"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-115234 → checkpoint-115856}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-115234 → checkpoint-115856}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-115234 → checkpoint-115856}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-115234 → checkpoint-115856}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-115234 → checkpoint-115856}/rng_state.pth +2 -2
- model-bin/finetune/base/{checkpoint-115234 → checkpoint-115856}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-115234 → checkpoint-115856}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-115234 → checkpoint-115856}/trainer_state.json +798 -3
- model-bin/finetune/base/{checkpoint-115234 → checkpoint-115856}/training_args.bin +0 -0
- model-bin/finetune/base/log/1630111579.4549904/events.out.tfevents.1630111579.86bb0ddabf9b.1042.21 +3 -0
- model-bin/finetune/base/log/1630111993.7382865/events.out.tfevents.1630111993.86bb0ddabf9b.1042.23 +3 -0
- model-bin/finetune/base/log/1630112410.354941/events.out.tfevents.1630112410.86bb0ddabf9b.1042.25 +3 -0
- model-bin/finetune/base/log/1630112831.6927667/events.out.tfevents.1630112831.86bb0ddabf9b.1042.27 +3 -0
- model-bin/finetune/base/log/1630113242.9622881/events.out.tfevents.1630113242.86bb0ddabf9b.1042.29 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630111579.86bb0ddabf9b.1042.20 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630111993.86bb0ddabf9b.1042.22 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630112410.86bb0ddabf9b.1042.24 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630112831.86bb0ddabf9b.1042.26 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630113242.86bb0ddabf9b.1042.28 +3 -0
model-bin/finetune/base/{checkpoint-115234 → checkpoint-115856}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-115234 → checkpoint-115856}/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 sha256:3f58f1dcc9bfb4099e2ea22006b4aa19dc101120c78d4e27ed04418e2aa3131c
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-115234 → checkpoint-115856}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-115234 → checkpoint-115856}/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 sha256:5c3434b3bbed70f641bae6fda0466c25c030ac8b24f7e064b84c605b9ad0e2f2
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-115234 → checkpoint-115856}/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 sha256:75bdff2a7c9917d8d2929d23e1aa6358f86a7a2889a1c230ec6a8044b485696e
|
| 3 |
+
size 14503
|
model-bin/finetune/base/{checkpoint-115234 → checkpoint-115856}/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:afc512941b00b221a55e7420c802cd31625bebe00c214122aedc10efbd598877
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-115234 → checkpoint-115856}/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:ee060aa484196bb13ded0b787197087803b23a7d6619688fd62bf478fba823e4
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-115234 → checkpoint-115856}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1743826049391605,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-101551",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -245976,11 +245976,806 @@
|
|
| 245976 |
"eval_steps_per_second": 0.681,
|
| 245977 |
"eval_wer": 0.18934867596075827,
|
| 245978 |
"step": 115234
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 245979 |
}
|
| 245980 |
],
|
| 245981 |
"max_steps": 620000,
|
| 245982 |
"num_train_epochs": 5000,
|
| 245983 |
-
"total_flos": 3.
|
| 245984 |
"trial_name": null,
|
| 245985 |
"trial_params": null
|
| 245986 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1743826049391605,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-101551",
|
| 4 |
+
"epoch": 933.995983935743,
|
| 5 |
+
"global_step": 115856,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 245976 |
"eval_steps_per_second": 0.681,
