"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-76532 → checkpoint-77155}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-76532 → checkpoint-77155}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-76532 → checkpoint-77155}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-76532 → checkpoint-77155}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-76532 → checkpoint-77155}/rng_state.pth +2 -2
- model-bin/finetune/base/{checkpoint-76532 → checkpoint-77155}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-76532 → checkpoint-77155}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-76532 → checkpoint-77155}/trainer_state.json +799 -4
- model-bin/finetune/base/{checkpoint-76532 → checkpoint-77155}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629889687.782439/events.out.tfevents.1629889687.7e498afd5545.7645.5 +3 -0
- model-bin/finetune/base/log/1629890176.1350722/events.out.tfevents.1629890176.7e498afd5545.7645.7 +3 -0
- model-bin/finetune/base/log/1629890651.9197931/events.out.tfevents.1629890651.7e498afd5545.7645.9 +3 -0
- model-bin/finetune/base/log/1629891122.8340254/events.out.tfevents.1629891122.7e498afd5545.7645.11 +3 -0
- model-bin/finetune/base/log/1629891582.6826005/events.out.tfevents.1629891582.7e498afd5545.7645.13 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629889687.7e498afd5545.7645.4 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629890175.7e498afd5545.7645.6 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629890651.7e498afd5545.7645.8 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629891122.7e498afd5545.7645.10 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629891582.7e498afd5545.7645.12 +3 -0
model-bin/finetune/base/{checkpoint-76532 → checkpoint-77155}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-76532 → checkpoint-77155}/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:e71760057eeaf95512c73bda8c72702fdcd07ff2723d1f896c7d0a878caeb038
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-76532 → checkpoint-77155}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-76532 → checkpoint-77155}/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 sha256:609b28af0de8e6654be054f551fe06d6ac03c459c4482a5a4552b7a7f7c75d86
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-76532 → checkpoint-77155}/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 sha256:6b164785d44f4581870a7236a75a3744bb07d4b54e54ff053a662784001b7aba
|
| 3 |
+
size 14503
|
model-bin/finetune/base/{checkpoint-76532 → checkpoint-77155}/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 sha256:86fe6f6b7182e301e95d9029914dff71904a56f874a0ac7549c42dca4b397c8c
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-76532 → checkpoint-77155}/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 sha256:5591339c9f642aace34467285535f0b0980daab6c067438026e89ebdc4a46cf7
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-76532 → checkpoint-77155}/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,
|
|
@@ -196737,11 +196737,806 @@
|
|
| 196737 |
"eval_steps_per_second": 0.625,
|
| 196738 |
"eval_wer": 0.18634209392503231,
|
| 196739 |
"step": 76532
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 196740 |
}
|
| 196741 |
],
|
| 196742 |
-
"max_steps":
|
| 196743 |
"num_train_epochs": 5000,
|
| 196744 |
-
"total_flos": 2.
|
| 196745 |
"trial_name": null,
|
| 196746 |
"trial_params": null
|
| 196747 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.18412114350410416,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-69565",
|
| 4 |
+
"epoch": 621.995983935743,
|
| 5 |
+
"global_step": 77155,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 196737 |
"eval_steps_per_second": 0.625,
|
| 196738 |
"eval_wer": 0.18634209392503231,
|
| 196739 |
"step": 76532
|
| 196740 |
+
},
|
| 196741 |
+
{
|
| 196742 |
+
