"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-125446 → checkpoint-126068}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-125446 → checkpoint-126068}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-125446 → checkpoint-126068}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-125446 → checkpoint-126068}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-125446 → checkpoint-126068}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-125446 → checkpoint-126068}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-125446 → checkpoint-126068}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-125446 → checkpoint-126068}/trainer_state.json +793 -4
- model-bin/finetune/base/{checkpoint-125446 → checkpoint-126068}/training_args.bin +0 -0
- model-bin/finetune/base/log/1630146705.8776062/events.out.tfevents.1630146705.86bb0ddabf9b.4092.121 +3 -0
- model-bin/finetune/base/log/1630147090.9692104/events.out.tfevents.1630147090.86bb0ddabf9b.4092.123 +3 -0
- model-bin/finetune/base/log/1630147594.250904/events.out.tfevents.1630147594.86bb0ddabf9b.4092.125 +3 -0
- model-bin/finetune/base/log/1630147976.392592/events.out.tfevents.1630147976.86bb0ddabf9b.4092.127 +3 -0
- model-bin/finetune/base/log/1630148364.6244335/events.out.tfevents.1630148364.86bb0ddabf9b.4092.129 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630146705.86bb0ddabf9b.4092.120 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630147090.86bb0ddabf9b.4092.122 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630147594.86bb0ddabf9b.4092.124 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630147976.86bb0ddabf9b.4092.126 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630148364.86bb0ddabf9b.4092.128 +3 -0
model-bin/finetune/base/{checkpoint-125446 → checkpoint-126068}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-125446 → checkpoint-126068}/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 sha256:16b005e9fb69b2e7c668f7d9d1ddcf9502c134575ee756c2b4dd2a09e4f3dfc7
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-125446 → checkpoint-126068}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-125446 → checkpoint-126068}/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 sha256:2cfc721d81f03e5e16e7e7c9c5142ddddbe2c39f6609dd7fe74fbe15bffd4bf7
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-125446 → checkpoint-126068}/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 sha256:499abeb83f7ca59edfcad7e7d1e42dccc904e7481abcc4da5c71266773800888
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-125446 → checkpoint-126068}/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 sha256:683ccffd7adf3cb2b0b65666beebc8e51b2ae5066df9b1b2e36f844178028c85
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-125446 → checkpoint-126068}/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 sha256:149ab5492f8d36be18b46436a0461744c2fef1b202cb12894209dd371443a41d
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-125446 → checkpoint-126068}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.17162025681719809,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-124947",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -258972,11 +258972,800 @@
|
|
| 258972 |
"eval_steps_per_second": 0.724,
|
| 258973 |
"eval_wer": 0.1797744687003036,
|
| 258974 |
"step": 125446
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 258975 |
}
|
| 258976 |
],
|
| 258977 |
-
"max_steps":
|
| 258978 |
"num_train_epochs": 5000,
|
| 258979 |
-
"total_flos": 3.
|
| 258980 |
"trial_name": null,
|
| 258981 |
"trial_params": null
|
| 258982 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.17162025681719809,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-124947",
|
| 4 |
+
"epoch": 1008.0,
|
| 5 |
+
"global_step": 126068,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 258972 |
"eval_steps_per_second": 0.724,
