"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-56372 → checkpoint-56994}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-56372 → checkpoint-56994}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-56372 → checkpoint-56994}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-56372 → checkpoint-56994}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-56372 → checkpoint-56994}/rng_state.pth +2 -2
- model-bin/finetune/base/{checkpoint-56372 → checkpoint-56994}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-56372 → checkpoint-56994}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-56372 → checkpoint-56994}/trainer_state.json +793 -4
- model-bin/finetune/base/{checkpoint-56372 → checkpoint-56994}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629778472.5431197/events.out.tfevents.1629778472.c435e1c5ee04.920.41 +3 -0
- model-bin/finetune/base/log/1629779133.0815535/events.out.tfevents.1629779133.c435e1c5ee04.920.43 +3 -0
- model-bin/finetune/base/log/1629779785.3045485/events.out.tfevents.1629779785.c435e1c5ee04.920.45 +3 -0
- model-bin/finetune/base/log/1629780436.4732363/events.out.tfevents.1629780436.c435e1c5ee04.920.47 +3 -0
- model-bin/finetune/base/log/1629781090.5327764/events.out.tfevents.1629781090.c435e1c5ee04.920.49 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629778472.c435e1c5ee04.920.40 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629779133.c435e1c5ee04.920.42 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629779785.c435e1c5ee04.920.44 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629780436.c435e1c5ee04.920.46 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629781090.c435e1c5ee04.920.48 +3 -0
model-bin/finetune/base/{checkpoint-56372 → checkpoint-56994}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-56372 → checkpoint-56994}/optimizer.pt
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 722165009
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:30cf6d09d0b50a9f67e11e8c67c49356a46eab3a11783da12312b2ad6c07427d
|
| 3 |
size 722165009
|
model-bin/finetune/base/{checkpoint-56372 → checkpoint-56994}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-56372 → checkpoint-56994}/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 sha256:25ab05e0b9cc5b5c4f58d78d8b2e9f01de46ec6c559978d368a049852d0476fd
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-56372 → checkpoint-56994}/rng_state.pth
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:fc67ac4d86710e96e4aff956d7725e59bf4ef61f39d585f959501871aaef7a35
|
| 3 |
+
size 14567
|
model-bin/finetune/base/{checkpoint-56372 → checkpoint-56994}/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 sha256:46c4f16ce8829f5786060a04c5b5a8cbb455321b5dfbceaa0c763f4777791489
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-56372 → checkpoint-56994}/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 sha256:7db297422c4c558e216638cad9666c70a88ca6d2485e276ec8d633302b42fd19
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-56372 → checkpoint-56994}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.18631571186315712,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-55501",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -171087,11 +171087,800 @@
|
|
| 171087 |
"eval_steps_per_second": 0.67,
|
| 171088 |
"eval_wer": 0.20751998847511344,
|
| 171089 |
"step": 56372
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 171090 |
}
|
| 171091 |
],
|
| 171092 |
-
"max_steps":
|
| 171093 |
"num_train_epochs": 5000,
|
| 171094 |
-
"total_flos": 1.
|
| 171095 |
"trial_name": null,
|
| 171096 |
"trial_params": null
|
| 171097 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.18631571186315712,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-55501",
|
| 4 |
+
"epoch": 462.99595141700405,
|
| 5 |
+
