"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-53883 → checkpoint-54505}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-53883 → checkpoint-54505}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-53883 → checkpoint-54505}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-53883 → checkpoint-54505}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-53883 → checkpoint-54505}/rng_state.pth +2 -2
- model-bin/finetune/base/{checkpoint-53883 → checkpoint-54505}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-53883 → checkpoint-54505}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-53883 → checkpoint-54505}/trainer_state.json +799 -4
- model-bin/finetune/base/{checkpoint-53883 → checkpoint-54505}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629764971.3905332/events.out.tfevents.1629764971.c435e1c5ee04.920.1 +3 -0
- model-bin/finetune/base/log/1629765618.347005/events.out.tfevents.1629765618.c435e1c5ee04.920.3 +3 -0
- model-bin/finetune/base/log/1629766288.11667/events.out.tfevents.1629766288.c435e1c5ee04.920.5 +3 -0
- model-bin/finetune/base/log/1629766954.1795254/events.out.tfevents.1629766954.c435e1c5ee04.920.7 +3 -0
- model-bin/finetune/base/log/1629767616.4417202/events.out.tfevents.1629767616.c435e1c5ee04.920.9 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629764971.c435e1c5ee04.920.0 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629765618.c435e1c5ee04.920.2 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629766288.c435e1c5ee04.920.4 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629766954.c435e1c5ee04.920.6 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629767616.c435e1c5ee04.920.8 +3 -0
model-bin/finetune/base/{checkpoint-53883 → checkpoint-54505}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-53883 → checkpoint-54505}/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 sha256:175a41a98f5203d0db6aa1fbd6a9ae7b059e2710fa700b5bea11a57ff3220cf0
|
| 3 |
size 722165009
|
model-bin/finetune/base/{checkpoint-53883 → checkpoint-54505}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-53883 → checkpoint-54505}/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:ec4d2a6a7040c89cd57e0b9cb66beef6221d858761995ab0704ac3e0af89527a
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-53883 → checkpoint-54505}/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:a41625b4db78a0c89b01c36cbddae265eafff10b0d61888fbd70eaf471ffd95f
|
| 3 |
+
size 14567
|
model-bin/finetune/base/{checkpoint-53883 → checkpoint-54505}/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 sha256:2e3fbcb39f68cfc7ed849162b4e354f6037864a19c979d4bf441c7c266c9efa4
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-53883 → checkpoint-54505}/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:f43a9c3edd8c53660caa0eb9f925a632bd24b739141d5b8e929015a7219a86ed
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-53883 → checkpoint-54505}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.18992848189928482,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-52515",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -167919,11 +167919,806 @@
|
|
| 167919 |
"eval_steps_per_second": 0.665,
|
| 167920 |
"eval_wer": 0.20484629294755877,
|
| 167921 |
"step": 53883
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 167922 |
}
|
| 167923 |
],
|
| 167924 |
-
"max_steps":
|
| 167925 |
"num_train_epochs": 5000,
|
| 167926 |
-
"total_flos": 1.
|
| 167927 |
"trial_name": null,
|
| 167928 |
"trial_params": null
|
| 167929 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.18992848189928482,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-52515",
|
| 4 |
+
"epoch": 438.99598393574297,
|
| 5 |
+
"global_step": 54505,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 167919 |
"eval_steps_per_second": 0.665,
|
| 167920 |
"eval_wer": 0.20484629294755877,
