"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-44550 → checkpoint-45173}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-44550 → checkpoint-45173}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-44550 → checkpoint-45173}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-44550 → checkpoint-45173}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-44550 → checkpoint-45173}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-44550 → checkpoint-45173}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-44550 → checkpoint-45173}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-44550 → checkpoint-45173}/trainer_state.json +792 -3
- model-bin/finetune/base/{checkpoint-44550 → checkpoint-45173}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629713095.9356778/events.out.tfevents.1629713095.74272264b15c.932.103 +3 -0
- model-bin/finetune/base/log/1629713745.6345289/events.out.tfevents.1629713745.74272264b15c.932.105 +3 -0
- model-bin/finetune/base/log/1629714384.1845772/events.out.tfevents.1629714384.74272264b15c.932.107 +3 -0
- model-bin/finetune/base/log/1629715035.40795/events.out.tfevents.1629715035.74272264b15c.932.109 +3 -0
- model-bin/finetune/base/log/1629715675.1128497/events.out.tfevents.1629715675.74272264b15c.932.111 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629713095.74272264b15c.932.102 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629713745.74272264b15c.932.104 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629714384.74272264b15c.932.106 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629715035.74272264b15c.932.108 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629715675.74272264b15c.932.110 +3 -0
model-bin/finetune/base/{checkpoint-44550 → checkpoint-45173}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-44550 → checkpoint-45173}/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:e03e84e5e2de2e89040568474dda156560785611486ddbbf44cd0911321d2e76
|
| 3 |
size 722165009
|
model-bin/finetune/base/{checkpoint-44550 → checkpoint-45173}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-44550 → checkpoint-45173}/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:f92f6dd5c4cd6b68322679888c0d028d8c925f50b8fcc9151246d447856a9a25
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-44550 → checkpoint-45173}/rng_state.pth
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 14503
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b2f28cddbdb9519b0485085cb45f99e6225363223e30b8696ba893fbb7357e30
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-44550 → checkpoint-45173}/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 sha256:0831c2a61563d74687eeabb1de647aeb187b95af92781ae63d6fd8ef10b4f0d4
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-44550 → checkpoint-45173}/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 sha256:089a9b8de54c0783e318cf9f990e00793f7e5952b3884eecf995cc74fb273731
