"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-60604 → checkpoint-61350}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-60604 → checkpoint-61350}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-60604 → checkpoint-61350}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-60604 → checkpoint-61350}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-60604 → checkpoint-61350}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-60604 → checkpoint-61350}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-60604 → checkpoint-61350}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-60604 → checkpoint-61350}/trainer_state.json +959 -5
- model-bin/finetune/base/{checkpoint-60604 → checkpoint-61350}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629801593.3967817/events.out.tfevents.1629801593.c435e1c5ee04.920.111 +3 -0
- model-bin/finetune/base/log/1629802228.319492/events.out.tfevents.1629802228.c435e1c5ee04.920.113 +3 -0
- model-bin/finetune/base/log/1629802982.729967/events.out.tfevents.1629802982.c435e1c5ee04.920.115 +3 -0
- model-bin/finetune/base/log/1629803633.8011563/events.out.tfevents.1629803633.c435e1c5ee04.920.117 +3 -0
- model-bin/finetune/base/log/1629804285.3739185/events.out.tfevents.1629804285.c435e1c5ee04.920.119 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629801593.c435e1c5ee04.920.110 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629802228.c435e1c5ee04.920.112 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629802982.c435e1c5ee04.920.114 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629803633.c435e1c5ee04.920.116 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629804285.c435e1c5ee04.920.118 +3 -0
model-bin/finetune/base/{checkpoint-60604 → checkpoint-61350}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-60604 → checkpoint-61350}/optimizer.pt
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 722165009
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:353d82c12d6735954a514be77e06ca9d192330f66e138929496fa6ba81bcfc02
|
| 3 |
size 722165009
|
model-bin/finetune/base/{checkpoint-60604 → checkpoint-61350}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-60604 → checkpoint-61350}/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:fce8f21fad01f8d884828fbcc6967cef29c49b990d1f4d92914cee1952071dc3
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-60604 → checkpoint-61350}/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 sha256:0eeb08b6e206cc952258c7cd42ad2583b0dbe57250380cbabcfa01e167250da1
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-60604 → checkpoint-61350}/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:cfc3a5c9b694edc68b9a3d30779c3730190586b5fec21d95cd916c9a9982b328
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-60604 → checkpoint-61350}/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:e510737aa08b522929d4f008035f55ac4839684a993beb48faf03dd344ad37fd
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-60604 → checkpoint-61350}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
-
"best_metric": 0.
|
| 3 |
-
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -176469,11 +176469,965 @@
|
|
| 176469 |
"eval_steps_per_second": 0.692,
|
| 176470 |
"eval_wer": 0.19932432432432431,
|
| 176471 |
"step": 60604
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 176472 |
}
|
| 176473 |
],
|
| 176474 |
"max_steps": 620000,
|
| 176475 |
"num_train_epochs": 5000,
|
| 176476 |
-
"total_flos": 1.
|
| 176477 |
"trial_name": null,
|
| 176478 |
"trial_params": null
|
| 176479 |
}
|
|
|
|
| 1 |
{
|
| 2 |
+
"best_metric": 0.18588425381903642,
|
| 3 |
+
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-60729",
|
| 4 |
+
"epoch": 494.0,
|
| 5 |
+
"global_step": 61350,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 176469 |
"eval_steps_per_second": 0.692,
|
| 176470 |
"eval_wer": 0.19932432432432431,
|
| 176471 |
"step": 60604
|
| 176472 |
+
},
|
| 176473 |
+
{
|
| 176474 |
+
"epoch": 484.01,
|
| 176475 |
+
"learning_rate": 9.045160256410256e-06,
|
| 176476 |
+
"loss": 0.395,
|
| 176477 |
+
"step": 60605
|
| 176478 |
+
},
|
| 176479 |
+
{
|
| 176480 |
+
"epoch": 484.05,
|
| 176481 |
+
"learning_rate": 9.04508012820513e-06,
|
| 176482 |
+
"loss": 0.3529,
|
| 176483 |
+
"step": 60610
|
| 176484 |
+
},
|
| 176485 |
+
{
|
| 176486 |
+
"epoch": 484.09,
|
| 176487 |
+
"learning_rate": 9.045e-06,
|
| 176488 |
+
"loss": 0.3603,
|
| 176489 |
+
"step": 60615
|