|
| 245977 |
"eval_wer": 0.18934867596075827,
|
| 245978 |
"step": 115234
|
| 245979 |
+
},
|
| 245980 |
+
{
|
| 245981 |
+
"epoch": 929.01,
|
| 245982 |
+
"learning_rate": 8.170064102564104e-06,
|
| 245983 |
+
"loss": 0.3867,
|
| 245984 |
+
"step": 115235
|
| 245985 |
+
},
|
| 245986 |
+
{
|
| 245987 |
+
"epoch": 929.05,
|
| 245988 |
+
"learning_rate": 8.169983974358975e-06,
|
| 245989 |
+
"loss": 0.2917,
|
| 245990 |
+
"step": 115240
|
| 245991 |
+
},
|
| 245992 |
+
{
|
| 245993 |
+
"epoch": 929.09,
|
| 245994 |
+
"learning_rate": 8.169903846153847e-06,
|
| 245995 |
+
"loss": 0.3074,
|
| 245996 |
+
"step": 115245
|
| 245997 |
+
},
|
| 245998 |
+
{
|
| 245999 |
+
"epoch": 929.13,
|
| 246000 |
+
"learning_rate": 8.169823717948718e-06,
|
| 246001 |
+
"loss": 0.3278,
|
| 246002 |
+
"step": 115250
|
| 246003 |
+
},
|
| 246004 |
+
{
|
| 246005 |
+
"epoch": 929.17,
|
| 246006 |
+
"learning_rate": 8.169743589743591e-06,
|
| 246007 |
+
"loss": 0.5108,
|
| 246008 |
+
"step": 115255
|
| 246009 |
+
},
|
| 246010 |
+
{
|
| 246011 |
+
"epoch": 929.21,
|
| 246012 |
+
"learning_rate": 8.169663461538462e-06,
|
| 246013 |
+
"loss": 1.2778,
|
| 246014 |
+
"step": 115260
|
| 246015 |
+
},
|
| 246016 |
+
{
|
| 246017 |
+
"epoch": 929.25,
|
| 246018 |
+
"learning_rate": 8.169583333333334e-06,
|
| 246019 |
+
"loss": 0.3347,
|
| 246020 |
+
"step": 115265
|
| 246021 |
+
},
|
| 246022 |
+
{
|
| 246023 |
+
"epoch": 929.29,
|
| 246024 |
+
"learning_rate": 8.169503205128205e-06,
|
| 246025 |
+
"loss": 0.2539,
|
| 246026 |
+
"step": 115270
|
| 246027 |
+
},
|
| 246028 |
+
{
|
| 246029 |
+
"epoch": 929.33,
|
| 246030 |
+
"learning_rate": 8.169423076923078e-06,
|
| 246031 |
+
"loss": 0.2978,
|
| 246032 |
+
"step": 115275
|
| 246033 |
+
},
|
| 246034 |
+
{
|
| 246035 |
+
"epoch": 929.37,
|
| 246036 |
+
"learning_rate": 8.16934294871795e-06,
|
| 246037 |
+
"loss": 0.5235,
|
| 246038 |
+
"step": 115280
|
| 246039 |
+
},
|
| 246040 |
+
{
|
| 246041 |
+
"epoch": 929.41,
|
| 246042 |
+
"learning_rate": 8.169262820512821e-06,
|
| 246043 |
+
"loss": 1.1577,
|
| 246044 |
+
"step": 115285
|
| 246045 |
+
},
|
| 246046 |
+
{
|
| 246047 |
+
"epoch": 929.45,
|
| 246048 |
+
"learning_rate": 8.169182692307694e-06,
|
| 246049 |
+
"loss": 0.341,
|
| 246050 |
+
"step": 115290
|
| 246051 |
+
},
|
| 246052 |
+
{
|
| 246053 |
+
"epoch": 929.49,
|
| 246054 |
+
"learning_rate": 8.169102564102565e-06,
|
| 246055 |
+
"loss": 0.3147,
|
| 246056 |
+
"step": 115295
|
| 246057 |
+
},
|
| 246058 |
+
{
|
| 246059 |
+
"epoch": 929.53,
|
| 246060 |
+
"learning_rate": 8.169022435897437e-06,
|
| 246061 |
+
"loss": 0.3392,
|
| 246062 |
+
"step": 115300
|
| 246063 |
+
},
|
| 246064 |
+
{
|
| 246065 |
+
"epoch": 929.57,
|
| 246066 |
+
"learning_rate": 8.168942307692308e-06,
|
| 246067 |
+
"loss": 0.5286,
|
| 246068 |
+
"step": 115305
|
| 246069 |
+
},
|
| 246070 |
+
{
|
| 246071 |
+
"epoch": 929.61,
|
| 246072 |
+
"learning_rate": 8.168862179487181e-06,
|
| 246073 |
+
"loss": 0.9704,
|
| 246074 |
+
"step": 115310
|
| 246075 |
+
},
|
| 246076 |
+
{
|
| 246077 |
+
"epoch": 929.65,
|
| 246078 |
+
"learning_rate": 8.16878205128205e-06,
|
| 246079 |
+
"loss": 0.2936,
|
| 246080 |
+
"step": 115315
|
| 246081 |
+
},
|
| 246082 |
+
{
|
| 246083 |
+
"epoch": 929.69,
|
| 246084 |
+
"learning_rate": 8.168701923076924e-06,
|
| 246085 |
+
"loss": 0.2851,
|
| 246086 |
+
"step": 115320
|
| 246087 |
+
},
|
| 246088 |
+
{
|
| 246089 |
+
"epoch": 929.73,
|
| 246090 |
+
"learning_rate": 8.168621794871797e-06,
|
| 246091 |
+
"loss": 0.4192,
|
| 246092 |
+
"step": 115325
|
| 246093 |
+
},
|
| 246094 |
+
{
|
| 246095 |
+
"epoch": 929.77,
|
| 246096 |
+
"learning_rate": 8.168541666666667e-06,
|
| 246097 |
+
"loss": 0.5268,
|
| 246098 |
+
"step": 115330
|
| 246099 |
+
},
|
| 246100 |
+
{
|
| 246101 |
+
"epoch": 929.81,
|
| 246102 |
+
"learning_rate": 8.16846153846154e-06,
|
| 246103 |
