"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-136018 → checkpoint-136640}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-136018 → checkpoint-136640}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-136018 → checkpoint-136640}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-136018 → checkpoint-136640}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-136018 → checkpoint-136640}/rng_state.pth +2 -2
- model-bin/finetune/base/{checkpoint-136018 → checkpoint-136640}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-136018 → checkpoint-136640}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-136018 → checkpoint-136640}/trainer_state.json +798 -3
- model-bin/finetune/base/{checkpoint-136018 → checkpoint-136640}/training_args.bin +0 -0
- model-bin/finetune/base/log/1630181516.414431/events.out.tfevents.1630181516.86bb0ddabf9b.4092.291 +3 -0
- model-bin/finetune/base/log/1630181907.2896452/events.out.tfevents.1630181907.86bb0ddabf9b.4092.293 +3 -0
- model-bin/finetune/base/log/1630182294.8279276/events.out.tfevents.1630182294.86bb0ddabf9b.4092.295 +3 -0
- model-bin/finetune/base/log/1630182688.064181/events.out.tfevents.1630182688.86bb0ddabf9b.4092.297 +3 -0
- model-bin/finetune/base/log/1630183080.6762185/events.out.tfevents.1630183080.86bb0ddabf9b.4092.299 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630181516.86bb0ddabf9b.4092.290 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630181907.86bb0ddabf9b.4092.292 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630182294.86bb0ddabf9b.4092.294 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630182688.86bb0ddabf9b.4092.296 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630183080.86bb0ddabf9b.4092.298 +3 -0
model-bin/finetune/base/{checkpoint-136018 → checkpoint-136640}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-136018 → checkpoint-136640}/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 sha256:56de4e35656eb741ff3ea00318b88884b8ca24c5196ca2cef940b211904819af
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-136018 → checkpoint-136640}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-136018 → checkpoint-136640}/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:fdbcb35f9616d9f7b966ddcc2e861edf71cf2755e4ef303326aa80d9a1cfe921
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-136018 → checkpoint-136640}/rng_state.pth
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:762b83a9c01114f1c7ec44e1f00e63f34066c4d63b37802d08eb80cc4d8194f5
|
| 3 |
+
size 14503
|
model-bin/finetune/base/{checkpoint-136018 → checkpoint-136640}/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:ac900979db688ef403a10bb8c56504634eaf5e168c8e0bb8562ff8478fde8c5c
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-136018 → checkpoint-136640}/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:e0f162b0763cec4419b76da81b5ab9dd5336248116fdab6cd95e903b341fee9f
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-136018 → checkpoint-136640}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1689111747851003,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-132910",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -272421,11 +272421,806 @@
|
|
| 272421 |
"eval_steps_per_second": 0.729,
|
| 272422 |
"eval_wer": 0.17515510027413073,
|
| 272423 |
"step": 136018
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 272424 |
}
|
| 272425 |
],
|
| 272426 |
"max_steps": 620000,
|
| 272427 |
"num_train_epochs": 5000,
|
| 272428 |
-
"total_flos": 3.
|
| 272429 |
"trial_name": null,
|
| 272430 |
"trial_params": null
|
| 272431 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1689111747851003,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-132910",
|
| 4 |
+
"epoch": 1100.995983935743,
|
| 5 |
+
"global_step": 136640,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 272421 |
"eval_steps_per_second": 0.729,
|