| 176490 |
+
},
|
| 176491 |
+
{
|
| 176492 |
+
"epoch": 484.13,
|
| 176493 |
+
"learning_rate": 9.044919871794872e-06,
|
| 176494 |
+
"loss": 0.3763,
|
| 176495 |
+
"step": 60620
|
| 176496 |
+
},
|
| 176497 |
+
{
|
| 176498 |
+
"epoch": 484.17,
|
| 176499 |
+
"learning_rate": 9.044839743589743e-06,
|
| 176500 |
+
"loss": 0.5586,
|
| 176501 |
+
"step": 60625
|
| 176502 |
+
},
|
| 176503 |
+
{
|
| 176504 |
+
"epoch": 484.21,
|
| 176505 |
+
"learning_rate": 9.044759615384616e-06,
|
| 176506 |
+
"loss": 1.1159,
|
| 176507 |
+
"step": 60630
|
| 176508 |
+
},
|
| 176509 |
+
{
|
| 176510 |
+
"epoch": 484.25,
|
| 176511 |
+
"learning_rate": 9.044679487179488e-06,
|
| 176512 |
+
"loss": 0.3736,
|
| 176513 |
+
"step": 60635
|
| 176514 |
+
},
|
| 176515 |
+
{
|
| 176516 |
+
"epoch": 484.29,
|
| 176517 |
+
"learning_rate": 9.044599358974359e-06,
|
| 176518 |
+
"loss": 0.289,
|
| 176519 |
+
"step": 60640
|
| 176520 |
+
},
|
| 176521 |
+
{
|
| 176522 |
+
"epoch": 484.33,
|
| 176523 |
+
"learning_rate": 9.044519230769232e-06,
|
| 176524 |
+
"loss": 0.3933,
|
| 176525 |
+
"step": 60645
|
| 176526 |
+
},
|
| 176527 |
+
{
|
| 176528 |
+
"epoch": 484.37,
|
| 176529 |
+
"learning_rate": 9.044439102564104e-06,
|
| 176530 |
+
"loss": 0.5288,
|
| 176531 |
+
"step": 60650
|
| 176532 |
+
},
|
| 176533 |
+
{
|
| 176534 |
+
"epoch": 484.41,
|
| 176535 |
+
"learning_rate": 9.044358974358975e-06,
|
| 176536 |
+
"loss": 1.1191,
|
| 176537 |
+
"step": 60655
|
| 176538 |
+
},
|
| 176539 |
+
{
|
| 176540 |
+
"epoch": 484.45,
|
| 176541 |
+
"learning_rate": 9.044278846153846e-06,
|
| 176542 |
+
"loss": 0.3084,
|
| 176543 |
+
"step": 60660
|
| 176544 |
+
},
|
| 176545 |
+
{
|
| 176546 |
+
"epoch": 484.49,
|
| 176547 |
+
"learning_rate": 9.04419871794872e-06,
|
| 176548 |
+
"loss": 0.3426,
|
| 176549 |
+
"step": 60665
|
| 176550 |
+
},
|
| 176551 |
+
{
|
| 176552 |
+
"epoch": 484.53,
|
| 176553 |
+
"learning_rate": 9.04411858974359e-06,
|
| 176554 |
+
"loss": 0.3757,
|
| 176555 |
+
"step": 60670
|
| 176556 |
+
},
|
| 176557 |
+
{
|
| 176558 |
+
"epoch": 484.57,
|
| 176559 |
+
"learning_rate": 9.044038461538462e-06,
|
| 176560 |
+
"loss": 0.5491,
|
| 176561 |
+
"step": 60675
|
| 176562 |
+
},
|
| 176563 |
+
{
|
| 176564 |
+
"epoch": 484.61,
|
| 176565 |
+
"learning_rate": 9.043958333333333e-06,
|
| 176566 |
+
"loss": 1.1088,
|
| 176567 |
+
"step": 60680
|
| 176568 |
+
},
|
| 176569 |
+
{
|
| 176570 |
+
"epoch": 484.65,
|
| 176571 |
+
"learning_rate": 9.043878205128206e-06,
|
| 176572 |
+
"loss": 0.3394,
|
| 176573 |
+
"step": 60685
|
| 176574 |
+
},
|
| 176575 |
+
{
|
| 176576 |
+
"epoch": 484.69,
|
| 176577 |
+
"learning_rate": 9.043798076923078e-06,
|
| 176578 |
+
"loss": 0.3385,
|
| 176579 |
+
"step": 60690
|
| 176580 |
+
},
|
| 176581 |
+
{
|
| 176582 |
+
"epoch": 484.73,
|
| 176583 |
+
"learning_rate": 9.04371794871795e-06,
|
| 176584 |
+
"loss": 0.3283,
|
| 176585 |
+
"step": 60695
|
| 176586 |
+
},
|
| 176587 |
+
{
|
| 176588 |
+
"epoch": 484.77,
|
| 176589 |
+
"learning_rate": 9.043637820512822e-06,
|
| 176590 |
+
"loss": 0.5318,
|
| 176591 |
+
"step": 60700
|
| 176592 |
+
},
|
| 176593 |
+
{
|
| 176594 |
+
"epoch": 484.81,
|
| 176595 |
+
"learning_rate": 9.043557692307694e-06,
|
| 176596 |
+
"loss": 1.11,
|
| 176597 |
+
"step": 60705
|
| 176598 |
+
},
|
| 176599 |
+
{
|
| 176600 |
+
"epoch": 484.85,
|
| 176601 |
+
"learning_rate": 9.043477564102565e-06,
|
| 176602 |
+
"loss": 0.3747,
|
| 176603 |
+
"step": 60710
|
| 176604 |
+
},
|
| 176605 |
+
{
|
| 176606 |
+
"epoch": 484.89,
|
| 176607 |
+
"learning_rate": 9.043397435897436e-06,
|
| 176608 |
+
"loss": 0.3217,
|
| 176609 |
+
"step": 60715
|
| 176610 |
+
},
|
| 176611 |
+
{
|
| 176612 |
+
"epoch": 484.93,
|
| 176613 |
+
"learning_rate": 9.04331730769231e-06,
|
| 176614 |
+
"loss": 0.3629,
|
| 176615 |
+
"step": 60720
|
| 176616 |
+
},
|
| 176617 |
+
{
|
| 176618 |
+
"epoch": 484.97,
|
| 176619 |
+
"learning_rate": 9.043237179487179e-06,
|
| 176620 |
+
"loss": 0.6506,
|
| 176621 |
+
"step": 60725
|
| 176622 |
+
},
|
| 176623 |
+
{
|
| 176624 |
+
"epoch": 485.0,
|
| 176625 |
+
"eval_loss": 0.40527278184890747,
|
| 176626 |
+
"eval_runtime": 38.5522,
|
| 176627 |
+
"eval_samples_per_second": 21.737,
|
| 176628 |
+
"eval_steps_per_second": 0.7,
|
| 176629 |
+
"eval_wer": 0.18588425381903642,
|
| 176630 |
+
"step": 60729
|
| 176631 |
+
},
|
| 176632 |
+
{
|
| 176633 |
+
"epoch": 489.01,
|
| 176634 |
+
"learning_rate": 9.043157051282052e-06,
|
| 176635 |