|
| 167921 |
"step": 53883
|
| 167922 |
+
},
|
| 167923 |
+
{
|
| 167924 |
+
"epoch": 431.02,
|
| 167925 |
+
"learning_rate": 9.152804487179488e-06,
|
| 167926 |
+
"loss": 0.419,
|
| 167927 |
+
"step": 53885
|
| 167928 |
+
},
|
| 167929 |
+
{
|
| 167930 |
+
"epoch": 431.06,
|
| 167931 |
+
"learning_rate": 9.152724358974359e-06,
|
| 167932 |
+
"loss": 0.3243,
|
| 167933 |
+
"step": 53890
|
| 167934 |
+
},
|
| 167935 |
+
{
|
| 167936 |
+
"epoch": 431.1,
|
| 167937 |
+
"learning_rate": 9.152644230769232e-06,
|
| 167938 |
+
"loss": 0.341,
|
| 167939 |
+
"step": 53895
|
| 167940 |
+
},
|
| 167941 |
+
{
|
| 167942 |
+
"epoch": 431.14,
|
| 167943 |
+
"learning_rate": 9.152564102564103e-06,
|
| 167944 |
+
"loss": 0.4508,
|
| 167945 |
+
"step": 53900
|
| 167946 |
+
},
|
| 167947 |
+
{
|
| 167948 |
+
"epoch": 431.18,
|
| 167949 |
+
"learning_rate": 9.152483974358975e-06,
|
| 167950 |
+
"loss": 0.7833,
|
| 167951 |
+
"step": 53905
|
| 167952 |
+
},
|
| 167953 |
+
{
|
| 167954 |
+
"epoch": 431.22,
|
| 167955 |
+
"learning_rate": 9.152403846153846e-06,
|
| 167956 |
+
"loss": 1.0565,
|
| 167957 |
+
"step": 53910
|
| 167958 |
+
},
|
| 167959 |
+
{
|
| 167960 |
+
"epoch": 431.25,
|
| 167961 |
+
"learning_rate": 9.15232371794872e-06,
|
| 167962 |
+
"loss": 0.3407,
|
| 167963 |
+
"step": 53915
|
| 167964 |
+
},
|
| 167965 |
+
{
|
| 167966 |
+
"epoch": 431.29,
|
| 167967 |
+
"learning_rate": 9.15224358974359e-06,
|
| 167968 |
+
"loss": 0.3398,
|
| 167969 |
+
"step": 53920
|
| 167970 |
+
},
|
| 167971 |
+
{
|
| 167972 |
+
"epoch": 431.33,
|
| 167973 |
+
"learning_rate": 9.152163461538462e-06,
|
| 167974 |
+
"loss": 0.3471,
|
| 167975 |
+
"step": 53925
|
| 167976 |
+
},
|
| 167977 |
+
{
|
| 167978 |
+
"epoch": 431.37,
|
| 167979 |
+
"learning_rate": 9.152083333333333e-06,
|
| 167980 |
+
"loss": 0.7131,
|
| 167981 |
+
"step": 53930
|
| 167982 |
+
},
|
| 167983 |
+
{
|
| 167984 |
+
"epoch": 431.41,
|
| 167985 |
+
"learning_rate": 9.152003205128206e-06,
|
| 167986 |
+
"loss": 0.996,
|
| 167987 |
+
"step": 53935
|
| 167988 |
+
},
|
| 167989 |
+
{
|
| 167990 |
+
"epoch": 431.45,
|
| 167991 |
+
"learning_rate": 9.151923076923078e-06,
|
| 167992 |
+
"loss": 0.3824,
|
| 167993 |
+
"step": 53940
|
| 167994 |
+
},
|
| 167995 |
+
{
|
| 167996 |
+
"epoch": 431.49,
|
| 167997 |
+
"learning_rate": 9.151842948717949e-06,
|
| 167998 |
+
"loss": 0.321,
|
| 167999 |
+
"step": 53945
|
| 168000 |
+
},
|
| 168001 |
+
{
|
| 168002 |
+
"epoch": 431.53,
|
| 168003 |
+
"learning_rate": 9.151762820512822e-06,
|
| 168004 |
+
"loss": 0.3952,
|
| 168005 |
+
"step": 53950
|
| 168006 |
+
},
|
| 168007 |
+
{
|
| 168008 |
+
"epoch": 431.57,
|
| 168009 |
+
"learning_rate": 9.151682692307693e-06,
|
| 168010 |
+
"loss": 0.6883,
|
| 168011 |
+
"step": 53955
|
| 168012 |
+
},
|
| 168013 |
+
{
|
| 168014 |
+
"epoch": 431.61,
|
| 168015 |
+
"learning_rate": 9.151618589743591e-06,
|
| 168016 |
+
"loss": 1.5506,
|
| 168017 |
+
"step": 53960
|
| 168018 |
+
},
|
| 168019 |
+
{
|
| 168020 |
+
"epoch": 431.65,
|
| 168021 |
+
"learning_rate": 9.151538461538462e-06,
|
| 168022 |
+
"loss": 0.3653,
|
| 168023 |
+
"step": 53965
|
| 168024 |
+
},
|
| 168025 |
+
{
|
| 168026 |
+
"epoch": 431.69,
|
| 168027 |
+
"learning_rate": 9.151458333333334e-06,
|
| 168028 |
+
"loss": 0.4062,
|
| 168029 |
+
"step": 53970
|
| 168030 |
+
},
|
| 168031 |
+
{
|
| 168032 |
+
"epoch": 431.73,
|
| 168033 |
+
"learning_rate": 9.151378205128207e-06,
|
| 168034 |
+
"loss": 0.3621,
|
| 168035 |
+
"step": 53975
|
| 168036 |
+
},
|
| 168037 |
+
{
|
| 168038 |
+
"epoch": 431.77,
|
| 168039 |
+
"learning_rate": 9.151298076923078e-06,
|
| 168040 |
+
"loss": 0.6849,
|
| 168041 |
+
"step": 53980
|
| 168042 |
+
},
|
| 168043 |
+
{
|
| 168044 |
+
"epoch": 431.81,
|
| 168045 |
+
"learning_rate": 9.15121794871795e-06,
|
| 168046 |
+
"loss": 1.0669,
|
| 168047 |
+
