |
+
"step": 75525
|
| 195453 |
+
},
|
| 195454 |
+
{
|
| 195455 |
+
"epoch": 603.94,
|
| 195456 |
+
"learning_rate": 8.806073717948719e-06,
|
| 195457 |
+
"loss": 0.3936,
|
| 195458 |
+
"step": 75530
|
| 195459 |
+
},
|
| 195460 |
+
{
|
| 195461 |
+
"epoch": 603.98,
|
| 195462 |
+
"learning_rate": 8.80599358974359e-06,
|
| 195463 |
+
"loss": 0.9524,
|
| 195464 |
+
"step": 75535
|
| 195465 |
+
},
|
| 195466 |
+
{
|
| 195467 |
+
"epoch": 604.0,
|
| 195468 |
+
"eval_loss": 0.4324324429035187,
|
| 195469 |
+
"eval_runtime": 41.0453,
|
| 195470 |
+
"eval_samples_per_second": 20.49,
|
| 195471 |
+
"eval_steps_per_second": 0.658,
|
| 195472 |
+
"eval_wer": 0.18972477064220183,
|
| 195473 |
+
"step": 75537
|
| 195474 |
+
},
|
| 195475 |
+
{
|
| 195476 |
+
"epoch": 604.02,
|
| 195477 |
+
"learning_rate": 8.805913461538462e-06,
|
| 195478 |
+
"loss": 0.3299,
|
| 195479 |
+
"step": 75540
|
| 195480 |
+
},
|
| 195481 |
+
{
|
| 195482 |
+
"epoch": 604.06,
|
| 195483 |
+
"learning_rate": 8.805833333333335e-06,
|
| 195484 |
+
"loss": 0.3726,
|
| 195485 |
+
"step": 75545
|
| 195486 |
+
},
|
| 195487 |
+
{
|
| 195488 |
+
"epoch": 604.1,
|
| 195489 |
+
"learning_rate": 8.805753205128206e-06,
|
| 195490 |
+
"loss": 0.3474,
|
| 195491 |
+
"step": 75550
|
| 195492 |
+
},
|
| 195493 |
+
{
|
| 195494 |
+
"epoch": 604.14,
|
| 195495 |
+
"learning_rate": 8.805673076923077e-06,
|
| 195496 |
+
"loss": 0.3935,
|
| 195497 |
+
"step": 75555
|
| 195498 |
+
},
|
| 195499 |
+
{
|
| 195500 |
+
"epoch": 604.18,
|
| 195501 |
+
"learning_rate": 8.805592948717949e-06,
|
| 195502 |
+
"loss": 0.7502,
|
| 195503 |
+
"step": 75560
|
| 195504 |
+
},
|
| 195505 |
+
{
|
| 195506 |
+
"epoch": 604.22,
|
| 195507 |
+
"learning_rate": 8.805512820512822e-06,
|
| 195508 |
+
"loss": 0.8715,
|
| 195509 |
+
"step": 75565
|
| 195510 |
+
},
|
| 195511 |
+
{
|
| 195512 |
+
"epoch": 604.26,
|
| 195513 |
+
"learning_rate": 8.805432692307693e-06,
|
| 195514 |
+
"loss": 0.4304,
|
| 195515 |
+
"step": 75570
|
| 195516 |
+
},
|
| 195517 |
+
{
|
| 195518 |
+
"epoch": 604.3,
|
| 195519 |
+
"learning_rate": 8.805352564102565e-06,
|
| 195520 |
+
"loss": 0.3175,
|
| 195521 |
+
"step": 75575
|
| 195522 |
+
},
|
| 195523 |
+
{
|
| 195524 |
+
"epoch": 604.34,
|
| 195525 |
+
"learning_rate": 8.805272435897436e-06,
|
| 195526 |
+
"loss": 0.3441,
|
| 195527 |
+
"step": 75580
|
| 195528 |
+
},
|
| 195529 |
+
{
|
| 195530 |
+
"epoch": 604.38,
|
| 195531 |
+
"learning_rate": 8.805192307692309e-06,
|
| 195532 |
+
"loss": 0.7054,
|
| 195533 |
+
"step": 75585
|
| 195534 |
+
},
|
| 195535 |
+
{
|
| 195536 |
+
"epoch": 604.42,
