"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-65705 → checkpoint-66326}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-65705 → checkpoint-66326}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-65705 → checkpoint-66326}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-65705 → checkpoint-66326}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-65705 → checkpoint-66326}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-65705 → checkpoint-66326}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-65705 → checkpoint-66326}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-65705 → checkpoint-66326}/trainer_state.json +793 -4
- model-bin/finetune/base/{checkpoint-65705 → checkpoint-66326}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629827645.253879/events.out.tfevents.1629827645.c435e1c5ee04.920.191 +3 -0
- model-bin/finetune/base/log/1629828281.4169908/events.out.tfevents.1629828281.c435e1c5ee04.920.193 +3 -0
- model-bin/finetune/base/log/1629828923.8657014/events.out.tfevents.1629828923.c435e1c5ee04.920.195 +3 -0
- model-bin/finetune/base/log/1629829555.331946/events.out.tfevents.1629829555.c435e1c5ee04.920.197 +3 -0
- model-bin/finetune/base/log/1629830197.8430533/events.out.tfevents.1629830197.c435e1c5ee04.920.199 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629827645.c435e1c5ee04.920.190 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629828281.c435e1c5ee04.920.192 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629828923.c435e1c5ee04.920.194 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629829555.c435e1c5ee04.920.196 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629830197.c435e1c5ee04.920.198 +3 -0
model-bin/finetune/base/{checkpoint-65705 → checkpoint-66326}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-65705 → checkpoint-66326}/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 sha256:85ea8ef6c64908f2115fcc83c3c7ba096536f425df00a3c9f1cbfc2255f20245
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-65705 → checkpoint-66326}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-65705 → checkpoint-66326}/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 sha256:989a241331db9f31c999d496faab12f6ba17620c47350944896a193165b684aa
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-65705 → checkpoint-66326}/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 sha256:06ffb88ff1db1b23a37760fd519fbca7feaf5222271d9129ca3900755d9cb924
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-65705 → checkpoint-66326}/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 sha256:1ae34bb3b632ce84a747aa987dc0f72facb05243dbdbd4095b0555d0a6525518
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-65705 → checkpoint-66326}/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 sha256:13a19303204503e39131fe6893ee9e90595d5035f7349c77cb80c388b6b78e06
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-65705 → checkpoint-66326}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.18588425381903642,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-60729",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -182964,11 +182964,800 @@
|
|
| 182964 |
"eval_steps_per_second": 0.677,
|
| 182965 |
"eval_wer": 0.18776816231546797,
|
| 182966 |
"step": 65705
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 182967 |
}
|
| 182968 |
],
|
| 182969 |
-
"max_steps":
|
| 182970 |
"num_train_epochs": 5000,
|
| 182971 |
-
"total_flos": 1.
|
| 182972 |
"trial_name": null,
|
| 182973 |
"trial_params": null
|
| 182974 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.18588425381903642,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-60729",
|
| 4 |
+
"epoch": 533.995983935743,
|
| 5 |
+
"global_step": 66326,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 182964 |
"eval_steps_per_second": 0.677,