"step": 53985
|
| 168048 |
+
},
|
| 168049 |
+
{
|
| 168050 |
+
"epoch": 431.85,
|
| 168051 |
+
"learning_rate": 9.15113782051282e-06,
|
| 168052 |
+
"loss": 0.3909,
|
| 168053 |
+
"step": 53990
|
| 168054 |
+
},
|
| 168055 |
+
{
|
| 168056 |
+
"epoch": 431.89,
|
| 168057 |
+
"learning_rate": 9.151057692307694e-06,
|
| 168058 |
+
"loss": 0.3625,
|
| 168059 |
+
"step": 53995
|
| 168060 |
+
},
|
| 168061 |
+
{
|
| 168062 |
+
"epoch": 431.93,
|
| 168063 |
+
"learning_rate": 9.150977564102565e-06,
|
| 168064 |
+
"loss": 0.4017,
|
| 168065 |
+
"step": 54000
|
| 168066 |
+
},
|
| 168067 |
+
{
|
| 168068 |
+
"epoch": 431.97,
|
| 168069 |
+
"learning_rate": 9.150897435897437e-06,
|
| 168070 |
+
"loss": 0.7696,
|
| 168071 |
+
"step": 54005
|
| 168072 |
+
},
|
| 168073 |
+
{
|
| 168074 |
+
"epoch": 432.0,
|
| 168075 |
+
"eval_loss": 0.37287962436676025,
|
| 168076 |
+
"eval_runtime": 38.0521,
|
| 168077 |
+
"eval_samples_per_second": 22.049,
|
| 168078 |
+
"eval_steps_per_second": 0.71,
|
| 168079 |
+
"eval_wer": 0.20122172932877608,
|
| 168080 |
+
"step": 54008
|
| 168081 |
+
},
|
| 168082 |
+
{
|
| 168083 |
+
"epoch": 435.02,
|
| 168084 |
+
"learning_rate": 9.150817307692308e-06,
|
| 168085 |
+
"loss": 0.3971,
|
| 168086 |
+
"step": 54010
|
| 168087 |
+
},
|
| 168088 |
+
{
|
| 168089 |
+
"epoch": 435.06,
|
| 168090 |
+
"learning_rate": 9.150737179487181e-06,
|
| 168091 |
+
"loss": 0.3162,
|
| 168092 |
+
"step": 54015
|
| 168093 |
+
},
|
| 168094 |
+
{
|
| 168095 |
+
"epoch": 435.1,
|
| 168096 |
+
"learning_rate": 9.150657051282052e-06,
|
| 168097 |
+
"loss": 0.3087,
|
| 168098 |
+
"step": 54020
|
| 168099 |
+
},
|
| 168100 |
+
{
|
| 168101 |
+
"epoch": 435.14,
|
| 168102 |
+
"learning_rate": 9.150576923076924e-06,
|
| 168103 |
+
"loss": 0.3867,
|
| 168104 |
+
"step": 54025
|
| 168105 |
+
},
|
| 168106 |
+
{
|
| 168107 |
+
"epoch": 435.18,
|
| 168108 |
+
"learning_rate": 9.150496794871797e-06,
|
| 168109 |
+
"loss": 0.6681,
|
| 168110 |
+
"step": 54030
|
| 168111 |
+
},
|
| 168112 |
+
{
|
| 168113 |
+
"epoch": 435.22,
|
| 168114 |
+
"learning_rate": 9.150416666666666e-06,
|
| 168115 |
+
"loss": 1.0952,
|
| 168116 |
+
"step": 54035
|
| 168117 |
+
},
|
| 168118 |
+
{
|
| 168119 |
+
"epoch": 435.26,
|
| 168120 |
+
"learning_rate": 9.15033653846154e-06,
|
| 168121 |
+
"loss": 0.3676,
|
| 168122 |
+
"step": 54040
|
| 168123 |
+
},
|
| 168124 |
+
{
|
| 168125 |
+
"epoch": 435.3,
|
| 168126 |
+
"learning_rate": 9.15025641025641e-06,
|
| 168127 |
+
"loss": 0.3552,
|
| 168128 |
+
"step": 54045
|
| 168129 |
+
},
|
| 168130 |
+
{
|
| 168131 |
+
"epoch": 435.34,
|
| 168132 |
+
"learning_rate": 9.150176282051282e-06,
|
| 168133 |
+
"loss": 0.3858,
|
| 168134 |
+
"step": 54050
|
| 168135 |
+
},
|
| 168136 |
+
{
|
| 168137 |
+
"epoch": 435.38,
|
| 168138 |
+
"learning_rate": 9.150096153846154e-06,
|
| 168139 |
+
"loss": 0.7447,
|
| 168140 |
+
"step": 54055
|
| 168141 |
+
},
|
| 168142 |
+
{
|
| 168143 |
+
"epoch": 435.42,
|
| 168144 |
+
"learning_rate": 9.150016025641027e-06,
|
| 168145 |
+
"loss": 1.1432,
|
| 168146 |
+
"step": 54060
|
| 168147 |
+
},
|
| 168148 |
+
{
|
| 168149 |
+
"epoch": 435.46,
|
| 168150 |
+
"learning_rate": 9.149935897435898e-06,
|
| 168151 |
+
"loss": 0.2896,
|
| 168152 |
+
"step": 54065
|
| 168153 |
+
},
|
| 168154 |
+
{
|
| 168155 |
+
"epoch": 435.5,
|
| 168156 |
+
"learning_rate": 9.14985576923077e-06,
|
| 168157 |
+
"loss": 0.3781,
|
| 168158 |
+
"step": 54070
|
| 168159 |
+
},
|
| 168160 |
+
{
|
| 168161 |
+
"epoch": 435.54,
|
| 168162 |
+
"learning_rate": 9.149775641025642e-06,
|
| 168163 |
+
"loss": 0.3899,
|
| 168164 |
+
"step": 54075
|
| 168165 |
+
},
|
| 168166 |
+
{
|
| 168167 |
+
"epoch": 435.58,
|
| 168168 |
+
"learning_rate": 9.149695512820514e-06,
|
| 168169 |
+
"loss": 0.6563,
|
| 168170 |
+
"step": 54080
|
| 168171 |
+
},
|
| 168172 |
+
{
|
| 168173 |
+