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-44550 → checkpoint-45173}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.2003221555132523,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-42807",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -156048,11 +156048,800 @@
|
|
| 156048 |
"eval_steps_per_second": 0.68,
|
| 156049 |
"eval_wer": 0.20374774774774776,
|
| 156050 |
"step": 44550
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 156051 |
}
|
| 156052 |
],
|
| 156053 |
"max_steps": 620000,
|
| 156054 |
"num_train_epochs": 5000,
|
| 156055 |
-
"total_flos": 1.
|
| 156056 |
"trial_name": null,
|
| 156057 |
"trial_params": null
|
| 156058 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.2003221555132523,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-42807",
|
| 4 |
+
"epoch": 363.99598393574297,
|
| 5 |
+
"global_step": 45173,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 156048 |
"eval_steps_per_second": 0.68,
|
| 156049 |
"eval_wer": 0.20374774774774776,
|
| 156050 |
"step": 44550
|
| 156051 |
+
},
|
| 156052 |
+
{
|
| 156053 |
+
"epoch": 356.04,
|
| 156054 |
+
"learning_rate": 9.302275641025641e-06,
|
| 156055 |
+
"loss": 0.3525,
|
| 156056 |
+
"step": 44555
|
| 156057 |
+
},
|
| 156058 |
+
{
|
| 156059 |
+
"epoch": 356.08,
|
| 156060 |
+
"learning_rate": 9.302195512820513e-06,
|
| 156061 |
+
"loss": 0.3678,
|
| 156062 |
+
"step": 44560
|
| 156063 |
+
},
|
| 156064 |
+
{
|
| 156065 |
+
"epoch": 356.12,
|
| 156066 |
+
"learning_rate": 9.302115384615386e-06,
|
| 156067 |
+
"loss": 0.3448,
|
| 156068 |
+
"step": 44565
|
| 156069 |
+
},
|
| 156070 |
+
{
|
| 156071 |
+
"epoch": 356.16,
|
| 156072 |
+
"learning_rate": 9.302035256410257e-06,
|
| 156073 |
+
"loss": 0.5647,
|
| 156074 |
+
"step": 44570
|
| 156075 |
+
},
|
| 156076 |
+
{
|
| 156077 |
+
"epoch": 356.2,
|
| 156078 |
+
"learning_rate": 9.301955128205128e-06,
|
| 156079 |
+
"loss": 1.2745,
|
| 156080 |
+
"step": 44575
|
| 156081 |
+
},
|
| 156082 |
+
{
|
| 156083 |
+
"epoch": 356.24,
|
| 156084 |
+
"learning_rate": 9.301875e-06,
|
| 156085 |
+
"loss": 0.4439,
|
| 156086 |
+
"step": 44580
|
| 156087 |
+
},
|
| 156088 |
+
{
|
| 156089 |
+
"epoch": 356.28,
|
| 156090 |
+
"learning_rate": 9.301794871794873e-06,
|
| 156091 |
+
"loss": 0.3479,
|
| 156092 |
+
"step": 44585
|
| 156093 |
+
},
|
| 156094 |
+
{
|
| 156095 |
+
"epoch": 356.32,
|
| 156096 |
+
"learning_rate": 9.301714743589744e-06,
|
| 156097 |
+
"loss": 0.4558,
|
| 156098 |
+
"step": 44590
|
| 156099 |
+
},
|
| 156100 |
+
{
|
| 156101 |
+
"epoch": 356.36,
|
| 156102 |
+
"learning_rate": 9.301634615384616e-06,
|
| 156103 |
+
"loss": 0.5289,
|
| 156104 |
+
"step": 44595
|
| 156105 |
+
},
|
| 156106 |
+
{
|
| 156107 |
+
"epoch": 356.4,
|
| 156108 |
+
"learning_rate": 9.301554487179489e-06,
|
| 156109 |
+
"loss": 1.3446,
|
| 156110 |
+
"step": 44600
|
| 156111 |
+
},
|
| 156112 |
+
{
|
| 156113 |
+
"epoch": 356.44,
|
| 156114 |
+
"learning_rate": 9.30147435897436e-06,
|
| 156115 |