+
"loss": 1.1848,
|
| 246104 |
+
"step": 115335
|
| 246105 |
+
},
|
| 246106 |
+
{
|
| 246107 |
+
"epoch": 929.85,
|
| 246108 |
+
"learning_rate": 8.168381410256411e-06,
|
| 246109 |
+
"loss": 0.2957,
|
| 246110 |
+
"step": 115340
|
| 246111 |
+
},
|
| 246112 |
+
{
|
| 246113 |
+
"epoch": 929.9,
|
| 246114 |
+
"learning_rate": 8.168301282051282e-06,
|
| 246115 |
+
"loss": 0.2697,
|
| 246116 |
+
"step": 115345
|
| 246117 |
+
},
|
| 246118 |
+
{
|
| 246119 |
+
"epoch": 929.94,
|
| 246120 |
+
"learning_rate": 8.168221153846154e-06,
|
| 246121 |
+
"loss": 0.3215,
|
| 246122 |
+
"step": 115350
|
| 246123 |
+
},
|
| 246124 |
+
{
|
| 246125 |
+
"epoch": 929.98,
|
| 246126 |
+
"learning_rate": 8.168141025641027e-06,
|
| 246127 |
+
"loss": 0.5917,
|
| 246128 |
+
"step": 115355
|
| 246129 |
+
},
|
| 246130 |
+
{
|
| 246131 |
+
"epoch": 930.0,
|
| 246132 |
+
"eval_loss": 0.43522909283638,
|
| 246133 |
+
"eval_runtime": 39.3482,
|
| 246134 |
+
"eval_samples_per_second": 21.424,
|
| 246135 |
+
"eval_steps_per_second": 0.686,
|
| 246136 |
+
"eval_wer": 0.18558214652767582,
|
| 246137 |
+
"step": 115358
|
| 246138 |
+
},
|
| 246139 |
+
{
|
| 246140 |
+
"epoch": 930.02,
|
| 246141 |
+
"learning_rate": 8.168060897435898e-06,
|
| 246142 |
+
"loss": 0.2904,
|
| 246143 |
+
"step": 115360
|
| 246144 |
+
},
|
| 246145 |
+
{
|
| 246146 |
+
"epoch": 930.06,
|
| 246147 |
+
"learning_rate": 8.16798076923077e-06,
|
| 246148 |
+
"loss": 0.2959,
|
| 246149 |
+
"step": 115365
|
| 246150 |
+
},
|
| 246151 |
+
{
|
| 246152 |
+
"epoch": 930.1,
|
| 246153 |
+
"learning_rate": 8.16790064102564e-06,
|
| 246154 |
+
"loss": 0.2997,
|
| 246155 |
+
"step": 115370
|
| 246156 |
+
},
|
| 246157 |
+
{
|
| 246158 |
+
"epoch": 930.14,
|
| 246159 |
+
"learning_rate": 8.167820512820514e-06,
|
| 246160 |
+
"loss": 0.3767,
|
| 246161 |
+
"step": 115375
|
| 246162 |
+
},
|
| 246163 |
+
{
|
| 246164 |
+
"epoch": 930.18,
|
| 246165 |
+
"learning_rate": 8.167740384615385e-06,
|
| 246166 |
+
"loss": 0.5481,
|
| 246167 |
+
"step": 115380
|
| 246168 |
+
},
|
| 246169 |
+
{
|
| 246170 |
+
"epoch": 930.22,
|
| 246171 |
+
"learning_rate": 8.167660256410257e-06,
|
| 246172 |
+
"loss": 0.9912,
|
| 246173 |
+
"step": 115385
|
| 246174 |
+
},
|
| 246175 |
+
{
|
| 246176 |
+
"epoch": 930.26,
|
| 246177 |
+
"learning_rate": 8.16758012820513e-06,
|
| 246178 |
+
"loss": 0.2959,
|
| 246179 |
+
"step": 115390
|
| 246180 |
+
},
|
| 246181 |
+
{
|
| 246182 |
+
"epoch": 930.3,
|
| 246183 |
+
"learning_rate": 8.167500000000001e-06,
|
| 246184 |
+
"loss": 0.3061,
|
| 246185 |
+
"step": 115395
|
| 246186 |
+
},
|
| 246187 |
+
{
|
| 246188 |
+
"epoch": 930.34,
|
| 246189 |
+
"learning_rate": 8.167419871794872e-06,
|
| 246190 |
+
"loss": 0.3396,
|
| 246191 |
+
"step": 115400
|
| 246192 |
+
},
|
| 246193 |
+
{
|
| 246194 |
+
"epoch": 930.38,
|
| 246195 |
+
"learning_rate": 8.167339743589744e-06,
|
| 246196 |
+
"loss": 0.5358,
|
| 246197 |
+
"step": 115405
|
| 246198 |
+
},
|
| 246199 |
+
{
|
| 246200 |
+
"epoch": 930.42,
|
| 246201 |
+
"learning_rate": 8.167259615384617e-06,
|
| 246202 |
+
"loss": 0.9547,
|
| 246203 |
+
"step": 115410
|
| 246204 |
+
},
|
| 246205 |
+
{
|
| 246206 |
+
"epoch": 930.46,
|
| 246207 |
+
"learning_rate": 8.167179487179488e-06,
|
| 246208 |
+
"loss": 0.2707,
|
| 246209 |
+
"step": 115415
|
| 246210 |
+
},
|
| 246211 |
+
{
|
| 246212 |
+
"epoch": 930.5,
|
| 246213 |
+
"learning_rate": 8.16709935897436e-06,
|
| 246214 |
+
"loss": 0.3489,
|
| 246215 |
+
"step": 115420
|
| 246216 |
+
},
|
| 246217 |
+
{
|
| 246218 |
+
"epoch": 930.54,
|
| 246219 |
+
"learning_rate": 8.167019230769233e-06,
|
| 246220 |
+
"loss": 0.3653,
|
| 246221 |
+
"step": 115425
|
| 246222 |
+
},
|
| 246223 |
+
{
|
| 246224 |
+
"epoch": 930.58,
|
| 246225 |
+
"learning_rate": 8.166939102564104e-06,
|
| 246226 |
+
"loss": 0.629,
|
| 246227 |
+
"step": 115430
|
| 246228 |
+
},
|