"epoch": 612.02,
|
| 196743 |
+
"learning_rate": 8.780193861066237e-06,
|
| 196744 |
+
"loss": 0.3461,
|
| 196745 |
+
"step": 76535
|
| 196746 |
+
},
|
| 196747 |
+
{
|
| 196748 |
+
"epoch": 612.06,
|
| 196749 |
+
"learning_rate": 8.780113085621971e-06,
|
| 196750 |
+
"loss": 0.3694,
|
| 196751 |
+
"step": 76540
|
| 196752 |
+
},
|
| 196753 |
+
{
|
| 196754 |
+
"epoch": 612.1,
|
| 196755 |
+
"learning_rate": 8.780032310177707e-06,
|
| 196756 |
+
"loss": 0.3309,
|
| 196757 |
+
"step": 76545
|
| 196758 |
+
},
|
| 196759 |
+
{
|
| 196760 |
+
"epoch": 612.14,
|
| 196761 |
+
"learning_rate": 8.779951534733441e-06,
|
| 196762 |
+
"loss": 0.3637,
|
| 196763 |
+
"step": 76550
|
| 196764 |
+
},
|
| 196765 |
+
{
|
| 196766 |
+
"epoch": 612.18,
|
| 196767 |
+
"learning_rate": 8.779870759289177e-06,
|
| 196768 |
+
"loss": 0.7028,
|
| 196769 |
+
"step": 76555
|
| 196770 |
+
},
|
| 196771 |
+
{
|
| 196772 |
+
"epoch": 612.22,
|
| 196773 |
+
"learning_rate": 8.779789983844911e-06,
|
| 196774 |
+
"loss": 0.9974,
|
| 196775 |
+
"step": 76560
|
| 196776 |
+
},
|
| 196777 |
+
{
|
| 196778 |
+
"epoch": 612.26,
|
| 196779 |
+
"learning_rate": 8.779709208400647e-06,
|
| 196780 |
+
"loss": 0.2751,
|
| 196781 |
+
"step": 76565
|
| 196782 |
+
},
|
| 196783 |
+
{
|
| 196784 |
+
"epoch": 612.3,
|
| 196785 |
+
"learning_rate": 8.779628432956381e-06,
|
| 196786 |
+
"loss": 0.3144,
|
| 196787 |
+
"step": 76570
|
| 196788 |
+
},
|
| 196789 |
+
{
|
| 196790 |
+
"epoch": 612.34,
|
| 196791 |
+
"learning_rate": 8.779547657512117e-06,
|
| 196792 |
+
"loss": 0.3936,
|
| 196793 |
+
"step": 76575
|
| 196794 |
+
},
|
| 196795 |
+
{
|
| 196796 |
+
"epoch": 612.38,
|
| 196797 |
+
"learning_rate": 8.779466882067853e-06,
|
| 196798 |
+
"loss": 0.843,
|
| 196799 |
+
"step": 76580
|
| 196800 |
+
},
|
| 196801 |
+
{
|
| 196802 |
+
"epoch": 612.42,
|
| 196803 |
+
"learning_rate": 8.779386106623587e-06,
|
| 196804 |
+
"loss": 0.9242,
|
| 196805 |
+
"step": 76585
|
| 196806 |
+
},
|
| 196807 |
+
{
|
| 196808 |
+
"epoch": 612.46,
|
| 196809 |
+
"learning_rate": 8.779305331179323e-06,
|
| 196810 |
+
"loss": 0.2976,
|
| 196811 |
+
"step": 76590
|
| 196812 |
+
},
|
| 196813 |
+
{
|
| 196814 |
+
"epoch": 612.5,
|
| 196815 |
+
"learning_rate": 8.779224555735057e-06,
|
| 196816 |
+
"loss": 0.2867,
|
| 196817 |
+
"step": 76595
|
| 196818 |
+
},
|
| 196819 |
+
{
|
| 196820 |
+
"epoch": 612.54,
|
| 196821 |
+
"learning_rate": 8.779143780290793e-06,
|
| 196822 |
+
"loss": 0.4087,
|
| 196823 |
+
"step": 76600
|
| 196824 |
+
},
|
| 196825 |
+
{
|
| 196826 |
+
"epoch": 612.58,
|
| 196827 |
+
"learning_rate": 8.779063004846527e-06,
|
| 196828 |
+
"loss": 0.853,
|
| 196829 |
+
"step": 76605
|
| 196830 |
+
},
|
| 196831 |
+
{
|
| 196832 |
+
"epoch": 612.62,
|
| 196833 |
+
"learning_rate": 8.778982229402263e-06,
|
| 196834 |
+
"loss": 0.9834,
|
| 196835 |
+
"step": 76610
|
| 196836 |
+
},
|
| 196837 |
+
{
|
| 196838 |
+
"epoch": 612.66,
|
| 196839 |
+
"learning_rate": 8.778901453957997e-06,
|
| 196840 |
+
"loss": 0.3562,
|
| 196841 |
+
"step": 76615
|
| 196842 |
+
},
|
| 196843 |
+
{
|
| 196844 |
+
"epoch": 612.7,
|
| 196845 |
+
"learning_rate": 8.778820678513733e-06,
|
| 196846 |
+
"loss": 0.2972,
|
| 196847 |
+
"step": 76620
|
| 196848 |
+
},
|
| 196849 |
+
{
|
| 196850 |
+
"epoch": 612.74,
|
| 196851 |
+
"learning_rate": 8.778739903069467e-06,
|
| 196852 |
+
"loss": 0.4146,
|
| 196853 |
+
"step": 76625
|
| 196854 |
+
},
|
| 196855 |
+
{
|
| 196856 |
+
"epoch": 612.78,
|
| 196857 |
+
"learning_rate": 8.778659127625203e-06,
|
| 196858 |
+
"loss": 0.7674,
|
| 196859 |
+
"step": 76630
|
| 196860 |
+
},
|
| 196861 |
+
{
|
| 196862 |
+
"epoch": 612.82,
|
| 196863 |
+
"learning_rate": 8.778578352180937e-06,
|
| 196864 |
+
"loss": 0.9026,
|
| 196865 |
+
"step": 76635
|
| 196866 |
+
},
|
| 196867 |
+
{
|
| 196868 |
+
"epoch": 612.86,
|