+
"loss": 0.3571,
|
| 156116 |
+
"step": 44605
|
| 156117 |
+
},
|
| 156118 |
+
{
|
| 156119 |
+
"epoch": 356.48,
|
| 156120 |
+
"learning_rate": 9.301394230769231e-06,
|
| 156121 |
+
"loss": 0.3611,
|
| 156122 |
+
"step": 44610
|
| 156123 |
+
},
|
| 156124 |
+
{
|
| 156125 |
+
"epoch": 356.52,
|
| 156126 |
+
"learning_rate": 9.301314102564103e-06,
|
| 156127 |
+
"loss": 0.5638,
|
| 156128 |
+
"step": 44615
|
| 156129 |
+
},
|
| 156130 |
+
{
|
| 156131 |
+
"epoch": 356.56,
|
| 156132 |
+
"learning_rate": 9.301233974358976e-06,
|
| 156133 |
+
"loss": 0.5012,
|
| 156134 |
+
"step": 44620
|
| 156135 |
+
},
|
| 156136 |
+
{
|
| 156137 |
+
"epoch": 356.6,
|
| 156138 |
+
"learning_rate": 9.301153846153847e-06,
|
| 156139 |
+
"loss": 1.3284,
|
| 156140 |
+
"step": 44625
|
| 156141 |
+
},
|
| 156142 |
+
{
|
| 156143 |
+
"epoch": 356.64,
|
| 156144 |
+
"learning_rate": 9.301073717948718e-06,
|
| 156145 |
+
"loss": 0.3341,
|
| 156146 |
+
"step": 44630
|
| 156147 |
+
},
|
| 156148 |
+
{
|
| 156149 |
+
"epoch": 356.68,
|
| 156150 |
+
"learning_rate": 9.300993589743591e-06,
|
| 156151 |
+
"loss": 0.3504,
|
| 156152 |
+
"step": 44635
|
| 156153 |
+
},
|
| 156154 |
+
{
|
| 156155 |
+
"epoch": 356.72,
|
| 156156 |
+
"learning_rate": 9.300913461538463e-06,
|
| 156157 |
+
"loss": 0.4351,
|
| 156158 |
+
"step": 44640
|
| 156159 |
+
},
|
| 156160 |
+
{
|
| 156161 |
+
"epoch": 356.76,
|
| 156162 |
+
"learning_rate": 9.300833333333334e-06,
|
| 156163 |
+
"loss": 0.5666,
|
| 156164 |
+
"step": 44645
|
| 156165 |
+
},
|
| 156166 |
+
{
|
| 156167 |
+
"epoch": 356.8,
|
| 156168 |
+
"learning_rate": 9.300753205128206e-06,
|
| 156169 |
+
"loss": 1.4891,
|
| 156170 |
+
"step": 44650
|
| 156171 |
+
},
|
| 156172 |
+
{
|
| 156173 |
+
"epoch": 356.84,
|
| 156174 |
+
"learning_rate": 9.300673076923079e-06,
|
| 156175 |
+
"loss": 0.3494,
|
| 156176 |
+
"step": 44655
|
| 156177 |
+
},
|
| 156178 |
+
{
|
| 156179 |
+
"epoch": 356.88,
|
| 156180 |
+
"learning_rate": 9.30059294871795e-06,
|
| 156181 |
+
"loss": 0.3815,
|
| 156182 |
+
"step": 44660
|
| 156183 |
+
},
|
| 156184 |
+
{
|
| 156185 |
+
"epoch": 356.92,
|
| 156186 |
+
"learning_rate": 9.300512820512821e-06,
|
| 156187 |
+
"loss": 0.6943,
|
| 156188 |
+
"step": 44665
|
| 156189 |
+
},
|
| 156190 |
+
{
|
| 156191 |
+
"epoch": 356.96,
|
| 156192 |
+
"learning_rate": 9.300432692307693e-06,
|
| 156193 |
+
"loss": 0.5394,
|
| 156194 |
+
"step": 44670
|
| 156195 |
+
},
|
| 156196 |
+
{
|
| 156197 |
+
"epoch": 357.0,
|
| 156198 |
+
"learning_rate": 9.300352564102566e-06,
|
| 156199 |
+
"loss": 1.4159,
|
| 156200 |
+
"step": 44675
|
| 156201 |
+
},
|
| 156202 |
+
{
|
| 156203 |
+
"epoch": 357.0,
|
| 156204 |
+
"eval_loss": 0.416872501373291,
|
| 156205 |
+
"eval_runtime": 39.9519,
|
| 156206 |
+
"eval_samples_per_second": 21.125,
|
| 156207 |
+
"eval_steps_per_second": 0.676,
|
| 156208 |
+
"eval_wer": 0.20372226198997853,
|
| 156209 |
+
"step": 44675
|
| 156210 |
+
},
|
| 156211 |
+