|
| 258973 |
"eval_wer": 0.1797744687003036,
|
| 258974 |
"step": 125446
|
| 258975 |
+
},
|
| 258976 |
+
{
|
| 258977 |
+
"epoch": 1011.03,
|
| 258978 |
+
"learning_rate": 8.006394230769231e-06,
|
| 258979 |
+
"loss": 0.3332,
|
| 258980 |
+
"step": 125450
|
| 258981 |
+
},
|
| 258982 |
+
{
|
| 258983 |
+
"epoch": 1011.07,
|
| 258984 |
+
"learning_rate": 8.006314102564104e-06,
|
| 258985 |
+
"loss": 0.2435,
|
| 258986 |
+
"step": 125455
|
| 258987 |
+
},
|
| 258988 |
+
{
|
| 258989 |
+
"epoch": 1011.11,
|
| 258990 |
+
"learning_rate": 8.006233974358974e-06,
|
| 258991 |
+
"loss": 0.2895,
|
| 258992 |
+
"step": 125460
|
| 258993 |
+
},
|
| 258994 |
+
{
|
| 258995 |
+
"epoch": 1011.15,
|
| 258996 |
+
"learning_rate": 8.006153846153847e-06,
|
| 258997 |
+
"loss": 0.3673,
|
| 258998 |
+
"step": 125465
|
| 258999 |
+
},
|
| 259000 |
+
{
|
| 259001 |
+
"epoch": 1011.19,
|
| 259002 |
+
"learning_rate": 8.00607371794872e-06,
|
| 259003 |
+
"loss": 0.761,
|
| 259004 |
+
"step": 125470
|
| 259005 |
+
},
|
| 259006 |
+
{
|
| 259007 |
+
"epoch": 1011.23,
|
| 259008 |
+
"learning_rate": 8.00599358974359e-06,
|
| 259009 |
+
"loss": 0.7003,
|
| 259010 |
+
"step": 125475
|
| 259011 |
+
},
|
| 259012 |
+
{
|
| 259013 |
+
"epoch": 1011.27,
|
| 259014 |
+
"learning_rate": 8.005913461538463e-06,
|
| 259015 |
+
"loss": 0.2643,
|
| 259016 |
+
"step": 125480
|
| 259017 |
+
},
|
| 259018 |
+
{
|
| 259019 |
+
"epoch": 1011.31,
|
| 259020 |
+
"learning_rate": 8.005833333333334e-06,
|
| 259021 |
+
"loss": 0.3928,
|
| 259022 |
+
"step": 125485
|
| 259023 |
+
},
|
| 259024 |
+
{
|
| 259025 |
+
"epoch": 1011.35,
|
| 259026 |
+
"learning_rate": 8.005753205128205e-06,
|
| 259027 |
+
"loss": 0.3336,
|
| 259028 |
+
"step": 125490
|
| 259029 |
+
},
|
| 259030 |
+
{
|
| 259031 |
+
"epoch": 1011.4,
|
| 259032 |
+
"learning_rate": 8.005673076923077e-06,
|
| 259033 |
+
"loss": 0.8739,
|
| 259034 |
+
"step": 125495
|
| 259035 |
+
},
|
| 259036 |
+
{
|
| 259037 |
+
"epoch": 1011.44,
|
| 259038 |
+
"learning_rate": 8.00559294871795e-06,
|
| 259039 |
+
"loss": 0.625,
|
| 259040 |
+
"step": 125500
|
| 259041 |
+
},
|
| 259042 |
+
{
|
| 259043 |
+
"epoch": 1011.48,
|
| 259044 |
+
"learning_rate": 8.005512820512821e-06,
|
| 259045 |
+
"loss": 0.3288,
|
| 259046 |
+
"step": 125505
|
| 259047 |
+
},
|
| 259048 |
+
{
|
| 259049 |
+
"epoch": 1011.52,
|
| 259050 |
+
"learning_rate": 8.005432692307692e-06,
|
| 259051 |
+
"loss": 0.3602,
|
| 259052 |
+
"step": 125510
|
| 259053 |
+
},
|
| 259054 |
+
{
|
| 259055 |
+
"epoch": 1011.56,
|
| 259056 |
+
"learning_rate": 8.005352564102564e-06,
|
| 259057 |
+
"loss": 0.4349,
|
| 259058 |
+
"step": 125515
|
| 259059 |
+
},
|
| 259060 |
+
{
|
| 259061 |
+
"epoch": 1011.6,
|
| 259062 |
+
"learning_rate": 8.005272435897437e-06,
|
| 259063 |
+
"loss": 0.8922,
|
| 259064 |
+
"step": 125520
|
| 259065 |
+
},
|
| 259066 |
+
{
|
| 259067 |
+
"epoch": 1011.64,
|
| 259068 |
+
"learning_rate": 8.005192307692308e-06,
|
| 259069 |
+
"loss": 0.6377,
|
| 259070 |
+
"step": 125525
|
| 259071 |
+
},
|
| 259072 |
+
{
|
| 259073 |
+
"epoch": 1011.68,
|
| 259074 |
+
"learning_rate": 8.00511217948718e-06,
|
| 259075 |
+
"loss": 0.3145,
|
| 259076 |
+
"step": 125530
|
| 259077 |
+
},
|
| 259078 |
+
{
|
| 259079 |
+
"epoch": 1011.72,
|
| 259080 |
+
"learning_rate": 8.005032051282053e-06,
|
| 259081 |
+
"loss": 0.3291,
|
| 259082 |
+
"step": 125535
|
| 259083 |
+
},
|
| 259084 |
+
{
|
| 259085 |
+
"epoch": 1011.76,
|
| 259086 |
+
"learning_rate": 8.004951923076924e-06,
|
| 259087 |
+
"loss": 0.422,
|
| 259088 |
+
"step": 125540
|
| 259089 |
+
},
|
| 259090 |
+
{
|
| 259091 |
+
"epoch": 1011.8,
|
| 259092 |
+
"learning_rate": 8.004871794871795e-06,
|
| 259093 |
+
"loss": 0.924,
|
| 259094 |
+
"step": 125545
|
| 259095 |
+
},
|
| 259096 |
+
{
|
| 259097 |
+
"epoch": 1011.84,
|
| 259098 |
+
"learning_rate": 8.004791666666667e-06,