+
"loss": 0.4361,
|
| 176636 |
+
"step": 60730
|
| 176637 |
+
},
|
| 176638 |
+
{
|
| 176639 |
+
"epoch": 489.05,
|
| 176640 |
+
"learning_rate": 9.043076923076923e-06,
|
| 176641 |
+
"loss": 0.3467,
|
| 176642 |
+
"step": 60735
|
| 176643 |
+
},
|
| 176644 |
+
{
|
| 176645 |
+
"epoch": 489.09,
|
| 176646 |
+
"learning_rate": 9.042996794871795e-06,
|
| 176647 |
+
"loss": 0.2973,
|
| 176648 |
+
"step": 60740
|
| 176649 |
+
},
|
| 176650 |
+
{
|
| 176651 |
+
"epoch": 489.13,
|
| 176652 |
+
"learning_rate": 9.042916666666668e-06,
|
| 176653 |
+
"loss": 0.3757,
|
| 176654 |
+
"step": 60745
|
| 176655 |
+
},
|
| 176656 |
+
{
|
| 176657 |
+
"epoch": 489.17,
|
| 176658 |
+
"learning_rate": 9.04283653846154e-06,
|
| 176659 |
+
"loss": 0.63,
|
| 176660 |
+
"step": 60750
|
| 176661 |
+
},
|
| 176662 |
+
{
|
| 176663 |
+
"epoch": 489.21,
|
| 176664 |
+
"learning_rate": 9.04275641025641e-06,
|
| 176665 |
+
"loss": 1.0862,
|
| 176666 |
+
"step": 60755
|
| 176667 |
+
},
|
| 176668 |
+
{
|
| 176669 |
+
"epoch": 489.25,
|
| 176670 |
+
"learning_rate": 9.042676282051282e-06,
|
| 176671 |
+
"loss": 0.3307,
|
| 176672 |
+
"step": 60760
|
| 176673 |
+
},
|
| 176674 |
+
{
|
| 176675 |
+
"epoch": 489.29,
|
| 176676 |
+
"learning_rate": 9.042596153846155e-06,
|
| 176677 |
+
"loss": 0.3111,
|
| 176678 |
+
"step": 60765
|
| 176679 |
+
},
|
| 176680 |
+
{
|
| 176681 |
+
"epoch": 489.33,
|
| 176682 |
+
"learning_rate": 9.042516025641026e-06,
|
| 176683 |
+
"loss": 0.3204,
|
| 176684 |
+
"step": 60770
|
| 176685 |
+
},
|
| 176686 |
+
{
|
| 176687 |
+
"epoch": 489.37,
|
| 176688 |
+
"learning_rate": 9.042435897435898e-06,
|
| 176689 |
+
"loss": 0.7555,
|
| 176690 |
+
"step": 60775
|
| 176691 |
+
},
|
| 176692 |
+
{
|
| 176693 |
+
"epoch": 489.41,
|
| 176694 |
+
"learning_rate": 9.042355769230769e-06,
|
| 176695 |
+
"loss": 1.2464,
|
| 176696 |
+
"step": 60780
|
| 176697 |
+
},
|
| 176698 |
+
{
|
| 176699 |
+
"epoch": 489.45,
|
| 176700 |
+
"learning_rate": 9.042275641025642e-06,
|
| 176701 |
+
"loss": 0.3736,
|
| 176702 |
+
"step": 60785
|
| 176703 |
+
},
|
| 176704 |
+
{
|
| 176705 |
+
"epoch": 489.49,
|
| 176706 |
+
"learning_rate": 9.042195512820513e-06,
|
| 176707 |
+
"loss": 0.343,
|
| 176708 |
+
"step": 60790
|
| 176709 |
+
},
|
| 176710 |
+
{
|
| 176711 |
+
"epoch": 489.53,
|
| 176712 |
+
"learning_rate": 9.042115384615385e-06,
|
| 176713 |
+
"loss": 0.4481,
|
| 176714 |
+
"step": 60795
|
| 176715 |
+
},
|
| 176716 |
+
{
|
| 176717 |
+
"epoch": 489.57,
|
| 176718 |
+
"learning_rate": 9.042035256410258e-06,
|
| 176719 |
+
"loss": 0.5257,
|
| 176720 |
+
"step": 60800
|
| 176721 |
+
},
|
| 176722 |
+
{
|
| 176723 |
+
"epoch": 489.61,
|
| 176724 |
+
"learning_rate": 9.04195512820513e-06,
|
| 176725 |
+
"loss": 1.198,
|
| 176726 |
+
"step": 60805
|
| 176727 |
+
},
|
| 176728 |
+
{
|
| 176729 |
+
"epoch": 489.65,
|
| 176730 |
+
"learning_rate": 9.041875e-06,
|
| 176731 |
+
"loss": 0.311,
|
| 176732 |
+
"step": 60810
|
| 176733 |
+
},
|
| 176734 |
+
{
|
| 176735 |
+
"epoch": 489.69,
|
| 176736 |
+
"learning_rate": 9.041794871794872e-06,
|
| 176737 |
+
"loss": 0.3372,
|
| 176738 |
+
"step": 60815
|
| 176739 |
+
},
|
| 176740 |
+
{
|
| 176741 |
+
"epoch": 489.73,
|
| 176742 |
+
"learning_rate": 9.041714743589745e-06,
|
| 176743 |
+
"loss": 0.3814,
|
| 176744 |
+
"step": 60820
|
| 176745 |
+
},
|
| 176746 |
+
{
|
| 176747 |
+
"epoch": 489.77,
|
| 176748 |
+
"learning_rate": 9.041634615384616e-06,
|
| 176749 |
+
"loss": 0.5157,
|
| 176750 |
+
"step": 60825
|
| 176751 |
+
},
|
| 176752 |
+
{
|
| 176753 |
+
"epoch": 489.81,
|
| 176754 |
+
"learning_rate": 9.041554487179488e-06,
|
| 176755 |
+
"loss": 1.1918,
|
| 176756 |
+
"step": 60830
|
| 176757 |
+
},
|
| 176758 |
+
{
|
| 176759 |
+
"epoch": 489.85,
|
| 176760 |
+
"learning_rate": 9.041474358974359e-06,
|
| 176761 |
+
"loss": 0.3764,
|
| 176762 |
+
"step": 60835
|
| 176763 |
+
},
|
| 176764 |
+
{
|
| 176765 |
+
"epoch": 489.9,
|
| 176766 |
+
"learning_rate": 9.041394230769232e-06,
|
| 176767 |
+
"loss": 0.3416,
|
| 176768 |
+
"step": 60840
|
| 176769 |
+
},
|
| 176770 |
+
{
|
| 176771 |
+
"epoch": 489.94,
|
| 176772 |
+
"learning_rate": 9.041314102564103e-06,
|
| 176773 |
+
"loss": 0.4076,
|
| 176774 |
+
"step": 60845
|
| 176775 |
+
},
|
| 176776 |
+
{
|
| 176777 |
+
"epoch": 489.98,
|
| 176778 |
+
"learning_rate": 9.041233974358975e-06,
|
| 176779 |
+
"loss": 0.7,
|
| 176780 |
+
"step": 60850
|
| 176781 |
+
},
|
| 176782 |
+
{
|
| 176783 |
+
