"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-72051 → checkpoint-72674}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-72051 → checkpoint-72674}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-72051 → checkpoint-72674}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-72051 → checkpoint-72674}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-72051 → checkpoint-72674}/rng_state.pth +2 -2
- model-bin/finetune/base/{checkpoint-72051 → checkpoint-72674}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-72051 → checkpoint-72674}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-72051 → checkpoint-72674}/trainer_state.json +792 -3
- model-bin/finetune/base/{checkpoint-72051 → checkpoint-72674}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629865712.3999202/events.out.tfevents.1629865712.7e498afd5545.905.43 +3 -0
- model-bin/finetune/base/log/1629866373.4153874/events.out.tfevents.1629866373.7e498afd5545.905.45 +3 -0
- model-bin/finetune/base/log/1629867035.5276291/events.out.tfevents.1629867035.7e498afd5545.905.47 +3 -0
- model-bin/finetune/base/log/1629867698.6983736/events.out.tfevents.1629867698.7e498afd5545.905.49 +3 -0
- model-bin/finetune/base/log/1629868357.748418/events.out.tfevents.1629868357.7e498afd5545.905.51 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629865712.7e498afd5545.905.42 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629866373.7e498afd5545.905.44 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629867035.7e498afd5545.905.46 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629867698.7e498afd5545.905.48 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629868357.7e498afd5545.905.50 +3 -0
model-bin/finetune/base/{checkpoint-72051 → checkpoint-72674}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-72051 → checkpoint-72674}/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 sha256:27f7e94c0d2f13a946288d827ec9755f75add3218235b06362dd344e338875de
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-72051 → checkpoint-72674}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-72051 → checkpoint-72674}/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 sha256:8602c7981b5ea5c0d6a2f7a98e772f3ecff4b6e6e11024389798cfbefe92c1d7
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-72051 → checkpoint-72674}/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:d09a749fbdf2cfa38549a5df81720966b7cc868c88d312d1e2998e3ad599f8ec
|
| 3 |
+
size 14503
|
model-bin/finetune/base/{checkpoint-72051 → checkpoint-72674}/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 sha256:06bd791a62086cb1e017bfe9d5222646f651e8165d1055ab3ef6d80ecdece72c
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-72051 → checkpoint-72674}/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:c835e72311c95e32d5f0b70d617e705a363bbb90fbd29eb2052c56576b97c7ab
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-72051 → checkpoint-72674}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.18412114350410416,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-69565",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -191037,11 +191037,800 @@
|
|
| 191037 |
"eval_steps_per_second": 0.704,
|
| 191038 |
"eval_wer": 0.19118408401745432,
|
| 191039 |
"step": 72051
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 191040 |
}
|
| 191041 |
],
|
| 191042 |
"max_steps": 625000,
|
| 191043 |
"num_train_epochs": 5000,
|
| 191044 |
-
"total_flos": 2.
|
| 191045 |
"trial_name": null,
|
| 191046 |
"trial_params": null
|
| 191047 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.18412114350410416,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-69565",
|
| 4 |
+
"epoch": 581.0,
|
| 5 |
+
"global_step": 72674,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 191037 |
"eval_steps_per_second": 0.704,
|