"epoch": 435.62,
|
| 168174 |
+
"learning_rate": 9.149615384615385e-06,
|
| 168175 |
+
"loss": 1.1815,
|
| 168176 |
+
"step": 54085
|
| 168177 |
+
},
|
| 168178 |
+
{
|
| 168179 |
+
"epoch": 435.66,
|
| 168180 |
+
"learning_rate": 9.149535256410256e-06,
|
| 168181 |
+
"loss": 0.4117,
|
| 168182 |
+
"step": 54090
|
| 168183 |
+
},
|
| 168184 |
+
{
|
| 168185 |
+
"epoch": 435.7,
|
| 168186 |
+
"learning_rate": 9.14945512820513e-06,
|
| 168187 |
+
"loss": 0.2979,
|
| 168188 |
+
"step": 54095
|
| 168189 |
+
},
|
| 168190 |
+
{
|
| 168191 |
+
"epoch": 435.74,
|
| 168192 |
+
"learning_rate": 9.149375e-06,
|
| 168193 |
+
"loss": 0.3699,
|
| 168194 |
+
"step": 54100
|
| 168195 |
+
},
|
| 168196 |
+
{
|
| 168197 |
+
"epoch": 435.78,
|
| 168198 |
+
"learning_rate": 9.149294871794872e-06,
|
| 168199 |
+
"loss": 0.6889,
|
| 168200 |
+
"step": 54105
|
| 168201 |
+
},
|
| 168202 |
+
{
|
| 168203 |
+
"epoch": 435.82,
|
| 168204 |
+
"learning_rate": 9.149214743589744e-06,
|
| 168205 |
+
"loss": 1.0284,
|
| 168206 |
+
"step": 54110
|
| 168207 |
+
},
|
| 168208 |
+
{
|
| 168209 |
+
"epoch": 435.86,
|
| 168210 |
+
"learning_rate": 9.149134615384617e-06,
|
| 168211 |
+
"loss": 0.3884,
|
| 168212 |
+
"step": 54115
|
| 168213 |
+
},
|
| 168214 |
+
{
|
| 168215 |
+
"epoch": 435.9,
|
| 168216 |
+
"learning_rate": 9.149054487179488e-06,
|
| 168217 |
+
"loss": 0.3676,
|
| 168218 |
+
"step": 54120
|
| 168219 |
+
},
|
| 168220 |
+
{
|
| 168221 |
+
"epoch": 435.94,
|
| 168222 |
+
"learning_rate": 9.14897435897436e-06,
|
| 168223 |
+
"loss": 0.3872,
|
| 168224 |
+
"step": 54125
|
| 168225 |
+
},
|
| 168226 |
+
{
|
| 168227 |
+
"epoch": 435.98,
|
| 168228 |
+
"learning_rate": 9.148894230769232e-06,
|
| 168229 |
+
"loss": 0.8216,
|
| 168230 |
+
"step": 54130
|
| 168231 |
+
},
|
| 168232 |
+
{
|
| 168233 |
+
"epoch": 436.0,
|
| 168234 |
+
"eval_loss": 0.44797709584236145,
|
| 168235 |
+
"eval_runtime": 40.441,
|
| 168236 |
+
"eval_samples_per_second": 20.771,
|
| 168237 |
+
"eval_steps_per_second": 0.668,
|
| 168238 |
+
"eval_wer": 0.19947140444901254,
|
| 168239 |
+
"step": 54132
|
| 168240 |
+
},
|
| 168241 |
+
{
|
| 168242 |
+
"epoch": 436.02,
|
| 168243 |
+
"learning_rate": 9.148814102564104e-06,
|
| 168244 |
+
"loss": 0.3841,
|
| 168245 |
+
"step": 54135
|
| 168246 |
+
},
|
| 168247 |
+
{
|
| 168248 |
+
"epoch": 436.06,
|
| 168249 |
+
"learning_rate": 9.148733974358975e-06,
|
| 168250 |
+
"loss": 0.364,
|
| 168251 |
+
"step": 54140
|
| 168252 |
+
},
|
| 168253 |
+
{
|
| 168254 |
+
"epoch": 436.1,
|
| 168255 |
+
"learning_rate": 9.148653846153846e-06,
|
| 168256 |
+
"loss": 0.384,
|
| 168257 |
+
"step": 54145
|
| 168258 |
+
},
|
| 168259 |
+
{
|
| 168260 |
+
"epoch": 436.14,
|
| 168261 |
+
"learning_rate": 9.14857371794872e-06,
|
| 168262 |
+
"loss": 0.374,
|
| 168263 |
+
"step": 54150
|
| 168264 |
+
},
|
| 168265 |
+
{
|
| 168266 |
+
"epoch": 436.18,
|
| 168267 |
+
"learning_rate": 9.14849358974359e-06,
|
| 168268 |
+
"loss": 0.8492,
|
| 168269 |
+
"step": 54155
|
| 168270 |
+
},
|
| 168271 |
+
{
|
| 168272 |
+
"epoch": 436.22,
|
| 168273 |
+
"learning_rate": 9.148413461538462e-06,
|
| 168274 |
+
"loss": 0.9104,
|
| 168275 |
+
"step": 54160
|
| 168276 |
+
},
|
| 168277 |
+
{
|
| 168278 |
+
"epoch": 436.27,
|
| 168279 |
+
"learning_rate": 9.148333333333335e-06,
|
| 168280 |
+
"loss": 0.3081,
|
| 168281 |
+
"step": 54165
|
| 168282 |
+
},
|
| 168283 |
+
{
|
| 168284 |
+
"epoch": 436.31,
|
| 168285 |
+
"learning_rate": 9.148253205128207e-06,
|
| 168286 |
+
"loss": 0.3592,
|
| 168287 |
+
"step": 54170
|
| 168288 |
+
},
|
| 168289 |
+
{
|
| 168290 |
+
"epoch": 436.35,
|
| 168291 |
+
"learning_rate": 9.148173076923078e-06,
|
| 168292 |
+
"loss": 0.4217,
|
| 168293 |
+
"step": 54175
|
| 168294 |
+
},
|
| 168295 |
+
{
|
| 168296 |
+
"epoch": 436.39,
|
| 168297 |
+
"learning_rate": 9.14809294871795e-06,
|