|
| 259099 |
+
"loss": 0.7038,
|
| 259100 |
+
"step": 125550
|
| 259101 |
+
},
|
| 259102 |
+
{
|
| 259103 |
+
"epoch": 1011.88,
|
| 259104 |
+
"learning_rate": 8.00471153846154e-06,
|
| 259105 |
+
"loss": 0.2517,
|
| 259106 |
+
"step": 125555
|
| 259107 |
+
},
|
| 259108 |
+
{
|
| 259109 |
+
"epoch": 1011.92,
|
| 259110 |
+
"learning_rate": 8.004631410256411e-06,
|
| 259111 |
+
"loss": 0.273,
|
| 259112 |
+
"step": 125560
|
| 259113 |
+
},
|
| 259114 |
+
{
|
| 259115 |
+
"epoch": 1011.96,
|
| 259116 |
+
"learning_rate": 8.004551282051282e-06,
|
| 259117 |
+
"loss": 0.4353,
|
| 259118 |
+
"step": 125565
|
| 259119 |
+
},
|
| 259120 |
+
{
|
| 259121 |
+
"epoch": 1012.0,
|
| 259122 |
+
"learning_rate": 8.004471153846155e-06,
|
| 259123 |
+
"loss": 1.0249,
|
| 259124 |
+
"step": 125570
|
| 259125 |
+
},
|
| 259126 |
+
{
|
| 259127 |
+
"epoch": 1012.0,
|
| 259128 |
+
"eval_loss": 0.4390537440776825,
|
| 259129 |
+
"eval_runtime": 36.2511,
|
| 259130 |
+
"eval_samples_per_second": 23.144,
|
| 259131 |
+
"eval_steps_per_second": 0.745,
|
| 259132 |
+
"eval_wer": 0.1779642141213156,
|
| 259133 |
+
"step": 125570
|
| 259134 |
+
},
|
| 259135 |
+
{
|
| 259136 |
+
"epoch": 1012.04,
|
| 259137 |
+
"learning_rate": 8.004391025641027e-06,
|
| 259138 |
+
"loss": 0.3656,
|
| 259139 |
+
"step": 125575
|
| 259140 |
+
},
|
| 259141 |
+
{
|
| 259142 |
+
"epoch": 1012.08,
|
| 259143 |
+
"learning_rate": 8.004310897435898e-06,
|
| 259144 |
+
"loss": 0.2624,
|
| 259145 |
+
"step": 125580
|
| 259146 |
+
},
|
| 259147 |
+
{
|
| 259148 |
+
"epoch": 1012.12,
|
| 259149 |
+
"learning_rate": 8.00423076923077e-06,
|
| 259150 |
+
"loss": 0.2921,
|
| 259151 |
+
"step": 125585
|
| 259152 |
+
},
|
| 259153 |
+
{
|
| 259154 |
+
"epoch": 1012.16,
|
| 259155 |
+
"learning_rate": 8.004150641025643e-06,
|
| 259156 |
+
"loss": 0.403,
|
| 259157 |
+
"step": 125590
|
| 259158 |
+
},
|
| 259159 |
+
{
|
| 259160 |
+
"epoch": 1012.2,
|
| 259161 |
+
"learning_rate": 8.004070512820514e-06,
|
| 259162 |
+
"loss": 1.3023,
|
| 259163 |
+
"step": 125595
|
| 259164 |
+
},
|
| 259165 |
+
{
|
| 259166 |
+
"epoch": 1012.24,
|
| 259167 |
+
"learning_rate": 8.003990384615385e-06,
|
| 259168 |
+
"loss": 0.3312,
|
| 259169 |
+
"step": 125600
|
| 259170 |
+
},
|
| 259171 |
+
{
|
| 259172 |
+
"epoch": 1012.28,
|
| 259173 |
+
"learning_rate": 8.003910256410257e-06,
|
| 259174 |
+
"loss": 0.2541,
|
| 259175 |
+
"step": 125605
|
| 259176 |
+
},
|
| 259177 |
+
{
|
| 259178 |
+
"epoch": 1012.32,
|
| 259179 |
+
"learning_rate": 8.00383012820513e-06,
|
| 259180 |
+
"loss": 0.3223,
|
| 259181 |
+
"step": 125610
|
| 259182 |
+
},
|
| 259183 |
+
{
|
| 259184 |
+
"epoch": 1012.36,
|
| 259185 |
+
"learning_rate": 8.00375e-06,
|
| 259186 |
+
"loss": 0.3871,
|
| 259187 |
+
"step": 125615
|
| 259188 |
+
},
|
| 259189 |
+
{
|
| 259190 |
+
"epoch": 1012.4,
|
| 259191 |
+
"learning_rate": 8.003669871794872e-06,
|
| 259192 |
+
"loss": 1.08,
|
| 259193 |
+
"step": 125620
|
| 259194 |
+
},
|
| 259195 |
+
{
|
| 259196 |
+
"epoch": 1012.44,
|
| 259197 |
+
"learning_rate": 8.003589743589745e-06,
|
| 259198 |
+
"loss": 0.3281,
|
| 259199 |
+
"step": 125625
|
| 259200 |
+
},
|
| 259201 |
+
{
|
| 259202 |
+
"epoch": 1012.48,
|
| 259203 |
+
"learning_rate": 8.003509615384615e-06,
|
| 259204 |
+
"loss": 0.2986,
|
| 259205 |
+
"step": 125630
|
| 259206 |
+
},
|
| 259207 |
+
{
|
| 259208 |
+
"epoch": 1012.52,
|
| 259209 |
+
"learning_rate": 8.003429487179488e-06,
|
| 259210 |
+
"loss": 0.2999,
|
| 259211 |
+
"step": 125635
|
| 259212 |
+
},
|
| 259213 |
+
{
|
| 259214 |
+
"epoch": 1012.56,
|
| 259215 |
+
"learning_rate": 8.00334935897436e-06,
|
| 259216 |
+
"loss": 0.3972,
|
| 259217 |
+
"step": 125640
|
| 259218 |
+
},
|
| 259219 |
+
{
|
| 259220 |
+
"epoch": 1012.6,
|
| 259221 |
+
"learning_rate": 8.003269230769231e-06,
|
| 259222 |
+
"loss": 1.2367,
|
| 259223 |
+
"step": 125645