| 246229 |
+
{
|
| 246230 |
+
"epoch": 930.62,
|
| 246231 |
+
"learning_rate": 8.166858974358975e-06,
|
| 246232 |
+
"loss": 1.0132,
|
| 246233 |
+
"step": 115435
|
| 246234 |
+
},
|
| 246235 |
+
{
|
| 246236 |
+
"epoch": 930.66,
|
| 246237 |
+
"learning_rate": 8.166778846153847e-06,
|
| 246238 |
+
"loss": 0.4032,
|
| 246239 |
+
"step": 115440
|
| 246240 |
+
},
|
| 246241 |
+
{
|
| 246242 |
+
"epoch": 930.7,
|
| 246243 |
+
"learning_rate": 8.16669871794872e-06,
|
| 246244 |
+
"loss": 0.2593,
|
| 246245 |
+
"step": 115445
|
| 246246 |
+
},
|
| 246247 |
+
{
|
| 246248 |
+
"epoch": 930.74,
|
| 246249 |
+
"learning_rate": 8.16661858974359e-06,
|
| 246250 |
+
"loss": 0.3794,
|
| 246251 |
+
"step": 115450
|
| 246252 |
+
},
|
| 246253 |
+
{
|
| 246254 |
+
"epoch": 930.78,
|
| 246255 |
+
"learning_rate": 8.166538461538462e-06,
|
| 246256 |
+
"loss": 0.6058,
|
| 246257 |
+
"step": 115455
|
| 246258 |
+
},
|
| 246259 |
+
{
|
| 246260 |
+
"epoch": 930.82,
|
| 246261 |
+
"learning_rate": 8.166458333333334e-06,
|
| 246262 |
+
"loss": 1.097,
|
| 246263 |
+
"step": 115460
|
| 246264 |
+
},
|
| 246265 |
+
{
|
| 246266 |
+
"epoch": 930.86,
|
| 246267 |
+
"learning_rate": 8.166378205128205e-06,
|
| 246268 |
+
"loss": 0.2487,
|
| 246269 |
+
"step": 115465
|
| 246270 |
+
},
|
| 246271 |
+
{
|
| 246272 |
+
"epoch": 930.9,
|
| 246273 |
+
"learning_rate": 8.166298076923076e-06,
|
| 246274 |
+
"loss": 0.2866,
|
| 246275 |
+
"step": 115470
|
| 246276 |
+
},
|
| 246277 |
+
{
|
| 246278 |
+
"epoch": 930.94,
|
| 246279 |
+
"learning_rate": 8.16621794871795e-06,
|
| 246280 |
+
"loss": 0.3734,
|
| 246281 |
+
"step": 115475
|
| 246282 |
+
},
|
| 246283 |
+
{
|
| 246284 |
+
"epoch": 930.98,
|
| 246285 |
+
"learning_rate": 8.166137820512821e-06,
|
| 246286 |
+
"loss": 0.686,
|
| 246287 |
+
"step": 115480
|
| 246288 |
+
},
|
| 246289 |
+
{
|
| 246290 |
+
"epoch": 931.0,
|
| 246291 |
+
"eval_loss": 0.41655367612838745,
|
| 246292 |
+
"eval_runtime": 39.4087,
|
| 246293 |
+
"eval_samples_per_second": 21.391,
|
| 246294 |
+
"eval_steps_per_second": 0.685,
|
| 246295 |
+
"eval_wer": 0.19136343708899606,
|
| 246296 |
+
"step": 115482
|
| 246297 |
+
},
|
| 246298 |
+
{
|
| 246299 |
+
"epoch": 923.02,
|
| 246300 |
+
"learning_rate": 8.166057692307692e-06,
|
| 246301 |
+
"loss": 0.3457,
|
| 246302 |
+
"step": 115485
|
| 246303 |
+
},
|
| 246304 |
+
{
|
| 246305 |
+
"epoch": 923.06,
|
| 246306 |
+
"learning_rate": 8.165977564102565e-06,
|
| 246307 |
+
"loss": 0.2669,
|
| 246308 |
+
"step": 115490
|
| 246309 |
+
},
|
| 246310 |
+
{
|
| 246311 |
+
"epoch": 923.1,
|
| 246312 |
+
"learning_rate": 8.165897435897437e-06,
|
| 246313 |
+
"loss": 0.283,
|
| 246314 |
+
"step": 115495
|
| 246315 |
+
},
|
| 246316 |
+
{
|
| 246317 |
+
"epoch": 923.14,
|
| 246318 |
+
"learning_rate": 8.165817307692308e-06,
|
| 246319 |
+
"loss": 0.3833,
|
| 246320 |
+
"step": 115500
|
| 246321 |
+
},
|
| 246322 |
+
{
|
| 246323 |
+
"epoch": 923.18,
|
| 246324 |
+
"learning_rate": 8.16573717948718e-06,
|
| 246325 |
+
"loss": 0.7343,
|
| 246326 |
+
"step": 115505
|
| 246327 |
+
},
|
| 246328 |
+
{
|
| 246329 |
+
"epoch": 923.22,
|
| 246330 |
+
"learning_rate": 8.165657051282052e-06,
|
| 246331 |
+
"loss": 0.9039,
|
| 246332 |
+
"step": 115510
|
| 246333 |
+
},
|
| 246334 |
+
{
|
| 246335 |
+
"epoch": 923.26,
|
| 246336 |
+
"learning_rate": 8.165576923076924e-06,
|
| 246337 |
+
"loss": 0.2882,
|
| 246338 |
+
"step": 115515
|
| 246339 |
+
},
|
| 246340 |
+
{
|
| 246341 |
+
"epoch": 923.3,
|
| 246342 |
+
"learning_rate": 8.165496794871795e-06,
|
| 246343 |
+
"loss": 0.3064,
|
| 246344 |
+
"step": 115520
|
| 246345 |
+
},
|
| 246346 |
+
{
|
| 246347 |
+
"epoch": 923.34,
|
| 246348 |
+
"learning_rate": 8.165416666666668e-06,
|
| 246349 |
+
"loss": 0.3276,
|
| 246350 |
+
"step": 115525
|
| 246351 |
+
},
|
| 246352 |
+
{
|
| 246353 |
+
"epoch": 923.38,
|
| 246354 |
+