|
| 195537 |
+
"learning_rate": 8.80511217948718e-06,
|
| 195538 |
+
"loss": 0.8604,
|
| 195539 |
+
"step": 75590
|
| 195540 |
+
},
|
| 195541 |
+
{
|
| 195542 |
+
"epoch": 604.46,
|
| 195543 |
+
"learning_rate": 8.805032051282052e-06,
|
| 195544 |
+
"loss": 0.2919,
|
| 195545 |
+
"step": 75595
|
| 195546 |
+
},
|
| 195547 |
+
{
|
| 195548 |
+
"epoch": 604.5,
|
| 195549 |
+
"learning_rate": 8.804951923076925e-06,
|
| 195550 |
+
"loss": 0.3158,
|
| 195551 |
+
"step": 75600
|
| 195552 |
+
},
|
| 195553 |
+
{
|
| 195554 |
+
"epoch": 604.54,
|
| 195555 |
+
"learning_rate": 8.804871794871796e-06,
|
| 195556 |
+
"loss": 0.4358,
|
| 195557 |
+
"step": 75605
|
| 195558 |
+
},
|
| 195559 |
+
{
|
| 195560 |
+
"epoch": 604.58,
|
| 195561 |
+
"learning_rate": 8.804791666666667e-06,
|
| 195562 |
+
"loss": 0.8265,
|
| 195563 |
+
"step": 75610
|
| 195564 |
+
},
|
| 195565 |
+
{
|
| 195566 |
+
"epoch": 604.62,
|
| 195567 |
+
"learning_rate": 8.804711538461539e-06,
|
| 195568 |
+
"loss": 0.8654,
|
| 195569 |
+
"step": 75615
|
| 195570 |
+
},
|
| 195571 |
+
{
|
| 195572 |
+
"epoch": 604.66,
|
| 195573 |
+
"learning_rate": 8.804631410256412e-06,
|
| 195574 |
+
"loss": 0.272,
|
| 195575 |
+
"step": 75620
|
| 195576 |
+
},
|
| 195577 |
+
{
|
| 195578 |
+
"epoch": 604.7,
|
| 195579 |
+
"learning_rate": 8.804551282051283e-06,
|
| 195580 |
+
"loss": 0.2909,
|
| 195581 |
+
"step": 75625
|
| 195582 |
+
},
|
| 195583 |
+
{
|
| 195584 |
+
"epoch": 604.74,
|
| 195585 |
+
"learning_rate": 8.804471153846155e-06,
|
| 195586 |
+
"loss": 0.4165,
|
| 195587 |
+
"step": 75630
|
| 195588 |
+
},
|
| 195589 |
+
{
|
| 195590 |
+
"epoch": 604.78,
|
| 195591 |
+
"learning_rate": 8.804391025641026e-06,
|
| 195592 |
+
"loss": 0.8363,
|
| 195593 |
+
"step": 75635
|
| 195594 |
+
},
|
| 195595 |
+
{
|
| 195596 |
+
"epoch": 604.82,
|
| 195597 |
+
"learning_rate": 8.804310897435899e-06,
|
| 195598 |
+
"loss": 0.8427,
|
| 195599 |
+
"step": 75640
|
| 195600 |
+
},
|
| 195601 |
+
{
|
| 195602 |
+
"epoch": 604.86,
|
| 195603 |
+
"learning_rate": 8.80423076923077e-06,
|
| 195604 |
+
"loss": 0.3652,
|
| 195605 |
+
"step": 75645
|
| 195606 |
+
},
|
| 195607 |
+
{
|
| 195608 |
+
"epoch": 604.9,
|
| 195609 |
+
"learning_rate": 8.804150641025642e-06,
|
| 195610 |
+
"loss": 0.3643,
|
| 195611 |
+
"step": 75650
|
| 195612 |
+
},
|
| 195613 |
+
{
|
| 195614 |
+
"epoch": 604.94,
|
| 195615 |
+
"learning_rate": 8.804070512820515e-06,
|
| 195616 |
+
"loss": 0.3866,
|
| 195617 |
+
"step": 75655
|
| 195618 |
+
},
|
| 195619 |
+
{
|
| 195620 |
+
"epoch": 604.98,
|
| 195621 |
+
"learning_rate": 8.803990384615384e-06,