| 168298 |
+
"loss": 0.888,
|
| 168299 |
+
"step": 54180
|
| 168300 |
+
},
|
| 168301 |
+
{
|
| 168302 |
+
"epoch": 436.43,
|
| 168303 |
+
"learning_rate": 9.148012820512822e-06,
|
| 168304 |
+
"loss": 1.0466,
|
| 168305 |
+
"step": 54185
|
| 168306 |
+
},
|
| 168307 |
+
{
|
| 168308 |
+
"epoch": 436.47,
|
| 168309 |
+
"learning_rate": 9.147932692307692e-06,
|
| 168310 |
+
"loss": 0.3215,
|
| 168311 |
+
"step": 54190
|
| 168312 |
+
},
|
| 168313 |
+
{
|
| 168314 |
+
"epoch": 436.51,
|
| 168315 |
+
"learning_rate": 9.147852564102565e-06,
|
| 168316 |
+
"loss": 0.3265,
|
| 168317 |
+
"step": 54195
|
| 168318 |
+
},
|
| 168319 |
+
{
|
| 168320 |
+
"epoch": 436.55,
|
| 168321 |
+
"learning_rate": 9.147772435897436e-06,
|
| 168322 |
+
"loss": 0.4732,
|
| 168323 |
+
"step": 54200
|
| 168324 |
+
},
|
| 168325 |
+
{
|
| 168326 |
+
"epoch": 436.59,
|
| 168327 |
+
"learning_rate": 9.147692307692308e-06,
|
| 168328 |
+
"loss": 0.9318,
|
| 168329 |
+
"step": 54205
|
| 168330 |
+
},
|
| 168331 |
+
{
|
| 168332 |
+
"epoch": 436.63,
|
| 168333 |
+
"learning_rate": 9.14761217948718e-06,
|
| 168334 |
+
"loss": 0.8966,
|
| 168335 |
+
"step": 54210
|
| 168336 |
+
},
|
| 168337 |
+
{
|
| 168338 |
+
"epoch": 436.67,
|
| 168339 |
+
"learning_rate": 9.147532051282052e-06,
|
| 168340 |
+
"loss": 0.3186,
|
| 168341 |
+
"step": 54215
|
| 168342 |
+
},
|
| 168343 |
+
{
|
| 168344 |
+
"epoch": 436.71,
|
| 168345 |
+
"learning_rate": 9.147451923076924e-06,
|
| 168346 |
+
"loss": 0.3307,
|
| 168347 |
+
"step": 54220
|
| 168348 |
+
},
|
| 168349 |
+
{
|
| 168350 |
+
"epoch": 436.75,
|
| 168351 |
+
"learning_rate": 9.147371794871795e-06,
|
| 168352 |
+
"loss": 0.4654,
|
| 168353 |
+
"step": 54225
|
| 168354 |
+
},
|
| 168355 |
+
{
|
| 168356 |
+
"epoch": 436.79,
|
| 168357 |
+
"learning_rate": 9.147291666666668e-06,
|
| 168358 |
+
"loss": 0.7942,
|
| 168359 |
+
"step": 54230
|
| 168360 |
+
},
|
| 168361 |
+
{
|
| 168362 |
+
"epoch": 436.83,
|
| 168363 |
+
"learning_rate": 9.14721153846154e-06,
|
| 168364 |
+
"loss": 0.9858,
|
| 168365 |
+
"step": 54235
|
| 168366 |
+
},
|
| 168367 |
+
{
|
| 168368 |
+
"epoch": 436.87,
|
| 168369 |
+
"learning_rate": 9.14713141025641e-06,
|
| 168370 |
+
"loss": 0.3137,
|
| 168371 |
+
"step": 54240
|
| 168372 |
+
},
|
| 168373 |
+
{
|
| 168374 |
+
"epoch": 436.91,
|
| 168375 |
+
"learning_rate": 9.147051282051282e-06,
|
| 168376 |
+
"loss": 0.3034,
|
| 168377 |
+
"step": 54245
|
| 168378 |
+
},
|
| 168379 |
+
{
|
| 168380 |
+
"epoch": 436.95,
|
| 168381 |
+
"learning_rate": 9.146971153846155e-06,
|
| 168382 |
+
"loss": 0.4034,
|
| 168383 |
+
"step": 54250
|
| 168384 |
+
},
|
| 168385 |
+
{
|
| 168386 |
+
"epoch": 436.99,
|
| 168387 |
+
"learning_rate": 9.146891025641026e-06,
|
| 168388 |
+
"loss": 0.9007,
|
| 168389 |
+
"step": 54255
|
| 168390 |
+
},
|
| 168391 |
+
{
|
| 168392 |
+
"epoch": 437.0,
|
| 168393 |
+
"eval_loss": 0.41687604784965515,
|
| 168394 |
+
"eval_runtime": 39.6761,
|
| 168395 |
+
"eval_samples_per_second": 21.171,
|
| 168396 |
+
"eval_steps_per_second": 0.681,
|
| 168397 |
+
"eval_wer": 0.19979597784902361,
|
| 168398 |
+
"step": 54256
|
| 168399 |
+
},
|
| 168400 |
+
{
|
| 168401 |
+
"epoch": 434.03,
|
| 168402 |
+
"learning_rate": 9.146810897435898e-06,
|
| 168403 |
+
"loss": 0.366,
|
| 168404 |
+
"step": 54260
|
| 168405 |
+
},
|
| 168406 |
+
{
|
| 168407 |
+
"epoch": 434.07,
|
| 168408 |
+
"learning_rate": 9.146730769230771e-06,
|
| 168409 |
+
"loss": 0.3047,
|
| 168410 |
+
"step": 54265
|
| 168411 |
+
},
|
| 168412 |
+
{
|
| 168413 |
+
"epoch": 434.11,
|
| 168414 |
+
"learning_rate": 9.146650641025642e-06,
|
| 168415 |
+
"loss": 0.4099,
|
| 168416 |
+
"step": 54270
|
| 168417 |
+
},
|
| 168418 |
+
{
|
| 168419 |
+
"epoch": 434.15,
|
| 168420 |
+
"learning_rate": 9.146570512820514e-06,
|
| 168421 |
+
"loss": 0.405,
|
| 168422 |
+
"step": 54275
|