"global_step": 56994,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 171087 |
"eval_steps_per_second": 0.67,
|
| 171088 |
"eval_wer": 0.20751998847511344,
|
| 171089 |
"step": 56372
|
| 171090 |
+
},
|
| 171091 |
+
{
|
| 171092 |
+
"epoch": 454.02,
|
| 171093 |
+
"learning_rate": 9.112916666666667e-06,
|
| 171094 |
+
"loss": 0.4064,
|
| 171095 |
+
"step": 56375
|
| 171096 |
+
},
|
| 171097 |
+
{
|
| 171098 |
+
"epoch": 454.06,
|
| 171099 |
+
"learning_rate": 9.112836538461539e-06,
|
| 171100 |
+
"loss": 0.319,
|
| 171101 |
+
"step": 56380
|
| 171102 |
+
},
|
| 171103 |
+
{
|
| 171104 |
+
"epoch": 454.1,
|
| 171105 |
+
"learning_rate": 9.112756410256412e-06,
|
| 171106 |
+
"loss": 0.3691,
|
| 171107 |
+
"step": 56385
|
| 171108 |
+
},
|
| 171109 |
+
{
|
| 171110 |
+
"epoch": 454.14,
|
| 171111 |
+
"learning_rate": 9.112676282051283e-06,
|
| 171112 |
+
"loss": 0.5006,
|
| 171113 |
+
"step": 56390
|
| 171114 |
+
},
|
| 171115 |
+
{
|
| 171116 |
+
"epoch": 454.18,
|
| 171117 |
+
"learning_rate": 9.112596153846154e-06,
|
| 171118 |
+
"loss": 0.8926,
|
| 171119 |
+
"step": 56395
|
| 171120 |
+
},
|
| 171121 |
+
{
|
| 171122 |
+
"epoch": 454.22,
|
| 171123 |
+
"learning_rate": 9.112516025641026e-06,
|
| 171124 |
+
"loss": 0.9786,
|
| 171125 |
+
"step": 56400
|
| 171126 |
+
},
|
| 171127 |
+
{
|
| 171128 |
+
"epoch": 454.27,
|
| 171129 |
+
"learning_rate": 9.112435897435899e-06,
|
| 171130 |
+
"loss": 0.3246,
|
| 171131 |
+
"step": 56405
|
| 171132 |
+
},
|
| 171133 |
+
{
|
| 171134 |
+
"epoch": 454.31,
|
| 171135 |
+
"learning_rate": 9.11235576923077e-06,
|
| 171136 |
+
"loss": 0.3865,
|
| 171137 |
+
"step": 56410
|
| 171138 |
+
},
|
| 171139 |
+
{
|
| 171140 |
+
"epoch": 454.35,
|
| 171141 |
+
"learning_rate": 9.112275641025641e-06,
|
| 171142 |
+
"loss": 0.3918,
|
| 171143 |
+
"step": 56415
|
| 171144 |
+
},
|
| 171145 |
+
{
|
| 171146 |
+
"epoch": 454.39,
|
| 171147 |
+
"learning_rate": 9.112195512820513e-06,
|
| 171148 |
+
"loss": 0.8994,
|
| 171149 |
+
"step": 56420
|
| 171150 |
+
},
|
| 171151 |
+
{
|
| 171152 |
+
"epoch": 454.43,
|
| 171153 |
+
"learning_rate": 9.112115384615386e-06,
|
| 171154 |
+
"loss": 1.0022,
|
| 171155 |
+
"step": 56425
|
| 171156 |
+
},
|
| 171157 |
+
{
|
| 171158 |
+
"epoch": 454.47,
|
| 171159 |
+
"learning_rate": 9.112035256410257e-06,
|
| 171160 |
+
"loss": 0.3168,
|
| 171161 |
+
"step": 56430
|
| 171162 |
+
},
|
| 171163 |
+
{
|
| 171164 |
+
"epoch": 454.51,
|
| 171165 |
+
"learning_rate": 9.111955128205129e-06,
|
| 171166 |
+
"loss": 0.5756,
|
| 171167 |
+
"step": 56435
|
| 171168 |
+
},
|
| 171169 |
+
{
|
| 171170 |
+
"epoch": 454.55,
|
| 171171 |
+
"learning_rate": 9.111875000000002e-06,
|
| 171172 |
+
"loss": 0.4405,
|
| 171173 |
+
"step": 56440
|
| 171174 |
+
},
|
| 171175 |
+
{
|
| 171176 |
+
"epoch": 454.59,
|
| 171177 |
+
"learning_rate": 9.111794871794873e-06,
|
| 171178 |
+
"loss": 0.8018,
|
| 171179 |
+
"step": 56445
|
| 171180 |
+
},
|
| 171181 |
+
{
|
| 171182 |
+
"epoch": 454.63,
|
| 171183 |
+
"learning_rate": 9.111714743589744e-06,
|
| 171184 |
+
"loss": 0.8747,
|
| 171185 |
+
"step": 56450
|
| 171186 |
+
},
|
| 171187 |
+
{
|
| 171188 |
+
"epoch": 454.67,
|
| 171189 |
+
"learning_rate": 9.111634615384616e-06,
|
| 171190 |
+
"loss": 0.2681,
|
| 171191 |
+
"step": 56455
|
| 171192 |
+
},
|
| 171193 |
+
{
|
| 171194 |
+
"epoch": 454.71,
|
| 171195 |
+
"learning_rate": 9.111554487179489e-06,
|
| 171196 |
+
"loss": 0.3752,
|
| 171197 |
+
"step": 56460
|
| 171198 |
+
},
|
| 171199 |
+
{
|
| 171200 |
+
"epoch": 454.75,
|
| 171201 |
+
"learning_rate": 9.11147435897436e-06,
|
| 171202 |
+
"loss": 0.4424,
|
| 171203 |
+
"step": 56465
|
| 171204 |
+
},
|
| 171205 |
+
{
|
| 171206 |
+
"epoch": 454.79,
|
| 171207 |
+
"learning_rate": 9.111394230769231e-06,
|
| 171208 |
+
"loss": 0.8706,
|
| 171209 |
+
"step": 56470
|
| 171210 |
+
},
|
| 171211 |
+
{
|
| 171212 |