| 272422 |
"eval_wer": 0.17515510027413073,
|
| 272423 |
"step": 136018
|
| 272424 |
+
},
|
| 272425 |
+
{
|
| 272426 |
+
"epoch": 1088.02,
|
| 272427 |
+
"learning_rate": 7.837067307692309e-06,
|
| 272428 |
+
"loss": 0.3298,
|
| 272429 |
+
"step": 136020
|
| 272430 |
+
},
|
| 272431 |
+
{
|
| 272432 |
+
"epoch": 1088.06,
|
| 272433 |
+
"learning_rate": 7.83698717948718e-06,
|
| 272434 |
+
"loss": 0.2669,
|
| 272435 |
+
"step": 136025
|
| 272436 |
+
},
|
| 272437 |
+
{
|
| 272438 |
+
"epoch": 1088.1,
|
| 272439 |
+
"learning_rate": 7.836907051282052e-06,
|
| 272440 |
+
"loss": 0.2738,
|
| 272441 |
+
"step": 136030
|
| 272442 |
+
},
|
| 272443 |
+
{
|
| 272444 |
+
"epoch": 1088.14,
|
| 272445 |
+
"learning_rate": 7.836826923076923e-06,
|
| 272446 |
+
"loss": 0.4134,
|
| 272447 |
+
"step": 136035
|
| 272448 |
+
},
|
| 272449 |
+
{
|
| 272450 |
+
"epoch": 1088.18,
|
| 272451 |
+
"learning_rate": 7.836746794871796e-06,
|
| 272452 |
+
"loss": 0.5786,
|
| 272453 |
+
"step": 136040
|
| 272454 |
+
},
|
| 272455 |
+
{
|
| 272456 |
+
"epoch": 1088.22,
|
| 272457 |
+
"learning_rate": 7.836666666666667e-06,
|
| 272458 |
+
"loss": 0.9578,
|
| 272459 |
+
"step": 136045
|
| 272460 |
+
},
|
| 272461 |
+
{
|
| 272462 |
+
"epoch": 1088.26,
|
| 272463 |
+
"learning_rate": 7.836586538461539e-06,
|
| 272464 |
+
"loss": 0.3027,
|
| 272465 |
+
"step": 136050
|
| 272466 |
+
},
|
| 272467 |
+
{
|
| 272468 |
+
"epoch": 1088.3,
|
| 272469 |
+
"learning_rate": 7.83650641025641e-06,
|
| 272470 |
+
"loss": 0.2726,
|
| 272471 |
+
"step": 136055
|
| 272472 |
+
},
|
| 272473 |
+
{
|
| 272474 |
+
"epoch": 1088.34,
|
| 272475 |
+
"learning_rate": 7.836426282051283e-06,
|
| 272476 |
+
"loss": 0.3082,
|
| 272477 |
+
"step": 136060
|
| 272478 |
+
},
|
| 272479 |
+
{
|
| 272480 |
+
"epoch": 1088.38,
|
| 272481 |
+
"learning_rate": 7.836346153846154e-06,
|
| 272482 |
+
"loss": 0.6455,
|
| 272483 |
+
"step": 136065
|
| 272484 |
+
},
|
| 272485 |
+
{
|
| 272486 |
+
"epoch": 1088.42,
|
| 272487 |
+
"learning_rate": 7.836266025641026e-06,
|
| 272488 |
+
"loss": 1.046,
|
| 272489 |
+
"step": 136070
|
| 272490 |
+
},
|
| 272491 |
+
{
|
| 272492 |
+
"epoch": 1088.46,
|
| 272493 |
+
"learning_rate": 7.836185897435899e-06,
|
| 272494 |
+
"loss": 0.3756,
|
| 272495 |
+
"step": 136075
|
| 272496 |
+
},
|
| 272497 |
+
{
|
| 272498 |
+
"epoch": 1088.5,
|
| 272499 |
+
"learning_rate": 7.83610576923077e-06,
|
| 272500 |
+
"loss": 0.3018,
|
| 272501 |
+
"step": 136080
|
| 272502 |
+
},
|
| 272503 |
+
{
|
| 272504 |
+
"epoch": 1088.54,
|
| 272505 |
+
"learning_rate": 7.836025641025642e-06,
|
| 272506 |
+
"loss": 0.3527,
|
| 272507 |
+
"step": 136085
|
| 272508 |
+
},
|
| 272509 |
+
{
|
| 272510 |
+
"epoch": 1088.58,
|
| 272511 |
+
"learning_rate": 7.835945512820513e-06,
|
| 272512 |
+
"loss": 0.6376,
|
| 272513 |
+
"step": 136090
|
| 272514 |
+
},
|
| 272515 |
+
{
|
| 272516 |
+
"epoch": 1088.62,
|
| 272517 |
+
"learning_rate": 7.835865384615386e-06,
|
| 272518 |
+
"loss": 0.9369,
|
| 272519 |
+
"step": 136095
|
| 272520 |
+
},
|
| 272521 |
+
{
|
| 272522 |
+
"epoch": 1088.66,
|
| 272523 |
+
"learning_rate": 7.835785256410257e-06,
|
| 272524 |
+
"loss": 0.2797,
|
| 272525 |
+
"step": 136100
|
| 272526 |
+
},
|
| 272527 |
+
{
|
| 272528 |
+
"epoch": 1088.7,
|
| 272529 |
+
"learning_rate": 7.835705128205129e-06,
|
| 272530 |
+
"loss": 0.2852,
|
| 272531 |
+
"step": 136105
|
| 272532 |
+
},
|
| 272533 |
+
{
|
| 272534 |
+
"epoch": 1088.74,
|
| 272535 |
+
"learning_rate": 7.835625000000002e-06,
|
| 272536 |
+
"loss": 0.299,
|
| 272537 |
+
"step": 136110
|
| 272538 |
+
},
|
| 272539 |
+
{
|
| 272540 |
+
"epoch": 1088.78,
|
| 272541 |
+
"learning_rate": 7.835544871794873e-06,
|
| 272542 |
+
"loss": 0.5605,
|
| 272543 |
+
"step": 136115
|
| 272544 |
+
},
|
| 272545 |
+
{
|
| 272546 |
+
"epoch": 1088.82,
|
| 272547 |
+
"learning_rate": 7.835464743589744e-06,
|
| 272548 |
+