| 168423 |
+
},
|
| 168424 |
+
{
|
| 168425 |
+
"epoch": 434.19,
|
| 168426 |
+
"learning_rate": 9.146490384615385e-06,
|
| 168427 |
+
"loss": 1.0881,
|
| 168428 |
+
"step": 54280
|
| 168429 |
+
},
|
| 168430 |
+
{
|
| 168431 |
+
"epoch": 434.23,
|
| 168432 |
+
"learning_rate": 9.146410256410258e-06,
|
| 168433 |
+
"loss": 0.6728,
|
| 168434 |
+
"step": 54285
|
| 168435 |
+
},
|
| 168436 |
+
{
|
| 168437 |
+
"epoch": 434.27,
|
| 168438 |
+
"learning_rate": 9.14633012820513e-06,
|
| 168439 |
+
"loss": 0.3143,
|
| 168440 |
+
"step": 54290
|
| 168441 |
+
},
|
| 168442 |
+
{
|
| 168443 |
+
"epoch": 434.31,
|
| 168444 |
+
"learning_rate": 9.14625e-06,
|
| 168445 |
+
"loss": 0.3771,
|
| 168446 |
+
"step": 54295
|
| 168447 |
+
},
|
| 168448 |
+
{
|
| 168449 |
+
"epoch": 434.35,
|
| 168450 |
+
"learning_rate": 9.146169871794872e-06,
|
| 168451 |
+
"loss": 0.4689,
|
| 168452 |
+
"step": 54300
|
| 168453 |
+
},
|
| 168454 |
+
{
|
| 168455 |
+
"epoch": 434.39,
|
| 168456 |
+
"learning_rate": 9.146089743589745e-06,
|
| 168457 |
+
"loss": 1.0307,
|
| 168458 |
+
"step": 54305
|
| 168459 |
+
},
|
| 168460 |
+
{
|
| 168461 |
+
"epoch": 434.43,
|
| 168462 |
+
"learning_rate": 9.146009615384615e-06,
|
| 168463 |
+
"loss": 0.6574,
|
| 168464 |
+
"step": 54310
|
| 168465 |
+
},
|
| 168466 |
+
{
|
| 168467 |
+
"epoch": 434.47,
|
| 168468 |
+
"learning_rate": 9.145929487179488e-06,
|
| 168469 |
+
"loss": 0.3558,
|
| 168470 |
+
"step": 54315
|
| 168471 |
+
},
|
| 168472 |
+
{
|
| 168473 |
+
"epoch": 434.51,
|
| 168474 |
+
"learning_rate": 9.145849358974361e-06,
|
| 168475 |
+
"loss": 0.3287,
|
| 168476 |
+
"step": 54320
|
| 168477 |
+
},
|
| 168478 |
+
{
|
| 168479 |
+
"epoch": 434.55,
|
| 168480 |
+
"learning_rate": 9.14576923076923e-06,
|
| 168481 |
+
"loss": 0.4354,
|
| 168482 |
+
"step": 54325
|
| 168483 |
+
},
|
| 168484 |
+
{
|
| 168485 |
+
"epoch": 434.59,
|
| 168486 |
+
"learning_rate": 9.145689102564104e-06,
|
| 168487 |
+
"loss": 0.8869,
|
| 168488 |
+
"step": 54330
|
| 168489 |
+
},
|
| 168490 |
+
{
|
| 168491 |
+
"epoch": 434.63,
|
| 168492 |
+
"learning_rate": 9.145608974358975e-06,
|
| 168493 |
+
"loss": 0.9042,
|
| 168494 |
+
"step": 54335
|
| 168495 |
+
},
|
| 168496 |
+
{
|
| 168497 |
+
"epoch": 434.67,
|
| 168498 |
+
"learning_rate": 9.145528846153846e-06,
|
| 168499 |
+
"loss": 0.3911,
|
| 168500 |
+
"step": 54340
|
| 168501 |
+
},
|
| 168502 |
+
{
|
| 168503 |
+
"epoch": 434.71,
|
| 168504 |
+
"learning_rate": 9.145448717948718e-06,
|
| 168505 |
+
"loss": 0.342,
|
| 168506 |
+
"step": 54345
|
| 168507 |
+
},
|
| 168508 |
+
{
|
| 168509 |
+
"epoch": 434.75,
|
| 168510 |
+
"learning_rate": 9.14536858974359e-06,
|
| 168511 |
+
"loss": 0.4601,
|
| 168512 |
+
"step": 54350
|
| 168513 |
+
},
|
| 168514 |
+
{
|
| 168515 |
+
"epoch": 434.79,
|
| 168516 |
+
"learning_rate": 9.145288461538462e-06,
|
| 168517 |
+
"loss": 1.0639,
|
| 168518 |
+
"step": 54355
|
| 168519 |
+
},
|
| 168520 |
+
{
|
| 168521 |
+
"epoch": 434.83,
|
| 168522 |
+
"learning_rate": 9.145208333333333e-06,
|
| 168523 |
+
"loss": 0.7023,
|
| 168524 |
+
"step": 54360
|
| 168525 |
+
},
|
| 168526 |
+
{
|
| 168527 |
+
"epoch": 434.87,
|
| 168528 |
+
"learning_rate": 9.145128205128207e-06,
|
| 168529 |
+
"loss": 0.3901,
|
| 168530 |
+
"step": 54365
|
| 168531 |
+
},
|
| 168532 |
+
{
|
| 168533 |
+
"epoch": 434.91,
|
| 168534 |
+
"learning_rate": 9.145048076923078e-06,
|
| 168535 |
+
"loss": 0.3836,
|
| 168536 |
+
"step": 54370
|
| 168537 |
+
},
|
| 168538 |
+
{
|
| 168539 |
+
"epoch": 434.95,
|
| 168540 |
+
"learning_rate": 9.14496794871795e-06,
|
| 168541 |
+
"loss": 0.4315,
|
| 168542 |
+
"step": 54375
|
| 168543 |
+
},
|
| 168544 |
+
{
|
| 168545 |
+
"epoch": 434.99,
|
| 168546 |
+
"learning_rate": 9.14488782051282e-06,
|
| 168547 |
+
"loss": 1.001,
|
| 168548 |
+
"step": 54380
|
| 168549 |
+
},
|
| 168550 |
+
{
|
| 168551 |