| 191038 |
"eval_wer": 0.19118408401745432,
|
| 191039 |
"step": 72051
|
| 191040 |
+
},
|
| 191041 |
+
{
|
| 191042 |
+
"epoch": 576.03,
|
| 191043 |
+
"learning_rate": 8.86173076923077e-06,
|
| 191044 |
+
"loss": 0.339,
|
| 191045 |
+
"step": 72055
|
| 191046 |
+
},
|
| 191047 |
+
{
|
| 191048 |
+
"epoch": 576.07,
|
| 191049 |
+
"learning_rate": 8.861650641025642e-06,
|
| 191050 |
+
"loss": 0.3217,
|
| 191051 |
+
"step": 72060
|
| 191052 |
+
},
|
| 191053 |
+
{
|
| 191054 |
+
"epoch": 576.11,
|
| 191055 |
+
"learning_rate": 8.861570512820514e-06,
|
| 191056 |
+
"loss": 0.3057,
|
| 191057 |
+
"step": 72065
|
| 191058 |
+
},
|
| 191059 |
+
{
|
| 191060 |
+
"epoch": 576.15,
|
| 191061 |
+
"learning_rate": 8.861490384615385e-06,
|
| 191062 |
+
"loss": 0.4795,
|
| 191063 |
+
"step": 72070
|
| 191064 |
+
},
|
| 191065 |
+
{
|
| 191066 |
+
"epoch": 576.19,
|
| 191067 |
+
"learning_rate": 8.861410256410257e-06,
|
| 191068 |
+
"loss": 0.9112,
|
| 191069 |
+
"step": 72075
|
| 191070 |
+
},
|
| 191071 |
+
{
|
| 191072 |
+
"epoch": 576.23,
|
| 191073 |
+
"learning_rate": 8.86133012820513e-06,
|
| 191074 |
+
"loss": 0.7722,
|
| 191075 |
+
"step": 72080
|
| 191076 |
+
},
|
| 191077 |
+
{
|
| 191078 |
+
"epoch": 576.27,
|
| 191079 |
+
"learning_rate": 8.861250000000001e-06,
|
| 191080 |
+
"loss": 0.3769,
|
| 191081 |
+
"step": 72085
|
| 191082 |
+
},
|
| 191083 |
+
{
|
| 191084 |
+
"epoch": 576.31,
|
| 191085 |
+
"learning_rate": 8.861169871794872e-06,
|
| 191086 |
+
"loss": 0.3351,
|
| 191087 |
+
"step": 72090
|
| 191088 |
+
},
|
| 191089 |
+
{
|
| 191090 |
+
"epoch": 576.35,
|
| 191091 |
+
"learning_rate": 8.861089743589745e-06,
|
| 191092 |
+
"loss": 0.3966,
|
| 191093 |
+
"step": 72095
|
| 191094 |
+
},
|
| 191095 |
+
{
|
| 191096 |
+
"epoch": 576.39,
|
| 191097 |
+
"learning_rate": 8.861009615384617e-06,
|
| 191098 |
+
"loss": 1.0172,
|
| 191099 |
+
"step": 72100
|
| 191100 |
+
},
|
| 191101 |
+
{
|
| 191102 |
+
"epoch": 576.43,
|
| 191103 |
+
"learning_rate": 8.860929487179488e-06,
|
| 191104 |
+
"loss": 0.7078,
|
| 191105 |
+
"step": 72105
|
| 191106 |
+
},
|
| 191107 |
+
{
|
| 191108 |
+
"epoch": 576.47,
|
| 191109 |
+
"learning_rate": 8.86084935897436e-06,
|
| 191110 |
+
"loss": 0.2888,
|
| 191111 |
+
"step": 72110
|
| 191112 |
+
},
|
| 191113 |
+
{
|
| 191114 |
+
"epoch": 576.51,
|
| 191115 |
+
"learning_rate": 8.860769230769232e-06,
|
| 191116 |
+
"loss": 0.3262,
|
| 191117 |
+
"step": 72115
|
| 191118 |
+
},
|
| 191119 |
+
{
|
| 191120 |
+
"epoch": 576.55,
|
| 191121 |
+
"learning_rate": 8.860689102564102e-06,
|
| 191122 |
+
"loss": 0.4345,
|
| 191123 |
+
"step": 72120
|
| 191124 |
+
},
|
| 191125 |
+
{
|
| 191126 |
+
"epoch": 576.59,
|
| 191127 |
+
"learning_rate": 8.860608974358975e-06,
|
| 191128 |
+
"loss": 0.9028,
|
| 191129 |
+
"step": 72125
|
| 191130 |
+
},
|
| 191131 |
+
{
|
| 191132 |
+
"epoch": 576.63,
|
| 191133 |
+
"learning_rate": 8.860528846153847e-06,
|
| 191134 |
+
"loss": 0.6099,
|
| 191135 |
+
"step": 72130
|
| 191136 |
+
},
|
| 191137 |
+
{
|
| 191138 |
+
"epoch": 576.67,
|
| 191139 |
+
"learning_rate": 8.860448717948718e-06,
|
| 191140 |
+
"loss": 0.3631,
|
| 191141 |
+
"step": 72135
|
| 191142 |
+
},
|
| 191143 |
+
{
|
| 191144 |
+
"epoch": 576.71,
|
| 191145 |
+
"learning_rate": 8.860368589743591e-06,
|
| 191146 |
+
"loss": 0.3006,
|
| 191147 |
+
"step": 72140
|
| 191148 |
+
},
|
| 191149 |
+
{
|
| 191150 |
+
"epoch": 576.75,
|
| 191151 |
+
"learning_rate": 8.860288461538462e-06,
|
| 191152 |
+
"loss": 0.3792,
|
| 191153 |
+
"step": 72145
|
| 191154 |
+
},
|
| 191155 |
+
{
|
| 191156 |
+
"epoch": 576.79,
|
| 191157 |
+
"learning_rate": 8.860208333333334e-06,
|
| 191158 |
+
"loss": 0.8652,
|
| 191159 |
+
"step": 72150
|
| 191160 |
+
},
|
| 191161 |
+
{
|
| 191162 |
+
"epoch": 576.83,
|
| 191163 |
+
"learning_rate": 8.860128205128205e-06,