"learning_rate": 8.16533653846154e-06,
|
| 246355 |
+
"loss": 0.7486,
|
| 246356 |
+
"step": 115530
|
| 246357 |
+
},
|
| 246358 |
+
{
|
| 246359 |
+
"epoch": 923.42,
|
| 246360 |
+
"learning_rate": 8.165256410256411e-06,
|
| 246361 |
+
"loss": 0.9101,
|
| 246362 |
+
"step": 115535
|
| 246363 |
+
},
|
| 246364 |
+
{
|
| 246365 |
+
"epoch": 923.46,
|
| 246366 |
+
"learning_rate": 8.165176282051282e-06,
|
| 246367 |
+
"loss": 0.2789,
|
| 246368 |
+
"step": 115540
|
| 246369 |
+
},
|
| 246370 |
+
{
|
| 246371 |
+
"epoch": 923.5,
|
| 246372 |
+
"learning_rate": 8.165096153846155e-06,
|
| 246373 |
+
"loss": 0.2705,
|
| 246374 |
+
"step": 115545
|
| 246375 |
+
},
|
| 246376 |
+
{
|
| 246377 |
+
"epoch": 923.54,
|
| 246378 |
+
"learning_rate": 8.165016025641027e-06,
|
| 246379 |
+
"loss": 0.4525,
|
| 246380 |
+
"step": 115550
|
| 246381 |
+
},
|
| 246382 |
+
{
|
| 246383 |
+
"epoch": 923.58,
|
| 246384 |
+
"learning_rate": 8.164935897435898e-06,
|
| 246385 |
+
"loss": 0.7603,
|
| 246386 |
+
"step": 115555
|
| 246387 |
+
},
|
| 246388 |
+
{
|
| 246389 |
+
"epoch": 923.62,
|
| 246390 |
+
"learning_rate": 8.16485576923077e-06,
|
| 246391 |
+
"loss": 0.9037,
|
| 246392 |
+
"step": 115560
|
| 246393 |
+
},
|
| 246394 |
+
{
|
| 246395 |
+
"epoch": 923.66,
|
| 246396 |
+
"learning_rate": 8.164775641025642e-06,
|
| 246397 |
+
"loss": 0.2632,
|
| 246398 |
+
"step": 115565
|
| 246399 |
+
},
|
| 246400 |
+
{
|
| 246401 |
+
"epoch": 923.7,
|
| 246402 |
+
"learning_rate": 8.164695512820514e-06,
|
| 246403 |
+
"loss": 0.3309,
|
| 246404 |
+
"step": 115570
|
| 246405 |
+
},
|
| 246406 |
+
{
|
| 246407 |
+
"epoch": 923.74,
|
| 246408 |
+
"learning_rate": 8.164615384615385e-06,
|
| 246409 |
+
"loss": 0.3608,
|
| 246410 |
+
"step": 115575
|
| 246411 |
+
},
|
| 246412 |
+
{
|
| 246413 |
+
"epoch": 923.78,
|
| 246414 |
+
"learning_rate": 8.164535256410258e-06,
|
| 246415 |
+
"loss": 0.7437,
|
| 246416 |
+
"step": 115580
|
| 246417 |
+
},
|
| 246418 |
+
{
|
| 246419 |
+
"epoch": 923.82,
|
| 246420 |
+
"learning_rate": 8.16445512820513e-06,
|
| 246421 |
+
"loss": 1.0031,
|
| 246422 |
+
"step": 115585
|
| 246423 |
+
},
|
| 246424 |
+
{
|
| 246425 |
+
"epoch": 923.86,
|
| 246426 |
+
"learning_rate": 8.164375000000001e-06,
|
| 246427 |
+
"loss": 0.2573,
|
| 246428 |
+
"step": 115590
|
| 246429 |
+
},
|
| 246430 |
+
{
|
| 246431 |
+
"epoch": 923.9,
|
| 246432 |
+
"learning_rate": 8.164294871794872e-06,
|
| 246433 |
+
"loss": 0.3339,
|
| 246434 |
+
"step": 115595
|
| 246435 |
+
},
|
| 246436 |
+
{
|
| 246437 |
+
"epoch": 923.94,
|
| 246438 |
+
"learning_rate": 8.164214743589745e-06,
|
| 246439 |
+
"loss": 0.3549,
|
| 246440 |
+
"step": 115600
|
| 246441 |
+
},
|
| 246442 |
+
{
|
| 246443 |
+
"epoch": 923.98,
|
| 246444 |
+
"learning_rate": 8.164134615384615e-06,
|
| 246445 |
+
"loss": 0.7648,
|
| 246446 |
+
"step": 115605
|
| 246447 |
+
},
|
| 246448 |
+
{
|
| 246449 |
+
"epoch": 924.0,
|
| 246450 |
+
"eval_loss": 0.3562644422054291,
|
| 246451 |
+
"eval_runtime": 39.7404,
|
| 246452 |
+
"eval_samples_per_second": 21.213,
|
| 246453 |
+
"eval_steps_per_second": 0.679,
|
| 246454 |
+
"eval_wer": 0.18705295860125715,
|
| 246455 |
+
"step": 115607
|
| 246456 |
+
},
|
| 246457 |
+
{
|
| 246458 |
+
"epoch": 924.02,
|
| 246459 |
+
"learning_rate": 8.164054487179488e-06,
|
| 246460 |
+
"loss": 0.3158,
|
| 246461 |
+
"step": 115610
|
| 246462 |
+
},
|
| 246463 |
+
{
|
| 246464 |
+
"epoch": 924.06,
|
| 246465 |
+
"learning_rate": 8.16397435897436e-06,
|
| 246466 |
+
"loss": 0.3048,
|
| 246467 |
+
"step": 115615
|
| 246468 |
+
},
|
| 246469 |
+
{
|
| 246470 |
+
"epoch": 924.1,
|
| 246471 |
+
"learning_rate": 8.16389423076923e-06,
|
| 246472 |
+
"loss": 0.3158,
|
| 246473 |
+
"step": 115620
|
| 246474 |
+
},
|
| 246475 |
+
{
|
| 246476 |
+
"epoch": 924.14,
|
| 246477 |
+
"learning_rate": 8.163814102564104e-06,
|
| 246478 |
+
"loss": 0.3699,
|
| 246479 |
+