|
| 259224 |
+
},
|
| 259225 |
+
{
|
| 259226 |
+
"epoch": 1012.64,
|
| 259227 |
+
"learning_rate": 8.003189102564102e-06,
|
| 259228 |
+
"loss": 0.314,
|
| 259229 |
+
"step": 125650
|
| 259230 |
+
},
|
| 259231 |
+
{
|
| 259232 |
+
"epoch": 1012.68,
|
| 259233 |
+
"learning_rate": 8.003108974358975e-06,
|
| 259234 |
+
"loss": 0.3581,
|
| 259235 |
+
"step": 125655
|
| 259236 |
+
},
|
| 259237 |
+
{
|
| 259238 |
+
"epoch": 1012.72,
|
| 259239 |
+
"learning_rate": 8.003028846153847e-06,
|
| 259240 |
+
"loss": 0.2869,
|
| 259241 |
+
"step": 125660
|
| 259242 |
+
},
|
| 259243 |
+
{
|
| 259244 |
+
"epoch": 1012.76,
|
| 259245 |
+
"learning_rate": 8.002948717948718e-06,
|
| 259246 |
+
"loss": 0.5337,
|
| 259247 |
+
"step": 125665
|
| 259248 |
+
},
|
| 259249 |
+
{
|
| 259250 |
+
"epoch": 1012.8,
|
| 259251 |
+
"learning_rate": 8.002868589743591e-06,
|
| 259252 |
+
"loss": 1.2025,
|
| 259253 |
+
"step": 125670
|
| 259254 |
+
},
|
| 259255 |
+
{
|
| 259256 |
+
"epoch": 1012.84,
|
| 259257 |
+
"learning_rate": 8.002788461538462e-06,
|
| 259258 |
+
"loss": 0.2866,
|
| 259259 |
+
"step": 125675
|
| 259260 |
+
},
|
| 259261 |
+
{
|
| 259262 |
+
"epoch": 1012.88,
|
| 259263 |
+
"learning_rate": 8.002708333333334e-06,
|
| 259264 |
+
"loss": 0.2661,
|
| 259265 |
+
"step": 125680
|
| 259266 |
+
},
|
| 259267 |
+
{
|
| 259268 |
+
"epoch": 1012.92,
|
| 259269 |
+
"learning_rate": 8.002628205128205e-06,
|
| 259270 |
+
"loss": 0.3406,
|
| 259271 |
+
"step": 125685
|
| 259272 |
+
},
|
| 259273 |
+
{
|
| 259274 |
+
"epoch": 1012.96,
|
| 259275 |
+
"learning_rate": 8.002548076923078e-06,
|
| 259276 |
+
"loss": 0.4782,
|
| 259277 |
+
"step": 125690
|
| 259278 |
+
},
|
| 259279 |
+
{
|
| 259280 |
+
"epoch": 1013.0,
|
| 259281 |
+
"eval_loss": 0.32623913884162903,
|
| 259282 |
+
"eval_runtime": 48.1801,
|
| 259283 |
+
"eval_samples_per_second": 17.414,
|
| 259284 |
+
"eval_steps_per_second": 0.56,
|
| 259285 |
+
"eval_wer": 0.18181818181818182,
|
| 259286 |
+
"step": 125694
|
| 259287 |
+
},
|
| 259288 |
+
{
|
| 259289 |
+
"epoch": 1005.01,
|
| 259290 |
+
"learning_rate": 8.00246794871795e-06,
|
| 259291 |
+
"loss": 0.5473,
|
| 259292 |
+
"step": 125695
|
| 259293 |
+
},
|
| 259294 |
+
{
|
| 259295 |
+
"epoch": 1005.05,
|
| 259296 |
+
"learning_rate": 8.002387820512821e-06,
|
| 259297 |
+
"loss": 0.3088,
|
| 259298 |
+
"step": 125700
|
| 259299 |
+
},
|
| 259300 |
+
{
|
| 259301 |
+
"epoch": 1005.09,
|
| 259302 |
+
"learning_rate": 8.002307692307692e-06,
|
| 259303 |
+
"loss": 0.304,
|
| 259304 |
+
"step": 125705
|
| 259305 |
+
},
|
| 259306 |
+
{
|
| 259307 |
+
"epoch": 1005.13,
|
| 259308 |
+
"learning_rate": 8.002227564102565e-06,
|
| 259309 |
+
"loss": 0.3238,
|
| 259310 |
+
"step": 125710
|
| 259311 |
+
},
|
| 259312 |
+
{
|
| 259313 |
+
"epoch": 1005.17,
|
| 259314 |
+
"learning_rate": 8.002147435897437e-06,
|
| 259315 |
+
"loss": 0.5224,
|
| 259316 |
+
"step": 125715
|
| 259317 |
+
},
|
| 259318 |
+
{
|
| 259319 |
+
"epoch": 1005.21,
|
| 259320 |
+
"learning_rate": 8.002067307692308e-06,
|
| 259321 |
+
"loss": 1.0523,
|
| 259322 |
+
"step": 125720
|
| 259323 |
+
},
|
| 259324 |
+
{
|
| 259325 |
+
"epoch": 1005.25,
|
| 259326 |
+
"learning_rate": 8.001987179487181e-06,
|
| 259327 |
+
"loss": 0.3129,
|
| 259328 |
+
"step": 125725
|
| 259329 |
+
},
|
| 259330 |
+
{
|
| 259331 |
+
"epoch": 1005.29,
|
| 259332 |
+
"learning_rate": 8.001907051282052e-06,
|
| 259333 |
+
"loss": 0.3159,
|
| 259334 |
+
"step": 125730
|
| 259335 |
+
},
|
| 259336 |
+
{
|
| 259337 |
+
"epoch": 1005.33,
|
| 259338 |
+
"learning_rate": 8.001826923076924e-06,
|
| 259339 |
+
"loss": 0.3222,
|
| 259340 |
+
"step": 125735
|
| 259341 |
+
},
|
| 259342 |
+
{
|
| 259343 |
+
"epoch": 1005.37,
|
| 259344 |
+
"learning_rate": 8.001746794871795e-06,
|
| 259345 |
+
"loss": 0.5416,
|
| 259346 |
+
"step": 125740
|
| 259347 |
+
},
|
| 259348 |
+
{
|