"loss": 0.9115,
|
| 272549 |
+
"step": 136120
|
| 272550 |
+
},
|
| 272551 |
+
{
|
| 272552 |
+
"epoch": 1088.86,
|
| 272553 |
+
"learning_rate": 7.835384615384616e-06,
|
| 272554 |
+
"loss": 0.2998,
|
| 272555 |
+
"step": 136125
|
| 272556 |
+
},
|
| 272557 |
+
{
|
| 272558 |
+
"epoch": 1088.9,
|
| 272559 |
+
"learning_rate": 7.835304487179489e-06,
|
| 272560 |
+
"loss": 0.289,
|
| 272561 |
+
"step": 136130
|
| 272562 |
+
},
|
| 272563 |
+
{
|
| 272564 |
+
"epoch": 1088.94,
|
| 272565 |
+
"learning_rate": 7.835224358974359e-06,
|
| 272566 |
+
"loss": 0.3251,
|
| 272567 |
+
"step": 136135
|
| 272568 |
+
},
|
| 272569 |
+
{
|
| 272570 |
+
"epoch": 1088.98,
|
| 272571 |
+
"learning_rate": 7.835144230769232e-06,
|
| 272572 |
+
"loss": 0.8695,
|
| 272573 |
+
"step": 136140
|
| 272574 |
+
},
|
| 272575 |
+
{
|
| 272576 |
+
"epoch": 1089.0,
|
| 272577 |
+
"eval_loss": 0.3172611892223358,
|
| 272578 |
+
"eval_runtime": 37.1674,
|
| 272579 |
+
"eval_samples_per_second": 22.681,
|
| 272580 |
+
"eval_steps_per_second": 0.726,
|
| 272581 |
+
"eval_wer": 0.17267363332622448,
|
| 272582 |
+
"step": 136143
|
| 272583 |
+
},
|
| 272584 |
+
{
|
| 272585 |
+
"epoch": 1089.02,
|
| 272586 |
+
"learning_rate": 7.835064102564103e-06,
|
| 272587 |
+
"loss": 0.3568,
|
| 272588 |
+
"step": 136145
|
| 272589 |
+
},
|
| 272590 |
+
{
|
| 272591 |
+
"epoch": 1089.06,
|
| 272592 |
+
"learning_rate": 7.834983974358974e-06,
|
| 272593 |
+
"loss": 0.3289,
|
| 272594 |
+
"step": 136150
|
| 272595 |
+
},
|
| 272596 |
+
{
|
| 272597 |
+
"epoch": 1089.1,
|
| 272598 |
+
"learning_rate": 7.834903846153846e-06,
|
| 272599 |
+
"loss": 0.2709,
|
| 272600 |
+
"step": 136155
|
| 272601 |
+
},
|
| 272602 |
+
{
|
| 272603 |
+
"epoch": 1089.14,
|
| 272604 |
+
"learning_rate": 7.834823717948719e-06,
|
| 272605 |
+
"loss": 0.3207,
|
| 272606 |
+
"step": 136160
|
| 272607 |
+
},
|
| 272608 |
+
{
|
| 272609 |
+
"epoch": 1089.18,
|
| 272610 |
+
"learning_rate": 7.83474358974359e-06,
|
| 272611 |
+
"loss": 0.6248,
|
| 272612 |
+
"step": 136165
|
| 272613 |
+
},
|
| 272614 |
+
{
|
| 272615 |
+
"epoch": 1089.22,
|
| 272616 |
+
"learning_rate": 7.834663461538461e-06,
|
| 272617 |
+
"loss": 1.1934,
|
| 272618 |
+
"step": 136170
|
| 272619 |
+
},
|
| 272620 |
+
{
|
| 272621 |
+
"epoch": 1089.26,
|
| 272622 |
+
"learning_rate": 7.834583333333334e-06,
|
| 272623 |
+
"loss": 0.3639,
|
| 272624 |
+
"step": 136175
|
| 272625 |
+
},
|
| 272626 |
+
{
|
| 272627 |
+
"epoch": 1089.3,
|
| 272628 |
+
"learning_rate": 7.834503205128206e-06,
|
| 272629 |
+
"loss": 0.2692,
|
| 272630 |
+
"step": 136180
|
| 272631 |
+
},
|
| 272632 |
+
{
|
| 272633 |
+
"epoch": 1089.34,
|
| 272634 |
+
"learning_rate": 7.834423076923077e-06,
|
| 272635 |
+
"loss": 0.313,
|
| 272636 |
+
"step": 136185
|
| 272637 |
+
},
|
| 272638 |
+
{
|
| 272639 |
+
"epoch": 1089.38,
|
| 272640 |
+
"learning_rate": 7.834342948717949e-06,
|
| 272641 |
+
"loss": 0.6021,
|
| 272642 |
+
"step": 136190
|
| 272643 |
+
},
|
| 272644 |
+
{
|
| 272645 |
+
"epoch": 1089.42,
|
| 272646 |
+
"learning_rate": 7.834262820512822e-06,
|
| 272647 |
+
"loss": 0.9874,
|
| 272648 |
+
"step": 136195
|
| 272649 |
+
},
|
| 272650 |
+
{
|
| 272651 |
+
"epoch": 1089.46,
|
| 272652 |
+
"learning_rate": 7.834182692307693e-06,
|
| 272653 |
+
"loss": 0.2489,
|
| 272654 |
+
"step": 136200
|
| 272655 |
+
},
|
| 272656 |
+
{
|
| 272657 |
+
"epoch": 1089.5,
|
| 272658 |
+
"learning_rate": 7.834102564102564e-06,
|
| 272659 |
+
"loss": 0.2623,
|
| 272660 |
+
"step": 136205
|
| 272661 |
+
},
|
| 272662 |
+
{
|
| 272663 |
+
"epoch": 1089.54,
|
| 272664 |
+
"learning_rate": 7.834022435897437e-06,
|
| 272665 |
+
"loss": 0.3548,
|
| 272666 |
+
"step": 136210
|
| 272667 |
+
},
|
| 272668 |
+
{
|
| 272669 |
+
"epoch": 1089.58,
|
| 272670 |
+
"learning_rate": 7.833942307692309e-06,
|
| 272671 |
+
"loss": 0.6559,
|
| 272672 |
+
"step": 136215
|
| 272673 |
+
},
|
| 272674 |