|
| 182965 |
"eval_wer": 0.18776816231546797,
|
| 182966 |
"step": 65705
|
| 182967 |
+
},
|
| 182968 |
+
{
|
| 182969 |
+
"epoch": 529.04,
|
| 182970 |
+
"learning_rate": 8.963381410256411e-06,
|
| 182971 |
+
"loss": 0.3291,
|
| 182972 |
+
"step": 65710
|
| 182973 |
+
},
|
| 182974 |
+
{
|
| 182975 |
+
"epoch": 529.08,
|
| 182976 |
+
"learning_rate": 8.963301282051284e-06,
|
| 182977 |
+
"loss": 0.325,
|
| 182978 |
+
"step": 65715
|
| 182979 |
+
},
|
| 182980 |
+
{
|
| 182981 |
+
"epoch": 529.12,
|
| 182982 |
+
"learning_rate": 8.963221153846154e-06,
|
| 182983 |
+
"loss": 0.3205,
|
| 182984 |
+
"step": 65720
|
| 182985 |
+
},
|
| 182986 |
+
{
|
| 182987 |
+
"epoch": 529.16,
|
| 182988 |
+
"learning_rate": 8.963141025641027e-06,
|
| 182989 |
+
"loss": 0.4565,
|
| 182990 |
+
"step": 65725
|
| 182991 |
+
},
|
| 182992 |
+
{
|
| 182993 |
+
"epoch": 529.2,
|
| 182994 |
+
"learning_rate": 8.963060897435898e-06,
|
| 182995 |
+
"loss": 1.14,
|
| 182996 |
+
"step": 65730
|
| 182997 |
+
},
|
| 182998 |
+
{
|
| 182999 |
+
"epoch": 529.24,
|
| 183000 |
+
"learning_rate": 8.96298076923077e-06,
|
| 183001 |
+
"loss": 0.2933,
|
| 183002 |
+
"step": 65735
|
| 183003 |
+
},
|
| 183004 |
+
{
|
| 183005 |
+
"epoch": 529.28,
|
| 183006 |
+
"learning_rate": 8.96290064102564e-06,
|
| 183007 |
+
"loss": 0.3287,
|
| 183008 |
+
"step": 65740
|
| 183009 |
+
},
|
| 183010 |
+
{
|
| 183011 |
+
"epoch": 529.32,
|
| 183012 |
+
"learning_rate": 8.962820512820514e-06,
|
| 183013 |
+
"loss": 0.3612,
|
| 183014 |
+
"step": 65745
|
| 183015 |
+
},
|
| 183016 |
+
{
|
| 183017 |
+
"epoch": 529.36,
|
| 183018 |
+
"learning_rate": 8.962740384615385e-06,
|
| 183019 |
+
"loss": 0.505,
|
| 183020 |
+
"step": 65750
|
| 183021 |
+
},
|
| 183022 |
+
{
|
| 183023 |
+
"epoch": 529.4,
|
| 183024 |
+
"learning_rate": 8.962660256410257e-06,
|
| 183025 |
+
"loss": 1.2446,
|
| 183026 |
+
"step": 65755
|
| 183027 |
+
},
|
| 183028 |
+
{
|
| 183029 |
+
"epoch": 529.44,
|
| 183030 |
+
"learning_rate": 8.96258012820513e-06,
|
| 183031 |
+
"loss": 0.4099,
|
| 183032 |
+
"step": 65760
|
| 183033 |
+
},
|
| 183034 |
+
{
|
| 183035 |
+
"epoch": 529.48,
|
| 183036 |
+
"learning_rate": 8.962500000000001e-06,
|
| 183037 |
+
"loss": 0.4326,
|
| 183038 |
+
"step": 65765
|
| 183039 |
+
},
|
| 183040 |
+
{
|
| 183041 |
+
"epoch": 529.52,
|
| 183042 |
+
"learning_rate": 8.962419871794872e-06,
|
| 183043 |
+
"loss": 0.29,
|
| 183044 |
+
"step": 65770
|
| 183045 |
+
},
|
| 183046 |
+
{
|
| 183047 |
+
"epoch": 529.56,
|
| 183048 |
+
"learning_rate": 8.962339743589744e-06,
|
| 183049 |
+
"loss": 0.501,
|
| 183050 |
+
"step": 65775
|
| 183051 |
+
},
|
| 183052 |
+
{
|
| 183053 |
+
"epoch": 529.6,
|
| 183054 |
+
"learning_rate": 8.962259615384617e-06,
|
| 183055 |
+
"loss": 1.2461,
|
| 183056 |
+
"step": 65780
|
| 183057 |
+
},
|
| 183058 |
+
{
|
| 183059 |
+
"epoch": 529.64,
|
| 183060 |
+
"learning_rate": 8.962179487179488e-06,
|
| 183061 |
+
"loss": 0.366,
|
| 183062 |
+
"step": 65785
|
| 183063 |
+
},
|
| 183064 |
+
{
|
| 183065 |
+
"epoch": 529.68,
|
| 183066 |
+
"learning_rate": 8.96209935897436e-06,
|
| 183067 |
+
"loss": 0.3331,
|
| 183068 |
+
"step": 65790
|
| 183069 |
+
},
|
| 183070 |
+
{
|
| 183071 |
+
"epoch": 529.72,
|
| 183072 |
+
"learning_rate": 8.962019230769231e-06,
|
| 183073 |
+
"loss": 0.3664,
|
| 183074 |
+
"step": 65795
|
| 183075 |
+
},
|
| 183076 |
+
{
|
| 183077 |
+
"epoch": 529.76,
|
| 183078 |
+
"learning_rate": 8.961939102564104e-06,
|
| 183079 |
+
"loss": 0.595,
|
| 183080 |
+
"step": 65800
|
| 183081 |
+
},
|
| 183082 |
+
{
|
| 183083 |
+
"epoch": 529.8,
|
| 183084 |
+
"learning_rate": 8.961858974358975e-06,
|
| 183085 |
+
"loss": 1.1922,
|
| 183086 |
+
"step": 65805
|
| 183087 |
+
},
|
| 183088 |
+
{
|
| 183089 |
+
"epoch": 529.84,
|
| 183090 |
+
"learning_rate": 8.961778846153847e-06,
|