+
"epoch": 454.83,
|
| 171213 |
+
"learning_rate": 9.111314102564103e-06,
|
| 171214 |
+
"loss": 1.0377,
|
| 171215 |
+
"step": 56475
|
| 171216 |
+
},
|
| 171217 |
+
{
|
| 171218 |
+
"epoch": 454.87,
|
| 171219 |
+
"learning_rate": 9.111233974358976e-06,
|
| 171220 |
+
"loss": 0.3031,
|
| 171221 |
+
"step": 56480
|
| 171222 |
+
},
|
| 171223 |
+
{
|
| 171224 |
+
"epoch": 454.91,
|
| 171225 |
+
"learning_rate": 9.111153846153847e-06,
|
| 171226 |
+
"loss": 0.3767,
|
| 171227 |
+
"step": 56485
|
| 171228 |
+
},
|
| 171229 |
+
{
|
| 171230 |
+
"epoch": 454.95,
|
| 171231 |
+
"learning_rate": 9.111073717948719e-06,
|
| 171232 |
+
"loss": 0.5039,
|
| 171233 |
+
"step": 56490
|
| 171234 |
+
},
|
| 171235 |
+
{
|
| 171236 |
+
"epoch": 454.99,
|
| 171237 |
+
"learning_rate": 9.110993589743592e-06,
|
| 171238 |
+
"loss": 0.9137,
|
| 171239 |
+
"step": 56495
|
| 171240 |
+
},
|
| 171241 |
+
{
|
| 171242 |
+
"epoch": 455.0,
|
| 171243 |
+
"eval_loss": 0.38671064376831055,
|
| 171244 |
+
"eval_runtime": 39.4284,
|
| 171245 |
+
"eval_samples_per_second": 21.304,
|
| 171246 |
+
"eval_steps_per_second": 0.685,
|
| 171247 |
+
"eval_wer": 0.1874216618742166,
|
| 171248 |
+
"step": 56496
|
| 171249 |
+
},
|
| 171250 |
+
{
|
| 171251 |
+
"epoch": 451.03,
|
| 171252 |
+
"learning_rate": 9.110913461538461e-06,
|
| 171253 |
+
"loss": 0.3197,
|
| 171254 |
+
"step": 56500
|
| 171255 |
+
},
|
| 171256 |
+
{
|
| 171257 |
+
"epoch": 451.07,
|
| 171258 |
+
"learning_rate": 9.110833333333334e-06,
|
| 171259 |
+
"loss": 0.3031,
|
| 171260 |
+
"step": 56505
|
| 171261 |
+
},
|
| 171262 |
+
{
|
| 171263 |
+
"epoch": 451.11,
|
| 171264 |
+
"learning_rate": 9.110753205128206e-06,
|
| 171265 |
+
"loss": 0.365,
|
| 171266 |
+
"step": 56510
|
| 171267 |
+
},
|
| 171268 |
+
{
|
| 171269 |
+
"epoch": 451.15,
|
| 171270 |
+
"learning_rate": 9.110673076923077e-06,
|
| 171271 |
+
"loss": 0.5421,
|
| 171272 |
+
"step": 56515
|
| 171273 |
+
},
|
| 171274 |
+
{
|
| 171275 |
+
"epoch": 451.19,
|
| 171276 |
+
"learning_rate": 9.110592948717948e-06,
|
| 171277 |
+
"loss": 1.0469,
|
| 171278 |
+
"step": 56520
|
| 171279 |
+
},
|
| 171280 |
+
{
|
| 171281 |
+
"epoch": 451.23,
|
| 171282 |
+
"learning_rate": 9.110512820512821e-06,
|
| 171283 |
+
"loss": 0.8389,
|
| 171284 |
+
"step": 56525
|
| 171285 |
+
},
|
| 171286 |
+
{
|
| 171287 |
+
"epoch": 451.27,
|
| 171288 |
+
"learning_rate": 9.110432692307693e-06,
|
| 171289 |
+
"loss": 0.3759,
|
| 171290 |
+
"step": 56530
|
| 171291 |
+
},
|
| 171292 |
+
{
|
| 171293 |
+
"epoch": 451.31,
|
| 171294 |
+
"learning_rate": 9.110352564102564e-06,
|
| 171295 |
+
"loss": 0.3245,
|
| 171296 |
+
"step": 56535
|
| 171297 |
+
},
|
| 171298 |
+
{
|
| 171299 |
+
"epoch": 451.35,
|
| 171300 |
+
"learning_rate": 9.110272435897437e-06,
|
| 171301 |
+
"loss": 0.4346,
|
| 171302 |
+
"step": 56540
|
| 171303 |
+
},
|
| 171304 |
+
{
|
| 171305 |
+
"epoch": 451.39,
|
| 171306 |
+
"learning_rate": 9.110192307692309e-06,
|
| 171307 |
+
"loss": 0.9923,
|
| 171308 |
+
"step": 56545
|
| 171309 |
+
},
|
| 171310 |
+
{
|
| 171311 |
+
"epoch": 451.43,
|
| 171312 |
+
"learning_rate": 9.11011217948718e-06,
|
| 171313 |
+
"loss": 0.6995,
|
| 171314 |
+
"step": 56550
|
| 171315 |
+
},
|
| 171316 |
+
{
|
| 171317 |
+
"epoch": 451.47,
|
| 171318 |
+
"learning_rate": 9.110032051282051e-06,
|
| 171319 |
+
"loss": 0.2944,
|
| 171320 |
+
"step": 56555
|
| 171321 |
+
},
|
| 171322 |
+
{
|
| 171323 |
+
"epoch": 451.51,
|
| 171324 |
+
"learning_rate": 9.109951923076924e-06,
|
| 171325 |
+
"loss": 0.3371,
|
| 171326 |
+
"step": 56560
|
| 171327 |
+
},
|
| 171328 |
+
{
|
| 171329 |
+
"epoch": 451.55,
|
| 171330 |
+
"learning_rate": 9.109871794871796e-06,
|
| 171331 |
+
"loss": 0.4263,
|
| 171332 |
+
"step": 56565
|
| 171333 |
+
},
|
| 171334 |
+
{
|
| 171335 |
+
"epoch": 451.59,
|
| 171336 |
+