+
"epoch": 435.0,
|
| 168552 |
+
"eval_loss": 0.3676057457923889,
|
| 168553 |
+
"eval_runtime": 40.0739,
|
| 168554 |
+
"eval_samples_per_second": 20.961,
|
| 168555 |
+
"eval_steps_per_second": 0.674,
|
| 168556 |
+
"eval_wer": 0.20047993019197208,
|
| 168557 |
+
"step": 54381
|
| 168558 |
+
},
|
| 168559 |
+
{
|
| 168560 |
+
"epoch": 438.03,
|
| 168561 |
+
"learning_rate": 9.144807692307694e-06,
|
| 168562 |
+
"loss": 0.3355,
|
| 168563 |
+
"step": 54385
|
| 168564 |
+
},
|
| 168565 |
+
{
|
| 168566 |
+
"epoch": 438.07,
|
| 168567 |
+
"learning_rate": 9.144727564102565e-06,
|
| 168568 |
+
"loss": 0.3302,
|
| 168569 |
+
"step": 54390
|
| 168570 |
+
},
|
| 168571 |
+
{
|
| 168572 |
+
"epoch": 438.11,
|
| 168573 |
+
"learning_rate": 9.144647435897436e-06,
|
| 168574 |
+
"loss": 0.3472,
|
| 168575 |
+
"step": 54395
|
| 168576 |
+
},
|
| 168577 |
+
{
|
| 168578 |
+
"epoch": 438.15,
|
| 168579 |
+
"learning_rate": 9.144567307692308e-06,
|
| 168580 |
+
"loss": 0.4144,
|
| 168581 |
+
"step": 54400
|
| 168582 |
+
},
|
| 168583 |
+
{
|
| 168584 |
+
"epoch": 438.19,
|
| 168585 |
+
"learning_rate": 9.14448717948718e-06,
|
| 168586 |
+
"loss": 0.8747,
|
| 168587 |
+
"step": 54405
|
| 168588 |
+
},
|
| 168589 |
+
{
|
| 168590 |
+
"epoch": 438.23,
|
| 168591 |
+
"learning_rate": 9.144407051282052e-06,
|
| 168592 |
+
"loss": 0.6965,
|
| 168593 |
+
"step": 54410
|
| 168594 |
+
},
|
| 168595 |
+
{
|
| 168596 |
+
"epoch": 438.27,
|
| 168597 |
+
"learning_rate": 9.144326923076924e-06,
|
| 168598 |
+
"loss": 0.3358,
|
| 168599 |
+
"step": 54415
|
| 168600 |
+
},
|
| 168601 |
+
{
|
| 168602 |
+
"epoch": 438.31,
|
| 168603 |
+
"learning_rate": 9.144246794871797e-06,
|
| 168604 |
+
"loss": 0.4455,
|
| 168605 |
+
"step": 54420
|
| 168606 |
+
},
|
| 168607 |
+
{
|
| 168608 |
+
"epoch": 438.35,
|
| 168609 |
+
"learning_rate": 9.144166666666668e-06,
|
| 168610 |
+
"loss": 0.4321,
|
| 168611 |
+
"step": 54425
|
| 168612 |
+
},
|
| 168613 |
+
{
|
| 168614 |
+
"epoch": 438.39,
|
| 168615 |
+
"learning_rate": 9.14408653846154e-06,
|
| 168616 |
+
"loss": 1.0387,
|
| 168617 |
+
"step": 54430
|
| 168618 |
+
},
|
| 168619 |
+
{
|
| 168620 |
+
"epoch": 438.43,
|
| 168621 |
+
"learning_rate": 9.14400641025641e-06,
|
| 168622 |
+
"loss": 0.7209,
|
| 168623 |
+
"step": 54435
|
| 168624 |
+
},
|
| 168625 |
+
{
|
| 168626 |
+
"epoch": 438.47,
|
| 168627 |
+
"learning_rate": 9.143926282051284e-06,
|
| 168628 |
+
"loss": 0.3056,
|
| 168629 |
+
"step": 54440
|
| 168630 |
+
},
|
| 168631 |
+
{
|
| 168632 |
+
"epoch": 438.51,
|
| 168633 |
+
"learning_rate": 9.143846153846155e-06,
|
| 168634 |
+
"loss": 0.36,
|
| 168635 |
+
"step": 54445
|
| 168636 |
+
},
|
| 168637 |
+
{
|
| 168638 |
+
"epoch": 438.55,
|
| 168639 |
+
"learning_rate": 9.143766025641026e-06,
|
| 168640 |
+
"loss": 0.5089,
|
| 168641 |
+
"step": 54450
|
| 168642 |
+
},
|
| 168643 |
+
{
|
| 168644 |
+
"epoch": 438.59,
|
| 168645 |
+
"learning_rate": 9.143685897435898e-06,
|
| 168646 |
+
"loss": 1.0719,
|
| 168647 |
+
"step": 54455
|
| 168648 |
+
},
|
| 168649 |
+
{
|
| 168650 |
+
"epoch": 438.63,
|
| 168651 |
+
"learning_rate": 9.14360576923077e-06,
|
| 168652 |
+
"loss": 0.7689,
|
| 168653 |
+
"step": 54460
|
| 168654 |
+
},
|
| 168655 |
+
{
|
| 168656 |
+
"epoch": 438.67,
|
| 168657 |
+
"learning_rate": 9.14352564102564e-06,
|
| 168658 |
+
"loss": 0.3371,
|
| 168659 |
+
"step": 54465
|
| 168660 |
+
},
|
| 168661 |
+
{
|
| 168662 |
+
"epoch": 438.71,
|
| 168663 |
+
"learning_rate": 9.143445512820514e-06,
|
| 168664 |
+
"loss": 0.3606,
|
| 168665 |
+
"step": 54470
|
| 168666 |
+
},
|
| 168667 |
+
{
|
| 168668 |
+
"epoch": 438.76,
|
| 168669 |
+
"learning_rate": 9.143365384615387e-06,
|
| 168670 |
+
"loss": 0.4898,
|
| 168671 |
+
"step": 54475
|
| 168672 |
+
},
|
| 168673 |
+
{
|
| 168674 |
+
"epoch": 438.8,
|
| 168675 |
+
"learning_rate": 9.143285256410256e-06,