|
| 191164 |
+
"loss": 0.7947,
|
| 191165 |
+
"step": 72155
|
| 191166 |
+
},
|
| 191167 |
+
{
|
| 191168 |
+
"epoch": 576.87,
|
| 191169 |
+
"learning_rate": 8.860048076923078e-06,
|
| 191170 |
+
"loss": 0.2886,
|
| 191171 |
+
"step": 72160
|
| 191172 |
+
},
|
| 191173 |
+
{
|
| 191174 |
+
"epoch": 576.91,
|
| 191175 |
+
"learning_rate": 8.85996794871795e-06,
|
| 191176 |
+
"loss": 0.365,
|
| 191177 |
+
"step": 72165
|
| 191178 |
+
},
|
| 191179 |
+
{
|
| 191180 |
+
"epoch": 576.95,
|
| 191181 |
+
"learning_rate": 8.85988782051282e-06,
|
| 191182 |
+
"loss": 0.4469,
|
| 191183 |
+
"step": 72170
|
| 191184 |
+
},
|
| 191185 |
+
{
|
| 191186 |
+
"epoch": 576.99,
|
| 191187 |
+
"learning_rate": 8.859807692307692e-06,
|
| 191188 |
+
"loss": 0.9515,
|
| 191189 |
+
"step": 72175
|
| 191190 |
+
},
|
| 191191 |
+
{
|
| 191192 |
+
"epoch": 577.0,
|
| 191193 |
+
"eval_loss": 0.4047023355960846,
|
| 191194 |
+
"eval_runtime": 39.665,
|
| 191195 |
+
"eval_samples_per_second": 21.177,
|
| 191196 |
+
"eval_steps_per_second": 0.681,
|
| 191197 |
+
"eval_wer": 0.194370455690735,
|
| 191198 |
+
"step": 72176
|
| 191199 |
+
},
|
| 191200 |
+
{
|
| 191201 |
+
"epoch": 577.03,
|
| 191202 |
+
"learning_rate": 8.859727564102565e-06,
|
| 191203 |
+
"loss": 0.33,
|
| 191204 |
+
"step": 72180
|
| 191205 |
+
},
|
| 191206 |
+
{
|
| 191207 |
+
"epoch": 577.07,
|
| 191208 |
+
"learning_rate": 8.859647435897437e-06,
|
| 191209 |
+
"loss": 0.2739,
|
| 191210 |
+
"step": 72185
|
| 191211 |
+
},
|
| 191212 |
+
{
|
| 191213 |
+
"epoch": 577.11,
|
| 191214 |
+
"learning_rate": 8.859567307692308e-06,
|
| 191215 |
+
"loss": 0.3062,
|
| 191216 |
+
"step": 72190
|
| 191217 |
+
},
|
| 191218 |
+
{
|
| 191219 |
+
"epoch": 577.15,
|
| 191220 |
+
"learning_rate": 8.859487179487181e-06,
|
| 191221 |
+
"loss": 0.47,
|
| 191222 |
+
"step": 72195
|
| 191223 |
+
},
|
| 191224 |
+
{
|
| 191225 |
+
"epoch": 577.19,
|
| 191226 |
+
"learning_rate": 8.859407051282052e-06,
|
| 191227 |
+
"loss": 1.0719,
|
| 191228 |
+
"step": 72200
|
| 191229 |
+
},
|
| 191230 |
+
{
|
| 191231 |
+
"epoch": 577.23,
|
| 191232 |
+
"learning_rate": 8.859326923076924e-06,
|
| 191233 |
+
"loss": 0.6248,
|
| 191234 |
+
"step": 72205
|
| 191235 |
+
},
|
| 191236 |
+
{
|
| 191237 |
+
"epoch": 577.27,
|
| 191238 |
+
"learning_rate": 8.859246794871795e-06,
|
| 191239 |
+
"loss": 0.3089,
|
| 191240 |
+
"step": 72210
|
| 191241 |
+
},
|
| 191242 |
+
{
|
| 191243 |
+
"epoch": 577.31,
|
| 191244 |
+
"learning_rate": 8.859166666666668e-06,
|
| 191245 |
+
"loss": 0.3345,
|
| 191246 |
+
"step": 72215
|
| 191247 |
+
},
|
| 191248 |
+
{
|
| 191249 |
+
"epoch": 577.35,
|
| 191250 |
+
"learning_rate": 8.85908653846154e-06,
|
| 191251 |
+
"loss": 0.402,
|
| 191252 |
+
"step": 72220
|
| 191253 |
+
},
|
| 191254 |
+
{
|
| 191255 |
+
"epoch": 577.39,
|
| 191256 |
+
"learning_rate": 8.85900641025641e-06,
|
| 191257 |
+
"loss": 0.9915,
|
| 191258 |
+
"step": 72225
|
| 191259 |
+
},
|
| 191260 |
+
{
|
| 191261 |
+
"epoch": 577.43,
|
| 191262 |
+
"learning_rate": 8.858926282051282e-06,
|
| 191263 |
+
"loss": 0.6853,
|
| 191264 |
+
"step": 72230
|
| 191265 |
+
},
|
| 191266 |
+
{
|
| 191267 |
+
"epoch": 577.47,
|
| 191268 |
+
"learning_rate": 8.858846153846155e-06,
|
| 191269 |
+
"loss": 0.2974,
|
| 191270 |
+
"step": 72235
|
| 191271 |
+
},
|
| 191272 |
+
{
|
| 191273 |
+
"epoch": 577.51,
|
| 191274 |
+
"learning_rate": 8.858766025641027e-06,
|
| 191275 |
+
"loss": 0.3472,
|
| 191276 |
+
"step": 72240
|
| 191277 |
+
},
|
| 191278 |
+
{
|
| 191279 |
+
"epoch": 577.55,
|
| 191280 |
+
"learning_rate": 8.858685897435898e-06,
|
| 191281 |
+
"loss": 0.4193,
|
| 191282 |
+
"step": 72245
|
| 191283 |
+
},
|
| 191284 |
+
{
|
| 191285 |
+
"epoch": 577.59,
|
| 191286 |
+
"learning_rate": 8.858605769230771e-06,
|
| 191287 |
+
"loss": 0.9081,
|
| 191288 |