+
{
|
| 272675 |
+
"epoch": 1089.62,
|
| 272676 |
+
"learning_rate": 7.83386217948718e-06,
|
| 272677 |
+
"loss": 1.001,
|
| 272678 |
+
"step": 136220
|
| 272679 |
+
},
|
| 272680 |
+
{
|
| 272681 |
+
"epoch": 1089.66,
|
| 272682 |
+
"learning_rate": 7.833782051282051e-06,
|
| 272683 |
+
"loss": 0.2933,
|
| 272684 |
+
"step": 136225
|
| 272685 |
+
},
|
| 272686 |
+
{
|
| 272687 |
+
"epoch": 1089.7,
|
| 272688 |
+
"learning_rate": 7.833701923076925e-06,
|
| 272689 |
+
"loss": 0.2641,
|
| 272690 |
+
"step": 136230
|
| 272691 |
+
},
|
| 272692 |
+
{
|
| 272693 |
+
"epoch": 1089.74,
|
| 272694 |
+
"learning_rate": 7.833621794871796e-06,
|
| 272695 |
+
"loss": 0.3023,
|
| 272696 |
+
"step": 136235
|
| 272697 |
+
},
|
| 272698 |
+
{
|
| 272699 |
+
"epoch": 1089.78,
|
| 272700 |
+
"learning_rate": 7.833541666666667e-06,
|
| 272701 |
+
"loss": 0.5769,
|
| 272702 |
+
"step": 136240
|
| 272703 |
+
},
|
| 272704 |
+
{
|
| 272705 |
+
"epoch": 1089.82,
|
| 272706 |
+
"learning_rate": 7.833461538461539e-06,
|
| 272707 |
+
"loss": 1.0256,
|
| 272708 |
+
"step": 136245
|
| 272709 |
+
},
|
| 272710 |
+
{
|
| 272711 |
+
"epoch": 1089.86,
|
| 272712 |
+
"learning_rate": 7.833381410256412e-06,
|
| 272713 |
+
"loss": 0.2867,
|
| 272714 |
+
"step": 136250
|
| 272715 |
+
},
|
| 272716 |
+
{
|
| 272717 |
+
"epoch": 1089.9,
|
| 272718 |
+
"learning_rate": 7.833301282051283e-06,
|
| 272719 |
+
"loss": 0.2768,
|
| 272720 |
+
"step": 136255
|
| 272721 |
+
},
|
| 272722 |
+
{
|
| 272723 |
+
"epoch": 1089.94,
|
| 272724 |
+
"learning_rate": 7.833221153846154e-06,
|
| 272725 |
+
"loss": 0.285,
|
| 272726 |
+
"step": 136260
|
| 272727 |
+
},
|
| 272728 |
+
{
|
| 272729 |
+
"epoch": 1089.98,
|
| 272730 |
+
"learning_rate": 7.833141025641027e-06,
|
| 272731 |
+
"loss": 0.774,
|
| 272732 |
+
"step": 136265
|
| 272733 |
+
},
|
| 272734 |
+
{
|
| 272735 |
+
"epoch": 1090.0,
|
| 272736 |
+
"eval_loss": 0.33900395035743713,
|
| 272737 |
+
"eval_runtime": 36.2364,
|
| 272738 |
+
"eval_samples_per_second": 23.264,
|
| 272739 |
+
"eval_steps_per_second": 0.745,
|
| 272740 |
+
"eval_wer": 0.17834966709059624,
|
| 272741 |
+
"step": 136268
|
| 272742 |
+
},
|
| 272743 |
+
{
|
| 272744 |
+
"epoch": 1098.02,
|
| 272745 |
+
"learning_rate": 7.833060897435899e-06,
|
| 272746 |
+
"loss": 0.3563,
|
| 272747 |
+
"step": 136270
|
| 272748 |
+
},
|
| 272749 |
+
{
|
| 272750 |
+
"epoch": 1098.06,
|
| 272751 |
+
"learning_rate": 7.83298076923077e-06,
|
| 272752 |
+
"loss": 0.2781,
|
| 272753 |
+
"step": 136275
|
| 272754 |
+
},
|
| 272755 |
+
{
|
| 272756 |
+
"epoch": 1098.1,
|
| 272757 |
+
"learning_rate": 7.832900641025641e-06,
|
| 272758 |
+
"loss": 0.2395,
|
| 272759 |
+
"step": 136280
|
| 272760 |
+
},
|
| 272761 |
+
{
|
| 272762 |
+
"epoch": 1098.14,
|
| 272763 |
+
"learning_rate": 7.832820512820515e-06,
|
| 272764 |
+
"loss": 0.3087,
|
| 272765 |
+
"step": 136285
|
| 272766 |
+
},
|
| 272767 |
+
{
|
| 272768 |
+
"epoch": 1098.18,
|
| 272769 |
+
"learning_rate": 7.832740384615384e-06,
|
| 272770 |
+
"loss": 0.6151,
|
| 272771 |
+
"step": 136290
|
| 272772 |
+
},
|
| 272773 |
+
{
|
| 272774 |
+
"epoch": 1098.22,
|
| 272775 |
+
"learning_rate": 7.832660256410257e-06,
|
| 272776 |
+
"loss": 0.9494,
|
| 272777 |
+
"step": 136295
|
| 272778 |
+
},
|
| 272779 |
+
{
|
| 272780 |
+
"epoch": 1098.26,
|
| 272781 |
+
"learning_rate": 7.832580128205129e-06,
|
| 272782 |
+
"loss": 0.2936,
|
| 272783 |
+
"step": 136300
|
| 272784 |
+
},
|
| 272785 |
+
{
|
| 272786 |
+
"epoch": 1098.3,
|
| 272787 |
+
"learning_rate": 7.8325e-06,
|
| 272788 |
+
"loss": 0.2773,
|
| 272789 |
+
"step": 136305
|
| 272790 |
+
},
|
| 272791 |
+
{
|
| 272792 |
+
"epoch": 1098.34,
|
| 272793 |
+
"learning_rate": 7.832419871794873e-06,
|
| 272794 |
+
"loss": 0.3448,
|
| 272795 |
+
"step": 136310
|
| 272796 |
+
},
|
| 272797 |
+
{
|
| 272798 |
+
"epoch": 1098.38,
|
| 272799 |
+
"learning_rate": 7.832339743589744e-06,
