"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-68818 → checkpoint-69440}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-68818 → checkpoint-69440}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-68818 → checkpoint-69440}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-68818 → checkpoint-69440}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-68818 → checkpoint-69440}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-68818 → checkpoint-69440}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-68818 → checkpoint-69440}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-68818 → checkpoint-69440}/trainer_state.json +798 -3
- model-bin/finetune/base/{checkpoint-68818 → checkpoint-69440}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629843978.3964214/events.out.tfevents.1629843978.c435e1c5ee04.920.241 +3 -0
- model-bin/finetune/base/log/1629844632.7215848/events.out.tfevents.1629844632.c435e1c5ee04.920.243 +3 -0
- model-bin/finetune/base/log/1629845266.9609022/events.out.tfevents.1629845266.c435e1c5ee04.920.245 +3 -0
- model-bin/finetune/base/log/1629845904.6603022/events.out.tfevents.1629845904.c435e1c5ee04.920.247 +3 -0
- model-bin/finetune/base/log/1629846554.1512308/events.out.tfevents.1629846554.c435e1c5ee04.920.249 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629843978.c435e1c5ee04.920.240 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629844632.c435e1c5ee04.920.242 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629845266.c435e1c5ee04.920.244 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629845904.c435e1c5ee04.920.246 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629846554.c435e1c5ee04.920.248 +3 -0
model-bin/finetune/base/{checkpoint-68818 → checkpoint-69440}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-68818 → checkpoint-69440}/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:f97b4fc6c219db9d3afa4880cdcccd79399ebe284feffb6e2efe98ae3ed6a268
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-68818 → checkpoint-69440}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-68818 → checkpoint-69440}/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 sha256:8f5e391fa4f938d866a965b45f82eebde6c0b705b2466903a9e82f6e9f4212c2
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-68818 → checkpoint-69440}/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:bb87bc92aafb41622424ca10817f949491f3f5e4b525515986b7a6fc2e6b3641
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-68818 → checkpoint-69440}/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 sha256:2cc9397ac385b0f2da90d122ad975747c5aef6a7097f14cfe46a57e0bc037043
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-68818 → checkpoint-69440}/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:dbc1e6d224edc3cba35b1ab2f1a230355e5f6ebaabb30a9e0c807f2453ae39b9
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-68818 → checkpoint-69440}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.18525332578545145,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-68444",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -186921,11 +186921,806 @@
|
|
| 186921 |
"eval_steps_per_second": 0.688,
|
| 186922 |
"eval_wer": 0.1943950177935943,
|
| 186923 |
"step": 68818
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 186924 |
}
|
| 186925 |
],
|
| 186926 |
"max_steps": 625000,
|
| 186927 |
"num_train_epochs": 5000,
|
| 186928 |
-
"total_flos": 1.
|
| 186929 |
"trial_name": null,
|
| 186930 |
"trial_params": null
|
| 186931 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.18525332578545145,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-68444",
|
| 4 |
+
"epoch": 555.0,
|
| 5 |
+
"global_step": 69440,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 186921 |
"eval_steps_per_second": 0.688,
|
| 186922 |
"eval_wer": 0.1943950177935943,