"step": 115625
|
| 246480 |
+
},
|
| 246481 |
+
{
|
| 246482 |
+
"epoch": 924.18,
|
| 246483 |
+
"learning_rate": 8.163733974358975e-06,
|
| 246484 |
+
"loss": 0.8154,
|
| 246485 |
+
"step": 115630
|
| 246486 |
+
},
|
| 246487 |
+
{
|
| 246488 |
+
"epoch": 924.22,
|
| 246489 |
+
"learning_rate": 8.163653846153847e-06,
|
| 246490 |
+
"loss": 0.8522,
|
| 246491 |
+
"step": 115635
|
| 246492 |
+
},
|
| 246493 |
+
{
|
| 246494 |
+
"epoch": 924.26,
|
| 246495 |
+
"learning_rate": 8.163573717948718e-06,
|
| 246496 |
+
"loss": 0.3013,
|
| 246497 |
+
"step": 115640
|
| 246498 |
+
},
|
| 246499 |
+
{
|
| 246500 |
+
"epoch": 924.3,
|
| 246501 |
+
"learning_rate": 8.163493589743591e-06,
|
| 246502 |
+
"loss": 0.2949,
|
| 246503 |
+
"step": 115645
|
| 246504 |
+
},
|
| 246505 |
+
{
|
| 246506 |
+
"epoch": 924.34,
|
| 246507 |
+
"learning_rate": 8.163413461538462e-06,
|
| 246508 |
+
"loss": 0.3306,
|
| 246509 |
+
"step": 115650
|
| 246510 |
+
},
|
| 246511 |
+
{
|
| 246512 |
+
"epoch": 924.38,
|
| 246513 |
+
"learning_rate": 8.163333333333334e-06,
|
| 246514 |
+
"loss": 0.6359,
|
| 246515 |
+
"step": 115655
|
| 246516 |
+
},
|
| 246517 |
+
{
|
| 246518 |
+
"epoch": 924.42,
|
| 246519 |
+
"learning_rate": 8.163253205128205e-06,
|
| 246520 |
+
"loss": 0.8363,
|
| 246521 |
+
"step": 115660
|
| 246522 |
+
},
|
| 246523 |
+
{
|
| 246524 |
+
"epoch": 924.46,
|
| 246525 |
+
"learning_rate": 8.163173076923078e-06,
|
| 246526 |
+
"loss": 0.2683,
|
| 246527 |
+
"step": 115665
|
| 246528 |
+
},
|
| 246529 |
+
{
|
| 246530 |
+
"epoch": 924.5,
|
| 246531 |
+
"learning_rate": 8.16309294871795e-06,
|
| 246532 |
+
"loss": 0.3217,
|
| 246533 |
+
"step": 115670
|
| 246534 |
+
},
|
| 246535 |
+
{
|
| 246536 |
+
"epoch": 924.54,
|
| 246537 |
+
"learning_rate": 8.16301282051282e-06,
|
| 246538 |
+
"loss": 0.3776,
|
| 246539 |
+
"step": 115675
|
| 246540 |
+
},
|
| 246541 |
+
{
|
| 246542 |
+
"epoch": 924.58,
|
| 246543 |
+
"learning_rate": 8.162932692307694e-06,
|
| 246544 |
+
"loss": 0.7414,
|
| 246545 |
+
"step": 115680
|
| 246546 |
+
},
|
| 246547 |
+
{
|
| 246548 |
+
"epoch": 924.62,
|
| 246549 |
+
"learning_rate": 8.162852564102565e-06,
|
| 246550 |
+
"loss": 0.8158,
|
| 246551 |
+
"step": 115685
|
| 246552 |
+
},
|
| 246553 |
+
{
|
| 246554 |
+
"epoch": 924.66,
|
| 246555 |
+
"learning_rate": 8.162772435897437e-06,
|
| 246556 |
+
"loss": 0.2372,
|
| 246557 |
+
"step": 115690
|
| 246558 |
+
},
|
| 246559 |
+
{
|
| 246560 |
+
"epoch": 924.7,
|
| 246561 |
+
"learning_rate": 8.162692307692308e-06,
|
| 246562 |
+
"loss": 0.2578,
|
| 246563 |
+
"step": 115695
|
| 246564 |
+
},
|
| 246565 |
+
{
|
| 246566 |
+
"epoch": 924.74,
|
| 246567 |
+
"learning_rate": 8.162612179487181e-06,
|
| 246568 |
+
"loss": 0.3898,
|
| 246569 |
+
"step": 115700
|
| 246570 |
+
},
|
| 246571 |
+
{
|
| 246572 |
+
"epoch": 924.78,
|
| 246573 |
+
"learning_rate": 8.162532051282052e-06,
|
| 246574 |
+
"loss": 0.7194,
|
| 246575 |
+
"step": 115705
|
| 246576 |
+
},
|
| 246577 |
+
{
|
| 246578 |
+
"epoch": 924.82,
|
| 246579 |
+
"learning_rate": 8.162451923076924e-06,
|
| 246580 |
+
"loss": 0.8795,
|
| 246581 |
+
"step": 115710
|
| 246582 |
+
},
|
| 246583 |
+
{
|
| 246584 |
+
"epoch": 924.86,
|
| 246585 |
+
"learning_rate": 8.162371794871795e-06,
|
| 246586 |
+
"loss": 0.298,
|
| 246587 |
+
"step": 115715
|
| 246588 |
+
},
|
| 246589 |
+
{
|
| 246590 |
+
"epoch": 924.9,
|
| 246591 |
+
"learning_rate": 8.162291666666668e-06,
|
| 246592 |
+
"loss": 0.2862,
|
| 246593 |
+
"step": 115720
|
| 246594 |
+
},
|
| 246595 |
+
{
|
| 246596 |
+
"epoch": 924.94,
|
| 246597 |
+
"learning_rate": 8.162211538461538e-06,
|
| 246598 |
+
"loss": 0.3337,
|
| 246599 |
+
"step": 115725
|
| 246600 |
+
},
|
| 246601 |
+
{
|
| 246602 |
+
"epoch": 924.98,
|
| 246603 |
+
"learning_rate": 8.16213141025641e-06,
|
| 246604 |
+
"loss": 0.8685,
|
| 246605 |
+
"step": 115730
|
| 246606 |
+
},
|