| 196869 |
+
"learning_rate": 8.778497576736673e-06,
|
| 196870 |
+
"loss": 0.2904,
|
| 196871 |
+
"step": 76640
|
| 196872 |
+
},
|
| 196873 |
+
{
|
| 196874 |
+
"epoch": 612.9,
|
| 196875 |
+
"learning_rate": 8.778416801292408e-06,
|
| 196876 |
+
"loss": 0.2887,
|
| 196877 |
+
"step": 76645
|
| 196878 |
+
},
|
| 196879 |
+
{
|
| 196880 |
+
"epoch": 612.94,
|
| 196881 |
+
"learning_rate": 8.778336025848143e-06,
|
| 196882 |
+
"loss": 0.4136,
|
| 196883 |
+
"step": 76650
|
| 196884 |
+
},
|
| 196885 |
+
{
|
| 196886 |
+
"epoch": 612.98,
|
| 196887 |
+
"learning_rate": 8.778255250403878e-06,
|
| 196888 |
+
"loss": 0.8516,
|
| 196889 |
+
"step": 76655
|
| 196890 |
+
},
|
| 196891 |
+
{
|
| 196892 |
+
"epoch": 613.0,
|
| 196893 |
+
"eval_loss": 0.39394015073776245,
|
| 196894 |
+
"eval_runtime": 42.2649,
|
| 196895 |
+
"eval_samples_per_second": 19.804,
|
| 196896 |
+
"eval_steps_per_second": 0.639,
|
| 196897 |
+
"eval_wer": 0.19633147248604366,
|
| 196898 |
+
"step": 76657
|
| 196899 |
+
},
|
| 196900 |
+
{
|
| 196901 |
+
"epoch": 618.02,
|
| 196902 |
+
"learning_rate": 8.778174474959613e-06,
|
| 196903 |
+
"loss": 0.3448,
|
| 196904 |
+
"step": 76660
|
| 196905 |
+
},
|
| 196906 |
+
{
|
| 196907 |
+
"epoch": 618.06,
|
| 196908 |
+
"learning_rate": 8.778093699515348e-06,
|
| 196909 |
+
"loss": 0.3466,
|
| 196910 |
+
"step": 76665
|
| 196911 |
+
},
|
| 196912 |
+
{
|
| 196913 |
+
"epoch": 618.1,
|
| 196914 |
+
"learning_rate": 8.778012924071083e-06,
|
| 196915 |
+
"loss": 0.313,
|
| 196916 |
+
"step": 76670
|
| 196917 |
+
},
|
| 196918 |
+
{
|
| 196919 |
+
"epoch": 618.14,
|
| 196920 |
+
"learning_rate": 8.777932148626818e-06,
|
| 196921 |
+
"loss": 0.4076,
|
| 196922 |
+
"step": 76675
|
| 196923 |
+
},
|
| 196924 |
+
{
|
| 196925 |
+
"epoch": 618.18,
|
| 196926 |
+
"learning_rate": 8.777851373182552e-06,
|
| 196927 |
+
"loss": 0.7918,
|
| 196928 |
+
"step": 76680
|
| 196929 |
+
},
|
| 196930 |
+
{
|
| 196931 |
+
"epoch": 618.22,
|
| 196932 |
+
"learning_rate": 8.777770597738288e-06,
|
| 196933 |
+
"loss": 0.8454,
|
| 196934 |
+
"step": 76685
|
| 196935 |
+
},
|
| 196936 |
+
{
|
| 196937 |
+
"epoch": 618.27,
|
| 196938 |
+
"learning_rate": 8.777689822294022e-06,
|
| 196939 |
+
"loss": 0.338,
|
| 196940 |
+
"step": 76690
|
| 196941 |
+
},
|
| 196942 |
+
{
|
| 196943 |
+
"epoch": 618.31,
|
| 196944 |
+
"learning_rate": 8.777609046849758e-06,
|
| 196945 |
+
"loss": 0.3607,
|
| 196946 |
+
"step": 76695
|
| 196947 |
+
},
|
| 196948 |
+
{
|
| 196949 |
+
"epoch": 618.35,
|
| 196950 |
+
"learning_rate": 8.777528271405492e-06,
|
| 196951 |
+
"loss": 0.4404,
|
| 196952 |
+
"step": 76700
|
| 196953 |
+
},
|
| 196954 |
+
{
|
| 196955 |
+
"epoch": 618.39,
|
| 196956 |
+
"learning_rate": 8.777447495961228e-06,
|
| 196957 |
+
"loss": 0.8371,
|
| 196958 |
+
"step": 76705
|
| 196959 |
+
},
|
| 196960 |
+
{
|
| 196961 |
+
"epoch": 618.43,
|
| 196962 |
+
"learning_rate": 8.777366720516962e-06,
|
| 196963 |
+
"loss": 0.9516,
|
| 196964 |
+
"step": 76710
|
| 196965 |
+
},
|
| 196966 |
+
{
|
| 196967 |
+
"epoch": 618.47,
|
| 196968 |
+
"learning_rate": 8.777285945072698e-06,
|
| 196969 |
+
"loss": 0.3076,
|
| 196970 |
+
"step": 76715
|
| 196971 |
+
},
|
| 196972 |
+
{
|
| 196973 |
+
"epoch": 618.51,
|
| 196974 |
+
"learning_rate": 8.777205169628434e-06,
|
| 196975 |
+
"loss": 0.2998,
|
| 196976 |
+
"step": 76720
|
| 196977 |
+
},
|
| 196978 |
+
{
|
| 196979 |
+
"epoch": 618.55,
|
| 196980 |
+
"learning_rate": 8.777124394184168e-06,
|
| 196981 |
+
"loss": 0.3559,
|
| 196982 |
+
"step": 76725
|
| 196983 |
+
},
|
| 196984 |
+
{
|
| 196985 |
+
"epoch": 618.59,
|
| 196986 |
+
"learning_rate": 8.777043618739904e-06,
|
| 196987 |
+
"loss": 0.7376,
|
| 196988 |
+
"step": 76730
|
| 196989 |
+
},
|
| 196990 |
+
{
|
| 196991 |
+
"epoch": 618.63,
|
| 196992 |
+
"learning_rate": 8.776962843295638e-06,