|
| 186923 |
"step": 68818
|
| 186924 |
+
},
|
| 186925 |
+
{
|
| 186926 |
+
"epoch": 550.02,
|
| 186927 |
+
"learning_rate": 8.913557692307693e-06,
|
| 186928 |
+
"loss": 0.4203,
|
| 186929 |
+
"step": 68820
|
| 186930 |
+
},
|
| 186931 |
+
{
|
| 186932 |
+
"epoch": 550.06,
|
| 186933 |
+
"learning_rate": 8.913477564102564e-06,
|
| 186934 |
+
"loss": 0.3611,
|
| 186935 |
+
"step": 68825
|
| 186936 |
+
},
|
| 186937 |
+
{
|
| 186938 |
+
"epoch": 550.1,
|
| 186939 |
+
"learning_rate": 8.913397435897436e-06,
|
| 186940 |
+
"loss": 0.343,
|
| 186941 |
+
"step": 68830
|
| 186942 |
+
},
|
| 186943 |
+
{
|
| 186944 |
+
"epoch": 550.14,
|
| 186945 |
+
"learning_rate": 8.913317307692309e-06,
|
| 186946 |
+
"loss": 0.3587,
|
| 186947 |
+
"step": 68835
|
| 186948 |
+
},
|
| 186949 |
+
{
|
| 186950 |
+
"epoch": 550.18,
|
| 186951 |
+
"learning_rate": 8.91323717948718e-06,
|
| 186952 |
+
"loss": 0.5963,
|
| 186953 |
+
"step": 68840
|
| 186954 |
+
},
|
| 186955 |
+
{
|
| 186956 |
+
"epoch": 550.22,
|
| 186957 |
+
"learning_rate": 8.913157051282052e-06,
|
| 186958 |
+
"loss": 1.1357,
|
| 186959 |
+
"step": 68845
|
| 186960 |
+
},
|
| 186961 |
+
{
|
| 186962 |
+
"epoch": 550.26,
|
| 186963 |
+
"learning_rate": 8.913076923076925e-06,
|
| 186964 |
+
"loss": 0.2899,
|
| 186965 |
+
"step": 68850
|
| 186966 |
+
},
|
| 186967 |
+
{
|
| 186968 |
+
"epoch": 550.3,
|
| 186969 |
+
"learning_rate": 8.912996794871796e-06,
|
| 186970 |
+
"loss": 0.3763,
|
| 186971 |
+
"step": 68855
|
| 186972 |
+
},
|
| 186973 |
+
{
|
| 186974 |
+
"epoch": 550.34,
|
| 186975 |
+
"learning_rate": 8.912916666666667e-06,
|
| 186976 |
+
"loss": 0.3944,
|
| 186977 |
+
"step": 68860
|
| 186978 |
+
},
|
| 186979 |
+
{
|
| 186980 |
+
"epoch": 550.38,
|
| 186981 |
+
"learning_rate": 8.912836538461539e-06,
|
| 186982 |
+
"loss": 0.7077,
|
| 186983 |
+
"step": 68865
|
| 186984 |
+
},
|
| 186985 |
+
{
|
| 186986 |
+
"epoch": 550.42,
|
| 186987 |
+
"learning_rate": 8.912756410256412e-06,
|
| 186988 |
+
"loss": 1.1302,
|
| 186989 |
+
"step": 68870
|
| 186990 |
+
},
|
| 186991 |
+
{
|
| 186992 |
+
"epoch": 550.46,
|
| 186993 |
+
"learning_rate": 8.912676282051283e-06,
|
| 186994 |
+
"loss": 0.2758,
|
| 186995 |
+
"step": 68875
|
| 186996 |
+
},
|
| 186997 |
+
{
|
| 186998 |
+
"epoch": 550.5,
|
| 186999 |
+
"learning_rate": 8.912596153846155e-06,
|
| 187000 |
+
"loss": 0.3103,
|
| 187001 |
+
"step": 68880
|
| 187002 |
+
},
|
| 187003 |
+
{
|
| 187004 |
+
"epoch": 550.54,
|
| 187005 |
+
"learning_rate": 8.912516025641026e-06,
|
| 187006 |
+
"loss": 0.4578,
|
| 187007 |
+
"step": 68885
|
| 187008 |
+
},
|
| 187009 |
+
{
|
| 187010 |
+
"epoch": 550.58,
|
| 187011 |
+
"learning_rate": 8.912435897435899e-06,
|
| 187012 |
+
"loss": 0.6749,
|
| 187013 |
+
"step": 68890
|
| 187014 |
+
},
|
| 187015 |
+
{
|
| 187016 |
+
"epoch": 550.62,
|
| 187017 |
+
"learning_rate": 8.91235576923077e-06,
|
| 187018 |
+
"loss": 1.0549,
|
| 187019 |
+
"step": 68895
|
| 187020 |
+
},
|
| 187021 |
+
{
|
| 187022 |
+
"epoch": 550.66,
|
| 187023 |
+
"learning_rate": 8.912275641025642e-06,
|
| 187024 |
+
"loss": 0.3558,
|
| 187025 |
+
"step": 68900
|
| 187026 |
+
},
|
| 187027 |
+
{
|
| 187028 |
+
"epoch": 550.7,
|
| 187029 |
+
"learning_rate": 8.912195512820515e-06,
|
| 187030 |
+
"loss": 0.3126,
|
| 187031 |
+
"step": 68905
|
| 187032 |
+
},
|
| 187033 |
+
{
|
| 187034 |
+
"epoch": 550.74,
|
| 187035 |
+
"learning_rate": 8.912115384615384e-06,
|
| 187036 |
+
"loss": 0.4069,
|
| 187037 |
+
"step": 68910
|
| 187038 |
+
},
|
| 187039 |
+
{
|
| 187040 |
+
"epoch": 550.78,
|
| 187041 |
+
"learning_rate": 8.912035256410257e-06,
|
| 187042 |
+
"loss": 0.6448,
|
| 187043 |
+
"step": 68915
|
| 187044 |
+
},
|
| 187045 |
+
{
|
| 187046 |
+
"epoch": 550.82,
|
| 187047 |
+
"learning_rate": 8.911955128205129e-06,
|
| 187048 |
+
"loss": 1.0045,
|
| 187049 |
+