|
| 272800 |
+
"loss": 0.6192,
|
| 272801 |
+
"step": 136315
|
| 272802 |
+
},
|
| 272803 |
+
{
|
| 272804 |
+
"epoch": 1098.42,
|
| 272805 |
+
"learning_rate": 7.832259615384616e-06,
|
| 272806 |
+
"loss": 0.8951,
|
| 272807 |
+
"step": 136320
|
| 272808 |
+
},
|
| 272809 |
+
{
|
| 272810 |
+
"epoch": 1098.46,
|
| 272811 |
+
"learning_rate": 7.832179487179487e-06,
|
| 272812 |
+
"loss": 0.2682,
|
| 272813 |
+
"step": 136325
|
| 272814 |
+
},
|
| 272815 |
+
{
|
| 272816 |
+
"epoch": 1098.5,
|
| 272817 |
+
"learning_rate": 7.83209935897436e-06,
|
| 272818 |
+
"loss": 0.2899,
|
| 272819 |
+
"step": 136330
|
| 272820 |
+
},
|
| 272821 |
+
{
|
| 272822 |
+
"epoch": 1098.54,
|
| 272823 |
+
"learning_rate": 7.832019230769232e-06,
|
| 272824 |
+
"loss": 0.2696,
|
| 272825 |
+
"step": 136335
|
| 272826 |
+
},
|
| 272827 |
+
{
|
| 272828 |
+
"epoch": 1098.58,
|
| 272829 |
+
"learning_rate": 7.831939102564103e-06,
|
| 272830 |
+
"loss": 0.5843,
|
| 272831 |
+
"step": 136340
|
| 272832 |
+
},
|
| 272833 |
+
{
|
| 272834 |
+
"epoch": 1098.62,
|
| 272835 |
+
"learning_rate": 7.831858974358974e-06,
|
| 272836 |
+
"loss": 0.9621,
|
| 272837 |
+
"step": 136345
|
| 272838 |
+
},
|
| 272839 |
+
{
|
| 272840 |
+
"epoch": 1098.66,
|
| 272841 |
+
"learning_rate": 7.831778846153847e-06,
|
| 272842 |
+
"loss": 0.2669,
|
| 272843 |
+
"step": 136350
|
| 272844 |
+
},
|
| 272845 |
+
{
|
| 272846 |
+
"epoch": 1098.7,
|
| 272847 |
+
"learning_rate": 7.831698717948719e-06,
|
| 272848 |
+
"loss": 0.2692,
|
| 272849 |
+
"step": 136355
|
| 272850 |
+
},
|
| 272851 |
+
{
|
| 272852 |
+
"epoch": 1098.74,
|
| 272853 |
+
"learning_rate": 7.83161858974359e-06,
|
| 272854 |
+
"loss": 0.332,
|
| 272855 |
+
"step": 136360
|
| 272856 |
+
},
|
| 272857 |
+
{
|
| 272858 |
+
"epoch": 1098.78,
|
| 272859 |
+
"learning_rate": 7.831538461538463e-06,
|
| 272860 |
+
"loss": 0.6186,
|
| 272861 |
+
"step": 136365
|
| 272862 |
+
},
|
| 272863 |
+
{
|
| 272864 |
+
"epoch": 1098.82,
|
| 272865 |
+
"learning_rate": 7.831458333333334e-06,
|
| 272866 |
+
"loss": 1.048,
|
| 272867 |
+
"step": 136370
|
| 272868 |
+
},
|
| 272869 |
+
{
|
| 272870 |
+
"epoch": 1098.86,
|
| 272871 |
+
"learning_rate": 7.831378205128206e-06,
|
| 272872 |
+
"loss": 0.2853,
|
| 272873 |
+
"step": 136375
|
| 272874 |
+
},
|
| 272875 |
+
{
|
| 272876 |
+
"epoch": 1098.9,
|
| 272877 |
+
"learning_rate": 7.831298076923077e-06,
|
| 272878 |
+
"loss": 0.2915,
|
| 272879 |
+
"step": 136380
|
| 272880 |
+
},
|
| 272881 |
+
{
|
| 272882 |
+
"epoch": 1098.94,
|
| 272883 |
+
"learning_rate": 7.83121794871795e-06,
|
| 272884 |
+
"loss": 0.327,
|
| 272885 |
+
"step": 136385
|
| 272886 |
+
},
|
| 272887 |
+
{
|
| 272888 |
+
"epoch": 1098.98,
|
| 272889 |
+
"learning_rate": 7.831137820512822e-06,
|
| 272890 |
+
"loss": 0.5615,
|
| 272891 |
+
"step": 136390
|
| 272892 |
+
},
|
| 272893 |
+
{
|
| 272894 |
+
"epoch": 1099.0,
|
| 272895 |
+
"eval_loss": 0.34387004375457764,
|
| 272896 |
+
"eval_runtime": 36.4397,
|
| 272897 |
+
"eval_samples_per_second": 23.134,
|
| 272898 |
+
"eval_steps_per_second": 0.741,
|
| 272899 |
+
"eval_wer": 0.18091042584434655,
|
| 272900 |
+
"step": 136392
|
| 272901 |
+
},
|
| 272902 |
+
{
|
| 272903 |
+
"epoch": 1099.02,
|
| 272904 |
+
"learning_rate": 7.831057692307693e-06,
|
| 272905 |
+
"loss": 0.3663,
|
| 272906 |
+
"step": 136395
|
| 272907 |
+
},
|
| 272908 |
+
{
|
| 272909 |
+
"epoch": 1099.06,
|
| 272910 |
+
"learning_rate": 7.830977564102564e-06,
|
| 272911 |
+
"loss": 0.2983,
|
| 272912 |
+
"step": 136400
|
| 272913 |
+
},
|
| 272914 |
+
{
|
| 272915 |
+
"epoch": 1099.1,
|
| 272916 |
+
"learning_rate": 7.830897435897437e-06,
|
| 272917 |
+
"loss": 0.2931,
|
| 272918 |
+
"step": 136405
|
| 272919 |
+
},
|
| 272920 |
+
{
|
| 272921 |
+
"epoch": 1099.14,
|
| 272922 |
+
"learning_rate": 7.830817307692309e-06,
|
| 272923 |
+
"loss": 0.3526,
|
| 272924 |
+
"step": 136410