| 259349 |
+
"epoch": 1005.41,
|
| 259350 |
+
"learning_rate": 8.001666666666668e-06,
|
| 259351 |
+
"loss": 0.9692,
|
| 259352 |
+
"step": 125745
|
| 259353 |
+
},
|
| 259354 |
+
{
|
| 259355 |
+
"epoch": 1005.45,
|
| 259356 |
+
"learning_rate": 8.001586538461538e-06,
|
| 259357 |
+
"loss": 0.2915,
|
| 259358 |
+
"step": 125750
|
| 259359 |
+
},
|
| 259360 |
+
{
|
| 259361 |
+
"epoch": 1005.49,
|
| 259362 |
+
"learning_rate": 8.001506410256411e-06,
|
| 259363 |
+
"loss": 0.243,
|
| 259364 |
+
"step": 125755
|
| 259365 |
+
},
|
| 259366 |
+
{
|
| 259367 |
+
"epoch": 1005.53,
|
| 259368 |
+
"learning_rate": 8.001426282051282e-06,
|
| 259369 |
+
"loss": 0.3572,
|
| 259370 |
+
"step": 125760
|
| 259371 |
+
},
|
| 259372 |
+
{
|
| 259373 |
+
"epoch": 1005.57,
|
| 259374 |
+
"learning_rate": 8.001346153846154e-06,
|
| 259375 |
+
"loss": 0.5895,
|
| 259376 |
+
"step": 125765
|
| 259377 |
+
},
|
| 259378 |
+
{
|
| 259379 |
+
"epoch": 1005.61,
|
| 259380 |
+
"learning_rate": 8.001266025641027e-06,
|
| 259381 |
+
"loss": 1.1071,
|
| 259382 |
+
"step": 125770
|
| 259383 |
+
},
|
| 259384 |
+
{
|
| 259385 |
+
"epoch": 1005.65,
|
| 259386 |
+
"learning_rate": 8.001185897435898e-06,
|
| 259387 |
+
"loss": 0.2927,
|
| 259388 |
+
"step": 125775
|
| 259389 |
+
},
|
| 259390 |
+
{
|
| 259391 |
+
"epoch": 1005.69,
|
| 259392 |
+
"learning_rate": 8.00110576923077e-06,
|
| 259393 |
+
"loss": 0.336,
|
| 259394 |
+
"step": 125780
|
| 259395 |
+
},
|
| 259396 |
+
{
|
| 259397 |
+
"epoch": 1005.73,
|
| 259398 |
+
"learning_rate": 8.00102564102564e-06,
|
| 259399 |
+
"loss": 0.3185,
|
| 259400 |
+
"step": 125785
|
| 259401 |
+
},
|
| 259402 |
+
{
|
| 259403 |
+
"epoch": 1005.77,
|
| 259404 |
+
"learning_rate": 8.000945512820514e-06,
|
| 259405 |
+
"loss": 0.5286,
|
| 259406 |
+
"step": 125790
|
| 259407 |
+
},
|
| 259408 |
+
{
|
| 259409 |
+
"epoch": 1005.81,
|
| 259410 |
+
"learning_rate": 8.000865384615385e-06,
|
| 259411 |
+
"loss": 1.0564,
|
| 259412 |
+
"step": 125795
|
| 259413 |
+
},
|
| 259414 |
+
{
|
| 259415 |
+
"epoch": 1005.85,
|
| 259416 |
+
"learning_rate": 8.000785256410257e-06,
|
| 259417 |
+
"loss": 0.351,
|
| 259418 |
+
"step": 125800
|
| 259419 |
+
},
|
| 259420 |
+
{
|
| 259421 |
+
"epoch": 1005.89,
|
| 259422 |
+
"learning_rate": 8.000705128205128e-06,
|
| 259423 |
+
"loss": 0.3298,
|
| 259424 |
+
"step": 125805
|
| 259425 |
+
},
|
| 259426 |
+
{
|
| 259427 |
+
"epoch": 1005.93,
|
| 259428 |
+
"learning_rate": 8.000625000000001e-06,
|
| 259429 |
+
"loss": 0.338,
|
| 259430 |
+
"step": 125810
|
| 259431 |
+
},
|
| 259432 |
+
{
|
| 259433 |
+
"epoch": 1005.97,
|
| 259434 |
+
"learning_rate": 8.000544871794872e-06,
|
| 259435 |
+
"loss": 0.488,
|
| 259436 |
+
"step": 125815
|
| 259437 |
+
},
|
| 259438 |
+
{
|
| 259439 |
+
"epoch": 1006.0,
|
| 259440 |
+
"eval_loss": 0.4103972613811493,
|
| 259441 |
+
"eval_runtime": 35.6053,
|
| 259442 |
+
"eval_samples_per_second": 23.62,
|
| 259443 |
+
"eval_steps_per_second": 0.758,
|
| 259444 |
+
"eval_wer": 0.17794615271443284,
|
| 259445 |
+
"step": 125819
|
| 259446 |
+
},
|
| 259447 |
+
{
|
| 259448 |
+
"epoch": 1014.01,
|
| 259449 |
+
"learning_rate": 8.000464743589744e-06,
|
| 259450 |
+
"loss": 0.2939,
|
| 259451 |
+
"step": 125820
|
| 259452 |
+
},
|
| 259453 |
+
{
|
| 259454 |
+
"epoch": 1014.05,
|
| 259455 |
+
"learning_rate": 8.000384615384617e-06,
|
| 259456 |
+
"loss": 0.3032,
|
| 259457 |
+
"step": 125825
|
| 259458 |
+
},
|
| 259459 |
+
{
|
| 259460 |
+
"epoch": 1014.09,
|
| 259461 |
+
"learning_rate": 8.000304487179488e-06,
|
| 259462 |
+
"loss": 0.2653,
|
| 259463 |
+
"step": 125830
|
| 259464 |
+
},
|
| 259465 |
+
{
|
| 259466 |
+
"epoch": 1014.13,
|
| 259467 |
+
"learning_rate": 8.00022435897436e-06,
|
| 259468 |
+
"loss": 0.3178,
|
| 259469 |
+
"step": 125835
|
| 259470 |
+
},
|
| 259471 |
+
{
|
| 259472 |
+
"epoch": 1014.17,
|
| 259473 |
+