"step": 68920
|
| 187050 |
+
},
|
| 187051 |
+
{
|
| 187052 |
+
"epoch": 550.86,
|
| 187053 |
+
"learning_rate": 8.911875e-06,
|
| 187054 |
+
"loss": 0.3323,
|
| 187055 |
+
"step": 68925
|
| 187056 |
+
},
|
| 187057 |
+
{
|
| 187058 |
+
"epoch": 550.9,
|
| 187059 |
+
"learning_rate": 8.911794871794871e-06,
|
| 187060 |
+
"loss": 0.3236,
|
| 187061 |
+
"step": 68930
|
| 187062 |
+
},
|
| 187063 |
+
{
|
| 187064 |
+
"epoch": 550.94,
|
| 187065 |
+
"learning_rate": 8.911714743589745e-06,
|
| 187066 |
+
"loss": 0.3786,
|
| 187067 |
+
"step": 68935
|
| 187068 |
+
},
|
| 187069 |
+
{
|
| 187070 |
+
"epoch": 550.98,
|
| 187071 |
+
"learning_rate": 8.911634615384616e-06,
|
| 187072 |
+
"loss": 0.6312,
|
| 187073 |
+
"step": 68940
|
| 187074 |
+
},
|
| 187075 |
+
{
|
| 187076 |
+
"epoch": 551.0,
|
| 187077 |
+
"eval_loss": 0.3938581943511963,
|
| 187078 |
+
"eval_runtime": 40.8487,
|
| 187079 |
+
"eval_samples_per_second": 20.515,
|
| 187080 |
+
"eval_steps_per_second": 0.661,
|
| 187081 |
+
"eval_wer": 0.19235865724381626,
|
| 187082 |
+
"step": 68943
|
| 187083 |
+
},
|
| 187084 |
+
{
|
| 187085 |
+
"epoch": 555.02,
|
| 187086 |
+
"learning_rate": 8.911554487179487e-06,
|
| 187087 |
+
"loss": 0.6145,
|
| 187088 |
+
"step": 68945
|
| 187089 |
+
},
|
| 187090 |
+
{
|
| 187091 |
+
"epoch": 555.06,
|
| 187092 |
+
"learning_rate": 8.91147435897436e-06,
|
| 187093 |
+
"loss": 0.3448,
|
| 187094 |
+
"step": 68950
|
| 187095 |
+
},
|
| 187096 |
+
{
|
| 187097 |
+
"epoch": 555.1,
|
| 187098 |
+
"learning_rate": 8.911394230769232e-06,
|
| 187099 |
+
"loss": 0.3279,
|
| 187100 |
+
"step": 68955
|
| 187101 |
+
},
|
| 187102 |
+
{
|
| 187103 |
+
"epoch": 555.14,
|
| 187104 |
+
"learning_rate": 8.911314102564103e-06,
|
| 187105 |
+
"loss": 0.3365,
|
| 187106 |
+
"step": 68960
|
| 187107 |
+
},
|
| 187108 |
+
{
|
| 187109 |
+
"epoch": 555.18,
|
| 187110 |
+
"learning_rate": 8.911233974358974e-06,
|
| 187111 |
+
"loss": 0.7031,
|
| 187112 |
+
"step": 68965
|
| 187113 |
+
},
|
| 187114 |
+
{
|
| 187115 |
+
"epoch": 555.22,
|
| 187116 |
+
"learning_rate": 8.911153846153847e-06,
|
| 187117 |
+
"loss": 1.028,
|
| 187118 |
+
"step": 68970
|
| 187119 |
+
},
|
| 187120 |
+
{
|
| 187121 |
+
"epoch": 555.26,
|
| 187122 |
+
"learning_rate": 8.911073717948719e-06,
|
| 187123 |
+
"loss": 0.3392,
|
| 187124 |
+
"step": 68975
|
| 187125 |
+
},
|
| 187126 |
+
{
|
| 187127 |
+
"epoch": 555.3,
|
| 187128 |
+
"learning_rate": 8.91099358974359e-06,
|
| 187129 |
+
"loss": 0.359,
|
| 187130 |
+
"step": 68980
|
| 187131 |
+
},
|
| 187132 |
+
{
|
| 187133 |
+
"epoch": 555.34,
|
| 187134 |
+
"learning_rate": 8.910913461538462e-06,
|
| 187135 |
+
"loss": 0.3914,
|
| 187136 |
+
"step": 68985
|
| 187137 |
+
},
|
| 187138 |
+
{
|
| 187139 |
+
"epoch": 555.38,
|
| 187140 |
+
"learning_rate": 8.910833333333335e-06,
|
| 187141 |
+
"loss": 0.6924,
|
| 187142 |
+
"step": 68990
|
| 187143 |
+
},
|
| 187144 |
+
{
|
| 187145 |
+
"epoch": 555.42,
|
| 187146 |
+
"learning_rate": 8.910753205128206e-06,
|
| 187147 |
+
"loss": 1.0995,
|
| 187148 |
+
"step": 68995
|
| 187149 |
+
},
|
| 187150 |
+
{
|
| 187151 |
+
"epoch": 555.46,
|
| 187152 |
+
"learning_rate": 8.910673076923077e-06,
|
| 187153 |
+
"loss": 0.3136,
|
| 187154 |
+
"step": 69000
|
| 187155 |
+
},
|
| 187156 |
+
{
|
| 187157 |
+
"epoch": 555.5,
|
| 187158 |
+
"learning_rate": 8.91059294871795e-06,
|
| 187159 |
+
"loss": 0.3497,
|
| 187160 |
+
"step": 69005
|
| 187161 |
+
},
|
| 187162 |
+
{
|
| 187163 |
+
"epoch": 555.54,
|
| 187164 |
+
"learning_rate": 8.910512820512822e-06,
|
| 187165 |
+
"loss": 0.3846,
|
| 187166 |
+
"step": 69010
|
| 187167 |
+
},
|
| 187168 |
+
{
|
| 187169 |
+
"epoch": 555.58,
|
| 187170 |
+
"learning_rate": 8.910432692307693e-06,
|
| 187171 |
+
"loss": 0.6693,
|
| 187172 |
+
"step": 69015
|
| 187173 |
+
},
|
| 187174 |
+
{
|
| 187175 |
+
"epoch": 555.62,