"learning_rate": 9.109791666666667e-06,
|
| 171337 |
+
"loss": 1.1097,
|
| 171338 |
+
"step": 56570
|
| 171339 |
+
},
|
| 171340 |
+
{
|
| 171341 |
+
"epoch": 451.63,
|
| 171342 |
+
"learning_rate": 9.109711538461538e-06,
|
| 171343 |
+
"loss": 0.6258,
|
| 171344 |
+
"step": 56575
|
| 171345 |
+
},
|
| 171346 |
+
{
|
| 171347 |
+
"epoch": 451.67,
|
| 171348 |
+
"learning_rate": 9.109631410256411e-06,
|
| 171349 |
+
"loss": 0.3704,
|
| 171350 |
+
"step": 56580
|
| 171351 |
+
},
|
| 171352 |
+
{
|
| 171353 |
+
"epoch": 451.71,
|
| 171354 |
+
"learning_rate": 9.109551282051283e-06,
|
| 171355 |
+
"loss": 0.2776,
|
| 171356 |
+
"step": 56585
|
| 171357 |
+
},
|
| 171358 |
+
{
|
| 171359 |
+
"epoch": 451.75,
|
| 171360 |
+
"learning_rate": 9.109471153846154e-06,
|
| 171361 |
+
"loss": 0.4297,
|
| 171362 |
+
"step": 56590
|
| 171363 |
+
},
|
| 171364 |
+
{
|
| 171365 |
+
"epoch": 451.79,
|
| 171366 |
+
"learning_rate": 9.109391025641027e-06,
|
| 171367 |
+
"loss": 1.1215,
|
| 171368 |
+
"step": 56595
|
| 171369 |
+
},
|
| 171370 |
+
{
|
| 171371 |
+
"epoch": 451.83,
|
| 171372 |
+
"learning_rate": 9.109310897435899e-06,
|
| 171373 |
+
"loss": 0.707,
|
| 171374 |
+
"step": 56600
|
| 171375 |
+
},
|
| 171376 |
+
{
|
| 171377 |
+
"epoch": 451.87,
|
| 171378 |
+
"learning_rate": 9.10923076923077e-06,
|
| 171379 |
+
"loss": 0.3058,
|
| 171380 |
+
"step": 56605
|
| 171381 |
+
},
|
| 171382 |
+
{
|
| 171383 |
+
"epoch": 451.91,
|
| 171384 |
+
"learning_rate": 9.109150641025641e-06,
|
| 171385 |
+
"loss": 0.3418,
|
| 171386 |
+
"step": 56610
|
| 171387 |
+
},
|
| 171388 |
+
{
|
| 171389 |
+
"epoch": 451.95,
|
| 171390 |
+
"learning_rate": 9.109070512820514e-06,
|
| 171391 |
+
"loss": 0.4603,
|
| 171392 |
+
"step": 56615
|
| 171393 |
+
},
|
| 171394 |
+
{
|
| 171395 |
+
"epoch": 451.99,
|
| 171396 |
+
"learning_rate": 9.108990384615384e-06,
|
| 171397 |
+
"loss": 1.0587,
|
| 171398 |
+
"step": 56620
|
| 171399 |
+
},
|
| 171400 |
+
{
|
| 171401 |
+
"epoch": 452.0,
|
| 171402 |
+
"eval_loss": 0.4021945893764496,
|
| 171403 |
+
"eval_runtime": 39.9783,
|
| 171404 |
+
"eval_samples_per_second": 21.011,
|
| 171405 |
+
"eval_steps_per_second": 0.675,
|
| 171406 |
+
"eval_wer": 0.19840631625118796,
|
| 171407 |
+
"step": 56621
|
| 171408 |
+
},
|
| 171409 |
+
{
|
| 171410 |
+
"epoch": 452.03,
|
| 171411 |
+
"learning_rate": 9.108910256410257e-06,
|
| 171412 |
+
"loss": 0.3565,
|
| 171413 |
+
"step": 56625
|
| 171414 |
+
},
|
| 171415 |
+
{
|
| 171416 |
+
"epoch": 452.07,
|
| 171417 |
+
"learning_rate": 9.10883012820513e-06,
|
| 171418 |
+
"loss": 0.2871,
|
| 171419 |
+
"step": 56630
|
| 171420 |
+
},
|
| 171421 |
+
{
|
| 171422 |
+
"epoch": 452.11,
|
| 171423 |
+
"learning_rate": 9.10875e-06,
|
| 171424 |
+
"loss": 0.3495,
|
| 171425 |
+
"step": 56635
|
| 171426 |
+
},
|
| 171427 |
+
{
|
| 171428 |
+
"epoch": 452.15,
|
| 171429 |
+
"learning_rate": 9.108669871794873e-06,
|
| 171430 |
+
"loss": 0.4982,
|
| 171431 |
+
"step": 56640
|
| 171432 |
+
},
|
| 171433 |
+
{
|
| 171434 |
+
"epoch": 452.19,
|
| 171435 |
+
"learning_rate": 9.108589743589744e-06,
|
| 171436 |
+
"loss": 0.9314,
|
| 171437 |
+
"step": 56645
|
| 171438 |
+
},
|
| 171439 |
+
{
|
| 171440 |
+
"epoch": 452.23,
|
| 171441 |
+
"learning_rate": 9.108509615384616e-06,
|
| 171442 |
+
"loss": 0.8041,
|
| 171443 |
+
"step": 56650
|
| 171444 |
+
},
|
| 171445 |
+
{
|
| 171446 |
+
"epoch": 452.27,
|
| 171447 |
+
"learning_rate": 9.108429487179487e-06,
|
| 171448 |
+
"loss": 0.3507,
|
| 171449 |
+
"step": 56655
|
| 171450 |
+
},
|
| 171451 |
+
{
|
| 171452 |
+
"epoch": 452.31,
|
| 171453 |
+
"learning_rate": 9.10834935897436e-06,
|
| 171454 |
+
"loss": 0.3273,
|
| 171455 |
+
"step": 56660
|
| 171456 |
+
},
|
| 171457 |
+
{
|
| 171458 |
+
"epoch": 452.35,
|
| 171459 |
+
"learning_rate": 9.108269230769231e-06,
|
| 171460 |
+
"loss": 0.4386,
|
| 171461 |