| 246607 |
+
{
|
| 246608 |
+
"epoch": 925.0,
|
| 246609 |
+
"eval_loss": 0.3646712303161621,
|
| 246610 |
+
"eval_runtime": 40.2916,
|
| 246611 |
+
"eval_samples_per_second": 20.922,
|
| 246612 |
+
"eval_steps_per_second": 0.67,
|
| 246613 |
+
"eval_wer": 0.18595250126326426,
|
| 246614 |
+
"step": 115732
|
| 246615 |
+
},
|
| 246616 |
+
{
|
| 246617 |
+
"epoch": 933.02,
|
| 246618 |
+
"learning_rate": 8.162051282051284e-06,
|
| 246619 |
+
"loss": 0.3357,
|
| 246620 |
+
"step": 115735
|
| 246621 |
+
},
|
| 246622 |
+
{
|
| 246623 |
+
"epoch": 933.06,
|
| 246624 |
+
"learning_rate": 8.161971153846154e-06,
|
| 246625 |
+
"loss": 0.2884,
|
| 246626 |
+
"step": 115740
|
| 246627 |
+
},
|
| 246628 |
+
{
|
| 246629 |
+
"epoch": 933.1,
|
| 246630 |
+
"learning_rate": 8.161891025641027e-06,
|
| 246631 |
+
"loss": 0.329,
|
| 246632 |
+
"step": 115745
|
| 246633 |
+
},
|
| 246634 |
+
{
|
| 246635 |
+
"epoch": 933.14,
|
| 246636 |
+
"learning_rate": 8.161810897435898e-06,
|
| 246637 |
+
"loss": 0.3759,
|
| 246638 |
+
"step": 115750
|
| 246639 |
+
},
|
| 246640 |
+
{
|
| 246641 |
+
"epoch": 933.18,
|
| 246642 |
+
"learning_rate": 8.16173076923077e-06,
|
| 246643 |
+
"loss": 0.7306,
|
| 246644 |
+
"step": 115755
|
| 246645 |
+
},
|
| 246646 |
+
{
|
| 246647 |
+
"epoch": 933.22,
|
| 246648 |
+
"learning_rate": 8.16165064102564e-06,
|
| 246649 |
+
"loss": 0.8451,
|
| 246650 |
+
"step": 115760
|
| 246651 |
+
},
|
| 246652 |
+
{
|
| 246653 |
+
"epoch": 933.27,
|
| 246654 |
+
"learning_rate": 8.161570512820514e-06,
|
| 246655 |
+
"loss": 0.2821,
|
| 246656 |
+
"step": 115765
|
| 246657 |
+
},
|
| 246658 |
+
{
|
| 246659 |
+
"epoch": 933.31,
|
| 246660 |
+
"learning_rate": 8.161490384615385e-06,
|
| 246661 |
+
"loss": 0.2711,
|
| 246662 |
+
"step": 115770
|
| 246663 |
+
},
|
| 246664 |
+
{
|
| 246665 |
+
"epoch": 933.35,
|
| 246666 |
+
"learning_rate": 8.161410256410256e-06,
|
| 246667 |
+
"loss": 0.3425,
|
| 246668 |
+
"step": 115775
|
| 246669 |
+
},
|
| 246670 |
+
{
|
| 246671 |
+
"epoch": 933.39,
|
| 246672 |
+
"learning_rate": 8.16133012820513e-06,
|
| 246673 |
+
"loss": 0.8055,
|
| 246674 |
+
"step": 115780
|
| 246675 |
+
},
|
| 246676 |
+
{
|
| 246677 |
+
"epoch": 933.43,
|
| 246678 |
+
"learning_rate": 8.16125e-06,
|
| 246679 |
+
"loss": 0.9612,
|
| 246680 |
+
"step": 115785
|
| 246681 |
+
},
|
| 246682 |
+
{
|
| 246683 |
+
"epoch": 933.47,
|
| 246684 |
+
"learning_rate": 8.161169871794872e-06,
|
| 246685 |
+
"loss": 0.2492,
|
| 246686 |
+
"step": 115790
|
| 246687 |
+
},
|
| 246688 |
+
{
|
| 246689 |
+
"epoch": 933.51,
|
| 246690 |
+
"learning_rate": 8.161089743589744e-06,
|
| 246691 |
+
"loss": 0.3836,
|
| 246692 |
+
"step": 115795
|
| 246693 |
+
},
|
| 246694 |
+
{
|
| 246695 |
+
"epoch": 933.55,
|
| 246696 |
+
"learning_rate": 8.161009615384617e-06,
|
| 246697 |
+
"loss": 0.4158,
|
| 246698 |
+
"step": 115800
|
| 246699 |
+
},
|
| 246700 |
+
{
|
| 246701 |
+
"epoch": 933.59,
|
| 246702 |
+
"learning_rate": 8.160929487179488e-06,
|
| 246703 |
+
"loss": 0.764,
|
| 246704 |
+
"step": 115805
|
| 246705 |
+
},
|
| 246706 |
+
{
|
| 246707 |
+
"epoch": 933.63,
|
| 246708 |
+
"learning_rate": 8.16084935897436e-06,
|
| 246709 |
+
"loss": 0.9013,
|
| 246710 |
+
"step": 115810
|
| 246711 |
+
},
|
| 246712 |
+
{
|
| 246713 |
+
"epoch": 933.67,
|
| 246714 |
+
"learning_rate": 8.16076923076923e-06,
|
| 246715 |
+
"loss": 0.2851,
|
| 246716 |
+
"step": 115815
|
| 246717 |
+
},
|
| 246718 |
+
{
|
| 246719 |
+
"epoch": 933.71,
|
| 246720 |
+
"learning_rate": 8.160689102564104e-06,
|
| 246721 |
+
"loss": 0.301,
|
| 246722 |
+
"step": 115820
|
| 246723 |
+
},
|
| 246724 |
+
{
|
| 246725 |
+
"epoch": 933.75,
|
| 246726 |
+
"learning_rate": 8.160608974358975e-06,
|
| 246727 |
+
"loss": 0.3565,
|
| 246728 |
+
"step": 115825
|
| 246729 |
+
},
|
| 246730 |
+
{
|
| 246731 |
+
"epoch": 933.79,
|
| 246732 |
+
"learning_rate": 8.160528846153846e-06,