| 183091 |
+
"loss": 0.3499,
|
| 183092 |
+
"step": 65810
|
| 183093 |
+
},
|
| 183094 |
+
{
|
| 183095 |
+
"epoch": 529.88,
|
| 183096 |
+
"learning_rate": 8.96169871794872e-06,
|
| 183097 |
+
"loss": 0.2735,
|
| 183098 |
+
"step": 65815
|
| 183099 |
+
},
|
| 183100 |
+
{
|
| 183101 |
+
"epoch": 529.92,
|
| 183102 |
+
"learning_rate": 8.961618589743591e-06,
|
| 183103 |
+
"loss": 0.3509,
|
| 183104 |
+
"step": 65820
|
| 183105 |
+
},
|
| 183106 |
+
{
|
| 183107 |
+
"epoch": 529.96,
|
| 183108 |
+
"learning_rate": 8.961538461538462e-06,
|
| 183109 |
+
"loss": 0.4441,
|
| 183110 |
+
"step": 65825
|
| 183111 |
+
},
|
| 183112 |
+
{
|
| 183113 |
+
"epoch": 530.0,
|
| 183114 |
+
"eval_loss": 0.38648757338523865,
|
| 183115 |
+
"eval_runtime": 40.1418,
|
| 183116 |
+
"eval_samples_per_second": 20.826,
|
| 183117 |
+
"eval_steps_per_second": 0.673,
|
| 183118 |
+
"eval_wer": 0.19017903573285788,
|
| 183119 |
+
"step": 65829
|
| 183120 |
+
},
|
| 183121 |
+
{
|
| 183122 |
+
"epoch": 530.01,
|
| 183123 |
+
"learning_rate": 8.961458333333334e-06,
|
| 183124 |
+
"loss": 0.4862,
|
| 183125 |
+
"step": 65830
|
| 183126 |
+
},
|
| 183127 |
+
{
|
| 183128 |
+
"epoch": 530.05,
|
| 183129 |
+
"learning_rate": 8.961378205128207e-06,
|
| 183130 |
+
"loss": 0.4615,
|
| 183131 |
+
"step": 65835
|
| 183132 |
+
},
|
| 183133 |
+
{
|
| 183134 |
+
"epoch": 530.09,
|
| 183135 |
+
"learning_rate": 8.961298076923076e-06,
|
| 183136 |
+
"loss": 0.3626,
|
| 183137 |
+
"step": 65840
|
| 183138 |
+
},
|
| 183139 |
+
{
|
| 183140 |
+
"epoch": 530.13,
|
| 183141 |
+
"learning_rate": 8.96121794871795e-06,
|
| 183142 |
+
"loss": 0.41,
|
| 183143 |
+
"step": 65845
|
| 183144 |
+
},
|
| 183145 |
+
{
|
| 183146 |
+
"epoch": 530.17,
|
| 183147 |
+
"learning_rate": 8.961137820512821e-06,
|
| 183148 |
+
"loss": 0.5743,
|
| 183149 |
+
"step": 65850
|
| 183150 |
+
},
|
| 183151 |
+
{
|
| 183152 |
+
"epoch": 530.21,
|
| 183153 |
+
"learning_rate": 8.961057692307692e-06,
|
| 183154 |
+
"loss": 1.2365,
|
| 183155 |
+
"step": 65855
|
| 183156 |
+
},
|
| 183157 |
+
{
|
| 183158 |
+
"epoch": 530.25,
|
| 183159 |
+
"learning_rate": 8.960977564102565e-06,
|
| 183160 |
+
"loss": 0.3229,
|
| 183161 |
+
"step": 65860
|
| 183162 |
+
},
|
| 183163 |
+
{
|
| 183164 |
+
"epoch": 530.29,
|
| 183165 |
+
"learning_rate": 8.960897435897437e-06,
|
| 183166 |
+
"loss": 0.3577,
|
| 183167 |
+
"step": 65865
|
| 183168 |
+
},
|
| 183169 |
+
{
|
| 183170 |
+
"epoch": 530.33,
|
| 183171 |
+
"learning_rate": 8.960817307692308e-06,
|
| 183172 |
+
"loss": 0.3767,
|
| 183173 |
+
"step": 65870
|
| 183174 |
+
},
|
| 183175 |
+
{
|
| 183176 |
+
"epoch": 530.37,
|
| 183177 |
+
"learning_rate": 8.96073717948718e-06,
|
| 183178 |
+
"loss": 0.4945,
|
| 183179 |
+
"step": 65875
|
| 183180 |
+
},
|
| 183181 |
+
{
|
| 183182 |
+
"epoch": 530.41,
|
| 183183 |
+
"learning_rate": 8.960657051282052e-06,
|
| 183184 |
+
"loss": 1.2582,
|
| 183185 |
+
"step": 65880
|
| 183186 |
+
},
|
| 183187 |
+
{
|
| 183188 |
+
"epoch": 530.45,
|
| 183189 |
+
"learning_rate": 8.960576923076924e-06,
|
| 183190 |
+
"loss": 0.3046,
|
| 183191 |
+
"step": 65885
|
| 183192 |
+
},
|
| 183193 |
+
{
|
| 183194 |
+
"epoch": 530.49,
|
| 183195 |
+
"learning_rate": 8.960496794871795e-06,
|
| 183196 |
+
"loss": 0.2987,
|
| 183197 |
+
"step": 65890
|
| 183198 |
+
},
|
| 183199 |
+
{
|
| 183200 |
+
"epoch": 530.53,
|
| 183201 |
+
"learning_rate": 8.960416666666666e-06,
|
| 183202 |
+
"loss": 0.3325,
|
| 183203 |
+
"step": 65895
|
| 183204 |
+
},
|
| 183205 |
+
{
|
| 183206 |
+
"epoch": 530.57,
|
| 183207 |
+
"learning_rate": 8.96033653846154e-06,
|
| 183208 |
+
"loss": 0.5671,
|
| 183209 |
+
"step": 65900
|
| 183210 |
+
},
|
| 183211 |
+
{
|
| 183212 |
+
"epoch": 530.61,
|
| 183213 |
+
"learning_rate": 8.960256410256411e-06,
|
| 183214 |
+
"loss": 1.1719,
|
| 183215 |
+