+
"step": 56665
|
| 171462 |
+
},
|
| 171463 |
+
{
|
| 171464 |
+
"epoch": 452.39,
|
| 171465 |
+
"learning_rate": 9.108189102564103e-06,
|
| 171466 |
+
"loss": 0.9502,
|
| 171467 |
+
"step": 56670
|
| 171468 |
+
},
|
| 171469 |
+
{
|
| 171470 |
+
"epoch": 452.43,
|
| 171471 |
+
"learning_rate": 9.108108974358974e-06,
|
| 171472 |
+
"loss": 0.7826,
|
| 171473 |
+
"step": 56675
|
| 171474 |
+
},
|
| 171475 |
+
{
|
| 171476 |
+
"epoch": 452.47,
|
| 171477 |
+
"learning_rate": 9.108028846153847e-06,
|
| 171478 |
+
"loss": 0.2844,
|
| 171479 |
+
"step": 56680
|
| 171480 |
+
},
|
| 171481 |
+
{
|
| 171482 |
+
"epoch": 452.51,
|
| 171483 |
+
"learning_rate": 9.107948717948718e-06,
|
| 171484 |
+
"loss": 0.3316,
|
| 171485 |
+
"step": 56685
|
| 171486 |
+
},
|
| 171487 |
+
{
|
| 171488 |
+
"epoch": 452.55,
|
| 171489 |
+
"learning_rate": 9.10786858974359e-06,
|
| 171490 |
+
"loss": 0.423,
|
| 171491 |
+
"step": 56690
|
| 171492 |
+
},
|
| 171493 |
+
{
|
| 171494 |
+
"epoch": 452.59,
|
| 171495 |
+
"learning_rate": 9.107788461538463e-06,
|
| 171496 |
+
"loss": 0.9898,
|
| 171497 |
+
"step": 56695
|
| 171498 |
+
},
|
| 171499 |
+
{
|
| 171500 |
+
"epoch": 452.63,
|
| 171501 |
+
"learning_rate": 9.107708333333334e-06,
|
| 171502 |
+
"loss": 0.696,
|
| 171503 |
+
"step": 56700
|
| 171504 |
+
},
|
| 171505 |
+
{
|
| 171506 |
+
"epoch": 452.67,
|
| 171507 |
+
"learning_rate": 9.107628205128206e-06,
|
| 171508 |
+
"loss": 0.2807,
|
| 171509 |
+
"step": 56705
|
| 171510 |
+
},
|
| 171511 |
+
{
|
| 171512 |
+
"epoch": 452.71,
|
| 171513 |
+
"learning_rate": 9.107548076923077e-06,
|
| 171514 |
+
"loss": 0.3262,
|
| 171515 |
+
"step": 56710
|
| 171516 |
+
},
|
| 171517 |
+
{
|
| 171518 |
+
"epoch": 452.75,
|
| 171519 |
+
"learning_rate": 9.10746794871795e-06,
|
| 171520 |
+
"loss": 0.4433,
|
| 171521 |
+
"step": 56715
|
| 171522 |
+
},
|
| 171523 |
+
{
|
| 171524 |
+
"epoch": 452.79,
|
| 171525 |
+
"learning_rate": 9.107387820512821e-06,
|
| 171526 |
+
"loss": 0.9766,
|
| 171527 |
+
"step": 56720
|
| 171528 |
+
},
|
| 171529 |
+
{
|
| 171530 |
+
"epoch": 452.83,
|
| 171531 |
+
"learning_rate": 9.107307692307693e-06,
|
| 171532 |
+
"loss": 0.8275,
|
| 171533 |
+
"step": 56725
|
| 171534 |
+
},
|
| 171535 |
+
{
|
| 171536 |
+
"epoch": 452.87,
|
| 171537 |
+
"learning_rate": 9.107227564102566e-06,
|
| 171538 |
+
"loss": 0.3268,
|
| 171539 |
+
"step": 56730
|
| 171540 |
+
},
|
| 171541 |
+
{
|
| 171542 |
+
"epoch": 452.91,
|
| 171543 |
+
"learning_rate": 9.107147435897437e-06,
|
| 171544 |
+
"loss": 0.3838,
|
| 171545 |
+
"step": 56735
|
| 171546 |
+
},
|
| 171547 |
+
{
|
| 171548 |
+
"epoch": 452.95,
|
| 171549 |
+
"learning_rate": 9.107067307692309e-06,
|
| 171550 |
+
"loss": 0.4729,
|
| 171551 |
+
"step": 56740
|
| 171552 |
+
},
|
| 171553 |
+
{
|
| 171554 |
+
"epoch": 452.99,
|
| 171555 |
+
"learning_rate": 9.10698717948718e-06,
|
| 171556 |
+
"loss": 0.9755,
|
| 171557 |
+
"step": 56745
|
| 171558 |
+
},
|
| 171559 |
+
{
|
| 171560 |
+
"epoch": 453.0,
|
| 171561 |
+
"eval_loss": 0.4409070611000061,
|
| 171562 |
+
"eval_runtime": 39.436,
|
| 171563 |
+
"eval_samples_per_second": 21.3,
|
| 171564 |
+
"eval_steps_per_second": 0.685,
|
| 171565 |
+
"eval_wer": 0.20334678021813835,
|
| 171566 |
+
"step": 56746
|
| 171567 |
+
},
|
| 171568 |
+
{
|
| 171569 |
+
"epoch": 453.03,
|
| 171570 |
+
"learning_rate": 9.106907051282053e-06,
|
| 171571 |
+
"loss": 0.3777,
|
| 171572 |
+
"step": 56750
|
| 171573 |
+
},
|
| 171574 |
+
{
|
| 171575 |
+
"epoch": 453.07,
|
| 171576 |
+
"learning_rate": 9.106826923076924e-06,
|
| 171577 |
+
"loss": 0.3611,
|
| 171578 |
+
"step": 56755
|
| 171579 |
+
},
|
| 171580 |
+
{
|
| 171581 |
+
"epoch": 453.11,
|
| 171582 |
+
"learning_rate": 9.106746794871796e-06,
|
| 171583 |
+
"loss": 0.36,
|
| 171584 |
+
"step": 56760
|
| 171585 |
+
},
|
| 171586 |
+
{
|