|
| 187176 |
+
"learning_rate": 8.910352564102564e-06,
|
| 187177 |
+
"loss": 1.1715,
|
| 187178 |
+
"step": 69020
|
| 187179 |
+
},
|
| 187180 |
+
{
|
| 187181 |
+
"epoch": 555.66,
|
| 187182 |
+
"learning_rate": 8.910272435897437e-06,
|
| 187183 |
+
"loss": 0.3312,
|
| 187184 |
+
"step": 69025
|
| 187185 |
+
},
|
| 187186 |
+
{
|
| 187187 |
+
"epoch": 555.7,
|
| 187188 |
+
"learning_rate": 8.910192307692309e-06,
|
| 187189 |
+
"loss": 0.373,
|
| 187190 |
+
"step": 69030
|
| 187191 |
+
},
|
| 187192 |
+
{
|
| 187193 |
+
"epoch": 555.74,
|
| 187194 |
+
"learning_rate": 8.91011217948718e-06,
|
| 187195 |
+
"loss": 0.3515,
|
| 187196 |
+
"step": 69035
|
| 187197 |
+
},
|
| 187198 |
+
{
|
| 187199 |
+
"epoch": 555.78,
|
| 187200 |
+
"learning_rate": 8.910032051282052e-06,
|
| 187201 |
+
"loss": 0.6376,
|
| 187202 |
+
"step": 69040
|
| 187203 |
+
},
|
| 187204 |
+
{
|
| 187205 |
+
"epoch": 555.82,
|
| 187206 |
+
"learning_rate": 8.909951923076925e-06,
|
| 187207 |
+
"loss": 1.0635,
|
| 187208 |
+
"step": 69045
|
| 187209 |
+
},
|
| 187210 |
+
{
|
| 187211 |
+
"epoch": 555.86,
|
| 187212 |
+
"learning_rate": 8.909871794871796e-06,
|
| 187213 |
+
"loss": 0.3108,
|
| 187214 |
+
"step": 69050
|
| 187215 |
+
},
|
| 187216 |
+
{
|
| 187217 |
+
"epoch": 555.9,
|
| 187218 |
+
"learning_rate": 8.909791666666667e-06,
|
| 187219 |
+
"loss": 0.3613,
|
| 187220 |
+
"step": 69055
|
| 187221 |
+
},
|
| 187222 |
+
{
|
| 187223 |
+
"epoch": 555.94,
|
| 187224 |
+
"learning_rate": 8.90971153846154e-06,
|
| 187225 |
+
"loss": 0.3515,
|
| 187226 |
+
"step": 69060
|
| 187227 |
+
},
|
| 187228 |
+
{
|
| 187229 |
+
"epoch": 555.98,
|
| 187230 |
+
"learning_rate": 8.90963141025641e-06,
|
| 187231 |
+
"loss": 0.7066,
|
| 187232 |
+
"step": 69065
|
| 187233 |
+
},
|
| 187234 |
+
{
|
| 187235 |
+
"epoch": 556.0,
|
| 187236 |
+
"eval_loss": 0.3735567331314087,
|
| 187237 |
+
"eval_runtime": 39.445,
|
| 187238 |
+
"eval_samples_per_second": 21.27,
|
| 187239 |
+
"eval_steps_per_second": 0.684,
|
| 187240 |
+
"eval_wer": 0.19350029815146094,
|
| 187241 |
+
"step": 69067
|
| 187242 |
+
},
|
| 187243 |
+
{
|
| 187244 |
+
"epoch": 556.02,
|
| 187245 |
+
"learning_rate": 8.909551282051283e-06,
|
| 187246 |
+
"loss": 0.372,
|
| 187247 |
+
"step": 69070
|
| 187248 |
+
},
|
| 187249 |
+
{
|
| 187250 |
+
"epoch": 556.06,
|
| 187251 |
+
"learning_rate": 8.909471153846154e-06,
|
| 187252 |
+
"loss": 0.3013,
|
| 187253 |
+
"step": 69075
|
| 187254 |
+
},
|
| 187255 |
+
{
|
| 187256 |
+
"epoch": 556.1,
|
| 187257 |
+
"learning_rate": 8.909391025641026e-06,
|
| 187258 |
+
"loss": 0.3405,
|
| 187259 |
+
"step": 69080
|
| 187260 |
+
},
|
| 187261 |
+
{
|
| 187262 |
+
"epoch": 556.15,
|
| 187263 |
+
"learning_rate": 8.909310897435897e-06,
|
| 187264 |
+
"loss": 0.4157,
|
| 187265 |
+
"step": 69085
|
| 187266 |
+
},
|
| 187267 |
+
{
|
| 187268 |
+
"epoch": 556.19,
|
| 187269 |
+
"learning_rate": 8.90923076923077e-06,
|
| 187270 |
+
"loss": 0.7011,
|
| 187271 |
+
"step": 69090
|
| 187272 |
+
},
|
| 187273 |
+
{
|
| 187274 |
+
"epoch": 556.23,
|
| 187275 |
+
"learning_rate": 8.909150641025642e-06,
|
| 187276 |
+
"loss": 1.0653,
|
| 187277 |
+
"step": 69095
|
| 187278 |
+
},
|
| 187279 |
+
{
|
| 187280 |
+
"epoch": 556.27,
|
| 187281 |
+
"learning_rate": 8.909070512820513e-06,
|
| 187282 |
+
"loss": 0.3359,
|
| 187283 |
+
"step": 69100
|
| 187284 |
+
},
|
| 187285 |
+
{
|
| 187286 |
+
"epoch": 556.31,
|
| 187287 |
+
"learning_rate": 8.908990384615386e-06,
|
| 187288 |
+
"loss": 0.3593,
|
| 187289 |
+
"step": 69105
|
| 187290 |
+
},
|
| 187291 |
+
{
|
| 187292 |
+
"epoch": 556.35,
|
| 187293 |
+
"learning_rate": 8.908910256410257e-06,
|
| 187294 |
+
"loss": 0.4316,
|
| 187295 |
+
"step": 69110
|
| 187296 |
+
},
|
| 187297 |
+
{
|
| 187298 |
+
"epoch": 556.39,
|
| 187299 |
+
"learning_rate": 8.908830128205129e-06,
|
| 187300 |