|
| 195622 |
+
"loss": 0.8886,
|
| 195623 |
+
"step": 75660
|
| 195624 |
+
},
|
| 195625 |
+
{
|
| 195626 |
+
"epoch": 605.0,
|
| 195627 |
+
"eval_loss": 0.3445565700531006,
|
| 195628 |
+
"eval_runtime": 40.5466,
|
| 195629 |
+
"eval_samples_per_second": 20.717,
|
| 195630 |
+
"eval_steps_per_second": 0.666,
|
| 195631 |
+
"eval_wer": 0.18799435028248587,
|
| 195632 |
+
"step": 75662
|
| 195633 |
+
},
|
| 195634 |
+
{
|
| 195635 |
+
"epoch": 605.02,
|
| 195636 |
+
"learning_rate": 8.803910256410257e-06,
|
| 195637 |
+
"loss": 0.3486,
|
| 195638 |
+
"step": 75665
|
| 195639 |
+
},
|
| 195640 |
+
{
|
| 195641 |
+
"epoch": 605.06,
|
| 195642 |
+
"learning_rate": 8.803830128205129e-06,
|
| 195643 |
+
"loss": 0.3072,
|
| 195644 |
+
"step": 75670
|
| 195645 |
+
},
|
| 195646 |
+
{
|
| 195647 |
+
"epoch": 605.1,
|
| 195648 |
+
"learning_rate": 8.80375e-06,
|
| 195649 |
+
"loss": 0.2955,
|
| 195650 |
+
"step": 75675
|
| 195651 |
+
},
|
| 195652 |
+
{
|
| 195653 |
+
"epoch": 605.14,
|
| 195654 |
+
"learning_rate": 8.803669871794872e-06,
|
| 195655 |
+
"loss": 0.3452,
|
| 195656 |
+
"step": 75680
|
| 195657 |
+
},
|
| 195658 |
+
{
|
| 195659 |
+
"epoch": 605.18,
|
| 195660 |
+
"learning_rate": 8.803589743589745e-06,
|
| 195661 |
+
"loss": 0.716,
|
| 195662 |
+
"step": 75685
|
| 195663 |
+
},
|
| 195664 |
+
{
|
| 195665 |
+
"epoch": 605.22,
|
| 195666 |
+
"learning_rate": 8.803509615384616e-06,
|
| 195667 |
+
"loss": 0.9065,
|
| 195668 |
+
"step": 75690
|
| 195669 |
+
},
|
| 195670 |
+
{
|
| 195671 |
+
"epoch": 605.26,
|
| 195672 |
+
"learning_rate": 8.803429487179487e-06,
|
| 195673 |
+
"loss": 0.2987,
|
| 195674 |
+
"step": 75695
|
| 195675 |
+
},
|
| 195676 |
+
{
|
| 195677 |
+
"epoch": 605.3,
|
| 195678 |
+
"learning_rate": 8.80334935897436e-06,
|
| 195679 |
+
"loss": 0.3722,
|
| 195680 |
+
"step": 75700
|
| 195681 |
+
},
|
| 195682 |
+
{
|
| 195683 |
+
"epoch": 605.34,
|
| 195684 |
+
"learning_rate": 8.803269230769232e-06,
|
| 195685 |
+
"loss": 0.4089,
|
| 195686 |
+
"step": 75705
|
| 195687 |
+
},
|
| 195688 |
+
{
|
| 195689 |
+
"epoch": 605.38,
|
| 195690 |
+
"learning_rate": 8.803189102564103e-06,
|
| 195691 |
+
"loss": 0.7393,
|
| 195692 |
+
"step": 75710
|
| 195693 |
+
},
|
| 195694 |
+
{
|
| 195695 |
+
"epoch": 605.42,
|
| 195696 |
+
"learning_rate": 8.803108974358974e-06,
|
| 195697 |
+
"loss": 1.0048,
|
| 195698 |
+
"step": 75715
|
| 195699 |
+
},
|
| 195700 |
+
{
|
| 195701 |
+
"epoch": 605.46,
|
| 195702 |
+
"learning_rate": 8.803028846153848e-06,
|
| 195703 |
+
"loss": 0.3025,
|
| 195704 |
+
"step": 75720
|
| 195705 |
+
},
|
| 195706 |
+