+
"step": 72250
|
| 191289 |
+
},
|
| 191290 |
+
{
|
| 191291 |
+
"epoch": 577.63,
|
| 191292 |
+
"learning_rate": 8.85852564102564e-06,
|
| 191293 |
+
"loss": 0.6596,
|
| 191294 |
+
"step": 72255
|
| 191295 |
+
},
|
| 191296 |
+
{
|
| 191297 |
+
"epoch": 577.67,
|
| 191298 |
+
"learning_rate": 8.858445512820514e-06,
|
| 191299 |
+
"loss": 0.2994,
|
| 191300 |
+
"step": 72260
|
| 191301 |
+
},
|
| 191302 |
+
{
|
| 191303 |
+
"epoch": 577.71,
|
| 191304 |
+
"learning_rate": 8.858365384615385e-06,
|
| 191305 |
+
"loss": 0.2584,
|
| 191306 |
+
"step": 72265
|
| 191307 |
+
},
|
| 191308 |
+
{
|
| 191309 |
+
"epoch": 577.75,
|
| 191310 |
+
"learning_rate": 8.858285256410256e-06,
|
| 191311 |
+
"loss": 0.4335,
|
| 191312 |
+
"step": 72270
|
| 191313 |
+
},
|
| 191314 |
+
{
|
| 191315 |
+
"epoch": 577.79,
|
| 191316 |
+
"learning_rate": 8.858205128205128e-06,
|
| 191317 |
+
"loss": 0.9398,
|
| 191318 |
+
"step": 72275
|
| 191319 |
+
},
|
| 191320 |
+
{
|
| 191321 |
+
"epoch": 577.83,
|
| 191322 |
+
"learning_rate": 8.858125e-06,
|
| 191323 |
+
"loss": 0.7375,
|
| 191324 |
+
"step": 72280
|
| 191325 |
+
},
|
| 191326 |
+
{
|
| 191327 |
+
"epoch": 577.87,
|
| 191328 |
+
"learning_rate": 8.858044871794872e-06,
|
| 191329 |
+
"loss": 0.3372,
|
| 191330 |
+
"step": 72285
|
| 191331 |
+
},
|
| 191332 |
+
{
|
| 191333 |
+
"epoch": 577.91,
|
| 191334 |
+
"learning_rate": 8.857964743589744e-06,
|
| 191335 |
+
"loss": 0.3376,
|
| 191336 |
+
"step": 72290
|
| 191337 |
+
},
|
| 191338 |
+
{
|
| 191339 |
+
"epoch": 577.95,
|
| 191340 |
+
"learning_rate": 8.857884615384617e-06,
|
| 191341 |
+
"loss": 0.4914,
|
| 191342 |
+
"step": 72295
|
| 191343 |
+
},
|
| 191344 |
+
{
|
| 191345 |
+
"epoch": 577.99,
|
| 191346 |
+
"learning_rate": 8.857804487179488e-06,
|
| 191347 |
+
"loss": 1.0205,
|
| 191348 |
+
"step": 72300
|
| 191349 |
+
},
|
| 191350 |
+
{
|
| 191351 |
+
"epoch": 578.0,
|
| 191352 |
+
"eval_loss": 0.447828084230423,
|
| 191353 |
+
"eval_runtime": 39.7038,
|
| 191354 |
+
"eval_samples_per_second": 21.157,
|
| 191355 |
+
"eval_steps_per_second": 0.68,
|
| 191356 |
+
"eval_wer": 0.1914162462528332,
|
| 191357 |
+
"step": 72301
|
| 191358 |
+
},
|
| 191359 |
+
{
|
| 191360 |
+
"epoch": 583.03,
|
| 191361 |
+
"learning_rate": 8.85772435897436e-06,
|
| 191362 |
+
"loss": 0.3688,
|
| 191363 |
+
"step": 72305
|
| 191364 |
+
},
|
| 191365 |
+
{
|
| 191366 |
+
"epoch": 583.07,
|
| 191367 |
+
"learning_rate": 8.85764423076923e-06,
|
| 191368 |
+
"loss": 0.3559,
|
| 191369 |
+
"step": 72310
|
| 191370 |
+
},
|
| 191371 |
+
{
|
| 191372 |
+
"epoch": 583.11,
|
| 191373 |
+
"learning_rate": 8.857564102564104e-06,
|
| 191374 |
+
"loss": 0.363,
|
| 191375 |
+
"step": 72315
|
| 191376 |
+
},
|
| 191377 |
+
{
|
| 191378 |
+
"epoch": 583.15,
|
| 191379 |
+
"learning_rate": 8.857483974358975e-06,
|
| 191380 |
+
"loss": 0.5013,
|
| 191381 |
+
"step": 72320
|
| 191382 |
+
},
|
| 191383 |
+
{
|
| 191384 |
+
"epoch": 583.19,
|
| 191385 |
+
"learning_rate": 8.857403846153846e-06,
|
| 191386 |
+
"loss": 1.1272,
|
| 191387 |
+
"step": 72325
|
| 191388 |
+
},
|
| 191389 |
+
{
|
| 191390 |
+
"epoch": 583.23,
|
| 191391 |
+
"learning_rate": 8.857323717948718e-06,
|
| 191392 |
+
"loss": 0.6567,
|
| 191393 |
+
"step": 72330
|
| 191394 |
+
},
|
| 191395 |
+
{
|
| 191396 |
+
"epoch": 583.27,
|
| 191397 |
+
"learning_rate": 8.857243589743591e-06,
|
| 191398 |
+
"loss": 0.3223,
|
| 191399 |
+
"step": 72335
|
| 191400 |
+
},
|
| 191401 |
+
{
|
| 191402 |
+
"epoch": 583.31,
|
| 191403 |
+
"learning_rate": 8.857163461538462e-06,
|
| 191404 |
+
"loss": 0.3139,
|
| 191405 |
+
"step": 72340
|
| 191406 |
+
},
|
| 191407 |
+
{
|
| 191408 |
+
"epoch": 583.35,
|
| 191409 |
+
"learning_rate": 8.857083333333334e-06,
|
| 191410 |
+
"loss": 0.3696,
|
| 191411 |
+
"step": 72345
|
| 191412 |
+
},
|
| 191413 |