|
| 196993 |
+
"loss": 0.9131,
|
| 196994 |
+
"step": 76735
|
| 196995 |
+
},
|
| 196996 |
+
{
|
| 196997 |
+
"epoch": 618.67,
|
| 196998 |
+
"learning_rate": 8.776882067851374e-06,
|
| 196999 |
+
"loss": 0.3395,
|
| 197000 |
+
"step": 76740
|
| 197001 |
+
},
|
| 197002 |
+
{
|
| 197003 |
+
"epoch": 618.71,
|
| 197004 |
+
"learning_rate": 8.776801292407108e-06,
|
| 197005 |
+
"loss": 0.3561,
|
| 197006 |
+
"step": 76745
|
| 197007 |
+
},
|
| 197008 |
+
{
|
| 197009 |
+
"epoch": 618.75,
|
| 197010 |
+
"learning_rate": 8.776720516962844e-06,
|
| 197011 |
+
"loss": 0.3623,
|
| 197012 |
+
"step": 76750
|
| 197013 |
+
},
|
| 197014 |
+
{
|
| 197015 |
+
"epoch": 618.79,
|
| 197016 |
+
"learning_rate": 8.776639741518578e-06,
|
| 197017 |
+
"loss": 0.8501,
|
| 197018 |
+
"step": 76755
|
| 197019 |
+
},
|
| 197020 |
+
{
|
| 197021 |
+
"epoch": 618.83,
|
| 197022 |
+
"learning_rate": 8.776558966074314e-06,
|
| 197023 |
+
"loss": 0.8904,
|
| 197024 |
+
"step": 76760
|
| 197025 |
+
},
|
| 197026 |
+
{
|
| 197027 |
+
"epoch": 618.87,
|
| 197028 |
+
"learning_rate": 8.776478190630048e-06,
|
| 197029 |
+
"loss": 0.3221,
|
| 197030 |
+
"step": 76765
|
| 197031 |
+
},
|
| 197032 |
+
{
|
| 197033 |
+
"epoch": 618.91,
|
| 197034 |
+
"learning_rate": 8.776397415185784e-06,
|
| 197035 |
+
"loss": 0.3619,
|
| 197036 |
+
"step": 76770
|
| 197037 |
+
},
|
| 197038 |
+
{
|
| 197039 |
+
"epoch": 618.95,
|
| 197040 |
+
"learning_rate": 8.776316639741518e-06,
|
| 197041 |
+
"loss": 0.3916,
|
| 197042 |
+
"step": 76775
|
| 197043 |
+
},
|
| 197044 |
+
{
|
| 197045 |
+
"epoch": 618.99,
|
| 197046 |
+
"learning_rate": 8.776235864297254e-06,
|
| 197047 |
+
"loss": 0.8502,
|
| 197048 |
+
"step": 76780
|
| 197049 |
+
},
|
| 197050 |
+
{
|
| 197051 |
+
"epoch": 619.0,
|
| 197052 |
+
"eval_loss": 0.41090697050094604,
|
| 197053 |
+
"eval_runtime": 42.6135,
|
| 197054 |
+
"eval_samples_per_second": 19.642,
|
| 197055 |
+
"eval_steps_per_second": 0.634,
|
| 197056 |
+
"eval_wer": 0.19282343246685146,
|
| 197057 |
+
"step": 76781
|
| 197058 |
+
},
|
| 197059 |
+
{
|
| 197060 |
+
"epoch": 614.03,
|
| 197061 |
+
"learning_rate": 8.77615508885299e-06,
|
| 197062 |
+
"loss": 0.3079,
|
| 197063 |
+
"step": 76785
|
| 197064 |
+
},
|
| 197065 |
+
{
|
| 197066 |
+
"epoch": 614.07,
|
| 197067 |
+
"learning_rate": 8.776074313408724e-06,
|
| 197068 |
+
"loss": 0.3059,
|
| 197069 |
+
"step": 76790
|
| 197070 |
+
},
|
| 197071 |
+
{
|
| 197072 |
+
"epoch": 614.11,
|
| 197073 |
+
"learning_rate": 8.77599353796446e-06,
|
| 197074 |
+
"loss": 0.3154,
|
| 197075 |
+
"step": 76795
|
| 197076 |
+
},
|
| 197077 |
+
{
|
| 197078 |
+
"epoch": 614.15,
|
| 197079 |
+
"learning_rate": 8.775912762520194e-06,
|
| 197080 |
+
"loss": 0.39,
|
| 197081 |
+
"step": 76800
|
| 197082 |
+
},
|
| 197083 |
+
{
|
| 197084 |
+
"epoch": 614.19,
|
| 197085 |
+
"learning_rate": 8.77583198707593e-06,
|
| 197086 |
+
"loss": 0.9246,
|
| 197087 |
+
"step": 76805
|
| 197088 |
+
},
|
| 197089 |
+
{
|
| 197090 |
+
"epoch": 614.23,
|
| 197091 |
+
"learning_rate": 8.775751211631664e-06,
|
| 197092 |
+
"loss": 0.6451,
|
| 197093 |
+
"step": 76810
|
| 197094 |
+
},
|
| 197095 |
+
{
|
| 197096 |
+
"epoch": 614.27,
|
| 197097 |
+
"learning_rate": 8.7756704361874e-06,
|
| 197098 |
+
"loss": 0.3489,
|
| 197099 |
+
"step": 76815
|
| 197100 |
+
},
|
| 197101 |
+
{
|
| 197102 |
+
"epoch": 614.31,
|
| 197103 |
+
"learning_rate": 8.775589660743134e-06,
|
| 197104 |
+
"loss": 0.3605,
|
| 197105 |
+
"step": 76820
|
| 197106 |
+
},
|
| 197107 |
+
{
|
| 197108 |
+
"epoch": 614.35,
|
| 197109 |
+
"learning_rate": 8.77550888529887e-06,
|
| 197110 |
+
"loss": 0.4545,
|
| 197111 |
+
"step": 76825
|
| 197112 |
+
},
|
| 197113 |
+
{
|
| 197114 |
+
"epoch": 614.39,
|
| 197115 |
+
"learning_rate": 8.775428109854604e-06,
|
| 197116 |
+
"loss": 0.9938,
|
| 197117 |
+
"step": 76830