+
"loss": 0.748,
|
| 187301 |
+
"step": 69115
|
| 187302 |
+
},
|
| 187303 |
+
{
|
| 187304 |
+
"epoch": 556.43,
|
| 187305 |
+
"learning_rate": 8.90875e-06,
|
| 187306 |
+
"loss": 0.8545,
|
| 187307 |
+
"step": 69120
|
| 187308 |
+
},
|
| 187309 |
+
{
|
| 187310 |
+
"epoch": 556.47,
|
| 187311 |
+
"learning_rate": 8.908669871794873e-06,
|
| 187312 |
+
"loss": 0.3329,
|
| 187313 |
+
"step": 69125
|
| 187314 |
+
},
|
| 187315 |
+
{
|
| 187316 |
+
"epoch": 556.51,
|
| 187317 |
+
"learning_rate": 8.908589743589744e-06,
|
| 187318 |
+
"loss": 0.2987,
|
| 187319 |
+
"step": 69130
|
| 187320 |
+
},
|
| 187321 |
+
{
|
| 187322 |
+
"epoch": 556.55,
|
| 187323 |
+
"learning_rate": 8.908509615384616e-06,
|
| 187324 |
+
"loss": 0.3832,
|
| 187325 |
+
"step": 69135
|
| 187326 |
+
},
|
| 187327 |
+
{
|
| 187328 |
+
"epoch": 556.59,
|
| 187329 |
+
"learning_rate": 8.908429487179487e-06,
|
| 187330 |
+
"loss": 0.8224,
|
| 187331 |
+
"step": 69140
|
| 187332 |
+
},
|
| 187333 |
+
{
|
| 187334 |
+
"epoch": 556.63,
|
| 187335 |
+
"learning_rate": 8.90834935897436e-06,
|
| 187336 |
+
"loss": 0.8839,
|
| 187337 |
+
"step": 69145
|
| 187338 |
+
},
|
| 187339 |
+
{
|
| 187340 |
+
"epoch": 556.67,
|
| 187341 |
+
"learning_rate": 8.908269230769232e-06,
|
| 187342 |
+
"loss": 0.3099,
|
| 187343 |
+
"step": 69150
|
| 187344 |
+
},
|
| 187345 |
+
{
|
| 187346 |
+
"epoch": 556.71,
|
| 187347 |
+
"learning_rate": 8.908189102564103e-06,
|
| 187348 |
+
"loss": 0.2932,
|
| 187349 |
+
"step": 69155
|
| 187350 |
+
},
|
| 187351 |
+
{
|
| 187352 |
+
"epoch": 556.75,
|
| 187353 |
+
"learning_rate": 8.908108974358976e-06,
|
| 187354 |
+
"loss": 0.3511,
|
| 187355 |
+
"step": 69160
|
| 187356 |
+
},
|
| 187357 |
+
{
|
| 187358 |
+
"epoch": 556.79,
|
| 187359 |
+
"learning_rate": 8.908028846153847e-06,
|
| 187360 |
+
"loss": 0.6736,
|
| 187361 |
+
"step": 69165
|
| 187362 |
+
},
|
| 187363 |
+
{
|
| 187364 |
+
"epoch": 556.83,
|
| 187365 |
+
"learning_rate": 8.907948717948719e-06,
|
| 187366 |
+
"loss": 0.8941,
|
| 187367 |
+
"step": 69170
|
| 187368 |
+
},
|
| 187369 |
+
{
|
| 187370 |
+
"epoch": 556.87,
|
| 187371 |
+
"learning_rate": 8.90786858974359e-06,
|
| 187372 |
+
"loss": 0.3152,
|
| 187373 |
+
"step": 69175
|
| 187374 |
+
},
|
| 187375 |
+
{
|
| 187376 |
+
"epoch": 556.91,
|
| 187377 |
+
"learning_rate": 8.907788461538463e-06,
|
| 187378 |
+
"loss": 0.355,
|
| 187379 |
+
"step": 69180
|
| 187380 |
+
},
|
| 187381 |
+
{
|
| 187382 |
+
"epoch": 556.95,
|
| 187383 |
+
"learning_rate": 8.907708333333333e-06,
|
| 187384 |
+
"loss": 0.4686,
|
| 187385 |
+
"step": 69185
|
| 187386 |
+
},
|
| 187387 |
+
{
|
| 187388 |
+
"epoch": 556.99,
|
| 187389 |
+
"learning_rate": 8.907628205128206e-06,
|
| 187390 |
+
"loss": 0.8468,
|
| 187391 |
+
"step": 69190
|
| 187392 |
+
},
|
| 187393 |
+
{
|
| 187394 |
+
"epoch": 557.0,
|
| 187395 |
+
"eval_loss": 0.3936476707458496,
|
| 187396 |
+
"eval_runtime": 39.1151,
|
| 187397 |
+
"eval_samples_per_second": 21.45,
|
| 187398 |
+
"eval_steps_per_second": 0.69,
|
| 187399 |
+
"eval_wer": 0.19881948595337717,
|
| 187400 |
+
"step": 69191
|
| 187401 |
+
},
|
| 187402 |
+
{
|
| 187403 |
+
"epoch": 557.03,
|
| 187404 |
+
"learning_rate": 8.907548076923077e-06,
|
| 187405 |
+
"loss": 0.3576,
|
| 187406 |
+
"step": 69195
|
| 187407 |
+
},
|
| 187408 |
+
{
|
| 187409 |
+
"epoch": 557.07,
|
| 187410 |
+
"learning_rate": 8.907467948717949e-06,
|
| 187411 |
+
"loss": 0.2662,
|
| 187412 |
+
"step": 69200
|
| 187413 |
+
},
|
| 187414 |
+
{
|
| 187415 |
+
"epoch": 557.11,
|
| 187416 |
+
"learning_rate": 8.907387820512822e-06,
|
| 187417 |
+
"loss": 0.3339,
|
| 187418 |
+
"step": 69205
|
| 187419 |
+
},
|
| 187420 |
+
{
|
| 187421 |
+
"epoch": 557.15,
|
| 187422 |
+
"learning_rate": 8.907307692307693e-06,
|
| 187423 |
+
"loss": 0.4627,
|
| 187424 |
+
"step": 69210
|
| 187425 |
+
},
|