"learning_rate": 8.000144230769231e-06,
|
| 259474 |
+
"loss": 0.5021,
|
| 259475 |
+
"step": 125840
|
| 259476 |
+
},
|
| 259477 |
+
{
|
| 259478 |
+
"epoch": 1014.21,
|
| 259479 |
+
"learning_rate": 8.000064102564104e-06,
|
| 259480 |
+
"loss": 1.0234,
|
| 259481 |
+
"step": 125845
|
| 259482 |
+
},
|
| 259483 |
+
{
|
| 259484 |
+
"epoch": 1014.25,
|
| 259485 |
+
"learning_rate": 7.999983974358975e-06,
|
| 259486 |
+
"loss": 0.3039,
|
| 259487 |
+
"step": 125850
|
| 259488 |
+
},
|
| 259489 |
+
{
|
| 259490 |
+
"epoch": 1014.29,
|
| 259491 |
+
"learning_rate": 7.999903846153847e-06,
|
| 259492 |
+
"loss": 0.3366,
|
| 259493 |
+
"step": 125855
|
| 259494 |
+
},
|
| 259495 |
+
{
|
| 259496 |
+
"epoch": 1014.33,
|
| 259497 |
+
"learning_rate": 7.999823717948718e-06,
|
| 259498 |
+
"loss": 0.3143,
|
| 259499 |
+
"step": 125860
|
| 259500 |
+
},
|
| 259501 |
+
{
|
| 259502 |
+
"epoch": 1014.37,
|
| 259503 |
+
"learning_rate": 7.999743589743591e-06,
|
| 259504 |
+
"loss": 0.6841,
|
| 259505 |
+
"step": 125865
|
| 259506 |
+
},
|
| 259507 |
+
{
|
| 259508 |
+
"epoch": 1014.41,
|
| 259509 |
+
"learning_rate": 7.999663461538462e-06,
|
| 259510 |
+
"loss": 1.2289,
|
| 259511 |
+
"step": 125870
|
| 259512 |
+
},
|
| 259513 |
+
{
|
| 259514 |
+
"epoch": 1014.45,
|
| 259515 |
+
"learning_rate": 7.999583333333334e-06,
|
| 259516 |
+
"loss": 0.3278,
|
| 259517 |
+
"step": 125875
|
| 259518 |
+
},
|
| 259519 |
+
{
|
| 259520 |
+
"epoch": 1014.49,
|
| 259521 |
+
"learning_rate": 7.999503205128207e-06,
|
| 259522 |
+
"loss": 0.2565,
|
| 259523 |
+
"step": 125880
|
| 259524 |
+
},
|
| 259525 |
+
{
|
| 259526 |
+
"epoch": 1014.53,
|
| 259527 |
+
"learning_rate": 7.999423076923078e-06,
|
| 259528 |
+
"loss": 0.3343,
|
| 259529 |
+
"step": 125885
|
| 259530 |
+
},
|
| 259531 |
+
{
|
| 259532 |
+
"epoch": 1014.57,
|
| 259533 |
+
"learning_rate": 7.99934294871795e-06,
|
| 259534 |
+
"loss": 0.5523,
|
| 259535 |
+
"step": 125890
|
| 259536 |
+
},
|
| 259537 |
+
{
|
| 259538 |
+
"epoch": 1014.61,
|
| 259539 |
+
"learning_rate": 7.999262820512821e-06,
|
| 259540 |
+
"loss": 1.0618,
|
| 259541 |
+
"step": 125895
|
| 259542 |
+
},
|
| 259543 |
+
{
|
| 259544 |
+
"epoch": 1014.65,
|
| 259545 |
+
"learning_rate": 7.999182692307694e-06,
|
| 259546 |
+
"loss": 0.3569,
|
| 259547 |
+
"step": 125900
|
| 259548 |
+
},
|
| 259549 |
+
{
|
| 259550 |
+
"epoch": 1014.69,
|
| 259551 |
+
"learning_rate": 7.999102564102564e-06,
|
| 259552 |
+
"loss": 0.3683,
|
| 259553 |
+
"step": 125905
|
| 259554 |
+
},
|
| 259555 |
+
{
|
| 259556 |
+
"epoch": 1014.73,
|
| 259557 |
+
"learning_rate": 7.999022435897437e-06,
|
| 259558 |
+
"loss": 0.3664,
|
| 259559 |
+
"step": 125910
|
| 259560 |
+
},
|
| 259561 |
+
{
|
| 259562 |
+
"epoch": 1014.77,
|
| 259563 |
+
"learning_rate": 7.998942307692308e-06,
|
| 259564 |
+
"loss": 0.5853,
|
| 259565 |
+
"step": 125915
|
| 259566 |
+
},
|
| 259567 |
+
{
|
| 259568 |
+
"epoch": 1014.81,
|
| 259569 |
+
"learning_rate": 7.99886217948718e-06,
|
| 259570 |
+
"loss": 1.0467,
|
| 259571 |
+
"step": 125920
|
| 259572 |
+
},
|
| 259573 |
+
{
|
| 259574 |
+
"epoch": 1014.85,
|
| 259575 |
+
"learning_rate": 7.998782051282052e-06,
|
| 259576 |
+
"loss": 0.3293,
|
| 259577 |
+
"step": 125925
|
| 259578 |
+
},
|
| 259579 |
+
{
|
| 259580 |
+
"epoch": 1014.89,
|
| 259581 |
+
"learning_rate": 7.998701923076924e-06,
|
| 259582 |
+
"loss": 0.2739,
|
| 259583 |
+
"step": 125930
|
| 259584 |
+
},
|
| 259585 |
+
{
|
| 259586 |
+
"epoch": 1014.93,
|
| 259587 |
+
"learning_rate": 7.998621794871795e-06,
|
| 259588 |
+
"loss": 0.3778,
|
| 259589 |
+
"step": 125935
|
| 259590 |
+
},
|
| 259591 |
+
{
|
| 259592 |
+
"epoch": 1014.97,
|
| 259593 |
+
"learning_rate": 7.998541666666666e-06,
|
| 259594 |
+
"loss": 0.5238,
|
| 259595 |
+
"step": 125940
|
| 259596 |
+
},
|
| 259597 |
+
{
|
| 259598 |
+
"epoch": 1015.0,
|
| 259599 |
+