|
| 197118 |
+
},
|
| 197119 |
+
{
|
| 197120 |
+
"epoch": 614.43,
|
| 197121 |
+
"learning_rate": 8.77534733441034e-06,
|
| 197122 |
+
"loss": 0.6444,
|
| 197123 |
+
"step": 76835
|
| 197124 |
+
},
|
| 197125 |
+
{
|
| 197126 |
+
"epoch": 614.47,
|
| 197127 |
+
"learning_rate": 8.775266558966074e-06,
|
| 197128 |
+
"loss": 0.2636,
|
| 197129 |
+
"step": 76840
|
| 197130 |
+
},
|
| 197131 |
+
{
|
| 197132 |
+
"epoch": 614.51,
|
| 197133 |
+
"learning_rate": 8.77518578352181e-06,
|
| 197134 |
+
"loss": 0.2717,
|
| 197135 |
+
"step": 76845
|
| 197136 |
+
},
|
| 197137 |
+
{
|
| 197138 |
+
"epoch": 614.55,
|
| 197139 |
+
"learning_rate": 8.775105008077545e-06,
|
| 197140 |
+
"loss": 0.4112,
|
| 197141 |
+
"step": 76850
|
| 197142 |
+
},
|
| 197143 |
+
{
|
| 197144 |
+
"epoch": 614.59,
|
| 197145 |
+
"learning_rate": 8.77502423263328e-06,
|
| 197146 |
+
"loss": 0.7701,
|
| 197147 |
+
"step": 76855
|
| 197148 |
+
},
|
| 197149 |
+
{
|
| 197150 |
+
"epoch": 614.63,
|
| 197151 |
+
"learning_rate": 8.774943457189015e-06,
|
| 197152 |
+
"loss": 0.7343,
|
| 197153 |
+
"step": 76860
|
| 197154 |
+
},
|
| 197155 |
+
{
|
| 197156 |
+
"epoch": 614.67,
|
| 197157 |
+
"learning_rate": 8.77486268174475e-06,
|
| 197158 |
+
"loss": 0.315,
|
| 197159 |
+
"step": 76865
|
| 197160 |
+
},
|
| 197161 |
+
{
|
| 197162 |
+
"epoch": 614.71,
|
| 197163 |
+
"learning_rate": 8.774781906300485e-06,
|
| 197164 |
+
"loss": 0.347,
|
| 197165 |
+
"step": 76870
|
| 197166 |
+
},
|
| 197167 |
+
{
|
| 197168 |
+
"epoch": 614.75,
|
| 197169 |
+
"learning_rate": 8.77470113085622e-06,
|
| 197170 |
+
"loss": 0.4801,
|
| 197171 |
+
"step": 76875
|
| 197172 |
+
},
|
| 197173 |
+
{
|
| 197174 |
+
"epoch": 614.79,
|
| 197175 |
+
"learning_rate": 8.774620355411955e-06,
|
| 197176 |
+
"loss": 0.9544,
|
| 197177 |
+
"step": 76880
|
| 197178 |
+
},
|
| 197179 |
+
{
|
| 197180 |
+
"epoch": 614.83,
|
| 197181 |
+
"learning_rate": 8.77453957996769e-06,
|
| 197182 |
+
"loss": 0.7629,
|
| 197183 |
+
"step": 76885
|
| 197184 |
+
},
|
| 197185 |
+
{
|
| 197186 |
+
"epoch": 614.87,
|
| 197187 |
+
"learning_rate": 8.774458804523425e-06,
|
| 197188 |
+
"loss": 0.3161,
|
| 197189 |
+
"step": 76890
|
| 197190 |
+
},
|
| 197191 |
+
{
|
| 197192 |
+
"epoch": 614.91,
|
| 197193 |
+
"learning_rate": 8.77437802907916e-06,
|
| 197194 |
+
"loss": 0.3741,
|
| 197195 |
+
"step": 76895
|
| 197196 |
+
},
|
| 197197 |
+
{
|
| 197198 |
+
"epoch": 614.95,
|
| 197199 |
+
"learning_rate": 8.774297253634895e-06,
|
| 197200 |
+
"loss": 0.3875,
|
| 197201 |
+
"step": 76900
|
| 197202 |
+
},
|
| 197203 |
+
{
|
| 197204 |
+
"epoch": 614.99,
|
| 197205 |
+
"learning_rate": 8.77421647819063e-06,
|
| 197206 |
+
"loss": 0.9842,
|
| 197207 |
+
"step": 76905
|
| 197208 |
+
},
|
| 197209 |
+
{
|
| 197210 |
+
"epoch": 615.0,
|
| 197211 |
+
"eval_loss": 0.3531261384487152,
|
| 197212 |
+
"eval_runtime": 40.1828,
|
| 197213 |
+
"eval_samples_per_second": 20.805,
|
| 197214 |
+
"eval_steps_per_second": 0.672,
|
| 197215 |
+
"eval_wer": 0.19276489580275902,
|
| 197216 |
+
"step": 76906
|
| 197217 |
+
},
|
| 197218 |
+
{
|
| 197219 |
+
"epoch": 615.03,
|
| 197220 |
+
"learning_rate": 8.774135702746365e-06,
|
| 197221 |
+
"loss": 0.3707,
|
| 197222 |
+
"step": 76910
|
| 197223 |
+
},
|
| 197224 |
+
{
|
| 197225 |
+
"epoch": 615.07,
|
| 197226 |
+
"learning_rate": 8.774054927302101e-06,
|
| 197227 |
+
"loss": 0.2975,
|
| 197228 |
+
"step": 76915
|
| 197229 |
+
},
|
| 197230 |
+
{
|
| 197231 |
+
"epoch": 615.11,
|
| 197232 |
+
"learning_rate": 8.773974151857835e-06,
|
| 197233 |
+
"loss": 0.3563,
|
| 197234 |
+
"step": 76920
|
| 197235 |
+
},
|
| 197236 |
+
{
|
| 197237 |
+
"epoch": 615.15,
|
| 197238 |
+
"learning_rate": 8.773893376413571e-06,
|
| 197239 |
+
"loss": 0.4086,
|
| 197240 |
+
"step": 76925
|
| 197241 |
+
},
|
| 197242 |
+
{
|
| 197243 |
+