| 187426 |
+
{
|
| 187427 |
+
"epoch": 557.19,
|
| 187428 |
+
"learning_rate": 8.907227564102564e-06,
|
| 187429 |
+
"loss": 1.0684,
|
| 187430 |
+
"step": 69215
|
| 187431 |
+
},
|
| 187432 |
+
{
|
| 187433 |
+
"epoch": 557.23,
|
| 187434 |
+
"learning_rate": 8.907147435897436e-06,
|
| 187435 |
+
"loss": 0.7947,
|
| 187436 |
+
"step": 69220
|
| 187437 |
+
},
|
| 187438 |
+
{
|
| 187439 |
+
"epoch": 557.27,
|
| 187440 |
+
"learning_rate": 8.907067307692309e-06,
|
| 187441 |
+
"loss": 0.3128,
|
| 187442 |
+
"step": 69225
|
| 187443 |
+
},
|
| 187444 |
+
{
|
| 187445 |
+
"epoch": 557.31,
|
| 187446 |
+
"learning_rate": 8.90698717948718e-06,
|
| 187447 |
+
"loss": 0.3003,
|
| 187448 |
+
"step": 69230
|
| 187449 |
+
},
|
| 187450 |
+
{
|
| 187451 |
+
"epoch": 557.35,
|
| 187452 |
+
"learning_rate": 8.906907051282051e-06,
|
| 187453 |
+
"loss": 0.4299,
|
| 187454 |
+
"step": 69235
|
| 187455 |
+
},
|
| 187456 |
+
{
|
| 187457 |
+
"epoch": 557.39,
|
| 187458 |
+
"learning_rate": 8.906826923076923e-06,
|
| 187459 |
+
"loss": 1.0473,
|
| 187460 |
+
"step": 69240
|
| 187461 |
+
},
|
| 187462 |
+
{
|
| 187463 |
+
"epoch": 557.43,
|
| 187464 |
+
"learning_rate": 8.906746794871796e-06,
|
| 187465 |
+
"loss": 0.5692,
|
| 187466 |
+
"step": 69245
|
| 187467 |
+
},
|
| 187468 |
+
{
|
| 187469 |
+
"epoch": 557.47,
|
| 187470 |
+
"learning_rate": 8.906666666666667e-06,
|
| 187471 |
+
"loss": 0.2719,
|
| 187472 |
+
"step": 69250
|
| 187473 |
+
},
|
| 187474 |
+
{
|
| 187475 |
+
"epoch": 557.51,
|
| 187476 |
+
"learning_rate": 8.906586538461539e-06,
|
| 187477 |
+
"loss": 0.3171,
|
| 187478 |
+
"step": 69255
|
| 187479 |
+
},
|
| 187480 |
+
{
|
| 187481 |
+
"epoch": 557.55,
|
| 187482 |
+
"learning_rate": 8.906506410256412e-06,
|
| 187483 |
+
"loss": 0.4372,
|
| 187484 |
+
"step": 69260
|
| 187485 |
+
},
|
| 187486 |
+
{
|
| 187487 |
+
"epoch": 557.59,
|
| 187488 |
+
"learning_rate": 8.906426282051283e-06,
|
| 187489 |
+
"loss": 0.9317,
|
| 187490 |
+
"step": 69265
|
| 187491 |
+
},
|
| 187492 |
+
{
|
| 187493 |
+
"epoch": 557.63,
|
| 187494 |
+
"learning_rate": 8.906346153846154e-06,
|
| 187495 |
+
"loss": 0.6513,
|
| 187496 |
+
"step": 69270
|
| 187497 |
+
},
|
| 187498 |
+
{
|
| 187499 |
+
"epoch": 557.67,
|
| 187500 |
+
"learning_rate": 8.906266025641026e-06,
|
| 187501 |
+
"loss": 0.272,
|
| 187502 |
+
"step": 69275
|
| 187503 |
+
},
|
| 187504 |
+
{
|
| 187505 |
+
"epoch": 557.71,
|
| 187506 |
+
"learning_rate": 8.906185897435899e-06,
|
| 187507 |
+
"loss": 0.3266,
|
| 187508 |
+
"step": 69280
|
| 187509 |
+
},
|
| 187510 |
+
{
|
| 187511 |
+
"epoch": 557.76,
|
| 187512 |
+
"learning_rate": 8.90610576923077e-06,
|
| 187513 |
+
"loss": 0.453,
|
| 187514 |
+
"step": 69285
|
| 187515 |
+
},
|
| 187516 |
+
{
|
| 187517 |
+
"epoch": 557.8,
|
| 187518 |
+
"learning_rate": 8.906025641025641e-06,
|
| 187519 |
+
"loss": 0.8466,
|
| 187520 |
+
"step": 69290
|
| 187521 |
+
},
|
| 187522 |
+
{
|
| 187523 |
+
"epoch": 557.84,
|
| 187524 |
+
"learning_rate": 8.905945512820513e-06,
|
| 187525 |
+
"loss": 0.6527,
|
| 187526 |
+
"step": 69295
|
| 187527 |
+
},
|
| 187528 |
+
{
|
| 187529 |
+
"epoch": 557.88,
|
| 187530 |
+
"learning_rate": 8.905865384615386e-06,
|
| 187531 |
+
"loss": 0.3322,
|
| 187532 |
+
"step": 69300
|
| 187533 |
+
},
|
| 187534 |
+
{
|
| 187535 |
+
"epoch": 557.92,
|
| 187536 |
+
"learning_rate": 8.905785256410257e-06,
|
| 187537 |
+
"loss": 0.3628,
|
| 187538 |
+
"step": 69305
|
| 187539 |
+
},
|
| 187540 |
+
{
|
| 187541 |
+
"epoch": 557.96,
|
| 187542 |
+
"learning_rate": 8.905705128205129e-06,
|
| 187543 |
+
"loss": 0.4183,
|
| 187544 |
+
"step": 69310
|
| 187545 |
+
},
|
| 187546 |
+
{
|
| 187547 |
+
"epoch": 558.0,
|
| 187548 |
+
"learning_rate": 8.905625000000002e-06,
|
| 187549 |
+
"loss": 1.0645,
|
| 187550 |
+
"step": 69315
|
| 187551 |
+
},
|
| 187552 |
+
{
|
| 187553 |
+