| 171587 |
+
"epoch": 453.15,
|
| 171588 |
+
"learning_rate": 9.106666666666667e-06,
|
| 171589 |
+
"loss": 0.4627,
|
| 171590 |
+
"step": 56765
|
| 171591 |
+
},
|
| 171592 |
+
{
|
| 171593 |
+
"epoch": 453.19,
|
| 171594 |
+
"learning_rate": 9.10658653846154e-06,
|
| 171595 |
+
"loss": 0.8728,
|
| 171596 |
+
"step": 56770
|
| 171597 |
+
},
|
| 171598 |
+
{
|
| 171599 |
+
"epoch": 453.23,
|
| 171600 |
+
"learning_rate": 9.10650641025641e-06,
|
| 171601 |
+
"loss": 0.6227,
|
| 171602 |
+
"step": 56775
|
| 171603 |
+
},
|
| 171604 |
+
{
|
| 171605 |
+
"epoch": 453.27,
|
| 171606 |
+
"learning_rate": 9.106426282051283e-06,
|
| 171607 |
+
"loss": 0.3291,
|
| 171608 |
+
"step": 56780
|
| 171609 |
+
},
|
| 171610 |
+
{
|
| 171611 |
+
"epoch": 453.31,
|
| 171612 |
+
"learning_rate": 9.106346153846156e-06,
|
| 171613 |
+
"loss": 0.2861,
|
| 171614 |
+
"step": 56785
|
| 171615 |
+
},
|
| 171616 |
+
{
|
| 171617 |
+
"epoch": 453.35,
|
| 171618 |
+
"learning_rate": 9.106266025641025e-06,
|
| 171619 |
+
"loss": 0.4064,
|
| 171620 |
+
"step": 56790
|
| 171621 |
+
},
|
| 171622 |
+
{
|
| 171623 |
+
"epoch": 453.39,
|
| 171624 |
+
"learning_rate": 9.106185897435899e-06,
|
| 171625 |
+
"loss": 0.9348,
|
| 171626 |
+
"step": 56795
|
| 171627 |
+
},
|
| 171628 |
+
{
|
| 171629 |
+
"epoch": 453.43,
|
| 171630 |
+
"learning_rate": 9.10610576923077e-06,
|
| 171631 |
+
"loss": 0.6904,
|
| 171632 |
+
"step": 56800
|
| 171633 |
+
},
|
| 171634 |
+
{
|
| 171635 |
+
"epoch": 453.47,
|
| 171636 |
+
"learning_rate": 9.106025641025641e-06,
|
| 171637 |
+
"loss": 0.3354,
|
| 171638 |
+
"step": 56805
|
| 171639 |
+
},
|
| 171640 |
+
{
|
| 171641 |
+
"epoch": 453.51,
|
| 171642 |
+
"learning_rate": 9.105945512820513e-06,
|
| 171643 |
+
"loss": 0.3073,
|
| 171644 |
+
"step": 56810
|
| 171645 |
+
},
|
| 171646 |
+
{
|
| 171647 |
+
"epoch": 453.55,
|
| 171648 |
+
"learning_rate": 9.105865384615386e-06,
|
| 171649 |
+
"loss": 0.3934,
|
| 171650 |
+
"step": 56815
|
| 171651 |
+
},
|
| 171652 |
+
{
|
| 171653 |
+
"epoch": 453.59,
|
| 171654 |
+
"learning_rate": 9.105785256410257e-06,
|
| 171655 |
+
"loss": 1.0902,
|
| 171656 |
+
"step": 56820
|
| 171657 |
+
},
|
| 171658 |
+
{
|
| 171659 |
+
"epoch": 453.63,
|
| 171660 |
+
"learning_rate": 9.105705128205128e-06,
|
| 171661 |
+
"loss": 0.7669,
|
| 171662 |
+
"step": 56825
|
| 171663 |
+
},
|
| 171664 |
+
{
|
| 171665 |
+
"epoch": 453.67,
|
| 171666 |
+
"learning_rate": 9.105625000000001e-06,
|
| 171667 |
+
"loss": 0.3124,
|
| 171668 |
+
"step": 56830
|
| 171669 |
+
},
|
| 171670 |
+
{
|
| 171671 |
+
"epoch": 453.71,
|
| 171672 |
+
"learning_rate": 9.105544871794873e-06,
|
| 171673 |
+
"loss": 0.3776,
|
| 171674 |
+
"step": 56835
|
| 171675 |
+
},
|
| 171676 |
+
{
|
| 171677 |
+
"epoch": 453.75,
|
| 171678 |
+
"learning_rate": 9.105464743589744e-06,
|
| 171679 |
+
"loss": 0.4458,
|
| 171680 |
+
"step": 56840
|
| 171681 |
+
},
|
| 171682 |
+
{
|
| 171683 |
+
"epoch": 453.79,
|
| 171684 |
+
"learning_rate": 9.105384615384616e-06,
|
| 171685 |
+
"loss": 0.8865,
|
| 171686 |
+
"step": 56845
|
| 171687 |
+
},
|
| 171688 |
+
{
|
| 171689 |
+
"epoch": 453.83,
|
| 171690 |
+
"learning_rate": 9.105304487179489e-06,
|
| 171691 |
+
"loss": 0.8748,
|
| 171692 |
+
"step": 56850
|
| 171693 |
+
},
|
| 171694 |
+
{
|
| 171695 |
+
"epoch": 453.87,
|
| 171696 |
+
"learning_rate": 9.10522435897436e-06,
|
| 171697 |
+
"loss": 0.5539,
|
| 171698 |
+
"step": 56855
|
| 171699 |
+
},
|
| 171700 |
+
{
|
| 171701 |
+
"epoch": 453.91,
|
| 171702 |
+
"learning_rate": 9.105144230769231e-06,
|
| 171703 |
+
"loss": 0.3233,
|
| 171704 |
+
"step": 56860
|
| 171705 |
+
},
|
| 171706 |
+
{
|
| 171707 |
+
"epoch": 453.95,
|
| 171708 |
+
"learning_rate": 9.105064102564103e-06,
|
| 171709 |
+
"loss": 0.4554,
|
| 171710 |
+
"step": 56865
|
| 171711 |
+
},
|
| 171712 |
+
{
|
| 171713 |
+
"epoch": 453.99,