+
{
|
| 191414 |
+
"epoch": 583.39,
|
| 191415 |
+
"learning_rate": 8.857003205128207e-06,
|
| 191416 |
+
"loss": 0.9257,
|
| 191417 |
+
"step": 72350
|
| 191418 |
+
},
|
| 191419 |
+
{
|
| 191420 |
+
"epoch": 583.43,
|
| 191421 |
+
"learning_rate": 8.856923076923078e-06,
|
| 191422 |
+
"loss": 0.6354,
|
| 191423 |
+
"step": 72355
|
| 191424 |
+
},
|
| 191425 |
+
{
|
| 191426 |
+
"epoch": 583.47,
|
| 191427 |
+
"learning_rate": 8.85684294871795e-06,
|
| 191428 |
+
"loss": 0.311,
|
| 191429 |
+
"step": 72360
|
| 191430 |
+
},
|
| 191431 |
+
{
|
| 191432 |
+
"epoch": 583.51,
|
| 191433 |
+
"learning_rate": 8.85676282051282e-06,
|
| 191434 |
+
"loss": 0.3108,
|
| 191435 |
+
"step": 72365
|
| 191436 |
+
},
|
| 191437 |
+
{
|
| 191438 |
+
"epoch": 583.55,
|
| 191439 |
+
"learning_rate": 8.856682692307694e-06,
|
| 191440 |
+
"loss": 0.3946,
|
| 191441 |
+
"step": 72370
|
| 191442 |
+
},
|
| 191443 |
+
{
|
| 191444 |
+
"epoch": 583.59,
|
| 191445 |
+
"learning_rate": 8.856602564102565e-06,
|
| 191446 |
+
"loss": 0.8928,
|
| 191447 |
+
"step": 72375
|
| 191448 |
+
},
|
| 191449 |
+
{
|
| 191450 |
+
"epoch": 583.63,
|
| 191451 |
+
"learning_rate": 8.856522435897436e-06,
|
| 191452 |
+
"loss": 0.688,
|
| 191453 |
+
"step": 72380
|
| 191454 |
+
},
|
| 191455 |
+
{
|
| 191456 |
+
"epoch": 583.67,
|
| 191457 |
+
"learning_rate": 8.85644230769231e-06,
|
| 191458 |
+
"loss": 0.3139,
|
| 191459 |
+
"step": 72385
|
| 191460 |
+
},
|
| 191461 |
+
{
|
| 191462 |
+
"epoch": 583.71,
|
| 191463 |
+
"learning_rate": 8.856362179487181e-06,
|
| 191464 |
+
"loss": 0.3488,
|
| 191465 |
+
"step": 72390
|
| 191466 |
+
},
|
| 191467 |
+
{
|
| 191468 |
+
"epoch": 583.76,
|
| 191469 |
+
"learning_rate": 8.856282051282052e-06,
|
| 191470 |
+
"loss": 0.4477,
|
| 191471 |
+
"step": 72395
|
| 191472 |
+
},
|
| 191473 |
+
{
|
| 191474 |
+
"epoch": 583.8,
|
| 191475 |
+
"learning_rate": 8.856201923076924e-06,
|
| 191476 |
+
"loss": 1.077,
|
| 191477 |
+
"step": 72400
|
| 191478 |
+
},
|
| 191479 |
+
{
|
| 191480 |
+
"epoch": 583.84,
|
| 191481 |
+
"learning_rate": 8.856121794871797e-06,
|
| 191482 |
+
"loss": 0.9005,
|
| 191483 |
+
"step": 72405
|
| 191484 |
+
},
|
| 191485 |
+
{
|
| 191486 |
+
"epoch": 583.88,
|
| 191487 |
+
"learning_rate": 8.856041666666666e-06,
|
| 191488 |
+
"loss": 0.2631,
|
| 191489 |
+
"step": 72410
|
| 191490 |
+
},
|
| 191491 |
+
{
|
| 191492 |
+
"epoch": 583.92,
|
| 191493 |
+
"learning_rate": 8.85596153846154e-06,
|
| 191494 |
+
"loss": 0.5206,
|
| 191495 |
+
"step": 72415
|
| 191496 |
+
},
|
| 191497 |
+
{
|
| 191498 |
+
"epoch": 583.96,
|
| 191499 |
+
"learning_rate": 8.85588141025641e-06,
|
| 191500 |
+
"loss": 0.4162,
|
| 191501 |
+
"step": 72420
|
| 191502 |
+
},
|
| 191503 |
+
{
|
| 191504 |
+
"epoch": 584.0,
|
| 191505 |
+
"learning_rate": 8.855801282051282e-06,
|
| 191506 |
+
"loss": 1.1406,
|
| 191507 |
+
"step": 72425
|
| 191508 |
+
},
|
| 191509 |
+
{
|
| 191510 |
+
"epoch": 584.0,
|
| 191511 |
+
"eval_loss": 0.3920002579689026,
|
| 191512 |
+
"eval_runtime": 40.716,
|
| 191513 |
+
"eval_samples_per_second": 20.631,
|
| 191514 |
+
"eval_steps_per_second": 0.663,
|
| 191515 |
+
"eval_wer": 0.18435091879075283,
|
| 191516 |
+
"step": 72425
|
| 191517 |
+
},
|
| 191518 |
+
{
|
| 191519 |
+
"epoch": 584.04,
|
| 191520 |
+
"learning_rate": 8.855721153846153e-06,
|
| 191521 |
+
"loss": 0.3607,
|
| 191522 |
+
"step": 72430
|
| 191523 |
+
},
|
| 191524 |
+
{
|
| 191525 |
+
"epoch": 584.08,
|
| 191526 |
+
"learning_rate": 8.855641025641027e-06,
|
| 191527 |
+
"loss": 0.3723,
|
| 191528 |
+
"step": 72435
|
| 191529 |
+
},
|
| 191530 |
+
{
|
| 191531 |
+
"epoch": 584.12,
|
| 191532 |
+
"learning_rate": 8.855560897435898e-06,
|
| 191533 |
+
"loss": 0.2966,
|
| 191534 |
+
"step": 72440
|
| 191535 |
+
},
|
| 191536 |
+
{
|
| 191537 |
+
"epoch": 584.16,
|
| 191538 |