"epoch": 558.0,
|
| 187554 |
+
"eval_loss": 0.4768179655075073,
|
| 187555 |
+
"eval_runtime": 41.3535,
|
| 187556 |
+
"eval_samples_per_second": 20.289,
|
| 187557 |
+
"eval_steps_per_second": 0.653,
|
| 187558 |
+
"eval_wer": 0.1948916076411247,
|
| 187559 |
+
"step": 69315
|
| 187560 |
+
},
|
| 187561 |
+
{
|
| 187562 |
+
"epoch": 554.04,
|
| 187563 |
+
"learning_rate": 8.905544871794873e-06,
|
| 187564 |
+
"loss": 0.3339,
|
| 187565 |
+
"step": 69320
|
| 187566 |
+
},
|
| 187567 |
+
{
|
| 187568 |
+
"epoch": 554.08,
|
| 187569 |
+
"learning_rate": 8.905464743589744e-06,
|
| 187570 |
+
"loss": 0.2938,
|
| 187571 |
+
"step": 69325
|
| 187572 |
+
},
|
| 187573 |
+
{
|
| 187574 |
+
"epoch": 554.12,
|
| 187575 |
+
"learning_rate": 8.905384615384616e-06,
|
| 187576 |
+
"loss": 0.3485,
|
| 187577 |
+
"step": 69330
|
| 187578 |
+
},
|
| 187579 |
+
{
|
| 187580 |
+
"epoch": 554.16,
|
| 187581 |
+
"learning_rate": 8.905304487179489e-06,
|
| 187582 |
+
"loss": 0.6218,
|
| 187583 |
+
"step": 69335
|
| 187584 |
+
},
|
| 187585 |
+
{
|
| 187586 |
+
"epoch": 554.2,
|
| 187587 |
+
"learning_rate": 8.905224358974358e-06,
|
| 187588 |
+
"loss": 1.2255,
|
| 187589 |
+
"step": 69340
|
| 187590 |
+
},
|
| 187591 |
+
{
|
| 187592 |
+
"epoch": 554.24,
|
| 187593 |
+
"learning_rate": 8.905144230769232e-06,
|
| 187594 |
+
"loss": 0.3385,
|
| 187595 |
+
"step": 69345
|
| 187596 |
+
},
|
| 187597 |
+
{
|
| 187598 |
+
"epoch": 554.28,
|
| 187599 |
+
"learning_rate": 8.905064102564105e-06,
|
| 187600 |
+
"loss": 0.3168,
|
| 187601 |
+
"step": 69350
|
| 187602 |
+
},
|
| 187603 |
+
{
|
| 187604 |
+
"epoch": 554.32,
|
| 187605 |
+
"learning_rate": 8.904983974358974e-06,
|
| 187606 |
+
"loss": 0.3309,
|
| 187607 |
+
"step": 69355
|
| 187608 |
+
},
|
| 187609 |
+
{
|
| 187610 |
+
"epoch": 554.36,
|
| 187611 |
+
"learning_rate": 8.904903846153847e-06,
|
| 187612 |
+
"loss": 0.5233,
|
| 187613 |
+
"step": 69360
|
| 187614 |
+
},
|
| 187615 |
+
{
|
| 187616 |
+
"epoch": 554.4,
|
| 187617 |
+
"learning_rate": 8.904823717948719e-06,
|
| 187618 |
+
"loss": 1.4065,
|
| 187619 |
+
"step": 69365
|
| 187620 |
+
},
|
| 187621 |
+
{
|
| 187622 |
+
"epoch": 554.44,
|
| 187623 |
+
"learning_rate": 8.90474358974359e-06,
|
| 187624 |
+
"loss": 0.3589,
|
| 187625 |
+
"step": 69370
|
| 187626 |
+
},
|
| 187627 |
+
{
|
| 187628 |
+
"epoch": 554.48,
|
| 187629 |
+
"learning_rate": 8.904663461538461e-06,
|
| 187630 |
+
"loss": 0.2869,
|
| 187631 |
+
"step": 69375
|
| 187632 |
+
},
|
| 187633 |
+
{
|
| 187634 |
+
"epoch": 554.52,
|
| 187635 |
+
"learning_rate": 8.904583333333334e-06,
|
| 187636 |
+
"loss": 0.3377,
|
| 187637 |
+
"step": 69380
|
| 187638 |
+
},
|
| 187639 |
+
{
|
| 187640 |
+
"epoch": 554.56,
|
| 187641 |
+
"learning_rate": 8.904503205128206e-06,
|
| 187642 |
+
"loss": 0.535,
|
| 187643 |
+
"step": 69385
|
| 187644 |
+
},
|
| 187645 |
+
{
|
| 187646 |
+
"epoch": 554.6,
|
| 187647 |
+
"learning_rate": 8.904423076923077e-06,
|
| 187648 |
+
"loss": 1.4467,
|
| 187649 |
+
"step": 69390
|
| 187650 |
+
},
|
| 187651 |
+
{
|
| 187652 |
+
"epoch": 554.64,
|
| 187653 |
+
"learning_rate": 8.904342948717948e-06,
|
| 187654 |
+
"loss": 0.5932,
|
| 187655 |
+
"step": 69395
|
| 187656 |
+
},
|
| 187657 |
+
{
|
| 187658 |
+
"epoch": 554.68,
|
| 187659 |
+
"learning_rate": 8.904262820512822e-06,
|
| 187660 |
+
"loss": 0.4281,
|
| 187661 |
+
"step": 69400
|
| 187662 |
+
},
|
| 187663 |
+
{
|
| 187664 |
+
"epoch": 554.72,
|
| 187665 |
+
"learning_rate": 8.904182692307693e-06,
|
| 187666 |
+
"loss": 0.3255,
|
| 187667 |
+
"step": 69405
|
| 187668 |
+
},
|
| 187669 |
+
{
|
| 187670 |
+
"epoch": 554.76,
|
| 187671 |
+
"learning_rate": 8.904102564102564e-06,
|
| 187672 |
+
"loss": 0.5435,
|
| 187673 |
+
"step": 69410
|
| 187674 |
+
},
|
| 187675 |
+
{
|
| 187676 |
+
"epoch": 554.8,
|
| 187677 |
+
"learning_rate": 8.904022435897437e-06,