"epoch": 615.19,
|
| 197244 |
+
"learning_rate": 8.773812600969305e-06,
|
| 197245 |
+
"loss": 0.9022,
|
| 197246 |
+
"step": 76930
|
| 197247 |
+
},
|
| 197248 |
+
{
|
| 197249 |
+
"epoch": 615.23,
|
| 197250 |
+
"learning_rate": 8.773731825525041e-06,
|
| 197251 |
+
"loss": 0.7052,
|
| 197252 |
+
"step": 76935
|
| 197253 |
+
},
|
| 197254 |
+
{
|
| 197255 |
+
"epoch": 615.27,
|
| 197256 |
+
"learning_rate": 8.773651050080775e-06,
|
| 197257 |
+
"loss": 0.3103,
|
| 197258 |
+
"step": 76940
|
| 197259 |
+
},
|
| 197260 |
+
{
|
| 197261 |
+
"epoch": 615.31,
|
| 197262 |
+
"learning_rate": 8.773570274636511e-06,
|
| 197263 |
+
"loss": 0.3271,
|
| 197264 |
+
"step": 76945
|
| 197265 |
+
},
|
| 197266 |
+
{
|
| 197267 |
+
"epoch": 615.35,
|
| 197268 |
+
"learning_rate": 8.773489499192245e-06,
|
| 197269 |
+
"loss": 0.4844,
|
| 197270 |
+
"step": 76950
|
| 197271 |
+
},
|
| 197272 |
+
{
|
| 197273 |
+
"epoch": 615.39,
|
| 197274 |
+
"learning_rate": 8.773408723747981e-06,
|
| 197275 |
+
"loss": 1.0046,
|
| 197276 |
+
"step": 76955
|
| 197277 |
+
},
|
| 197278 |
+
{
|
| 197279 |
+
"epoch": 615.43,
|
| 197280 |
+
"learning_rate": 8.773327948303715e-06,
|
| 197281 |
+
"loss": 0.7013,
|
| 197282 |
+
"step": 76960
|
| 197283 |
+
},
|
| 197284 |
+
{
|
| 197285 |
+
"epoch": 615.47,
|
| 197286 |
+
"learning_rate": 8.773247172859451e-06,
|
| 197287 |
+
"loss": 0.3092,
|
| 197288 |
+
"step": 76965
|
| 197289 |
+
},
|
| 197290 |
+
{
|
| 197291 |
+
"epoch": 615.51,
|
| 197292 |
+
"learning_rate": 8.773166397415187e-06,
|
| 197293 |
+
"loss": 0.406,
|
| 197294 |
+
"step": 76970
|
| 197295 |
+
},
|
| 197296 |
+
{
|
| 197297 |
+
"epoch": 615.55,
|
| 197298 |
+
"learning_rate": 8.773085621970921e-06,
|
| 197299 |
+
"loss": 0.4009,
|
| 197300 |
+
"step": 76975
|
| 197301 |
+
},
|
| 197302 |
+
{
|
| 197303 |
+
"epoch": 615.59,
|
| 197304 |
+
"learning_rate": 8.773004846526657e-06,
|
| 197305 |
+
"loss": 0.8637,
|
| 197306 |
+
"step": 76980
|
| 197307 |
+
},
|
| 197308 |
+
{
|
| 197309 |
+
"epoch": 615.63,
|
| 197310 |
+
"learning_rate": 8.772924071082391e-06,
|
| 197311 |
+
"loss": 0.6035,
|
| 197312 |
+
"step": 76985
|
| 197313 |
+
},
|
| 197314 |
+
{
|
| 197315 |
+
"epoch": 615.67,
|
| 197316 |
+
"learning_rate": 8.772843295638127e-06,
|
| 197317 |
+
"loss": 0.2557,
|
| 197318 |
+
"step": 76990
|
| 197319 |
+
},
|
| 197320 |
+
{
|
| 197321 |
+
"epoch": 615.71,
|
| 197322 |
+
"learning_rate": 8.772762520193861e-06,
|
| 197323 |
+
"loss": 0.329,
|
| 197324 |
+
"step": 76995
|
| 197325 |
+
},
|
| 197326 |
+
{
|
| 197327 |
+
"epoch": 615.75,
|
| 197328 |
+
"learning_rate": 8.772681744749597e-06,
|
| 197329 |
+
"loss": 0.4156,
|
| 197330 |
+
"step": 77000
|
| 197331 |
+
},
|
| 197332 |
+
{
|
| 197333 |
+
"epoch": 615.79,
|
| 197334 |
+
"learning_rate": 8.772600969305331e-06,
|
| 197335 |
+
"loss": 1.0505,
|
| 197336 |
+
"step": 77005
|
| 197337 |
+
},
|
| 197338 |
+
{
|
| 197339 |
+
"epoch": 615.83,
|
| 197340 |
+
"learning_rate": 8.772520193861067e-06,
|
| 197341 |
+
"loss": 0.6839,
|
| 197342 |
+
"step": 77010
|
| 197343 |
+
},
|
| 197344 |
+
{
|
| 197345 |
+
"epoch": 615.87,
|
| 197346 |
+
"learning_rate": 8.772439418416801e-06,
|
| 197347 |
+
"loss": 0.2779,
|
| 197348 |
+
"step": 77015
|
| 197349 |
+
},
|
| 197350 |
+
{
|
| 197351 |
+
"epoch": 615.91,
|
| 197352 |
+
"learning_rate": 8.772358642972537e-06,
|
| 197353 |
+
"loss": 0.3321,
|
| 197354 |
+
"step": 77020
|
| 197355 |
+
},
|
| 197356 |
+
{
|
| 197357 |
+
"epoch": 615.95,
|
| 197358 |
+
"learning_rate": 8.772277867528271e-06,
|
| 197359 |
+
"loss": 0.4941,
|
| 197360 |
+
"step": 77025
|
| 197361 |
+
},
|
| 197362 |
+
{
|
| 197363 |
+
"epoch": 615.99,
|
| 197364 |
+
"learning_rate": 8.772197092084007e-06,
|
| 197365 |
+
"loss": 0.882,
|
| 197366 |
+
"step": 77030
|
| 197367 |
+
},
|
| 197368 |
+
{
|
| 197369 |
+
"epoch": 616.0,
|