"step": 65905
|
| 183216 |
+
},
|
| 183217 |
+
{
|
| 183218 |
+
"epoch": 530.65,
|
| 183219 |
+
"learning_rate": 8.960176282051282e-06,
|
| 183220 |
+
"loss": 0.3255,
|
| 183221 |
+
"step": 65910
|
| 183222 |
+
},
|
| 183223 |
+
{
|
| 183224 |
+
"epoch": 530.69,
|
| 183225 |
+
"learning_rate": 8.960096153846155e-06,
|
| 183226 |
+
"loss": 0.3273,
|
| 183227 |
+
"step": 65915
|
| 183228 |
+
},
|
| 183229 |
+
{
|
| 183230 |
+
"epoch": 530.73,
|
| 183231 |
+
"learning_rate": 8.960016025641027e-06,
|
| 183232 |
+
"loss": 0.4071,
|
| 183233 |
+
"step": 65920
|
| 183234 |
+
},
|
| 183235 |
+
{
|
| 183236 |
+
"epoch": 530.77,
|
| 183237 |
+
"learning_rate": 8.959935897435898e-06,
|
| 183238 |
+
"loss": 0.4977,
|
| 183239 |
+
"step": 65925
|
| 183240 |
+
},
|
| 183241 |
+
{
|
| 183242 |
+
"epoch": 530.81,
|
| 183243 |
+
"learning_rate": 8.95985576923077e-06,
|
| 183244 |
+
"loss": 1.1886,
|
| 183245 |
+
"step": 65930
|
| 183246 |
+
},
|
| 183247 |
+
{
|
| 183248 |
+
"epoch": 530.85,
|
| 183249 |
+
"learning_rate": 8.959775641025642e-06,
|
| 183250 |
+
"loss": 0.3619,
|
| 183251 |
+
"step": 65935
|
| 183252 |
+
},
|
| 183253 |
+
{
|
| 183254 |
+
"epoch": 530.89,
|
| 183255 |
+
"learning_rate": 8.959695512820514e-06,
|
| 183256 |
+
"loss": 0.296,
|
| 183257 |
+
"step": 65940
|
| 183258 |
+
},
|
| 183259 |
+
{
|
| 183260 |
+
"epoch": 530.93,
|
| 183261 |
+
"learning_rate": 8.959615384615385e-06,
|
| 183262 |
+
"loss": 0.3815,
|
| 183263 |
+
"step": 65945
|
| 183264 |
+
},
|
| 183265 |
+
{
|
| 183266 |
+
"epoch": 530.97,
|
| 183267 |
+
"learning_rate": 8.959535256410257e-06,
|
| 183268 |
+
"loss": 0.5965,
|
| 183269 |
+
"step": 65950
|
| 183270 |
+
},
|
| 183271 |
+
{
|
| 183272 |
+
"epoch": 531.0,
|
| 183273 |
+
"eval_loss": 0.4443589448928833,
|
| 183274 |
+
"eval_runtime": 40.3659,
|
| 183275 |
+
"eval_samples_per_second": 20.711,
|
| 183276 |
+
"eval_steps_per_second": 0.669,
|
| 183277 |
+
"eval_wer": 0.2028847561868502,
|
| 183278 |
+
"step": 65953
|
| 183279 |
+
},
|
| 183280 |
+
{
|
| 183281 |
+
"epoch": 527.02,
|
| 183282 |
+
"learning_rate": 8.95945512820513e-06,
|
| 183283 |
+
"loss": 0.4132,
|
| 183284 |
+
"step": 65955
|
| 183285 |
+
},
|
| 183286 |
+
{
|
| 183287 |
+
"epoch": 527.06,
|
| 183288 |
+
"learning_rate": 8.959375000000001e-06,
|
| 183289 |
+
"loss": 0.3103,
|
| 183290 |
+
"step": 65960
|
| 183291 |
+
},
|
| 183292 |
+
{
|
| 183293 |
+
"epoch": 527.1,
|
| 183294 |
+
"learning_rate": 8.959294871794872e-06,
|
| 183295 |
+
"loss": 0.3276,
|
| 183296 |
+
"step": 65965
|
| 183297 |
+
},
|
| 183298 |
+
{
|
| 183299 |
+
"epoch": 527.14,
|
| 183300 |
+
"learning_rate": 8.959214743589745e-06,
|
| 183301 |
+
"loss": 0.4584,
|
| 183302 |
+
"step": 65970
|
| 183303 |
+
},
|
| 183304 |
+
{
|
| 183305 |
+
"epoch": 527.18,
|
| 183306 |
+
"learning_rate": 8.959134615384617e-06,
|
| 183307 |
+
"loss": 0.6109,
|
| 183308 |
+
"step": 65975
|
| 183309 |
+
},
|
| 183310 |
+
{
|
| 183311 |
+
"epoch": 527.22,
|
| 183312 |
+
"learning_rate": 8.959054487179488e-06,
|
| 183313 |
+
"loss": 1.1637,
|
| 183314 |
+
"step": 65980
|
| 183315 |
+
},
|
| 183316 |
+
{
|
| 183317 |
+
"epoch": 527.26,
|
| 183318 |
+
"learning_rate": 8.95897435897436e-06,
|
| 183319 |
+
"loss": 0.2906,
|
| 183320 |
+
"step": 65985
|
| 183321 |
+
},
|
| 183322 |
+
{
|
| 183323 |
+
"epoch": 527.3,
|
| 183324 |
+
"learning_rate": 8.958894230769232e-06,
|
| 183325 |
+
"loss": 0.3802,
|
| 183326 |
+
"step": 65990
|
| 183327 |
+
},
|
| 183328 |
+
{
|
| 183329 |
+
"epoch": 527.34,
|
| 183330 |
+
"learning_rate": 8.958814102564102e-06,
|
| 183331 |
+
"loss": 0.3634,
|
| 183332 |
+
"step": 65995
|
| 183333 |
+
},
|
| 183334 |
+
{
|
| 183335 |
+
"epoch": 527.38,
|
| 183336 |
+
"learning_rate": 8.958733974358975e-06,
|
| 183337 |
+
"loss": 0.6906,
|
| 183338 |
+
"step": 66000
|
| 183339 |
+
},
|
| 183340 |
+
{
|