+
"learning_rate": 8.85548076923077e-06,
|
| 191539 |
+
"loss": 0.513,
|
| 191540 |
+
"step": 72445
|
| 191541 |
+
},
|
| 191542 |
+
{
|
| 191543 |
+
"epoch": 584.2,
|
| 191544 |
+
"learning_rate": 8.855400641025642e-06,
|
| 191545 |
+
"loss": 1.1743,
|
| 191546 |
+
"step": 72450
|
| 191547 |
+
},
|
| 191548 |
+
{
|
| 191549 |
+
"epoch": 584.24,
|
| 191550 |
+
"learning_rate": 8.855320512820514e-06,
|
| 191551 |
+
"loss": 0.3608,
|
| 191552 |
+
"step": 72455
|
| 191553 |
+
},
|
| 191554 |
+
{
|
| 191555 |
+
"epoch": 584.28,
|
| 191556 |
+
"learning_rate": 8.855240384615385e-06,
|
| 191557 |
+
"loss": 0.2819,
|
| 191558 |
+
"step": 72460
|
| 191559 |
+
},
|
| 191560 |
+
{
|
| 191561 |
+
"epoch": 584.32,
|
| 191562 |
+
"learning_rate": 8.855160256410256e-06,
|
| 191563 |
+
"loss": 0.358,
|
| 191564 |
+
"step": 72465
|
| 191565 |
+
},
|
| 191566 |
+
{
|
| 191567 |
+
"epoch": 584.36,
|
| 191568 |
+
"learning_rate": 8.85508012820513e-06,
|
| 191569 |
+
"loss": 0.501,
|
| 191570 |
+
"step": 72470
|
| 191571 |
+
},
|
| 191572 |
+
{
|
| 191573 |
+
"epoch": 584.4,
|
| 191574 |
+
"learning_rate": 8.855e-06,
|
| 191575 |
+
"loss": 1.1884,
|
| 191576 |
+
"step": 72475
|
| 191577 |
+
},
|
| 191578 |
+
{
|
| 191579 |
+
"epoch": 584.44,
|
| 191580 |
+
"learning_rate": 8.854919871794872e-06,
|
| 191581 |
+
"loss": 0.3289,
|
| 191582 |
+
"step": 72480
|
| 191583 |
+
},
|
| 191584 |
+
{
|
| 191585 |
+
"epoch": 584.48,
|
| 191586 |
+
"learning_rate": 8.854839743589745e-06,
|
| 191587 |
+
"loss": 0.2994,
|
| 191588 |
+
"step": 72485
|
| 191589 |
+
},
|
| 191590 |
+
{
|
| 191591 |
+
"epoch": 584.52,
|
| 191592 |
+
"learning_rate": 8.854759615384617e-06,
|
| 191593 |
+
"loss": 0.3836,
|
| 191594 |
+
"step": 72490
|
| 191595 |
+
},
|
| 191596 |
+
{
|
| 191597 |
+
"epoch": 584.56,
|
| 191598 |
+
"learning_rate": 8.854679487179488e-06,
|
| 191599 |
+
"loss": 0.4344,
|
| 191600 |
+
"step": 72495
|
| 191601 |
+
},
|
| 191602 |
+
{
|
| 191603 |
+
"epoch": 584.6,
|
| 191604 |
+
"learning_rate": 8.85459935897436e-06,
|
| 191605 |
+
"loss": 1.2481,
|
| 191606 |
+
"step": 72500
|
| 191607 |
+
},
|
| 191608 |
+
{
|
| 191609 |
+
"epoch": 584.64,
|
| 191610 |
+
"learning_rate": 8.854519230769232e-06,
|
| 191611 |
+
"loss": 0.4116,
|
| 191612 |
+
"step": 72505
|
| 191613 |
+
},
|
| 191614 |
+
{
|
| 191615 |
+
"epoch": 584.68,
|
| 191616 |
+
"learning_rate": 8.854439102564104e-06,
|
| 191617 |
+
"loss": 0.2749,
|
| 191618 |
+
"step": 72510
|
| 191619 |
+
},
|
| 191620 |
+
{
|
| 191621 |
+
"epoch": 584.72,
|
| 191622 |
+
"learning_rate": 8.854358974358975e-06,
|
| 191623 |
+
"loss": 0.3268,
|
| 191624 |
+
"step": 72515
|
| 191625 |
+
},
|
| 191626 |
+
{
|
| 191627 |
+
"epoch": 584.76,
|
| 191628 |
+
"learning_rate": 8.854278846153846e-06,
|
| 191629 |
+
"loss": 0.4482,
|
| 191630 |
+
"step": 72520
|
| 191631 |
+
},
|
| 191632 |
+
{
|
| 191633 |
+
"epoch": 584.8,
|
| 191634 |
+
"learning_rate": 8.85419871794872e-06,
|
| 191635 |
+
"loss": 1.2621,
|
| 191636 |
+
"step": 72525
|
| 191637 |
+
},
|
| 191638 |
+
{
|
| 191639 |
+
"epoch": 584.84,
|
| 191640 |
+
"learning_rate": 8.85411858974359e-06,
|
| 191641 |
+
"loss": 0.3344,
|
| 191642 |
+
"step": 72530
|
| 191643 |
+
},
|
| 191644 |
+
{
|
| 191645 |
+
"epoch": 584.88,
|
| 191646 |
+
"learning_rate": 8.854038461538462e-06,
|
| 191647 |
+
"loss": 0.3034,
|
| 191648 |
+
"step": 72535
|
| 191649 |
+
},
|
| 191650 |
+
{
|
| 191651 |
+
"epoch": 584.92,
|
| 191652 |
+
"learning_rate": 8.853958333333335e-06,
|
| 191653 |
+
"loss": 0.3423,
|
| 191654 |
+
"step": 72540
|
| 191655 |
+
},
|
| 191656 |
+
{
|
| 191657 |
+
"epoch": 584.96,
|
| 191658 |
+
"learning_rate": 8.853878205128207e-06,
|
| 191659 |
+
"loss": 0.5005,
|
| 191660 |
+
"step": 72545
|
| 191661 |
+
},
|
| 191662 |
+
{
|
| 191663 |
+
"epoch": 585.0,
|
| 191664 |
+
"eval_loss": 0.39568430185317993,