| 197370 |
+
"eval_loss": 0.37907248735427856,
|
| 197371 |
+
"eval_runtime": 42.2976,
|
| 197372 |
+
"eval_samples_per_second": 19.765,
|
| 197373 |
+
"eval_steps_per_second": 0.638,
|
| 197374 |
+
"eval_wer": 0.187862130441242,
|
| 197375 |
+
"step": 77031
|
| 197376 |
+
},
|
| 197377 |
+
{
|
| 197378 |
+
"epoch": 621.03,
|
| 197379 |
+
"learning_rate": 8.772116316639743e-06,
|
| 197380 |
+
"loss": 0.3323,
|
| 197381 |
+
"step": 77035
|
| 197382 |
+
},
|
| 197383 |
+
{
|
| 197384 |
+
"epoch": 621.07,
|
| 197385 |
+
"learning_rate": 8.772035541195477e-06,
|
| 197386 |
+
"loss": 0.3167,
|
| 197387 |
+
"step": 77040
|
| 197388 |
+
},
|
| 197389 |
+
{
|
| 197390 |
+
"epoch": 621.11,
|
| 197391 |
+
"learning_rate": 8.771954765751213e-06,
|
| 197392 |
+
"loss": 0.3232,
|
| 197393 |
+
"step": 77045
|
| 197394 |
+
},
|
| 197395 |
+
{
|
| 197396 |
+
"epoch": 621.15,
|
| 197397 |
+
"learning_rate": 8.771873990306947e-06,
|
| 197398 |
+
"loss": 0.3957,
|
| 197399 |
+
"step": 77050
|
| 197400 |
+
},
|
| 197401 |
+
{
|
| 197402 |
+
"epoch": 621.19,
|
| 197403 |
+
"learning_rate": 8.771793214862682e-06,
|
| 197404 |
+
"loss": 1.009,
|
| 197405 |
+
"step": 77055
|
| 197406 |
+
},
|
| 197407 |
+
{
|
| 197408 |
+
"epoch": 621.23,
|
| 197409 |
+
"learning_rate": 8.771712439418417e-06,
|
| 197410 |
+
"loss": 0.6908,
|
| 197411 |
+
"step": 77060
|
| 197412 |
+
},
|
| 197413 |
+
{
|
| 197414 |
+
"epoch": 621.27,
|
| 197415 |
+
"learning_rate": 8.771631663974152e-06,
|
| 197416 |
+
"loss": 0.2909,
|
| 197417 |
+
"step": 77065
|
| 197418 |
+
},
|
| 197419 |
+
{
|
| 197420 |
+
"epoch": 621.31,
|
| 197421 |
+
"learning_rate": 8.771550888529887e-06,
|
| 197422 |
+
"loss": 0.3166,
|
| 197423 |
+
"step": 77070
|
| 197424 |
+
},
|
| 197425 |
+
{
|
| 197426 |
+
"epoch": 621.35,
|
| 197427 |
+
"learning_rate": 8.771470113085622e-06,
|
| 197428 |
+
"loss": 0.4271,
|
| 197429 |
+
"step": 77075
|
| 197430 |
+
},
|
| 197431 |
+
{
|
| 197432 |
+
"epoch": 621.39,
|
| 197433 |
+
"learning_rate": 8.771389337641357e-06,
|
| 197434 |
+
"loss": 1.1016,
|
| 197435 |
+
"step": 77080
|
| 197436 |
+
},
|
| 197437 |
+
{
|
| 197438 |
+
"epoch": 621.43,
|
| 197439 |
+
"learning_rate": 8.771308562197092e-06,
|
| 197440 |
+
"loss": 0.7708,
|
| 197441 |
+
"step": 77085
|
| 197442 |
+
},
|
| 197443 |
+
{
|
| 197444 |
+
"epoch": 621.47,
|
| 197445 |
+
"learning_rate": 8.771227786752828e-06,
|
| 197446 |
+
"loss": 0.2747,
|
| 197447 |
+
"step": 77090
|
| 197448 |
+
},
|
| 197449 |
+
{
|
| 197450 |
+
"epoch": 621.51,
|
| 197451 |
+
"learning_rate": 8.771147011308562e-06,
|
| 197452 |
+
"loss": 0.3577,
|
| 197453 |
+
"step": 77095
|
| 197454 |
+
},
|
| 197455 |
+
{
|
| 197456 |
+
"epoch": 621.55,
|
| 197457 |
+
"learning_rate": 8.771066235864298e-06,
|
| 197458 |
+
"loss": 0.4163,
|
| 197459 |
+
"step": 77100
|
| 197460 |
+
},
|
| 197461 |
+
{
|
| 197462 |
+
"epoch": 621.59,
|
| 197463 |
+
"learning_rate": 8.770985460420032e-06,
|
| 197464 |
+
"loss": 1.0869,
|
| 197465 |
+
"step": 77105
|
| 197466 |
+
},
|
| 197467 |
+
{
|
| 197468 |
+
"epoch": 621.63,
|
| 197469 |
+
"learning_rate": 8.770904684975768e-06,
|
| 197470 |
+
"loss": 0.6872,
|
| 197471 |
+
"step": 77110
|
| 197472 |
+
},
|
| 197473 |
+
{
|
| 197474 |
+
"epoch": 621.67,
|
| 197475 |
+
"learning_rate": 8.770823909531502e-06,
|
| 197476 |
+
"loss": 0.3605,
|
| 197477 |
+
"step": 77115
|
| 197478 |
+
},
|
| 197479 |
+
{
|
| 197480 |
+
"epoch": 621.71,
|
| 197481 |
+
"learning_rate": 8.770743134087238e-06,
|
| 197482 |
+
"loss": 0.2861,
|
| 197483 |
+
"step": 77120
|
| 197484 |
+
},
|
| 197485 |
+
{
|
| 197486 |
+
"epoch": 621.76,
|
| 197487 |
+
"learning_rate": 8.770662358642972e-06,
|
| 197488 |
+
"loss": 0.3662,
|
| 197489 |
+
"step": 77125
|
| 197490 |
+
},
|
| 197491 |
+
{
|
| 197492 |
+
"epoch": 621.8,
|
| 197493 |
+
"learning_rate": 8.770581583198708e-06,