|
| 187678 |
+
"loss": 1.2416,
|
| 187679 |
+
"step": 69415
|
| 187680 |
+
},
|
| 187681 |
+
{
|
| 187682 |
+
"epoch": 554.84,
|
| 187683 |
+
"learning_rate": 8.903942307692309e-06,
|
| 187684 |
+
"loss": 0.6299,
|
| 187685 |
+
"step": 69420
|
| 187686 |
+
},
|
| 187687 |
+
{
|
| 187688 |
+
"epoch": 554.88,
|
| 187689 |
+
"learning_rate": 8.90386217948718e-06,
|
| 187690 |
+
"loss": 0.3255,
|
| 187691 |
+
"step": 69425
|
| 187692 |
+
},
|
| 187693 |
+
{
|
| 187694 |
+
"epoch": 554.92,
|
| 187695 |
+
"learning_rate": 8.903782051282051e-06,
|
| 187696 |
+
"loss": 0.3651,
|
| 187697 |
+
"step": 69430
|
| 187698 |
+
},
|
| 187699 |
+
{
|
| 187700 |
+
"epoch": 554.96,
|
| 187701 |
+
"learning_rate": 8.903701923076924e-06,
|
| 187702 |
+
"loss": 0.4798,
|
| 187703 |
+
"step": 69435
|
| 187704 |
+
},
|
| 187705 |
+
{
|
| 187706 |
+
"epoch": 555.0,
|
| 187707 |
+
"learning_rate": 8.903621794871796e-06,
|
| 187708 |
+
"loss": 1.206,
|
| 187709 |
+
"step": 69440
|
| 187710 |
+
},
|
| 187711 |
+
{
|
| 187712 |
+
"epoch": 555.0,
|
| 187713 |
+
"eval_loss": 0.34815290570259094,
|
| 187714 |
+
"eval_runtime": 40.671,
|
| 187715 |
+
"eval_samples_per_second": 20.629,
|
| 187716 |
+
"eval_steps_per_second": 0.664,
|
| 187717 |
+
"eval_wer": 0.18589017831870933,
|
| 187718 |
+
"step": 69440
|
| 187719 |
}
|
| 187720 |
],
|
| 187721 |
"max_steps": 625000,
|
| 187722 |
"num_train_epochs": 5000,
|
| 187723 |
+
"total_flos": 1.9540963521520874e+20,
|
| 187724 |
"trial_name": null,
|
| 187725 |
"trial_params": null
|
| 187726 |
}
|
model-bin/finetune/base/{checkpoint-68818 → checkpoint-69440}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629843978.3964214/events.out.tfevents.1629843978.c435e1c5ee04.920.241
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f5f924f05c1761d6e622e1fe364132cc7ce7a60ed587f21962642aed9c6dea3c
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629844632.7215848/events.out.tfevents.1629844632.c435e1c5ee04.920.243
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:0c66b00ffda2d8831a8a42b631b2a7b9edaed066b9afb6bf70a72fc55207e8b1
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629845266.9609022/events.out.tfevents.1629845266.c435e1c5ee04.920.245
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c2600798a7c11f3f5bc24a7889aca81841c8038630b09e3505897c3e25c760ce
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629845904.6603022/events.out.tfevents.1629845904.c435e1c5ee04.920.247
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:fff7009ac8fdf816d297df477743834bfdefea6c6649b21a6e1c53abff5c3578
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629846554.1512308/events.out.tfevents.1629846554.c435e1c5ee04.920.249
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:36d230724e1655ef4bdeeba22e814367809fafcc7e4d7253811e6d0448d06690
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629843978.c435e1c5ee04.920.240
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:05432f54c615f59447fc156df5739ef2da6178de5b7a1694bcdc6a38347fc914
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629844632.c435e1c5ee04.920.242
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e0909f46e291559be47da07e3c7249cbbaa3d97aca8858064f634c2eaecc6e0d
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629845266.c435e1c5ee04.920.244
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9465a9c811dc5c646b4caa1afdca954c1dee5fd1f750e1b92640e3e4eb19515a
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629845904.c435e1c5ee04.920.246
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b779365eed4809a3b611dd3419710d7f88c6fd81ca818fc2378f5bd0b67bc4eb
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629846554.c435e1c5ee04.920.248
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:628bd5086daaad6d03bd552a731725168c88677e588962a70ed38ebe2787a540
|
| 3 |
+
size 8622
|