|
| 171714 |
+
"learning_rate": 9.104983974358976e-06,
|
| 171715 |
+
"loss": 0.9328,
|
| 171716 |
+
"step": 56870
|
| 171717 |
+
},
|
| 171718 |
+
{
|
| 171719 |
+
"epoch": 454.0,
|
| 171720 |
+
"eval_loss": 0.44104063510894775,
|
| 171721 |
+
"eval_runtime": 40.1551,
|
| 171722 |
+
"eval_samples_per_second": 20.894,
|
| 171723 |
+
"eval_steps_per_second": 0.672,
|
| 171724 |
+
"eval_wer": 0.2034719710669078,
|
| 171725 |
+
"step": 56871
|
| 171726 |
+
},
|
| 171727 |
+
{
|
| 171728 |
+
"epoch": 462.03,
|
| 171729 |
+
"learning_rate": 9.104903846153847e-06,
|
| 171730 |
+
"loss": 0.4305,
|
| 171731 |
+
"step": 56875
|
| 171732 |
+
},
|
| 171733 |
+
{
|
| 171734 |
+
"epoch": 462.07,
|
| 171735 |
+
"learning_rate": 9.104823717948718e-06,
|
| 171736 |
+
"loss": 0.3532,
|
| 171737 |
+
"step": 56880
|
| 171738 |
+
},
|
| 171739 |
+
{
|
| 171740 |
+
"epoch": 462.11,
|
| 171741 |
+
"learning_rate": 9.104743589743591e-06,
|
| 171742 |
+
"loss": 0.3565,
|
| 171743 |
+
"step": 56885
|
| 171744 |
+
},
|
| 171745 |
+
{
|
| 171746 |
+
"epoch": 462.15,
|
| 171747 |
+
"learning_rate": 9.104663461538463e-06,
|
| 171748 |
+
"loss": 0.5439,
|
| 171749 |
+
"step": 56890
|
| 171750 |
+
},
|
| 171751 |
+
{
|
| 171752 |
+
"epoch": 462.19,
|
| 171753 |
+
"learning_rate": 9.104583333333334e-06,
|
| 171754 |
+
"loss": 0.975,
|
| 171755 |
+
"step": 56895
|
| 171756 |
+
},
|
| 171757 |
+
{
|
| 171758 |
+
"epoch": 462.23,
|
| 171759 |
+
"learning_rate": 9.104503205128206e-06,
|
| 171760 |
+
"loss": 0.6971,
|
| 171761 |
+
"step": 56900
|
| 171762 |
+
},
|
| 171763 |
+
{
|
| 171764 |
+
"epoch": 462.28,
|
| 171765 |
+
"learning_rate": 9.104423076923079e-06,
|
| 171766 |
+
"loss": 0.287,
|
| 171767 |
+
"step": 56905
|
| 171768 |
+
},
|
| 171769 |
+
{
|
| 171770 |
+
"epoch": 462.32,
|
| 171771 |
+
"learning_rate": 9.104342948717948e-06,
|
| 171772 |
+
"loss": 0.3621,
|
| 171773 |
+
"step": 56910
|
| 171774 |
+
},
|
| 171775 |
+
{
|
| 171776 |
+
"epoch": 462.36,
|
| 171777 |
+
"learning_rate": 9.104262820512821e-06,
|
| 171778 |
+
"loss": 0.4668,
|
| 171779 |
+
"step": 56915
|
| 171780 |
+
},
|
| 171781 |
+
{
|
| 171782 |
+
"epoch": 462.4,
|
| 171783 |
+
"learning_rate": 9.104182692307693e-06,
|
| 171784 |
+
"loss": 0.9692,
|
| 171785 |
+
"step": 56920
|
| 171786 |
+
},
|
| 171787 |
+
{
|
| 171788 |
+
"epoch": 462.44,
|
| 171789 |
+
"learning_rate": 9.104102564102566e-06,
|
| 171790 |
+
"loss": 0.6832,
|
| 171791 |
+
"step": 56925
|
| 171792 |
+
},
|
| 171793 |
+
{
|
| 171794 |
+
"epoch": 462.48,
|
| 171795 |
+
"learning_rate": 9.104022435897437e-06,
|
| 171796 |
+
"loss": 0.3668,
|
| 171797 |
+
"step": 56930
|
| 171798 |
+
},
|
| 171799 |
+
{
|
| 171800 |
+
"epoch": 462.52,
|
| 171801 |
+
"learning_rate": 9.103942307692308e-06,
|
| 171802 |
+
"loss": 0.3313,
|
| 171803 |
+
"step": 56935
|
| 171804 |
+
},
|
| 171805 |
+
{
|
| 171806 |
+
"epoch": 462.56,
|
| 171807 |
+
"learning_rate": 9.103862179487181e-06,
|
| 171808 |
+
"loss": 0.4229,
|
| 171809 |
+
"step": 56940
|
| 171810 |
+
},
|
| 171811 |
+
{
|
| 171812 |
+
"epoch": 462.6,
|
| 171813 |
+
"learning_rate": 9.103782051282051e-06,
|
| 171814 |
+
"loss": 0.9475,
|
| 171815 |
+
"step": 56945
|
| 171816 |
+
},
|
| 171817 |
+
{
|
| 171818 |
+
"epoch": 462.64,
|
| 171819 |
+
"learning_rate": 9.103701923076924e-06,
|
| 171820 |
+
"loss": 0.7394,
|
| 171821 |
+
"step": 56950
|
| 171822 |
+
},
|
| 171823 |
+
{
|
| 171824 |
+
"epoch": 462.68,
|
| 171825 |
+
"learning_rate": 9.103621794871796e-06,
|
| 171826 |
+
"loss": 0.306,
|
| 171827 |
+
"step": 56955
|
| 171828 |
+
},
|
| 171829 |
+
{
|
| 171830 |
+
"epoch": 462.72,
|
| 171831 |
+
"learning_rate": 9.103541666666667e-06,
|
| 171832 |
+
"loss": 0.2872,
|
| 171833 |
+
"step": 56960
|
| 171834 |
+
},
|
| 171835 |
+
{
|
| 171836 |
+
"epoch": 462.76,
|
| 171837 |
+
"learning_rate": 9.103461538461538e-06,