|
| 272925 |
+
},
|
| 272926 |
+
{
|
| 272927 |
+
"epoch": 1099.18,
|
| 272928 |
+
"learning_rate": 7.83073717948718e-06,
|
| 272929 |
+
"loss": 0.7211,
|
| 272930 |
+
"step": 136415
|
| 272931 |
+
},
|
| 272932 |
+
{
|
| 272933 |
+
"epoch": 1099.22,
|
| 272934 |
+
"learning_rate": 7.830657051282053e-06,
|
| 272935 |
+
"loss": 0.8186,
|
| 272936 |
+
"step": 136420
|
| 272937 |
+
},
|
| 272938 |
+
{
|
| 272939 |
+
"epoch": 1099.27,
|
| 272940 |
+
"learning_rate": 7.830576923076923e-06,
|
| 272941 |
+
"loss": 0.2835,
|
| 272942 |
+
"step": 136425
|
| 272943 |
+
},
|
| 272944 |
+
{
|
| 272945 |
+
"epoch": 1099.31,
|
| 272946 |
+
"learning_rate": 7.830496794871796e-06,
|
| 272947 |
+
"loss": 0.3151,
|
| 272948 |
+
"step": 136430
|
| 272949 |
+
},
|
| 272950 |
+
{
|
| 272951 |
+
"epoch": 1099.35,
|
| 272952 |
+
"learning_rate": 7.830416666666667e-06,
|
| 272953 |
+
"loss": 0.3508,
|
| 272954 |
+
"step": 136435
|
| 272955 |
+
},
|
| 272956 |
+
{
|
| 272957 |
+
"epoch": 1099.39,
|
| 272958 |
+
"learning_rate": 7.830336538461539e-06,
|
| 272959 |
+
"loss": 0.8427,
|
| 272960 |
+
"step": 136440
|
| 272961 |
+
},
|
| 272962 |
+
{
|
| 272963 |
+
"epoch": 1099.43,
|
| 272964 |
+
"learning_rate": 7.83025641025641e-06,
|
| 272965 |
+
"loss": 0.8308,
|
| 272966 |
+
"step": 136445
|
| 272967 |
+
},
|
| 272968 |
+
{
|
| 272969 |
+
"epoch": 1099.47,
|
| 272970 |
+
"learning_rate": 7.830176282051283e-06,
|
| 272971 |
+
"loss": 0.2817,
|
| 272972 |
+
"step": 136450
|
| 272973 |
+
},
|
| 272974 |
+
{
|
| 272975 |
+
"epoch": 1099.51,
|
| 272976 |
+
"learning_rate": 7.830096153846154e-06,
|
| 272977 |
+
"loss": 0.2922,
|
| 272978 |
+
"step": 136455
|
| 272979 |
+
},
|
| 272980 |
+
{
|
| 272981 |
+
"epoch": 1099.55,
|
| 272982 |
+
"learning_rate": 7.830016025641026e-06,
|
| 272983 |
+
"loss": 0.3556,
|
| 272984 |
+
"step": 136460
|
| 272985 |
+
},
|
| 272986 |
+
{
|
| 272987 |
+
"epoch": 1099.59,
|
| 272988 |
+
"learning_rate": 7.829935897435899e-06,
|
| 272989 |
+
"loss": 0.6501,
|
| 272990 |
+
"step": 136465
|
| 272991 |
+
},
|
| 272992 |
+
{
|
| 272993 |
+
"epoch": 1099.63,
|
| 272994 |
+
"learning_rate": 7.82985576923077e-06,
|
| 272995 |
+
"loss": 0.716,
|
| 272996 |
+
"step": 136470
|
| 272997 |
+
},
|
| 272998 |
+
{
|
| 272999 |
+
"epoch": 1099.67,
|
| 273000 |
+
"learning_rate": 7.829775641025641e-06,
|
| 273001 |
+
"loss": 0.3043,
|
| 273002 |
+
"step": 136475
|
| 273003 |
+
},
|
| 273004 |
+
{
|
| 273005 |
+
"epoch": 1099.71,
|
| 273006 |
+
"learning_rate": 7.829695512820513e-06,
|
| 273007 |
+
"loss": 0.2674,
|
| 273008 |
+
"step": 136480
|
| 273009 |
+
},
|
| 273010 |
+
{
|
| 273011 |
+
"epoch": 1099.75,
|
| 273012 |
+
"learning_rate": 7.829615384615386e-06,
|
| 273013 |
+
"loss": 0.4306,
|
| 273014 |
+
"step": 136485
|
| 273015 |
+
},
|
| 273016 |
+
{
|
| 273017 |
+
"epoch": 1099.79,
|
| 273018 |
+
"learning_rate": 7.829535256410257e-06,
|
| 273019 |
+
"loss": 0.663,
|
| 273020 |
+
"step": 136490
|
| 273021 |
+
},
|
| 273022 |
+
{
|
| 273023 |
+
"epoch": 1099.83,
|
| 273024 |
+
"learning_rate": 7.829455128205129e-06,
|
| 273025 |
+
"loss": 0.882,
|
| 273026 |
+
"step": 136495
|
| 273027 |
+
},
|
| 273028 |
+
{
|
| 273029 |
+
"epoch": 1099.87,
|
| 273030 |
+
"learning_rate": 7.829375e-06,
|
| 273031 |
+
"loss": 0.3191,
|
| 273032 |
+
"step": 136500
|
| 273033 |
+
},
|
| 273034 |
+
{
|
| 273035 |
+
"epoch": 1099.91,
|
| 273036 |
+
"learning_rate": 7.829294871794873e-06,
|
| 273037 |
+
"loss": 0.3352,
|
| 273038 |
+
"step": 136505
|
| 273039 |
+
},
|
| 273040 |
+
{
|
| 273041 |
+
"epoch": 1099.95,
|
| 273042 |
+
"learning_rate": 7.829214743589744e-06,
|
| 273043 |
+
"loss": 0.4366,
|
| 273044 |
+
"step": 136510
|
| 273045 |
+
},
|
| 273046 |
+
{
|
| 273047 |
+
"epoch": 1099.99,
|
| 273048 |
+
"learning_rate": 7.829134615384616e-06,
|
| 273049 |
+
"loss": 0.8065,
|
| 273050 |
+
"step": 136515
|
| 273051 |
+
},
|
| 273052 |
+