"eval_loss": 0.36781224608421326,
|
| 259600 |
+
"eval_runtime": 36.1928,
|
| 259601 |
+
"eval_samples_per_second": 23.237,
|
| 259602 |
+
"eval_steps_per_second": 0.746,
|
| 259603 |
+
"eval_wer": 0.18371311169219276,
|
| 259604 |
+
"step": 125943
|
| 259605 |
+
},
|
| 259606 |
+
{
|
| 259607 |
+
"epoch": 1007.02,
|
| 259608 |
+
"learning_rate": 7.99846153846154e-06,
|
| 259609 |
+
"loss": 0.3979,
|
| 259610 |
+
"step": 125945
|
| 259611 |
+
},
|
| 259612 |
+
{
|
| 259613 |
+
"epoch": 1007.06,
|
| 259614 |
+
"learning_rate": 7.998381410256411e-06,
|
| 259615 |
+
"loss": 0.3103,
|
| 259616 |
+
"step": 125950
|
| 259617 |
+
},
|
| 259618 |
+
{
|
| 259619 |
+
"epoch": 1007.1,
|
| 259620 |
+
"learning_rate": 7.998301282051282e-06,
|
| 259621 |
+
"loss": 0.2781,
|
| 259622 |
+
"step": 125955
|
| 259623 |
+
},
|
| 259624 |
+
{
|
| 259625 |
+
"epoch": 1007.14,
|
| 259626 |
+
"learning_rate": 7.998221153846154e-06,
|
| 259627 |
+
"loss": 0.3187,
|
| 259628 |
+
"step": 125960
|
| 259629 |
+
},
|
| 259630 |
+
{
|
| 259631 |
+
"epoch": 1007.18,
|
| 259632 |
+
"learning_rate": 7.998141025641027e-06,
|
| 259633 |
+
"loss": 0.6326,
|
| 259634 |
+
"step": 125965
|
| 259635 |
+
},
|
| 259636 |
+
{
|
| 259637 |
+
"epoch": 1007.22,
|
| 259638 |
+
"learning_rate": 7.998060897435898e-06,
|
| 259639 |
+
"loss": 1.0426,
|
| 259640 |
+
"step": 125970
|
| 259641 |
+
},
|
| 259642 |
+
{
|
| 259643 |
+
"epoch": 1007.26,
|
| 259644 |
+
"learning_rate": 7.99798076923077e-06,
|
| 259645 |
+
"loss": 0.2652,
|
| 259646 |
+
"step": 125975
|
| 259647 |
+
},
|
| 259648 |
+
{
|
| 259649 |
+
"epoch": 1007.3,
|
| 259650 |
+
"learning_rate": 7.997900641025642e-06,
|
| 259651 |
+
"loss": 0.2786,
|
| 259652 |
+
"step": 125980
|
| 259653 |
+
},
|
| 259654 |
+
{
|
| 259655 |
+
"epoch": 1007.34,
|
| 259656 |
+
"learning_rate": 7.997820512820514e-06,
|
| 259657 |
+
"loss": 0.406,
|
| 259658 |
+
"step": 125985
|
| 259659 |
+
},
|
| 259660 |
+
{
|
| 259661 |
+
"epoch": 1007.38,
|
| 259662 |
+
"learning_rate": 7.997740384615385e-06,
|
| 259663 |
+
"loss": 0.6165,
|
| 259664 |
+
"step": 125990
|
| 259665 |
+
},
|
| 259666 |
+
{
|
| 259667 |
+
"epoch": 1007.42,
|
| 259668 |
+
"learning_rate": 7.997660256410257e-06,
|
| 259669 |
+
"loss": 1.0945,
|
| 259670 |
+
"step": 125995
|
| 259671 |
+
},
|
| 259672 |
+
{
|
| 259673 |
+
"epoch": 1007.46,
|
| 259674 |
+
"learning_rate": 7.99758012820513e-06,
|
| 259675 |
+
"loss": 0.2789,
|
| 259676 |
+
"step": 126000
|
| 259677 |
+
},
|
| 259678 |
+
{
|
| 259679 |
+
"epoch": 1007.5,
|
| 259680 |
+
"learning_rate": 7.997500000000001e-06,
|
| 259681 |
+
"loss": 0.2985,
|
| 259682 |
+
"step": 126005
|
| 259683 |
+
},
|
| 259684 |
+
{
|
| 259685 |
+
"epoch": 1007.54,
|
| 259686 |
+
"learning_rate": 7.997419871794872e-06,
|
| 259687 |
+
"loss": 0.2813,
|
| 259688 |
+
"step": 126010
|
| 259689 |
+
},
|
| 259690 |
+
{
|
| 259691 |
+
"epoch": 1007.58,
|
| 259692 |
+
"learning_rate": 7.997339743589744e-06,
|
| 259693 |
+
"loss": 0.6877,
|
| 259694 |
+
"step": 126015
|
| 259695 |
+
},
|
| 259696 |
+
{
|
| 259697 |
+
"epoch": 1007.62,
|
| 259698 |
+
"learning_rate": 7.997259615384617e-06,
|
| 259699 |
+
"loss": 0.9805,
|
| 259700 |
+
"step": 126020
|
| 259701 |
+
},
|
| 259702 |
+
{
|
| 259703 |
+
"epoch": 1007.66,
|
| 259704 |
+
"learning_rate": 7.997179487179488e-06,
|
| 259705 |
+
"loss": 0.2643,
|
| 259706 |
+
"step": 126025
|
| 259707 |
+
},
|
| 259708 |
+
{
|
| 259709 |
+
"epoch": 1007.7,
|
| 259710 |
+
"learning_rate": 7.99709935897436e-06,
|
| 259711 |
+
"loss": 0.2754,
|
| 259712 |
+
"step": 126030
|
| 259713 |
+
},
|
| 259714 |
+
{
|
| 259715 |
+
"epoch": 1007.74,
|
| 259716 |
+
"learning_rate": 7.997019230769232e-06,
|
| 259717 |
+
"loss": 0.3053,
|
| 259718 |
+
"step": 126035
|
| 259719 |
+
},
|
| 259720 |
+
{
|
| 259721 |
+
"epoch": 1007.78,
|
| 259722 |
+
"learning_rate": 7.996939102564104e-06,
|
| 259723 |
+