|
| 191665 |
+
"eval_runtime": 39.2591,
|
| 191666 |
+
"eval_samples_per_second": 21.396,
|
| 191667 |
+
"eval_steps_per_second": 0.688,
|
| 191668 |
+
"eval_wer": 0.1896831245394252,
|
| 191669 |
+
"step": 72549
|
| 191670 |
+
},
|
| 191671 |
+
{
|
| 191672 |
+
"epoch": 580.01,
|
| 191673 |
+
"learning_rate": 8.853798076923078e-06,
|
| 191674 |
+
"loss": 0.4587,
|
| 191675 |
+
"step": 72550
|
| 191676 |
+
},
|
| 191677 |
+
{
|
| 191678 |
+
"epoch": 580.05,
|
| 191679 |
+
"learning_rate": 8.85371794871795e-06,
|
| 191680 |
+
"loss": 0.3441,
|
| 191681 |
+
"step": 72555
|
| 191682 |
+
},
|
| 191683 |
+
{
|
| 191684 |
+
"epoch": 580.09,
|
| 191685 |
+
"learning_rate": 8.853637820512822e-06,
|
| 191686 |
+
"loss": 0.3167,
|
| 191687 |
+
"step": 72560
|
| 191688 |
+
},
|
| 191689 |
+
{
|
| 191690 |
+
"epoch": 580.13,
|
| 191691 |
+
"learning_rate": 8.853557692307692e-06,
|
| 191692 |
+
"loss": 0.4195,
|
| 191693 |
+
"step": 72565
|
| 191694 |
+
},
|
| 191695 |
+
{
|
| 191696 |
+
"epoch": 580.17,
|
| 191697 |
+
"learning_rate": 8.853477564102565e-06,
|
| 191698 |
+
"loss": 0.5473,
|
| 191699 |
+
"step": 72570
|
| 191700 |
+
},
|
| 191701 |
+
{
|
| 191702 |
+
"epoch": 580.21,
|
| 191703 |
+
"learning_rate": 8.853397435897436e-06,
|
| 191704 |
+
"loss": 1.2351,
|
| 191705 |
+
"step": 72575
|
| 191706 |
+
},
|
| 191707 |
+
{
|
| 191708 |
+
"epoch": 580.25,
|
| 191709 |
+
"learning_rate": 8.853317307692308e-06,
|
| 191710 |
+
"loss": 0.4162,
|
| 191711 |
+
"step": 72580
|
| 191712 |
+
},
|
| 191713 |
+
{
|
| 191714 |
+
"epoch": 580.29,
|
| 191715 |
+
"learning_rate": 8.85323717948718e-06,
|
| 191716 |
+
"loss": 0.3726,
|
| 191717 |
+
"step": 72585
|
| 191718 |
+
},
|
| 191719 |
+
{
|
| 191720 |
+
"epoch": 580.33,
|
| 191721 |
+
"learning_rate": 8.853157051282052e-06,
|
| 191722 |
+
"loss": 0.311,
|
| 191723 |
+
"step": 72590
|
| 191724 |
+
},
|
| 191725 |
+
{
|
| 191726 |
+
"epoch": 580.37,
|
| 191727 |
+
"learning_rate": 8.853076923076924e-06,
|
| 191728 |
+
"loss": 0.6207,
|
| 191729 |
+
"step": 72595
|
| 191730 |
+
},
|
| 191731 |
+
{
|
| 191732 |
+
"epoch": 580.41,
|
| 191733 |
+
"learning_rate": 8.852996794871795e-06,
|
| 191734 |
+
"loss": 1.2596,
|
| 191735 |
+
"step": 72600
|
| 191736 |
+
},
|
| 191737 |
+
{
|
| 191738 |
+
"epoch": 580.45,
|
| 191739 |
+
"learning_rate": 8.852916666666668e-06,
|
| 191740 |
+
"loss": 0.3445,
|
| 191741 |
+
"step": 72605
|
| 191742 |
+
},
|
| 191743 |
+
{
|
| 191744 |
+
"epoch": 580.49,
|
| 191745 |
+
"learning_rate": 8.85283653846154e-06,
|
| 191746 |
+
"loss": 0.3112,
|
| 191747 |
+
"step": 72610
|
| 191748 |
+
},
|
| 191749 |
+
{
|
| 191750 |
+
"epoch": 580.53,
|
| 191751 |
+
"learning_rate": 8.85275641025641e-06,
|
| 191752 |
+
"loss": 0.3731,
|
| 191753 |
+
"step": 72615
|
| 191754 |
+
},
|
| 191755 |
+
{
|
| 191756 |
+
"epoch": 580.57,
|
| 191757 |
+
"learning_rate": 8.852676282051282e-06,
|
| 191758 |
+
"loss": 0.5238,
|
| 191759 |
+
"step": 72620
|
| 191760 |
+
},
|
| 191761 |
+
{
|
| 191762 |
+
"epoch": 580.61,
|
| 191763 |
+
"learning_rate": 8.852596153846155e-06,
|
| 191764 |
+
"loss": 1.196,
|
| 191765 |
+
"step": 72625
|
| 191766 |
+
},
|
| 191767 |
+
{
|
| 191768 |
+
"epoch": 580.65,
|
| 191769 |
+
"learning_rate": 8.852516025641026e-06,
|
| 191770 |
+
"loss": 0.2666,
|
| 191771 |
+
"step": 72630
|
| 191772 |
+
},
|
| 191773 |
+
{
|
| 191774 |
+
"epoch": 580.69,
|
| 191775 |
+
"learning_rate": 8.852435897435898e-06,
|
| 191776 |
+
"loss": 0.3653,
|
| 191777 |
+
"step": 72635
|
| 191778 |
+
},
|
| 191779 |
+
{
|
| 191780 |
+
"epoch": 580.73,
|
| 191781 |
+
"learning_rate": 8.852355769230771e-06,
|
| 191782 |
+
"loss": 0.3944,
|
| 191783 |
+
"step": 72640
|
| 191784 |
+
},
|
| 191785 |
+
{
|
| 191786 |
+
"epoch": 580.77,
|
| 191787 |
+
"learning_rate": 8.852275641025642e-06,
|
| 191788 |
+