{
|
| 273053 |
+
"epoch": 1100.0,
|
| 273054 |
+
"eval_loss": 0.37556183338165283,
|
| 273055 |
+
"eval_runtime": 36.8499,
|
| 273056 |
+
"eval_samples_per_second": 22.877,
|
| 273057 |
+
"eval_steps_per_second": 0.733,
|
| 273058 |
+
"eval_wer": 0.1698699421965318,
|
| 273059 |
+
"step": 136516
|
| 273060 |
+
},
|
| 273061 |
+
{
|
| 273062 |
+
"epoch": 1100.03,
|
| 273063 |
+
"learning_rate": 7.829054487179489e-06,
|
| 273064 |
+
"loss": 0.3477,
|
| 273065 |
+
"step": 136520
|
| 273066 |
+
},
|
| 273067 |
+
{
|
| 273068 |
+
"epoch": 1100.07,
|
| 273069 |
+
"learning_rate": 7.82897435897436e-06,
|
| 273070 |
+
"loss": 0.2823,
|
| 273071 |
+
"step": 136525
|
| 273072 |
+
},
|
| 273073 |
+
{
|
| 273074 |
+
"epoch": 1100.11,
|
| 273075 |
+
"learning_rate": 7.828894230769231e-06,
|
| 273076 |
+
"loss": 0.2641,
|
| 273077 |
+
"step": 136530
|
| 273078 |
+
},
|
| 273079 |
+
{
|
| 273080 |
+
"epoch": 1100.15,
|
| 273081 |
+
"learning_rate": 7.828814102564103e-06,
|
| 273082 |
+
"loss": 0.4155,
|
| 273083 |
+
"step": 136535
|
| 273084 |
+
},
|
| 273085 |
+
{
|
| 273086 |
+
"epoch": 1100.19,
|
| 273087 |
+
"learning_rate": 7.828733974358976e-06,
|
| 273088 |
+
"loss": 0.8419,
|
| 273089 |
+
"step": 136540
|
| 273090 |
+
},
|
| 273091 |
+
{
|
| 273092 |
+
"epoch": 1100.23,
|
| 273093 |
+
"learning_rate": 7.828653846153847e-06,
|
| 273094 |
+
"loss": 0.5977,
|
| 273095 |
+
"step": 136545
|
| 273096 |
+
},
|
| 273097 |
+
{
|
| 273098 |
+
"epoch": 1100.27,
|
| 273099 |
+
"learning_rate": 7.828573717948719e-06,
|
| 273100 |
+
"loss": 0.2601,
|
| 273101 |
+
"step": 136550
|
| 273102 |
+
},
|
| 273103 |
+
{
|
| 273104 |
+
"epoch": 1100.31,
|
| 273105 |
+
"learning_rate": 7.82849358974359e-06,
|
| 273106 |
+
"loss": 0.3252,
|
| 273107 |
+
"step": 136555
|
| 273108 |
+
},
|
| 273109 |
+
{
|
| 273110 |
+
"epoch": 1100.35,
|
| 273111 |
+
"learning_rate": 7.828413461538463e-06,
|
| 273112 |
+
"loss": 0.4857,
|
| 273113 |
+
"step": 136560
|
| 273114 |
+
},
|
| 273115 |
+
{
|
| 273116 |
+
"epoch": 1100.39,
|
| 273117 |
+
"learning_rate": 7.828333333333334e-06,
|
| 273118 |
+
"loss": 0.9063,
|
| 273119 |
+
"step": 136565
|
| 273120 |
+
},
|
| 273121 |
+
{
|
| 273122 |
+
"epoch": 1100.43,
|
| 273123 |
+
"learning_rate": 7.828253205128206e-06,
|
| 273124 |
+
"loss": 0.7182,
|
| 273125 |
+
"step": 136570
|
| 273126 |
+
},
|
| 273127 |
+
{
|
| 273128 |
+
"epoch": 1100.47,
|
| 273129 |
+
"learning_rate": 7.828173076923079e-06,
|
| 273130 |
+
"loss": 0.343,
|
| 273131 |
+
"step": 136575
|
| 273132 |
+
},
|
| 273133 |
+
{
|
| 273134 |
+
"epoch": 1100.51,
|
| 273135 |
+
"learning_rate": 7.828092948717948e-06,
|
| 273136 |
+
"loss": 0.298,
|
| 273137 |
+
"step": 136580
|
| 273138 |
+
},
|
| 273139 |
+
{
|
| 273140 |
+
"epoch": 1100.55,
|
| 273141 |
+
"learning_rate": 7.828012820512821e-06,
|
| 273142 |
+
"loss": 0.3371,
|
| 273143 |
+
"step": 136585
|
| 273144 |
+
},
|
| 273145 |
+
{
|
| 273146 |
+
"epoch": 1100.59,
|
| 273147 |
+
"learning_rate": 7.827932692307693e-06,
|
| 273148 |
+
"loss": 0.9165,
|
| 273149 |
+
"step": 136590
|
| 273150 |
+
},
|
| 273151 |
+
{
|
| 273152 |
+
"epoch": 1100.63,
|
| 273153 |
+
"learning_rate": 7.827852564102564e-06,
|
| 273154 |
+
"loss": 0.6564,
|
| 273155 |
+
"step": 136595
|
| 273156 |
+
},
|
| 273157 |
+
{
|
| 273158 |
+
"epoch": 1100.67,
|
| 273159 |
+
"learning_rate": 7.827772435897436e-06,
|
| 273160 |
+
"loss": 0.2597,
|
| 273161 |
+
"step": 136600
|
| 273162 |
+
},
|
| 273163 |
+
{
|
| 273164 |
+
"epoch": 1100.71,
|
| 273165 |
+
"learning_rate": 7.827692307692309e-06,
|
| 273166 |
+
"loss": 0.3239,
|
| 273167 |
+
"step": 136605
|
| 273168 |
+
},
|
| 273169 |
+
{
|
| 273170 |
+
"epoch": 1100.76,
|
| 273171 |
+
"learning_rate": 7.82761217948718e-06,
|
| 273172 |
+
"loss": 0.3628,
|
| 273173 |
+
"step": 136610
|
| 273174 |
+
},
|
| 273175 |
+
{
|
| 273176 |
+
"epoch": 1100.8,
|
| 273177 |
+