"epoch": 490.0,
|
| 176784 |
+
"eval_loss": 0.4080774784088135,
|
| 176785 |
+
"eval_runtime": 40.3984,
|
| 176786 |
+
"eval_samples_per_second": 20.743,
|
| 176787 |
+
"eval_steps_per_second": 0.668,
|
| 176788 |
+
"eval_wer": 0.19368847712796758,
|
| 176789 |
+
"step": 60853
|
| 176790 |
+
},
|
| 176791 |
+
{
|
| 176792 |
+
"epoch": 486.02,
|
| 176793 |
+
"learning_rate": 9.041153846153848e-06,
|
| 176794 |
+
"loss": 0.4242,
|
| 176795 |
+
"step": 60855
|
| 176796 |
+
},
|
| 176797 |
+
{
|
| 176798 |
+
"epoch": 486.06,
|
| 176799 |
+
"learning_rate": 9.04107371794872e-06,
|
| 176800 |
+
"loss": 0.3067,
|
| 176801 |
+
"step": 60860
|
| 176802 |
+
},
|
| 176803 |
+
{
|
| 176804 |
+
"epoch": 486.1,
|
| 176805 |
+
"learning_rate": 9.04099358974359e-06,
|
| 176806 |
+
"loss": 0.4052,
|
| 176807 |
+
"step": 60865
|
| 176808 |
+
},
|
| 176809 |
+
{
|
| 176810 |
+
"epoch": 486.14,
|
| 176811 |
+
"learning_rate": 9.040913461538462e-06,
|
| 176812 |
+
"loss": 0.4187,
|
| 176813 |
+
"step": 60870
|
| 176814 |
+
},
|
| 176815 |
+
{
|
| 176816 |
+
"epoch": 486.18,
|
| 176817 |
+
"learning_rate": 9.040833333333335e-06,
|
| 176818 |
+
"loss": 0.6364,
|
| 176819 |
+
"step": 60875
|
| 176820 |
+
},
|
| 176821 |
+
{
|
| 176822 |
+
"epoch": 486.22,
|
| 176823 |
+
"learning_rate": 9.040753205128205e-06,
|
| 176824 |
+
"loss": 1.0251,
|
| 176825 |
+
"step": 60880
|
| 176826 |
+
},
|
| 176827 |
+
{
|
| 176828 |
+
"epoch": 486.26,
|
| 176829 |
+
"learning_rate": 9.040673076923078e-06,
|
| 176830 |
+
"loss": 0.3019,
|
| 176831 |
+
"step": 60885
|
| 176832 |
+
},
|
| 176833 |
+
{
|
| 176834 |
+
"epoch": 486.3,
|
| 176835 |
+
"learning_rate": 9.040592948717949e-06,
|
| 176836 |
+
"loss": 0.366,
|
| 176837 |
+
"step": 60890
|
| 176838 |
+
},
|
| 176839 |
+
{
|
| 176840 |
+
"epoch": 486.34,
|
| 176841 |
+
"learning_rate": 9.04051282051282e-06,
|
| 176842 |
+
"loss": 0.3735,
|
| 176843 |
+
"step": 60895
|
| 176844 |
+
},
|
| 176845 |
+
{
|
| 176846 |
+
"epoch": 486.38,
|
| 176847 |
+
"learning_rate": 9.040432692307693e-06,
|
| 176848 |
+
"loss": 0.6687,
|
| 176849 |
+
"step": 60900
|
| 176850 |
+
},
|
| 176851 |
+
{
|
| 176852 |
+
"epoch": 486.42,
|
| 176853 |
+
"learning_rate": 9.040352564102565e-06,
|
| 176854 |
+
"loss": 1.0873,
|
| 176855 |
+
"step": 60905
|
| 176856 |
+
},
|
| 176857 |
+
{
|
| 176858 |
+
"epoch": 486.46,
|
| 176859 |
+
"learning_rate": 9.040272435897436e-06,
|
| 176860 |
+
"loss": 0.2771,
|
| 176861 |
+
"step": 60910
|
| 176862 |
+
},
|
| 176863 |
+
{
|
| 176864 |
+
"epoch": 486.5,
|
| 176865 |
+
"learning_rate": 9.040192307692308e-06,
|
| 176866 |
+
"loss": 0.3668,
|
| 176867 |
+
"step": 60915
|
| 176868 |
+
},
|
| 176869 |
+
{
|
| 176870 |
+
"epoch": 486.54,
|
| 176871 |
+
"learning_rate": 9.04011217948718e-06,
|
| 176872 |
+
"loss": 0.3356,
|
| 176873 |
+
"step": 60920
|
| 176874 |
+
},
|
| 176875 |
+
{
|
| 176876 |
+
"epoch": 486.58,
|
| 176877 |
+
"learning_rate": 9.040032051282052e-06,
|
| 176878 |
+
"loss": 0.6527,
|
| 176879 |
+
"step": 60925
|
| 176880 |
+
},
|
| 176881 |
+
{
|
| 176882 |
+
"epoch": 486.62,
|
| 176883 |
+
"learning_rate": 9.039951923076923e-06,
|
| 176884 |
+
"loss": 0.9712,
|
| 176885 |
+
"step": 60930
|
| 176886 |
+
},
|
| 176887 |
+
{
|
| 176888 |
+
"epoch": 486.66,
|
| 176889 |
+
"learning_rate": 9.039871794871795e-06,
|
| 176890 |
+
"loss": 0.3367,
|
| 176891 |
+
"step": 60935
|
| 176892 |
+
},
|
| 176893 |
+
{
|
| 176894 |
+
"epoch": 486.7,
|
| 176895 |
+
"learning_rate": 9.039791666666668e-06,
|
| 176896 |
+
"loss": 0.3588,
|
| 176897 |
+
"step": 60940
|
| 176898 |
+
},
|
| 176899 |
+
{
|
| 176900 |
+
"epoch": 486.74,
|
| 176901 |
+
"learning_rate": 9.039711538461539e-06,
|
| 176902 |
+
"loss": 0.3842,
|
| 176903 |
+
"step": 60945
|
| 176904 |
+
},
|
| 176905 |
+
{
|
| 176906 |
+
"epoch": 486.78,
|
| 176907 |
+
"learning_rate": 9.03963141025641e-06,
|
| 176908 |
+
"loss": 0.6641,
|
| 176909 |
+
"step": 60950
|
| 176910 |
+
},
|
| 176911 |
+
{
|
| 176912 |
+
"epoch": 486.82,
|
| 176913 |
+
"learning_rate": 9.039551282051284e-06,
|
| 176914 |
+
"loss": 1.1598,
|
| 176915 |
+
"step": 60955
|
| 176916 |
+
},
|
| 176917 |
+
{
|
| 176918 |
+
"epoch": 486.86,
|
| 176919 |
+
"learning_rate": 9.039471153846155e-06,
|
| 176920 |
+
"loss": 0.3243,
|
| 176921 |
+
"step": 60960
|
| 176922 |
+
},
|
| 176923 |
+
{
|
| 176924 |
+
"epoch": 486.9,
|
| 176925 |
+
"learning_rate": 9.039391025641026e-06,
|
| 176926 |
+
"loss": 0.3483,
|
| 176927 |
+
"step": 60965
|