|
| 171838 |
+
"loss": 0.4024,
|
| 171839 |
+
"step": 56965
|
| 171840 |
+
},
|
| 171841 |
+
{
|
| 171842 |
+
"epoch": 462.8,
|
| 171843 |
+
"learning_rate": 9.103381410256411e-06,
|
| 171844 |
+
"loss": 0.9516,
|
| 171845 |
+
"step": 56970
|
| 171846 |
+
},
|
| 171847 |
+
{
|
| 171848 |
+
"epoch": 462.84,
|
| 171849 |
+
"learning_rate": 9.103301282051283e-06,
|
| 171850 |
+
"loss": 0.7913,
|
| 171851 |
+
"step": 56975
|
| 171852 |
+
},
|
| 171853 |
+
{
|
| 171854 |
+
"epoch": 462.88,
|
| 171855 |
+
"learning_rate": 9.103221153846154e-06,
|
| 171856 |
+
"loss": 0.3061,
|
| 171857 |
+
"step": 56980
|
| 171858 |
+
},
|
| 171859 |
+
{
|
| 171860 |
+
"epoch": 462.92,
|
| 171861 |
+
"learning_rate": 9.103141025641027e-06,
|
| 171862 |
+
"loss": 0.4049,
|
| 171863 |
+
"step": 56985
|
| 171864 |
+
},
|
| 171865 |
+
{
|
| 171866 |
+
"epoch": 462.96,
|
| 171867 |
+
"learning_rate": 9.103060897435898e-06,
|
| 171868 |
+
"loss": 0.4995,
|
| 171869 |
+
"step": 56990
|
| 171870 |
+
},
|
| 171871 |
+
{
|
| 171872 |
+
"epoch": 463.0,
|
| 171873 |
+
"eval_loss": 0.38458994030952454,
|
| 171874 |
+
"eval_runtime": 40.5268,
|
| 171875 |
+
"eval_samples_per_second": 20.702,
|
| 171876 |
+
"eval_steps_per_second": 0.666,
|
| 171877 |
+
"eval_wer": 0.1988261188554659,
|
| 171878 |
+
"step": 56994
|
| 171879 |
}
|
| 171880 |
],
|
| 171881 |
+
"max_steps": 615000,
|
| 171882 |
"num_train_epochs": 5000,
|
| 171883 |
+
"total_flos": 1.6039691475592757e+20,
|
| 171884 |
"trial_name": null,
|
| 171885 |
"trial_params": null
|
| 171886 |
}
|
model-bin/finetune/base/{checkpoint-56372 → checkpoint-56994}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629778472.5431197/events.out.tfevents.1629778472.c435e1c5ee04.920.41
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:fd4455d0a891e6f5a00d31a4462d63094040cdb10d0a46c1db5a0d46ff972af5
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629779133.0815535/events.out.tfevents.1629779133.c435e1c5ee04.920.43
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9680d8d37e919800b1254384b37840a50341d557714bd38ba236c6e0fe996ff5
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629779785.3045485/events.out.tfevents.1629779785.c435e1c5ee04.920.45
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ee94580bf491928c6738e5c944d6013c9cec35d37b3a9625920f7229985d6965
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629780436.4732363/events.out.tfevents.1629780436.c435e1c5ee04.920.47
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:37c6d96d0c2be205fe016a35592a3ef051169be88525e373fd581cd556266145
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629781090.5327764/events.out.tfevents.1629781090.c435e1c5ee04.920.49
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:49f019551c5dd52a891e2dc4a6fc21d09861d20be5eb27f09e533c9a9c82802e
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629778472.c435e1c5ee04.920.40
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1751399f588b9effc54dd7e7367165c47a52e73cefa571f4b33aecfd699afb1b
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629779133.c435e1c5ee04.920.42
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c240a440bb6c8ff1993bc3eb5c21851b0c0c31e07baf833620b89c9822e43809
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629779785.c435e1c5ee04.920.44
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:dd2cd03fefff7ac84ffb1d9abe4c882573c8539fcb787644c74f6cf152de04c3
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629780436.c435e1c5ee04.920.46
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a4b19575163b50308202cef4a37d406146eb02794863b51c5696db008e67af57
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629781090.c435e1c5ee04.920.48
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8533a5c3626250d1ea47e29be6f2c15af0ee72559fa9d25af58f76c29fbc2c60
|
| 3 |
+
size 8462
|