{
|
| 195707 |
+
"epoch": 605.5,
|
| 195708 |
+
"learning_rate": 8.802948717948719e-06,
|
| 195709 |
+
"loss": 0.2752,
|
| 195710 |
+
"step": 75725
|
| 195711 |
+
},
|
| 195712 |
+
{
|
| 195713 |
+
"epoch": 605.54,
|
| 195714 |
+
"learning_rate": 8.80286858974359e-06,
|
| 195715 |
+
"loss": 0.4407,
|
| 195716 |
+
"step": 75730
|
| 195717 |
+
},
|
| 195718 |
+
{
|
| 195719 |
+
"epoch": 605.58,
|
| 195720 |
+
"learning_rate": 8.802788461538462e-06,
|
| 195721 |
+
"loss": 0.7918,
|
| 195722 |
+
"step": 75735
|
| 195723 |
+
},
|
| 195724 |
+
{
|
| 195725 |
+
"epoch": 605.62,
|
| 195726 |
+
"learning_rate": 8.802708333333335e-06,
|
| 195727 |
+
"loss": 0.8038,
|
| 195728 |
+
"step": 75740
|
| 195729 |
+
},
|
| 195730 |
+
{
|
| 195731 |
+
"epoch": 605.66,
|
| 195732 |
+
"learning_rate": 8.802628205128206e-06,
|
| 195733 |
+
"loss": 0.2619,
|
| 195734 |
+
"step": 75745
|
| 195735 |
+
},
|
| 195736 |
+
{
|
| 195737 |
+
"epoch": 605.7,
|
| 195738 |
+
"learning_rate": 8.802548076923077e-06,
|
| 195739 |
+
"loss": 0.2717,
|
| 195740 |
+
"step": 75750
|
| 195741 |
+
},
|
| 195742 |
+
{
|
| 195743 |
+
"epoch": 605.74,
|
| 195744 |
+
"learning_rate": 8.80246794871795e-06,
|
| 195745 |
+
"loss": 0.3758,
|
| 195746 |
+
"step": 75755
|
| 195747 |
+
},
|
| 195748 |
+
{
|
| 195749 |
+
"epoch": 605.78,
|
| 195750 |
+
"learning_rate": 8.802387820512822e-06,
|
| 195751 |
+
"loss": 0.862,
|
| 195752 |
+
"step": 75760
|
| 195753 |
+
},
|
| 195754 |
+
{
|
| 195755 |
+
"epoch": 605.82,
|
| 195756 |
+
"learning_rate": 8.802307692307693e-06,
|
| 195757 |
+
"loss": 0.8831,
|
| 195758 |
+
"step": 75765
|
| 195759 |
+
},
|
| 195760 |
+
{
|
| 195761 |
+
"epoch": 605.86,
|
| 195762 |
+
"learning_rate": 8.802227564102564e-06,
|
| 195763 |
+
"loss": 0.2855,
|
| 195764 |
+
"step": 75770
|
| 195765 |
+
},
|
| 195766 |
+
{
|
| 195767 |
+
"epoch": 605.9,
|
| 195768 |
+
"learning_rate": 8.802147435897438e-06,
|
| 195769 |
+
"loss": 0.2964,
|
| 195770 |
+
"step": 75775
|
| 195771 |
+
},
|
| 195772 |
+
{
|
| 195773 |
+
"epoch": 605.94,
|
| 195774 |
+
"learning_rate": 8.802067307692307e-06,
|
| 195775 |
+
"loss": 0.3788,
|
| 195776 |
+
"step": 75780
|
| 195777 |
+
},
|
| 195778 |
+
{
|
| 195779 |
+
"epoch": 605.98,
|
| 195780 |
+
"learning_rate": 8.80198717948718e-06,
|
| 195781 |
+
"loss": 0.7917,
|
| 195782 |
+
"step": 75785
|
| 195783 |
+
},
|
| 195784 |
+
{
|
| 195785 |
+
"epoch": 606.0,
|
| 195786 |
+
"eval_loss": 0.40229326486587524,
|
| 195787 |
+
"eval_runtime": 40.811,
|
| 195788 |
+
"eval_samples_per_second": 20.583,
|
| 195789 |
+
"eval_steps_per_second": 0.662,
|
| 195790 |
+