| 176928 |
+
},
|
| 176929 |
+
{
|
| 176930 |
+
"epoch": 486.94,
|
| 176931 |
+
"learning_rate": 9.039310897435898e-06,
|
| 176932 |
+
"loss": 0.4508,
|
| 176933 |
+
"step": 60970
|
| 176934 |
+
},
|
| 176935 |
+
{
|
| 176936 |
+
"epoch": 486.98,
|
| 176937 |
+
"learning_rate": 9.03923076923077e-06,
|
| 176938 |
+
"loss": 0.9019,
|
| 176939 |
+
"step": 60975
|
| 176940 |
+
},
|
| 176941 |
+
{
|
| 176942 |
+
"epoch": 487.0,
|
| 176943 |
+
"eval_loss": 0.3954157531261444,
|
| 176944 |
+
"eval_runtime": 40.0837,
|
| 176945 |
+
"eval_samples_per_second": 20.906,
|
| 176946 |
+
"eval_steps_per_second": 0.674,
|
| 176947 |
+
"eval_wer": 0.19892081085022606,
|
| 176948 |
+
"step": 60978
|
| 176949 |
+
},
|
| 176950 |
+
{
|
| 176951 |
+
"epoch": 491.02,
|
| 176952 |
+
"learning_rate": 9.039150641025642e-06,
|
| 176953 |
+
"loss": 0.341,
|
| 176954 |
+
"step": 60980
|
| 176955 |
+
},
|
| 176956 |
+
{
|
| 176957 |
+
"epoch": 491.06,
|
| 176958 |
+
"learning_rate": 9.039070512820513e-06,
|
| 176959 |
+
"loss": 0.3259,
|
| 176960 |
+
"step": 60985
|
| 176961 |
+
},
|
| 176962 |
+
{
|
| 176963 |
+
"epoch": 491.1,
|
| 176964 |
+
"learning_rate": 9.038990384615385e-06,
|
| 176965 |
+
"loss": 0.3462,
|
| 176966 |
+
"step": 60990
|
| 176967 |
+
},
|
| 176968 |
+
{
|
| 176969 |
+
"epoch": 491.14,
|
| 176970 |
+
"learning_rate": 9.038910256410258e-06,
|
| 176971 |
+
"loss": 0.3834,
|
| 176972 |
+
"step": 60995
|
| 176973 |
+
},
|
| 176974 |
+
{
|
| 176975 |
+
"epoch": 491.18,
|
| 176976 |
+
"learning_rate": 9.038830128205129e-06,
|
| 176977 |
+
"loss": 0.6381,
|
| 176978 |
+
"step": 61000
|
| 176979 |
+
},
|
| 176980 |
+
{
|
| 176981 |
+
"epoch": 491.22,
|
| 176982 |
+
"learning_rate": 9.03875e-06,
|
| 176983 |
+
"loss": 1.1421,
|
| 176984 |
+
"step": 61005
|
| 176985 |
+
},
|
| 176986 |
+
{
|
| 176987 |
+
"epoch": 491.26,
|
| 176988 |
+
"learning_rate": 9.038669871794874e-06,
|
| 176989 |
+
"loss": 0.3287,
|
| 176990 |
+
"step": 61010
|
| 176991 |
+
},
|
| 176992 |
+
{
|
| 176993 |
+
"epoch": 491.3,
|
| 176994 |
+
"learning_rate": 9.038589743589745e-06,
|
| 176995 |
+
"loss": 0.3466,
|
| 176996 |
+
"step": 61015
|
| 176997 |
+
},
|
| 176998 |
+
{
|
| 176999 |
+
"epoch": 491.34,
|
| 177000 |
+
"learning_rate": 9.038509615384616e-06,
|
| 177001 |
+
"loss": 0.4107,
|
| 177002 |
+
"step": 61020
|
| 177003 |
+
},
|
| 177004 |
+
{
|
| 177005 |
+
"epoch": 491.38,
|
| 177006 |
+
"learning_rate": 9.038429487179488e-06,
|
| 177007 |
+
"loss": 0.651,
|
| 177008 |
+
"step": 61025
|
| 177009 |
+
},
|
| 177010 |
+
{
|
| 177011 |
+
"epoch": 491.42,
|
| 177012 |
+
"learning_rate": 9.03834935897436e-06,
|
| 177013 |
+
"loss": 1.0088,
|
| 177014 |
+
"step": 61030
|
| 177015 |
+
},
|
| 177016 |
+
{
|
| 177017 |
+
"epoch": 491.46,
|
| 177018 |
+
"learning_rate": 9.03826923076923e-06,
|
| 177019 |
+
"loss": 0.3308,
|
| 177020 |
+
"step": 61035
|
| 177021 |
+
},
|
| 177022 |
+
{
|
| 177023 |
+
"epoch": 491.5,
|
| 177024 |
+
"learning_rate": 9.038189102564103e-06,
|
| 177025 |
+
"loss": 0.3197,
|
| 177026 |
+
"step": 61040
|
| 177027 |
+
},
|
| 177028 |
+
{
|
| 177029 |
+
"epoch": 491.54,
|
| 177030 |
+
"learning_rate": 9.038108974358976e-06,
|
| 177031 |
+
"loss": 0.3684,
|
| 177032 |
+
"step": 61045
|
| 177033 |
+
},
|
| 177034 |
+
{
|
| 177035 |
+
"epoch": 491.58,
|
| 177036 |
+
"learning_rate": 9.038028846153846e-06,
|
| 177037 |
+
"loss": 0.6528,
|
| 177038 |
+
"step": 61050
|
| 177039 |
+
},
|
| 177040 |
+
{
|
| 177041 |
+
"epoch": 491.62,
|
| 177042 |
+
"learning_rate": 9.03794871794872e-06,
|
| 177043 |
+
"loss": 1.0219,
|
| 177044 |
+
"step": 61055
|
| 177045 |
+
},
|
| 177046 |
+
{
|
| 177047 |
+
"epoch": 491.66,
|
| 177048 |
+
"learning_rate": 9.03786858974359e-06,
|
| 177049 |
+
"loss": 0.4409,
|
| 177050 |
+
"step": 61060
|
| 177051 |
+
},
|
| 177052 |
+
{
|
| 177053 |
+
"epoch": 491.7,
|
| 177054 |
+
"learning_rate": 9.037788461538462e-06,
|
| 177055 |
+
"loss": 0.3411,
|
| 177056 |
+
"step": 61065
|
| 177057 |
+
},
|
| 177058 |
+
{
|
| 177059 |
+
"epoch": 491.74,
|
| 177060 |
+
"learning_rate": 9.037708333333333e-06,
|
| 177061 |
+
"loss": 0.415,
|
| 177062 |
+
"step": 61070
|
| 177063 |
+
},
|
| 177064 |
+
{
|
| 177065 |
+
"epoch": 491.78,
|
| 177066 |
+
"learning_rate": 9.037628205128206e-06,
|
| 177067 |
+
"loss": 0.6908,
|
| 177068 |
+
"step": 61075
|
| 177069 |
+
},
|
| 177070 |
+
{
|
| 177071 |
+
"epoch": 491.82,
|
| 177072 |
+
"learning_rate": 9.037548076923078e-06,
|
| 177073 |
+
"loss": 1.0197,