"loss": 0.5596,
|
| 191789 |
+
"step": 72645
|
| 191790 |
+
},
|
| 191791 |
+
{
|
| 191792 |
+
"epoch": 580.81,
|
| 191793 |
+
"learning_rate": 8.852195512820514e-06,
|
| 191794 |
+
"loss": 1.1513,
|
| 191795 |
+
"step": 72650
|
| 191796 |
+
},
|
| 191797 |
+
{
|
| 191798 |
+
"epoch": 580.85,
|
| 191799 |
+
"learning_rate": 8.852115384615385e-06,
|
| 191800 |
+
"loss": 0.3255,
|
| 191801 |
+
"step": 72655
|
| 191802 |
+
},
|
| 191803 |
+
{
|
| 191804 |
+
"epoch": 580.89,
|
| 191805 |
+
"learning_rate": 8.852035256410258e-06,
|
| 191806 |
+
"loss": 0.25,
|
| 191807 |
+
"step": 72660
|
| 191808 |
+
},
|
| 191809 |
+
{
|
| 191810 |
+
"epoch": 580.93,
|
| 191811 |
+
"learning_rate": 8.85195512820513e-06,
|
| 191812 |
+
"loss": 0.4041,
|
| 191813 |
+
"step": 72665
|
| 191814 |
+
},
|
| 191815 |
+
{
|
| 191816 |
+
"epoch": 580.97,
|
| 191817 |
+
"learning_rate": 8.851875e-06,
|
| 191818 |
+
"loss": 0.6177,
|
| 191819 |
+
"step": 72670
|
| 191820 |
+
},
|
| 191821 |
+
{
|
| 191822 |
+
"epoch": 581.0,
|
| 191823 |
+
"eval_loss": 0.3960975706577301,
|
| 191824 |
+
"eval_runtime": 40.3396,
|
| 191825 |
+
"eval_samples_per_second": 20.823,
|
| 191826 |
+
"eval_steps_per_second": 0.669,
|
| 191827 |
+
"eval_wer": 0.19356979573777744,
|
| 191828 |
+
"step": 72674
|
| 191829 |
}
|
| 191830 |
],
|
| 191831 |
"max_steps": 625000,
|
| 191832 |
"num_train_epochs": 5000,
|
| 191833 |
+
"total_flos": 2.0451382170982056e+20,
|
| 191834 |
"trial_name": null,
|
| 191835 |
"trial_params": null
|
| 191836 |
}
|
model-bin/finetune/base/{checkpoint-72051 → checkpoint-72674}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629865712.3999202/events.out.tfevents.1629865712.7e498afd5545.905.43
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8fe9358d32b91a31bd70c3b3f5de7f554b01d7c8e0997e77719220418e3a4d0b
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629866373.4153874/events.out.tfevents.1629866373.7e498afd5545.905.45
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:24d11bc9642f9b18f8b6e5803de4413f541482a4466333368fa876b6cc3709d3
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629867035.5276291/events.out.tfevents.1629867035.7e498afd5545.905.47
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:410a7f0161ac6cc4e8ebb05f5a9482de7b36e630250a0f7e7ebaa919175d729d
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629867698.6983736/events.out.tfevents.1629867698.7e498afd5545.905.49
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d3e789b2e73958b508e4dc81d29e447645345b6cabef52a495c211a76e64910e
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629868357.748418/events.out.tfevents.1629868357.7e498afd5545.905.51
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ca7151d4eeeeace9d594753ddac35893e62970b4085ae20b385b897f6982beb1
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629865712.7e498afd5545.905.42
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:57bb040a5525dec5c7b9f9a950eb9fb3bd287a1af563c2b933e8b972d14f9326
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629866373.7e498afd5545.905.44
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9154ee54589c0587718af3f4a26a91428cc1ba2de06727df37fd75ec34586f4c
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629867035.7e498afd5545.905.46
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:858dfb7a5ae08a765b032454f21eac0366f1cd4b4bce7d2e0686d09489f449dd
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629867698.7e498afd5545.905.48
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:39db8ce2bf39668b96bbf7ee7e4378a20f42e93e9e02e90dd6395941f8451209
|
| 3 |
+
size 8462
|
model-bin/finetune/base/log/events.out.tfevents.1629868357.7e498afd5545.905.50
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:5e1f96f05af17f0df8c92986868f317df4cb46095842d83a3ce3e7150543ed85
|
| 3 |
+
size 8622
|