|
| 246733 |
+
"loss": 0.7047,
|
| 246734 |
+
"step": 115830
|
| 246735 |
+
},
|
| 246736 |
+
{
|
| 246737 |
+
"epoch": 933.83,
|
| 246738 |
+
"learning_rate": 8.16044871794872e-06,
|
| 246739 |
+
"loss": 0.8573,
|
| 246740 |
+
"step": 115835
|
| 246741 |
+
},
|
| 246742 |
+
{
|
| 246743 |
+
"epoch": 933.87,
|
| 246744 |
+
"learning_rate": 8.160368589743591e-06,
|
| 246745 |
+
"loss": 0.2863,
|
| 246746 |
+
"step": 115840
|
| 246747 |
+
},
|
| 246748 |
+
{
|
| 246749 |
+
"epoch": 933.91,
|
| 246750 |
+
"learning_rate": 8.160288461538462e-06,
|
| 246751 |
+
"loss": 0.3732,
|
| 246752 |
+
"step": 115845
|
| 246753 |
+
},
|
| 246754 |
+
{
|
| 246755 |
+
"epoch": 933.95,
|
| 246756 |
+
"learning_rate": 8.160208333333334e-06,
|
| 246757 |
+
"loss": 0.3254,
|
| 246758 |
+
"step": 115850
|
| 246759 |
+
},
|
| 246760 |
+
{
|
| 246761 |
+
"epoch": 933.99,
|
| 246762 |
+
"learning_rate": 8.160128205128207e-06,
|
| 246763 |
+
"loss": 0.9797,
|
| 246764 |
+
"step": 115855
|
| 246765 |
+
},
|
| 246766 |
+
{
|
| 246767 |
+
"epoch": 934.0,
|
| 246768 |
+
"eval_loss": 0.3789617717266083,
|
| 246769 |
+
"eval_runtime": 38.7963,
|
| 246770 |
+
"eval_samples_per_second": 21.729,
|
| 246771 |
+
"eval_steps_per_second": 0.696,
|
| 246772 |
+
"eval_wer": 0.18200635765506026,
|
| 246773 |
+
"step": 115856
|
| 246774 |
}
|
| 246775 |
],
|
| 246776 |
"max_steps": 620000,
|
| 246777 |
"num_train_epochs": 5000,
|
| 246778 |
+
"total_flos": 3.260533347890663e+20,
|
| 246779 |
"trial_name": null,
|
| 246780 |
"trial_params": null
|
| 246781 |
}
|
model-bin/finetune/base/{checkpoint-115234 → checkpoint-115856}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1630111579.4549904/events.out.tfevents.1630111579.86bb0ddabf9b.1042.21
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:95f58fb51cee41f386b4585a19326a53730b6bbbe585f543021ae5acb2745a3d
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630111993.7382865/events.out.tfevents.1630111993.86bb0ddabf9b.1042.23
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c0e96cd5ba4ef52b500e089c11e28969334c8d3b6697fdd3b72683eff05f9ced
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630112410.354941/events.out.tfevents.1630112410.86bb0ddabf9b.1042.25
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:7d81916fe5e9e04cc0b52c08662d5a18dbfedbd842cd82ac1eb07dbac9e2df63
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630112831.6927667/events.out.tfevents.1630112831.86bb0ddabf9b.1042.27
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:74ea8d125465164f93c42375f7c13be612bb857413dc210089422061ce65f4cc
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630113242.9622881/events.out.tfevents.1630113242.86bb0ddabf9b.1042.29
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:aa708c4f927a9fea7e4cfbaebae65791d8502dd2927d07c41b3157e77465ff92
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1630111579.86bb0ddabf9b.1042.20
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:35302831d47931f68582a51c4526c141f50d4d6b1e276ad00936a09295de0963
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630111993.86bb0ddabf9b.1042.22
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:bc032fe9fac2fa6c55dcd8415dca2b76cc28eb5bda89ca632972ce40e718d2d3
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630112410.86bb0ddabf9b.1042.24
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6d551c6f77ece7a937dbec5a99834b948cde0ba0597facf1986c4c20d90ae6cd
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630112831.86bb0ddabf9b.1042.26
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:3a421a6d4c1906dca310ad16c2e52dcfaa60b63c6fd17b734f81fa4ee96da6db
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630113242.86bb0ddabf9b.1042.28
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:75e263b83127d33b8ce511ead8e112737571dd3bc684457225a97b0fe4950b4f
|
| 3 |
+
size 8622
|