| 183341 |
+
"epoch": 527.42,
|
| 183342 |
+
"learning_rate": 8.958653846153847e-06,
|
| 183343 |
+
"loss": 1.0945,
|
| 183344 |
+
"step": 66005
|
| 183345 |
+
},
|
| 183346 |
+
{
|
| 183347 |
+
"epoch": 527.46,
|
| 183348 |
+
"learning_rate": 8.958573717948718e-06,
|
| 183349 |
+
"loss": 0.3012,
|
| 183350 |
+
"step": 66010
|
| 183351 |
+
},
|
| 183352 |
+
{
|
| 183353 |
+
"epoch": 527.5,
|
| 183354 |
+
"learning_rate": 8.958493589743591e-06,
|
| 183355 |
+
"loss": 0.3196,
|
| 183356 |
+
"step": 66015
|
| 183357 |
+
},
|
| 183358 |
+
{
|
| 183359 |
+
"epoch": 527.54,
|
| 183360 |
+
"learning_rate": 8.958413461538462e-06,
|
| 183361 |
+
"loss": 0.3637,
|
| 183362 |
+
"step": 66020
|
| 183363 |
+
},
|
| 183364 |
+
{
|
| 183365 |
+
"epoch": 527.58,
|
| 183366 |
+
"learning_rate": 8.958333333333334e-06,
|
| 183367 |
+
"loss": 0.6842,
|
| 183368 |
+
"step": 66025
|
| 183369 |
+
},
|
| 183370 |
+
{
|
| 183371 |
+
"epoch": 527.62,
|
| 183372 |
+
"learning_rate": 8.958253205128205e-06,
|
| 183373 |
+
"loss": 0.9815,
|
| 183374 |
+
"step": 66030
|
| 183375 |
+
},
|
| 183376 |
+
{
|
| 183377 |
+
"epoch": 527.66,
|
| 183378 |
+
"learning_rate": 8.958173076923078e-06,
|
| 183379 |
+
"loss": 0.2745,
|
| 183380 |
+
"step": 66035
|
| 183381 |
+
},
|
| 183382 |
+
{
|
| 183383 |
+
"epoch": 527.7,
|
| 183384 |
+
"learning_rate": 8.95809294871795e-06,
|
| 183385 |
+
"loss": 0.4007,
|
| 183386 |
+
"step": 66040
|
| 183387 |
+
},
|
| 183388 |
+
{
|
| 183389 |
+
"epoch": 527.74,
|
| 183390 |
+
"learning_rate": 8.95801282051282e-06,
|
| 183391 |
+
"loss": 0.3229,
|
| 183392 |
+
"step": 66045
|
| 183393 |
+
},
|
| 183394 |
+
{
|
| 183395 |
+
"epoch": 527.78,
|
| 183396 |
+
"learning_rate": 8.957932692307692e-06,
|
| 183397 |
+
"loss": 0.7089,
|
| 183398 |
+
"step": 66050
|
| 183399 |
+
},
|
| 183400 |
+
{
|
| 183401 |
+
"epoch": 527.82,
|
| 183402 |
+
"learning_rate": 8.957852564102565e-06,
|
| 183403 |
+
"loss": 1.0354,
|
| 183404 |
+
"step": 66055
|
| 183405 |
+
},
|
| 183406 |
+
{
|
| 183407 |
+
"epoch": 527.86,
|
| 183408 |
+
"learning_rate": 8.957772435897437e-06,
|
| 183409 |
+
"loss": 0.3453,
|
| 183410 |
+
"step": 66060
|
| 183411 |
+
},
|
| 183412 |
+
{
|
| 183413 |
+
"epoch": 527.9,
|
| 183414 |
+
"learning_rate": 8.957692307692308e-06,
|
| 183415 |
+
"loss": 0.2743,
|
| 183416 |
+
"step": 66065
|
| 183417 |
+
},
|
| 183418 |
+
{
|
| 183419 |
+
"epoch": 527.94,
|
| 183420 |
+
"learning_rate": 8.957612179487181e-06,
|
| 183421 |
+
"loss": 0.4068,
|
| 183422 |
+
"step": 66070
|
| 183423 |
+
},
|
| 183424 |
+
{
|
| 183425 |
+
"epoch": 527.98,
|
| 183426 |
+
"learning_rate": 8.957532051282052e-06,
|
| 183427 |
+
"loss": 0.7323,
|
| 183428 |
+
"step": 66075
|
| 183429 |
+
},
|
| 183430 |
+
{
|
| 183431 |
+
"epoch": 528.0,
|
| 183432 |
+
"eval_loss": 0.41770240664482117,
|
| 183433 |
+
"eval_runtime": 39.8837,
|
| 183434 |
+
"eval_samples_per_second": 20.986,
|
| 183435 |
+
"eval_steps_per_second": 0.677,
|
| 183436 |
+
"eval_wer": 0.19590305828043855,
|
| 183437 |
+
"step": 66078
|
| 183438 |
+
},
|
| 183439 |
+
{
|
| 183440 |
+
"epoch": 532.02,
|
| 183441 |
+
"learning_rate": 8.957451923076924e-06,
|
| 183442 |
+
"loss": 0.3931,
|
| 183443 |
+
"step": 66080
|
| 183444 |
+
},
|
| 183445 |
+
{
|
| 183446 |
+
"epoch": 532.06,
|
| 183447 |
+
"learning_rate": 8.957371794871795e-06,
|
| 183448 |
+
"loss": 0.3411,
|
| 183449 |
+
"step": 66085
|
| 183450 |
+
},
|
| 183451 |
+
{
|
| 183452 |
+
"epoch": 532.1,
|
| 183453 |
+
"learning_rate": 8.957291666666668e-06,
|
| 183454 |
+
"loss": 0.3569,
|
| 183455 |
+
"step": 66090
|
| 183456 |
+
},
|
| 183457 |
+
{
|
| 183458 |
+
"epoch": 532.14,
|
| 183459 |
+
"learning_rate": 8.95721153846154e-06,
|
| 183460 |
+
"loss": 0.3359,
|
| 183461 |
+
"step": 66095
|
| 183462 |
+
},
|
| 183463 |
+
{
|
| 183464 |
+
"epoch": 532.18,
|
| 183465 |
+