"learning_rate": 7.827532051282051e-06,
|
| 273178 |
+
"loss": 0.9045,
|
| 273179 |
+
"step": 136615
|
| 273180 |
+
},
|
| 273181 |
+
{
|
| 273182 |
+
"epoch": 1100.84,
|
| 273183 |
+
"learning_rate": 7.827451923076924e-06,
|
| 273184 |
+
"loss": 0.6694,
|
| 273185 |
+
"step": 136620
|
| 273186 |
+
},
|
| 273187 |
+
{
|
| 273188 |
+
"epoch": 1100.88,
|
| 273189 |
+
"learning_rate": 7.827371794871796e-06,
|
| 273190 |
+
"loss": 0.3575,
|
| 273191 |
+
"step": 136625
|
| 273192 |
+
},
|
| 273193 |
+
{
|
| 273194 |
+
"epoch": 1100.92,
|
| 273195 |
+
"learning_rate": 7.827291666666667e-06,
|
| 273196 |
+
"loss": 0.2864,
|
| 273197 |
+
"step": 136630
|
| 273198 |
+
},
|
| 273199 |
+
{
|
| 273200 |
+
"epoch": 1100.96,
|
| 273201 |
+
"learning_rate": 7.827211538461538e-06,
|
| 273202 |
+
"loss": 0.39,
|
| 273203 |
+
"step": 136635
|
| 273204 |
+
},
|
| 273205 |
+
{
|
| 273206 |
+
"epoch": 1101.0,
|
| 273207 |
+
"learning_rate": 7.827131410256411e-06,
|
| 273208 |
+
"loss": 0.9859,
|
| 273209 |
+
"step": 136640
|
| 273210 |
+
},
|
| 273211 |
+
{
|
| 273212 |
+
"epoch": 1101.0,
|
| 273213 |
+
"eval_loss": 0.3771774172782898,
|
| 273214 |
+
"eval_runtime": 35.9527,
|
| 273215 |
+
"eval_samples_per_second": 23.447,
|
| 273216 |
+
"eval_steps_per_second": 0.751,
|
| 273217 |
+
"eval_wer": 0.17527468420659242,
|
| 273218 |
+
"step": 136640
|
| 273219 |
}
|
| 273220 |
],
|
| 273221 |
"max_steps": 620000,
|
| 273222 |
"num_train_epochs": 5000,
|
| 273223 |
+
"total_flos": 3.845067191515764e+20,
|
| 273224 |
"trial_name": null,
|
| 273225 |
"trial_params": null
|
| 273226 |
}
|
model-bin/finetune/base/{checkpoint-136018 → checkpoint-136640}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1630181516.414431/events.out.tfevents.1630181516.86bb0ddabf9b.4092.291
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:970c1deba0b430c44985cdc7bc8dd942cad2fd84cd25963db9004ab1b3d1d4e9
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630181907.2896452/events.out.tfevents.1630181907.86bb0ddabf9b.4092.293
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:90d5f2b0e2dc4f421ca7ccf1a31d7f7b4c657358de8e8e632a51d41cb444aae6
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630182294.8279276/events.out.tfevents.1630182294.86bb0ddabf9b.4092.295
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d13ed41df75de9a1726d915099b29bec50c291b145eb6683f8e1334c6d304606
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630182688.064181/events.out.tfevents.1630182688.86bb0ddabf9b.4092.297
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:854cc968caf50850d9d8951d6f30106487a0c6922aa6e54a6b272e06997490a1
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630183080.6762185/events.out.tfevents.1630183080.86bb0ddabf9b.4092.299
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:bdb3093e1452f6b4d1a65b38c75877d8a9ee65e0914c13028473c22035599f13
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1630181516.86bb0ddabf9b.4092.290
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:97171a7c86be53283f68683dde94c0d013a5e0389ed7fe4d0c99af66ae4fd5c8
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630181907.86bb0ddabf9b.4092.292
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:afbda275c0ac3660846647390b6c2033e7a751b4007308214afdca82e90674ad
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630182294.86bb0ddabf9b.4092.294
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d64a3fbb36724e52f5c6c358fce919ae20c5312a15cc1b4a6f68d4476c3d0a3d
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630182688.86bb0ddabf9b.4092.296
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:de30debd537820d3cf4abacfac6fd20602efe211ba45206e53f78394c990cdb1
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630183080.86bb0ddabf9b.4092.298
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6f062491ee084dcf5dff27a2db02a7ce531e227275377fde372f222ac9b9eb92
|
| 3 |
+
size 8622
|