"loss": 0.6208,
|
| 259724 |
+
"step": 126040
|
| 259725 |
+
},
|
| 259726 |
+
{
|
| 259727 |
+
"epoch": 1007.82,
|
| 259728 |
+
"learning_rate": 7.996858974358975e-06,
|
| 259729 |
+
"loss": 1.0308,
|
| 259730 |
+
"step": 126045
|
| 259731 |
+
},
|
| 259732 |
+
{
|
| 259733 |
+
"epoch": 1007.86,
|
| 259734 |
+
"learning_rate": 7.996778846153847e-06,
|
| 259735 |
+
"loss": 0.2589,
|
| 259736 |
+
"step": 126050
|
| 259737 |
+
},
|
| 259738 |
+
{
|
| 259739 |
+
"epoch": 1007.9,
|
| 259740 |
+
"learning_rate": 7.99669871794872e-06,
|
| 259741 |
+
"loss": 0.2949,
|
| 259742 |
+
"step": 126055
|
| 259743 |
+
},
|
| 259744 |
+
{
|
| 259745 |
+
"epoch": 1007.94,
|
| 259746 |
+
"learning_rate": 7.99661858974359e-06,
|
| 259747 |
+
"loss": 0.316,
|
| 259748 |
+
"step": 126060
|
| 259749 |
+
},
|
| 259750 |
+
{
|
| 259751 |
+
"epoch": 1007.98,
|
| 259752 |
+
"learning_rate": 7.996538461538462e-06,
|
| 259753 |
+
"loss": 0.5452,
|
| 259754 |
+
"step": 126065
|
| 259755 |
+
},
|
| 259756 |
+
{
|
| 259757 |
+
"epoch": 1008.0,
|
| 259758 |
+
"eval_loss": 0.4000602066516876,
|
| 259759 |
+
"eval_runtime": 36.515,
|
| 259760 |
+
"eval_samples_per_second": 23.032,
|
| 259761 |
+
"eval_steps_per_second": 0.739,
|
| 259762 |
+
"eval_wer": 0.18030087568295786,
|
| 259763 |
+
"step": 126068
|
| 259764 |
}
|
| 259765 |
],
|
| 259766 |
+
"max_steps": 625000,
|
| 259767 |
"num_train_epochs": 5000,
|
| 259768 |
+
"total_flos": 3.547605377048925e+20,
|
| 259769 |
"trial_name": null,
|
| 259770 |
"trial_params": null
|
| 259771 |
}
|
model-bin/finetune/base/{checkpoint-125446 → checkpoint-126068}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1630146705.8776062/events.out.tfevents.1630146705.86bb0ddabf9b.4092.121
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:5022a9069cb186dbe772a5b12206d0f5add6f2140b964a82d4f5988b8716ce8f
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630147090.9692104/events.out.tfevents.1630147090.86bb0ddabf9b.4092.123
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:982f99c0e956f481e5fc347ef161d31d63e408957501f6c1ec947beadfa90553
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630147594.250904/events.out.tfevents.1630147594.86bb0ddabf9b.4092.125
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d8b93e9700ca21911643d51dac49e768b32cf23e330ef4a59fb1d41200ceb663
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630147976.392592/events.out.tfevents.1630147976.86bb0ddabf9b.4092.127
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f30bfb48fce824788b39b728a038285ad4abede0cc55779a95d5f92ddccfea8c
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630148364.6244335/events.out.tfevents.1630148364.86bb0ddabf9b.4092.129
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:08ee3ab483c8ec102883e3bd12c6b718989b5c7649f48dc2055817ec21d331e5
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1630146705.86bb0ddabf9b.4092.120
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a5bbf253060e791d99008d225cae68311e731bfbefaca264082bfef66b0920dd
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630147090.86bb0ddabf9b.4092.122
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:67bd7e33213924bc59d9ed9c4eadea66e0729ba0d99643999e8961d3ab56b493
|
| 3 |
+
size 8462
|
model-bin/finetune/base/log/events.out.tfevents.1630147594.86bb0ddabf9b.4092.124
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:5511a402159ad90b2f655ce38138505d72085dc9f06b2370a379b8bde9112476
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630147976.86bb0ddabf9b.4092.126
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:bfc30050d1853a87afb646d8049f4ae0412bf14a163343bff1e2bab7c2cb7daf
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630148364.86bb0ddabf9b.4092.128
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:bcb1d439a5450a64819aa55a96fbcc2bc3bd14c2128b8ebea5e7a7d1c13a698c
|
| 3 |
+
size 8622
|