"learning_rate": 8.95713141025641e-06,
|
| 183466 |
+
"loss": 0.7029,
|
| 183467 |
+
"step": 66100
|
| 183468 |
+
},
|
| 183469 |
+
{
|
| 183470 |
+
"epoch": 532.22,
|
| 183471 |
+
"learning_rate": 8.957051282051282e-06,
|
| 183472 |
+
"loss": 1.1022,
|
| 183473 |
+
"step": 66105
|
| 183474 |
+
},
|
| 183475 |
+
{
|
| 183476 |
+
"epoch": 532.26,
|
| 183477 |
+
"learning_rate": 8.956971153846155e-06,
|
| 183478 |
+
"loss": 0.3636,
|
| 183479 |
+
"step": 66110
|
| 183480 |
+
},
|
| 183481 |
+
{
|
| 183482 |
+
"epoch": 532.3,
|
| 183483 |
+
"learning_rate": 8.956891025641027e-06,
|
| 183484 |
+
"loss": 0.3155,
|
| 183485 |
+
"step": 66115
|
| 183486 |
+
},
|
| 183487 |
+
{
|
| 183488 |
+
"epoch": 532.34,
|
| 183489 |
+
"learning_rate": 8.956810897435898e-06,
|
| 183490 |
+
"loss": 0.4378,
|
| 183491 |
+
"step": 66120
|
| 183492 |
+
},
|
| 183493 |
+
{
|
| 183494 |
+
"epoch": 532.38,
|
| 183495 |
+
"learning_rate": 8.956730769230771e-06,
|
| 183496 |
+
"loss": 0.6137,
|
| 183497 |
+
"step": 66125
|
| 183498 |
+
},
|
| 183499 |
+
{
|
| 183500 |
+
"epoch": 532.42,
|
| 183501 |
+
"learning_rate": 8.95665064102564e-06,
|
| 183502 |
+
"loss": 1.0293,
|
| 183503 |
+
"step": 66130
|
| 183504 |
+
},
|
| 183505 |
+
{
|
| 183506 |
+
"epoch": 532.46,
|
| 183507 |
+
"learning_rate": 8.956570512820514e-06,
|
| 183508 |
+
"loss": 0.3335,
|
| 183509 |
+
"step": 66135
|
| 183510 |
+
},
|
| 183511 |
+
{
|
| 183512 |
+
"epoch": 532.5,
|
| 183513 |
+
"learning_rate": 8.956490384615385e-06,
|
| 183514 |
+
"loss": 0.3447,
|
| 183515 |
+
"step": 66140
|
| 183516 |
+
},
|
| 183517 |
+
{
|
| 183518 |
+
"epoch": 532.54,
|
| 183519 |
+
"learning_rate": 8.956410256410258e-06,
|
| 183520 |
+
"loss": 0.3888,
|
| 183521 |
+
"step": 66145
|
| 183522 |
+
},
|
| 183523 |
+
{
|
| 183524 |
+
"epoch": 532.58,
|
| 183525 |
+
"learning_rate": 8.956330128205128e-06,
|
| 183526 |
+
"loss": 0.6986,
|
| 183527 |
+
"step": 66150
|
| 183528 |
+
},
|
| 183529 |
+
{
|
| 183530 |
+
"epoch": 532.62,
|
| 183531 |
+
"learning_rate": 8.956250000000001e-06,
|
| 183532 |
+
"loss": 0.9883,
|
| 183533 |
+
"step": 66155
|
| 183534 |
+
},
|
| 183535 |
+
{
|
| 183536 |
+
"epoch": 532.66,
|
| 183537 |
+
"learning_rate": 8.956169871794874e-06,
|
| 183538 |
+
"loss": 0.3206,
|
| 183539 |
+
"step": 66160
|
| 183540 |
+
},
|
| 183541 |
+
{
|
| 183542 |
+
"epoch": 532.7,
|
| 183543 |
+
"learning_rate": 8.956089743589744e-06,
|
| 183544 |
+
"loss": 0.3433,
|
| 183545 |
+
"step": 66165
|
| 183546 |
+
},
|
| 183547 |
+
{
|
| 183548 |
+
"epoch": 532.74,
|
| 183549 |
+
"learning_rate": 8.956009615384617e-06,
|
| 183550 |
+
"loss": 0.4522,
|
| 183551 |
+
"step": 66170
|
| 183552 |
+
},
|
| 183553 |
+
{
|
| 183554 |
+
"epoch": 532.78,
|
| 183555 |
+
"learning_rate": 8.955929487179488e-06,
|
| 183556 |
+
"loss": 0.7144,
|
| 183557 |
+
"step": 66175
|
| 183558 |
+
},
|
| 183559 |
+
{
|
| 183560 |
+
"epoch": 532.82,
|
| 183561 |
+
"learning_rate": 8.95584935897436e-06,
|
| 183562 |
+
"loss": 1.0969,
|
| 183563 |
+
"step": 66180
|
| 183564 |
+
},
|
| 183565 |
+
{
|
| 183566 |
+
"epoch": 532.86,
|
| 183567 |
+
"learning_rate": 8.95576923076923e-06,
|
| 183568 |
+
"loss": 0.2837,
|
| 183569 |
+
"step": 66185
|
| 183570 |
+
},
|
| 183571 |
+
{
|
| 183572 |
+
"epoch": 532.9,
|
| 183573 |
+
"learning_rate": 8.955689102564104e-06,
|
| 183574 |
+
"loss": 0.2845,
|
| 183575 |
+
"step": 66190
|
| 183576 |
+
},
|
| 183577 |
+
{
|
| 183578 |
+
"epoch": 532.94,
|
| 183579 |
+
"learning_rate": 8.955608974358975e-06,
|
| 183580 |
+
"loss": 0.4446,
|
| 183581 |
+
"step": 66195
|
| 183582 |
+
},
|
| 183583 |
+
{
|
| 183584 |
+
"epoch": 532.98,
|
| 183585 |
+
"learning_rate": 8.955528846153846e-06,
|
| 183586 |
+
"loss": 0.6778,
|
| 183587 |
+
"step": 66200
|
| 183588 |
+
},
|
| 183589 |
+
{
|
| 183590 |
+
"epoch": 533.0,
|
| 183591 |
+
"eval_loss": 0.4082199037075043,