|
| 197494 |
+
"loss": 0.9206,
|
| 197495 |
+
"step": 77130
|
| 197496 |
+
},
|
| 197497 |
+
{
|
| 197498 |
+
"epoch": 621.84,
|
| 197499 |
+
"learning_rate": 8.770500807754442e-06,
|
| 197500 |
+
"loss": 0.7002,
|
| 197501 |
+
"step": 77135
|
| 197502 |
+
},
|
| 197503 |
+
{
|
| 197504 |
+
"epoch": 621.88,
|
| 197505 |
+
"learning_rate": 8.770420032310178e-06,
|
| 197506 |
+
"loss": 0.2666,
|
| 197507 |
+
"step": 77140
|
| 197508 |
+
},
|
| 197509 |
+
{
|
| 197510 |
+
"epoch": 621.92,
|
| 197511 |
+
"learning_rate": 8.770339256865914e-06,
|
| 197512 |
+
"loss": 0.3497,
|
| 197513 |
+
"step": 77145
|
| 197514 |
+
},
|
| 197515 |
+
{
|
| 197516 |
+
"epoch": 621.96,
|
| 197517 |
+
"learning_rate": 8.770258481421648e-06,
|
| 197518 |
+
"loss": 0.4444,
|
| 197519 |
+
"step": 77150
|
| 197520 |
+
},
|
| 197521 |
+
{
|
| 197522 |
+
"epoch": 622.0,
|
| 197523 |
+
"learning_rate": 8.770177705977384e-06,
|
| 197524 |
+
"loss": 1.0565,
|
| 197525 |
+
"step": 77155
|
| 197526 |
+
},
|
| 197527 |
+
{
|
| 197528 |
+
"epoch": 622.0,
|
| 197529 |
+
"eval_loss": 0.3980950117111206,
|
| 197530 |
+
"eval_runtime": 43.2712,
|
| 197531 |
+
"eval_samples_per_second": 19.32,
|
| 197532 |
+
"eval_steps_per_second": 0.624,
|
| 197533 |
+
"eval_wer": 0.18589790672052883,
|
| 197534 |
+
"step": 77155
|
| 197535 |
}
|
| 197536 |
],
|
| 197537 |
+
"max_steps": 620000,
|
| 197538 |
"num_train_epochs": 5000,
|
| 197539 |
+
"total_flos": 2.1712747856892222e+20,
|
| 197540 |
"trial_name": null,
|
| 197541 |
"trial_params": null
|
| 197542 |
}
|
model-bin/finetune/base/{checkpoint-76532 → checkpoint-77155}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629889687.782439/events.out.tfevents.1629889687.7e498afd5545.7645.5
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:63bbb99ff92ce267f623f335fda5574eed7582921fe6b4153ac1943861743b37
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629890176.1350722/events.out.tfevents.1629890176.7e498afd5545.7645.7
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:281e723948064c5a320ce1cb692d9a7c969e63870e8af9b4b85223b71004ef26
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629890651.9197931/events.out.tfevents.1629890651.7e498afd5545.7645.9
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1c5f5833da50935c68156bd4995dfc510ae2cc3fe68dfa3db1d2c9cf3aa25b25
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629891122.8340254/events.out.tfevents.1629891122.7e498afd5545.7645.11
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c525ed9acaa2b75060c4ed61e1ee1be3fa4cbb57e33e0f8b0c47d9836ddeca74
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629891582.6826005/events.out.tfevents.1629891582.7e498afd5545.7645.13
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:97ce8c95c060fd1d36e41554a7a14cd8a02320f51ef42c29a503c6eece636a96
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629889687.7e498afd5545.7645.4
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:4eb4f6f7ff969594bfcf1dc7d8aa298a58053b00f0efc3ae568d57996759302f
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629890175.7e498afd5545.7645.6
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b7e30792658b366537c0877beb43a3dce9eed32f785ca3ad28abb179e5c26126
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629890651.7e498afd5545.7645.8
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:02f070a58906f8135dcd14a5fe0a28b73a656c0e20880a55e62af2805c2e7390
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629891122.7e498afd5545.7645.10
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e722c47588628dc7831664e11533e538c2eb7b7b24482e5ed9756890c8dc740d
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629891582.7e498afd5545.7645.12
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8283bf2528bb0da669cef05253a06eaa151d048dd99db67c7e254cb7aad6b5cb
|
| 3 |
+
size 8622
|