|
| 177074 |
+
"step": 61080
|
| 177075 |
+
},
|
| 177076 |
+
{
|
| 177077 |
+
"epoch": 491.86,
|
| 177078 |
+
"learning_rate": 9.037467948717949e-06,
|
| 177079 |
+
"loss": 0.3332,
|
| 177080 |
+
"step": 61085
|
| 177081 |
+
},
|
| 177082 |
+
{
|
| 177083 |
+
"epoch": 491.9,
|
| 177084 |
+
"learning_rate": 9.03738782051282e-06,
|
| 177085 |
+
"loss": 0.3519,
|
| 177086 |
+
"step": 61090
|
| 177087 |
+
},
|
| 177088 |
+
{
|
| 177089 |
+
"epoch": 491.94,
|
| 177090 |
+
"learning_rate": 9.037307692307693e-06,
|
| 177091 |
+
"loss": 0.3885,
|
| 177092 |
+
"step": 61095
|
| 177093 |
+
},
|
| 177094 |
+
{
|
| 177095 |
+
"epoch": 491.98,
|
| 177096 |
+
"learning_rate": 9.037227564102565e-06,
|
| 177097 |
+
"loss": 0.6986,
|
| 177098 |
+
"step": 61100
|
| 177099 |
+
},
|
| 177100 |
+
{
|
| 177101 |
+
"epoch": 492.0,
|
| 177102 |
+
"eval_loss": 0.4045012295246124,
|
| 177103 |
+
"eval_runtime": 38.0364,
|
| 177104 |
+
"eval_samples_per_second": 22.216,
|
| 177105 |
+
"eval_steps_per_second": 0.71,
|
| 177106 |
+
"eval_wer": 0.1972527878295547,
|
| 177107 |
+
"step": 61102
|
| 177108 |
+
},
|
| 177109 |
+
{
|
| 177110 |
+
"epoch": 492.02,
|
| 177111 |
+
"learning_rate": 9.037147435897436e-06,
|
| 177112 |
+
"loss": 0.3484,
|
| 177113 |
+
"step": 61105
|
| 177114 |
+
},
|
| 177115 |
+
{
|
| 177116 |
+
"epoch": 492.06,
|
| 177117 |
+
"learning_rate": 9.03706730769231e-06,
|
| 177118 |
+
"loss": 0.29,
|
| 177119 |
+
"step": 61110
|
| 177120 |
+
},
|
| 177121 |
+
{
|
| 177122 |
+
"epoch": 492.1,
|
| 177123 |
+
"learning_rate": 9.03698717948718e-06,
|
| 177124 |
+
"loss": 0.3342,
|
| 177125 |
+
"step": 61115
|
| 177126 |
+
},
|
| 177127 |
+
{
|
| 177128 |
+
"epoch": 492.14,
|
| 177129 |
+
"learning_rate": 9.036907051282052e-06,
|
| 177130 |
+
"loss": 0.4317,
|
| 177131 |
+
"step": 61120
|
| 177132 |
+
},
|
| 177133 |
+
{
|
| 177134 |
+
"epoch": 492.18,
|
| 177135 |
+
"learning_rate": 9.036826923076923e-06,
|
| 177136 |
+
"loss": 0.6594,
|
| 177137 |
+
"step": 61125
|
| 177138 |
+
},
|
| 177139 |
+
{
|
| 177140 |
+
"epoch": 492.22,
|
| 177141 |
+
"learning_rate": 9.036746794871796e-06,
|
| 177142 |
+
"loss": 1.2607,
|
| 177143 |
+
"step": 61130
|
| 177144 |
+
},
|
| 177145 |
+
{
|
| 177146 |
+
"epoch": 492.27,
|
| 177147 |
+
"learning_rate": 9.036666666666668e-06,
|
| 177148 |
+
"loss": 0.529,
|
| 177149 |
+
"step": 61135
|
| 177150 |
+
},
|
| 177151 |
+
{
|
| 177152 |
+
"epoch": 492.31,
|
| 177153 |
+
"learning_rate": 9.036586538461539e-06,
|
| 177154 |
+
"loss": 0.3166,
|
| 177155 |
+
"step": 61140
|
| 177156 |
+
},
|
| 177157 |
+
{
|
| 177158 |
+
"epoch": 492.35,
|
| 177159 |
+
"learning_rate": 9.036506410256412e-06,
|
| 177160 |
+
"loss": 0.4587,
|
| 177161 |
+
"step": 61145
|
| 177162 |
+
},
|
| 177163 |
+
{
|
| 177164 |
+
"epoch": 492.39,
|
| 177165 |
+
"learning_rate": 9.036426282051283e-06,
|
| 177166 |
+
"loss": 0.7595,
|
| 177167 |
+
"step": 61150
|
| 177168 |
+
},
|
| 177169 |
+
{
|
| 177170 |
+
"epoch": 492.43,
|
| 177171 |
+
"learning_rate": 9.036346153846155e-06,
|
| 177172 |
+
"loss": 0.799,
|
| 177173 |
+
"step": 61155
|
| 177174 |
+
},
|
| 177175 |
+
{
|
| 177176 |
+
"epoch": 492.47,
|
| 177177 |
+
"learning_rate": 9.036266025641026e-06,
|
| 177178 |
+
"loss": 0.2994,
|
| 177179 |
+
"step": 61160
|
| 177180 |
+
},
|
| 177181 |
+
{
|
| 177182 |
+
"epoch": 492.51,
|
| 177183 |
+
"learning_rate": 9.0361858974359e-06,
|
| 177184 |
+
"loss": 0.379,
|
| 177185 |
+
"step": 61165
|
| 177186 |
+
},
|
| 177187 |
+
{
|
| 177188 |
+
"epoch": 492.55,
|
| 177189 |
+
"learning_rate": 9.036105769230769e-06,
|
| 177190 |
+
"loss": 0.4963,
|
| 177191 |
+
"step": 61170
|
| 177192 |
+
},
|
| 177193 |
+
{
|
| 177194 |
+
"epoch": 492.59,
|
| 177195 |
+
"learning_rate": 9.036025641025642e-06,
|
| 177196 |
+
"loss": 0.9508,
|
| 177197 |
+
"step": 61175
|
| 177198 |
+
},
|
| 177199 |
+
{
|
| 177200 |
+
"epoch": 492.63,
|
| 177201 |
+
"learning_rate": 9.035945512820513e-06,
|
| 177202 |
+
"loss": 0.9842,
|
| 177203 |
+
"step": 61180
|
| 177204 |
+
},
|
| 177205 |
+
{
|
| 177206 |
+
"epoch": 492.67,
|
| 177207 |
+
"learning_rate": 9.035865384615385e-06,
|
| 177208 |
+
"loss": 0.3276,
|
| 177209 |
+
"step": 61185
|
| 177210 |
+
},
|
| 177211 |
+
{
|
| 177212 |
+
"epoch": 492.71,
|
| 177213 |
+
"learning_rate": 9.035785256410256e-06,
|
| 177214 |
+
"loss": 0.323,
|
| 177215 |
+
"step": 61190
|
| 177216 |
+
},
|
| 177217 |
+
{
|
| 177218 |
+
"epoch": 492.75,
|
| 177219 |
+