|
| 183592 |
+
"eval_runtime": 39.496,
|
| 183593 |
+
"eval_samples_per_second": 21.192,
|
| 183594 |
+
"eval_steps_per_second": 0.684,
|
| 183595 |
+
"eval_wer": 0.190224570673712,
|
| 183596 |
+
"step": 66202
|
| 183597 |
+
},
|
| 183598 |
+
{
|
| 183599 |
+
"epoch": 533.02,
|
| 183600 |
+
"learning_rate": 8.955448717948718e-06,
|
| 183601 |
+
"loss": 0.3439,
|
| 183602 |
+
"step": 66205
|
| 183603 |
+
},
|
| 183604 |
+
{
|
| 183605 |
+
"epoch": 533.06,
|
| 183606 |
+
"learning_rate": 8.955368589743591e-06,
|
| 183607 |
+
"loss": 0.2805,
|
| 183608 |
+
"step": 66210
|
| 183609 |
+
},
|
| 183610 |
+
{
|
| 183611 |
+
"epoch": 533.1,
|
| 183612 |
+
"learning_rate": 8.955288461538462e-06,
|
| 183613 |
+
"loss": 0.3489,
|
| 183614 |
+
"step": 66215
|
| 183615 |
+
},
|
| 183616 |
+
{
|
| 183617 |
+
"epoch": 533.14,
|
| 183618 |
+
"learning_rate": 8.955208333333334e-06,
|
| 183619 |
+
"loss": 0.3674,
|
| 183620 |
+
"step": 66220
|
| 183621 |
+
},
|
| 183622 |
+
{
|
| 183623 |
+
"epoch": 533.18,
|
| 183624 |
+
"learning_rate": 8.955128205128207e-06,
|
| 183625 |
+
"loss": 0.7902,
|
| 183626 |
+
"step": 66225
|
| 183627 |
+
},
|
| 183628 |
+
{
|
| 183629 |
+
"epoch": 533.22,
|
| 183630 |
+
"learning_rate": 8.955048076923078e-06,
|
| 183631 |
+
"loss": 0.8694,
|
| 183632 |
+
"step": 66230
|
| 183633 |
+
},
|
| 183634 |
+
{
|
| 183635 |
+
"epoch": 533.27,
|
| 183636 |
+
"learning_rate": 8.95496794871795e-06,
|
| 183637 |
+
"loss": 0.3494,
|
| 183638 |
+
"step": 66235
|
| 183639 |
+
},
|
| 183640 |
+
{
|
| 183641 |
+
"epoch": 533.31,
|
| 183642 |
+
"learning_rate": 8.95488782051282e-06,
|
| 183643 |
+
"loss": 0.3249,
|
| 183644 |
+
"step": 66240
|
| 183645 |
+
},
|
| 183646 |
+
{
|
| 183647 |
+
"epoch": 533.35,
|
| 183648 |
+
"learning_rate": 8.954807692307694e-06,
|
| 183649 |
+
"loss": 0.3455,
|
| 183650 |
+
"step": 66245
|
| 183651 |
+
},
|
| 183652 |
+
{
|
| 183653 |
+
"epoch": 533.39,
|
| 183654 |
+
"learning_rate": 8.954727564102565e-06,
|
| 183655 |
+
"loss": 0.7674,
|
| 183656 |
+
"step": 66250
|
| 183657 |
+
},
|
| 183658 |
+
{
|
| 183659 |
+
"epoch": 533.43,
|
| 183660 |
+
"learning_rate": 8.954647435897436e-06,
|
| 183661 |
+
"loss": 0.9707,
|
| 183662 |
+
"step": 66255
|
| 183663 |
+
},
|
| 183664 |
+
{
|
| 183665 |
+
"epoch": 533.47,
|
| 183666 |
+
"learning_rate": 8.95456730769231e-06,
|
| 183667 |
+
"loss": 0.3004,
|
| 183668 |
+
"step": 66260
|
| 183669 |
+
},
|
| 183670 |
+
{
|
| 183671 |
+
"epoch": 533.51,
|
| 183672 |
+
"learning_rate": 8.954487179487181e-06,
|
| 183673 |
+
"loss": 0.2967,
|
| 183674 |
+
"step": 66265
|
| 183675 |
+
},
|
| 183676 |
+
{
|
| 183677 |
+
"epoch": 533.55,
|
| 183678 |
+
"learning_rate": 8.954407051282052e-06,
|
| 183679 |
+
"loss": 0.412,
|
| 183680 |
+
"step": 66270
|
| 183681 |
+
},
|
| 183682 |
+
{
|
| 183683 |
+
"epoch": 533.59,
|
| 183684 |
+
"learning_rate": 8.954326923076924e-06,
|
| 183685 |
+
"loss": 0.7517,
|
| 183686 |
+
"step": 66275
|
| 183687 |
+
},
|
| 183688 |
+
{
|
| 183689 |
+
"epoch": 533.63,
|
| 183690 |
+
"learning_rate": 8.954246794871797e-06,
|
| 183691 |
+
"loss": 0.7993,
|
| 183692 |
+
"step": 66280
|
| 183693 |
+
},
|
| 183694 |
+
{
|
| 183695 |
+
"epoch": 533.67,
|
| 183696 |
+
"learning_rate": 8.954166666666666e-06,
|
| 183697 |
+
"loss": 0.3173,
|
| 183698 |
+
"step": 66285
|
| 183699 |
+
},
|
| 183700 |
+
{
|
| 183701 |
+
"epoch": 533.71,
|
| 183702 |
+
"learning_rate": 8.95408653846154e-06,
|
| 183703 |
+
"loss": 0.3012,
|
| 183704 |
+
"step": 66290
|
| 183705 |
+
},
|
| 183706 |
+
{
|
| 183707 |
+
"epoch": 533.75,
|
| 183708 |
+
"learning_rate": 8.95400641025641e-06,
|
| 183709 |
+
"loss": 0.4471,
|
| 183710 |
+
"step": 66295
|
| 183711 |
+
},
|
| 183712 |
+
{
|
| 183713 |
+
"epoch": 533.79,
|
| 183714 |
+
"learning_rate": 8.953926282051282e-06,
|
| 183715 |
+
"loss": 0.7729,