|
| 168676 |
+
"loss": 0.8839,
|
| 168677 |
+
"step": 54480
|
| 168678 |
+
},
|
| 168679 |
+
{
|
| 168680 |
+
"epoch": 438.84,
|
| 168681 |
+
"learning_rate": 9.14320512820513e-06,
|
| 168682 |
+
"loss": 0.7543,
|
| 168683 |
+
"step": 54485
|
| 168684 |
+
},
|
| 168685 |
+
{
|
| 168686 |
+
"epoch": 438.88,
|
| 168687 |
+
"learning_rate": 9.143125e-06,
|
| 168688 |
+
"loss": 0.3785,
|
| 168689 |
+
"step": 54490
|
| 168690 |
+
},
|
| 168691 |
+
{
|
| 168692 |
+
"epoch": 438.92,
|
| 168693 |
+
"learning_rate": 9.143044871794872e-06,
|
| 168694 |
+
"loss": 0.3476,
|
| 168695 |
+
"step": 54495
|
| 168696 |
+
},
|
| 168697 |
+
{
|
| 168698 |
+
"epoch": 438.96,
|
| 168699 |
+
"learning_rate": 9.142964743589743e-06,
|
| 168700 |
+
"loss": 0.4472,
|
| 168701 |
+
"step": 54500
|
| 168702 |
+
},
|
| 168703 |
+
{
|
| 168704 |
+
"epoch": 439.0,
|
| 168705 |
+
"learning_rate": 9.142884615384616e-06,
|
| 168706 |
+
"loss": 1.1279,
|
| 168707 |
+
"step": 54505
|
| 168708 |
+
},
|
| 168709 |
+
{
|
| 168710 |
+
"epoch": 439.0,
|
| 168711 |
+
"eval_loss": 0.4378458261489868,
|
| 168712 |
+
"eval_runtime": 39.018,
|
| 168713 |
+
"eval_samples_per_second": 21.529,
|
| 168714 |
+
"eval_steps_per_second": 0.692,
|
| 168715 |
+
"eval_wer": 0.19587552664646316,
|
| 168716 |
+
"step": 54505
|
| 168717 |
}
|
| 168718 |
],
|
| 168719 |
+
"max_steps": 620000,
|
| 168720 |
"num_train_epochs": 5000,
|
| 168721 |
+
"total_flos": 1.5338222874079727e+20,
|
| 168722 |
"trial_name": null,
|
| 168723 |
"trial_params": null
|
| 168724 |
}
|
model-bin/finetune/base/{checkpoint-53883 → checkpoint-54505}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629764971.3905332/events.out.tfevents.1629764971.c435e1c5ee04.920.1
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:29684233bd0fc4fa53d622430a8437859d228e1ebd196acb25eeeea72e1a8f1c
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629765618.347005/events.out.tfevents.1629765618.c435e1c5ee04.920.3
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:0549caa70a6959dff71a87146413f73853d214adadf9196b1be635a8986d790d
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629766288.11667/events.out.tfevents.1629766288.c435e1c5ee04.920.5
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c0cc0b65297bee44aa7440209a58938fbbf58bd65000cd58866a923c9ebdf11a
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629766954.1795254/events.out.tfevents.1629766954.c435e1c5ee04.920.7
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:da279e002f3b92c5bbaf9d3330dcf1c5dada61a83139b6d77bcff44ebd4c700a
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629767616.4417202/events.out.tfevents.1629767616.c435e1c5ee04.920.9
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:dc5cc1ea273768db4c516a02293a473944e45c923b417341b4901a28d7e11f7c
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629764971.c435e1c5ee04.920.0
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:715a5a3034e8b68c78630095bbb82afa35e33b2999fc887e193a2b427caf914c
|
| 3 |
+
size 8630
|
model-bin/finetune/base/log/events.out.tfevents.1629765618.c435e1c5ee04.920.2
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:7dc7447d2a605ddad8639d873081d8f1f91dd34fe28727938fdba49261e39f2e
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629766288.c435e1c5ee04.920.4
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:95b4ff4f8f7e5cc21e69b1a1df95a4c09ee29051c93d79b474b3291a209da371
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629766954.c435e1c5ee04.920.6
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f06cf4074fc3d80ff4404a5ce32673db1e7bc7194839e0f0e7041ce5d46a615c
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629767616.c435e1c5ee04.920.8
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e530f4e120d4d9a33abdc0271cce83168eca7ef2c15f6b01fe615709673f4362
|
| 3 |
+
size 8622
|