"learning_rate": 9.035705128205129e-06,
|
| 177220 |
+
"loss": 0.3674,
|
| 177221 |
+
"step": 61195
|
| 177222 |
+
},
|
| 177223 |
+
{
|
| 177224 |
+
"epoch": 492.79,
|
| 177225 |
+
"learning_rate": 9.035625e-06,
|
| 177226 |
+
"loss": 0.8951,
|
| 177227 |
+
"step": 61200
|
| 177228 |
+
},
|
| 177229 |
+
{
|
| 177230 |
+
"epoch": 492.83,
|
| 177231 |
+
"learning_rate": 9.035544871794872e-06,
|
| 177232 |
+
"loss": 0.9872,
|
| 177233 |
+
"step": 61205
|
| 177234 |
+
},
|
| 177235 |
+
{
|
| 177236 |
+
"epoch": 492.87,
|
| 177237 |
+
"learning_rate": 9.035464743589745e-06,
|
| 177238 |
+
"loss": 0.3025,
|
| 177239 |
+
"step": 61210
|
| 177240 |
+
},
|
| 177241 |
+
{
|
| 177242 |
+
"epoch": 492.91,
|
| 177243 |
+
"learning_rate": 9.035384615384616e-06,
|
| 177244 |
+
"loss": 0.3818,
|
| 177245 |
+
"step": 61215
|
| 177246 |
+
},
|
| 177247 |
+
{
|
| 177248 |
+
"epoch": 492.95,
|
| 177249 |
+
"learning_rate": 9.035304487179488e-06,
|
| 177250 |
+
"loss": 0.4108,
|
| 177251 |
+
"step": 61220
|
| 177252 |
+
},
|
| 177253 |
+
{
|
| 177254 |
+
"epoch": 492.99,
|
| 177255 |
+
"learning_rate": 9.035224358974359e-06,
|
| 177256 |
+
"loss": 0.8702,
|
| 177257 |
+
"step": 61225
|
| 177258 |
+
},
|
| 177259 |
+
{
|
| 177260 |
+
"epoch": 493.0,
|
| 177261 |
+
"eval_loss": 0.41527244448661804,
|
| 177262 |
+
"eval_runtime": 41.5752,
|
| 177263 |
+
"eval_samples_per_second": 20.325,
|
| 177264 |
+
"eval_steps_per_second": 0.649,
|
| 177265 |
+
"eval_wer": 0.19053151321521927,
|
| 177266 |
+
"step": 61226
|
| 177267 |
+
},
|
| 177268 |
+
{
|
| 177269 |
+
"epoch": 493.03,
|
| 177270 |
+
"learning_rate": 9.035144230769232e-06,
|
| 177271 |
+
"loss": 0.403,
|
| 177272 |
+
"step": 61230
|
| 177273 |
+
},
|
| 177274 |
+
{
|
| 177275 |
+
"epoch": 493.07,
|
| 177276 |
+
"learning_rate": 9.035064102564103e-06,
|
| 177277 |
+
"loss": 0.3474,
|
| 177278 |
+
"step": 61235
|
| 177279 |
+
},
|
| 177280 |
+
{
|
| 177281 |
+
"epoch": 493.11,
|
| 177282 |
+
"learning_rate": 9.034983974358975e-06,
|
| 177283 |
+
"loss": 0.3855,
|
| 177284 |
+
"step": 61240
|
| 177285 |
+
},
|
| 177286 |
+
{
|
| 177287 |
+
"epoch": 493.15,
|
| 177288 |
+
"learning_rate": 9.034903846153848e-06,
|
| 177289 |
+
"loss": 0.4023,
|
| 177290 |
+
"step": 61245
|
| 177291 |
+
},
|
| 177292 |
+
{
|
| 177293 |
+
"epoch": 493.19,
|
| 177294 |
+
"learning_rate": 9.034823717948719e-06,
|
| 177295 |
+
"loss": 1.1158,
|
| 177296 |
+
"step": 61250
|
| 177297 |
+
},
|
| 177298 |
+
{
|
| 177299 |
+
"epoch": 493.23,
|
| 177300 |
+
"learning_rate": 9.03474358974359e-06,
|
| 177301 |
+
"loss": 0.7991,
|
| 177302 |
+
"step": 61255
|
| 177303 |
+
},
|
| 177304 |
+
{
|
| 177305 |
+
"epoch": 493.27,
|
| 177306 |
+
"learning_rate": 9.034663461538462e-06,
|
| 177307 |
+
"loss": 0.3647,
|
| 177308 |
+
"step": 61260
|
| 177309 |
+
},
|
| 177310 |
+
{
|
| 177311 |
+
"epoch": 493.31,
|
| 177312 |
+
"learning_rate": 9.034583333333335e-06,
|
| 177313 |
+
"loss": 0.342,
|
| 177314 |
+
"step": 61265
|
| 177315 |
+
},
|
| 177316 |
+
{
|
| 177317 |
+
"epoch": 493.35,
|
| 177318 |
+
"learning_rate": 9.034503205128206e-06,
|
| 177319 |
+
"loss": 0.4179,
|
| 177320 |
+
"step": 61270
|
| 177321 |
+
},
|
| 177322 |
+
{
|
| 177323 |
+
"epoch": 493.4,
|
| 177324 |
+
"learning_rate": 9.034423076923078e-06,
|
| 177325 |
+
"loss": 0.968,
|
| 177326 |
+
"step": 61275
|
| 177327 |
+
},
|
| 177328 |
+
{
|
| 177329 |
+
"epoch": 493.44,
|
| 177330 |
+
"learning_rate": 9.034342948717949e-06,
|
| 177331 |
+
"loss": 0.6878,
|
| 177332 |
+
"step": 61280
|
| 177333 |
+
},
|
| 177334 |
+
{
|
| 177335 |
+
"epoch": 493.48,
|
| 177336 |
+
"learning_rate": 9.034262820512822e-06,
|
| 177337 |
+
"loss": 0.3062,
|
| 177338 |
+
"step": 61285
|
| 177339 |
+
},
|
| 177340 |
+
{
|
| 177341 |
+
"epoch": 493.52,
|
| 177342 |
+
"learning_rate": 9.034182692307693e-06,
|
| 177343 |
+
"loss": 0.2974,
|
| 177344 |
+
"step": 61290
|
| 177345 |
+
},
|
| 177346 |
+
{
|
| 177347 |
+
"epoch": 493.56,
|
| 177348 |
+
"learning_rate": 9.034102564102565e-06,
|
| 177349 |
+
"loss": 0.4139,
|
| 177350 |
+
"step": 61295
|
| 177351 |
+
},
|
| 177352 |
+
{
|
| 177353 |
+
"epoch": 493.6,
|
| 177354 |
+
"learning_rate": 9.034022435897438e-06,
|
| 177355 |
+
"loss": 0.9779,
|
| 177356 |
+
"step": 61300
|
| 177357 |
+
},
|
| 177358 |
+
{
|
| 177359 |
+
"epoch": 493.64,
|
| 177360 |
+
"learning_rate": 9.033942307692309e-06,
|
| 177361 |
+
"loss": 0.5961,
|
| 177362 |
+
"step": 61305
|
| 177363 |
+
},
|
| 177364 |
+
{
|
| 177365 |
+