|
| 183716 |
+
"step": 66300
|
| 183717 |
+
},
|
| 183718 |
+
{
|
| 183719 |
+
"epoch": 533.83,
|
| 183720 |
+
"learning_rate": 8.953846153846153e-06,
|
| 183721 |
+
"loss": 0.8857,
|
| 183722 |
+
"step": 66305
|
| 183723 |
+
},
|
| 183724 |
+
{
|
| 183725 |
+
"epoch": 533.87,
|
| 183726 |
+
"learning_rate": 8.953766025641027e-06,
|
| 183727 |
+
"loss": 0.3209,
|
| 183728 |
+
"step": 66310
|
| 183729 |
+
},
|
| 183730 |
+
{
|
| 183731 |
+
"epoch": 533.91,
|
| 183732 |
+
"learning_rate": 8.953685897435898e-06,
|
| 183733 |
+
"loss": 0.3262,
|
| 183734 |
+
"step": 66315
|
| 183735 |
+
},
|
| 183736 |
+
{
|
| 183737 |
+
"epoch": 533.95,
|
| 183738 |
+
"learning_rate": 8.95360576923077e-06,
|
| 183739 |
+
"loss": 0.4383,
|
| 183740 |
+
"step": 66320
|
| 183741 |
+
},
|
| 183742 |
+
{
|
| 183743 |
+
"epoch": 533.99,
|
| 183744 |
+
"learning_rate": 8.953525641025642e-06,
|
| 183745 |
+
"loss": 0.9138,
|
| 183746 |
+
"step": 66325
|
| 183747 |
+
},
|
| 183748 |
+
{
|
| 183749 |
+
"epoch": 534.0,
|
| 183750 |
+
"eval_loss": 0.3869166672229767,
|
| 183751 |
+
"eval_runtime": 39.8868,
|
| 183752 |
+
"eval_samples_per_second": 20.984,
|
| 183753 |
+
"eval_steps_per_second": 0.677,
|
| 183754 |
+
"eval_wer": 0.19705408178220724,
|
| 183755 |
+
"step": 66326
|
| 183756 |
}
|
| 183757 |
],
|
| 183758 |
+
"max_steps": 620000,
|
| 183759 |
"num_train_epochs": 5000,
|
| 183760 |
+
"total_flos": 1.866428893671497e+20,
|
| 183761 |
"trial_name": null,
|
| 183762 |
"trial_params": null
|
| 183763 |
}
|
model-bin/finetune/base/{checkpoint-65705 → checkpoint-66326}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629827645.253879/events.out.tfevents.1629827645.c435e1c5ee04.920.191
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:872c038887a161a464f61dfc1bd7fb3419cf63fe36b08086a4be082eeea4b456
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629828281.4169908/events.out.tfevents.1629828281.c435e1c5ee04.920.193
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:38cc1698243689a4e07596c20d358915a55461f481aa14dbf3a6aa3f68d4412b
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629828923.8657014/events.out.tfevents.1629828923.c435e1c5ee04.920.195
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1c2834245a6b5685d069560b77886515d3846cd31764fafd250f4c67f885ec68
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629829555.331946/events.out.tfevents.1629829555.c435e1c5ee04.920.197
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f3f3cf89b5faea2d9873af60bceb7caf01fd9fb18a1b596c30476ddd252b9758
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629830197.8430533/events.out.tfevents.1629830197.c435e1c5ee04.920.199
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:935bbeba559dcbc45d49319c829d68c04f8a629414eb471bbf5108379f88da1c
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629827645.c435e1c5ee04.920.190
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9c9db1b9680624628069187065587324bf6e43d3dec54350e4e84aecc3bc0f17
|
| 3 |
+
size 8462
|
model-bin/finetune/base/log/events.out.tfevents.1629828281.c435e1c5ee04.920.192
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c89f6ad475e5e639d42a3763c7212290b119596b392a58764c8515aacb20a05a
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629828923.c435e1c5ee04.920.194
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:83f9437a06d2c9b772fb5c7b77a362493630d69a53034dc7802d1d022a6d1128
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629829555.c435e1c5ee04.920.196
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2eb0097ac856f24766d524b13034bca29928832090f5611adb8c6adcd0bced64
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629830197.c435e1c5ee04.920.198
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b3ae473a659ee7593575954eb1040e59cf7ccd057b8594dc6fece1639d9365af
|
| 3 |
+
size 8622
|