"eval_wer": 0.1924281241104469,
|
| 195791 |
+
"step": 75787
|
| 195792 |
}
|
| 195793 |
],
|
| 195794 |
"max_steps": 625000,
|
| 195795 |
"num_train_epochs": 5000,
|
| 195796 |
+
"total_flos": 2.1327406074811505e+20,
|
| 195797 |
"trial_name": null,
|
| 195798 |
"trial_params": null
|
| 195799 |
}
|
model-bin/finetune/base/{checkpoint-75162 → checkpoint-75787}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629882392.1236827/events.out.tfevents.1629882392.7e498afd5545.905.93
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:40faae71c44c577de7af7caf69225cd70fa91b8106891c9da59aeafbc4240253
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629883056.4519293/events.out.tfevents.1629883056.7e498afd5545.905.95
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:939b06a3739cc17a7e4511ffb94973599bbcb123803382e8d793b74decdad891
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629883697.070274/events.out.tfevents.1629883697.7e498afd5545.905.97
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:bd4a1c43c8d0f8393356223e62609d12fe99feeffdaa48864cd0affc5cd80930
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629884340.6409886/events.out.tfevents.1629884340.7e498afd5545.905.99
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:3047e66a47c520ca0d38a0885389812710a8a99fe0f4c1eee86b29e1ed75e104
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629885006.4519074/events.out.tfevents.1629885006.7e498afd5545.905.101
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:082cef36521c91b17fff53cd50db092c8004bc2f4c943a8df79734e4914e2dbf
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629882392.7e498afd5545.905.92
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e16d61233274ee5c7a30f50978aade662d5dab45638811ad3f304ccc8670268a
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629883056.7e498afd5545.905.94
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:90b55c10715a3dd6e440f8c50d728d3f72f21cd59abdfaebc6a4eb38fc95b3af
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629883697.7e498afd5545.905.96
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a9f668ebd3dee0e05fcda258dc1e05f0bfad91436085cd5c6757bb21b1221574
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629884340.7e498afd5545.905.98
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e943e824b462a7e1503eebe32d0a77c274a8edcba1936795e22fd327cc802000
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629885006.7e498afd5545.905.100
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:45aa3d1626d97711f20adbacc8607e2484e99e05409cd57762d9b59f5f3c47cb
|
| 3 |
+
size 8622
|