"epoch": 493.68,
|
| 177366 |
+
"learning_rate": 9.03386217948718e-06,
|
| 177367 |
+
"loss": 0.3516,
|
| 177368 |
+
"step": 61310
|
| 177369 |
+
},
|
| 177370 |
+
{
|
| 177371 |
+
"epoch": 493.72,
|
| 177372 |
+
"learning_rate": 9.033782051282052e-06,
|
| 177373 |
+
"loss": 0.369,
|
| 177374 |
+
"step": 61315
|
| 177375 |
+
},
|
| 177376 |
+
{
|
| 177377 |
+
"epoch": 493.76,
|
| 177378 |
+
"learning_rate": 9.033701923076925e-06,
|
| 177379 |
+
"loss": 0.3593,
|
| 177380 |
+
"step": 61320
|
| 177381 |
+
},
|
| 177382 |
+
{
|
| 177383 |
+
"epoch": 493.8,
|
| 177384 |
+
"learning_rate": 9.033621794871795e-06,
|
| 177385 |
+
"loss": 1.1128,
|
| 177386 |
+
"step": 61325
|
| 177387 |
+
},
|
| 177388 |
+
{
|
| 177389 |
+
"epoch": 493.84,
|
| 177390 |
+
"learning_rate": 9.033541666666668e-06,
|
| 177391 |
+
"loss": 0.7125,
|
| 177392 |
+
"step": 61330
|
| 177393 |
+
},
|
| 177394 |
+
{
|
| 177395 |
+
"epoch": 493.88,
|
| 177396 |
+
"learning_rate": 9.033461538461539e-06,
|
| 177397 |
+
"loss": 0.2982,
|
| 177398 |
+
"step": 61335
|
| 177399 |
+
},
|
| 177400 |
+
{
|
| 177401 |
+
"epoch": 493.92,
|
| 177402 |
+
"learning_rate": 9.03338141025641e-06,
|
| 177403 |
+
"loss": 0.3721,
|
| 177404 |
+
"step": 61340
|
| 177405 |
+
},
|
| 177406 |
+
{
|
| 177407 |
+
"epoch": 493.96,
|
| 177408 |
+
"learning_rate": 9.033301282051283e-06,
|
| 177409 |
+
"loss": 0.4503,
|
| 177410 |
+
"step": 61345
|
| 177411 |
+
},
|
| 177412 |
+
{
|
| 177413 |
+
"epoch": 494.0,
|
| 177414 |
+
"learning_rate": 9.033221153846155e-06,
|
| 177415 |
+
"loss": 1.2852,
|
| 177416 |
+
"step": 61350
|
| 177417 |
+
},
|
| 177418 |
+
{
|
| 177419 |
+
"epoch": 494.0,
|
| 177420 |
+
"eval_loss": 0.44114968180656433,
|
| 177421 |
+
"eval_runtime": 39.8577,
|
| 177422 |
+
"eval_samples_per_second": 21.175,
|
| 177423 |
+
"eval_steps_per_second": 0.677,
|
| 177424 |
+
"eval_wer": 0.19714999265462024,
|
| 177425 |
+
"step": 61350
|
| 177426 |
}
|
| 177427 |
],
|
| 177428 |
"max_steps": 620000,
|
| 177429 |
"num_train_epochs": 5000,
|
| 177430 |
+
"total_flos": 1.726617692128464e+20,
|
| 177431 |
"trial_name": null,
|
| 177432 |
"trial_params": null
|
| 177433 |
}
|
model-bin/finetune/base/{checkpoint-60604 → checkpoint-61350}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629801593.3967817/events.out.tfevents.1629801593.c435e1c5ee04.920.111
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:dd7ebb6b9ad8bdcb543b4d0b20493ae5f5c2d74faaa9abb297525498ba670dd2
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629802228.319492/events.out.tfevents.1629802228.c435e1c5ee04.920.113
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9bdebc67ff479d3c7a2dbca1af927fed594257ef4b3272741471a6cbe4ed290e
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629802982.729967/events.out.tfevents.1629802982.c435e1c5ee04.920.115
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:803609220a3c1c6b285ba8193e913ac1accfacd5683e8618ab6d512c6788b3bb
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629803633.8011563/events.out.tfevents.1629803633.c435e1c5ee04.920.117
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:fbf42755bccd1212cd17d5ec34bb0ccb56632d0e7a48f49fa8e1910d9aeee5b4
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629804285.3739185/events.out.tfevents.1629804285.c435e1c5ee04.920.119
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8cdc9f3670826ec32a43a56c0da8c1873056c8b9dac6ba0d35260eaf6cfbe0db
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629801593.c435e1c5ee04.920.110
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f6fff149151cfa3dbdc11f79036d4a2096438833e33402587b49702d1d38487d
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629802228.c435e1c5ee04.920.112
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:90c340b035d4304c7511ab2db433a6d0437c8906b59f6c9e2777fb8f1b16487c
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629802982.c435e1c5ee04.920.114
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:3f0aac5616423719fba7b23a3b8af9d3d46a2fa430725705fe9a065c0cd4b5a0
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629803633.c435e1c5ee04.920.116
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b55ce2b849ac588f7a9a4ae4c1f5e343fea811d7d3a475be3d235173366d6579
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629804285.c435e1c5ee04.920.118
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:49e5c45499277e77336a08ac9e7a8c75f0b85b40bf898b671ff93389db7c45d6
|
| 3 |
+
size 8622
|