{
|
| 156212 |
+
"epoch": 357.04,
|
| 156213 |
+
"learning_rate": 9.300272435897435e-06,
|
| 156214 |
+
"loss": 0.415,
|
| 156215 |
+
"step": 44680
|
| 156216 |
+
},
|
| 156217 |
+
{
|
| 156218 |
+
"epoch": 357.08,
|
| 156219 |
+
"learning_rate": 9.300192307692308e-06,
|
| 156220 |
+
"loss": 0.3118,
|
| 156221 |
+
"step": 44685
|
| 156222 |
+
},
|
| 156223 |
+
{
|
| 156224 |
+
"epoch": 357.12,
|
| 156225 |
+
"learning_rate": 9.300112179487182e-06,
|
| 156226 |
+
"loss": 0.3974,
|
| 156227 |
+
"step": 44690
|
| 156228 |
+
},
|
| 156229 |
+
{
|
| 156230 |
+
"epoch": 357.16,
|
| 156231 |
+
"learning_rate": 9.300032051282051e-06,
|
| 156232 |
+
"loss": 0.5174,
|
| 156233 |
+
"step": 44695
|
| 156234 |
+
},
|
| 156235 |
+
{
|
| 156236 |
+
"epoch": 357.2,
|
| 156237 |
+
"learning_rate": 9.299951923076924e-06,
|
| 156238 |
+
"loss": 1.329,
|
| 156239 |
+
"step": 44700
|
| 156240 |
+
},
|
| 156241 |
+
{
|
| 156242 |
+
"epoch": 357.24,
|
| 156243 |
+
"learning_rate": 9.299871794871796e-06,
|
| 156244 |
+
"loss": 0.4483,
|
| 156245 |
+
"step": 44705
|
| 156246 |
+
},
|
| 156247 |
+
{
|
| 156248 |
+
"epoch": 357.28,
|
| 156249 |
+
"learning_rate": 9.299791666666667e-06,
|
| 156250 |
+
"loss": 0.3715,
|
| 156251 |
+
"step": 44710
|
| 156252 |
+
},
|
| 156253 |
+
{
|
| 156254 |
+
"epoch": 357.32,
|
| 156255 |
+
"learning_rate": 9.299711538461538e-06,
|
| 156256 |
+
"loss": 0.3451,
|
| 156257 |
+
"step": 44715
|
| 156258 |
+
},
|
| 156259 |
+
{
|
| 156260 |
+
"epoch": 357.36,
|
| 156261 |
+
"learning_rate": 9.299631410256411e-06,
|
| 156262 |
+
"loss": 0.5133,
|
| 156263 |
+
"step": 44720
|
| 156264 |
+
},
|
| 156265 |
+
{
|
| 156266 |
+
"epoch": 357.4,
|
| 156267 |
+
"learning_rate": 9.299551282051283e-06,
|
| 156268 |
+
"loss": 1.3289,
|
| 156269 |
+
"step": 44725
|
| 156270 |
+
},
|
| 156271 |
+
{
|
| 156272 |
+
"epoch": 357.44,
|
| 156273 |
+
"learning_rate": 9.299471153846154e-06,
|
| 156274 |
+
"loss": 0.4257,
|
| 156275 |
+
"step": 44730
|
| 156276 |
+
},
|
| 156277 |
+
{
|
| 156278 |
+
"epoch": 357.48,
|
| 156279 |
+
"learning_rate": 9.299391025641027e-06,
|
| 156280 |
+
"loss": 0.2918,
|
| 156281 |
+
"step": 44735
|
| 156282 |
+
},
|
| 156283 |
+
{
|
| 156284 |
+
"epoch": 357.52,
|
| 156285 |
+
"learning_rate": 9.299310897435898e-06,
|
| 156286 |
+
"loss": 0.4379,
|
| 156287 |
+
"step": 44740
|
| 156288 |
+
},
|
| 156289 |
+
{
|
| 156290 |
+
"epoch": 357.56,
|
| 156291 |
+
"learning_rate": 9.29923076923077e-06,
|
| 156292 |
+
"loss": 0.5659,
|
| 156293 |
+
"step": 44745
|
| 156294 |
+
},
|
| 156295 |
+
{
|
| 156296 |
+
"epoch": 357.6,
|
| 156297 |
+
"learning_rate": 9.299150641025641e-06,
|
| 156298 |
+
"loss": 1.3485,
|
| 156299 |
+
"step": 44750
|
| 156300 |
+
},
|
| 156301 |
+
{
|
| 156302 |
+
"epoch": 357.64,
|
| 156303 |
+
"learning_rate": 9.299070512820514e-06,
|
| 156304 |
+
"loss": 0.4277,
|
| 156305 |
+
"step": 44755
|
| 156306 |
+
},
|
| 156307 |
+
{
|
| 156308 |
+
"epoch": 357.68,
|
| 156309 |
+
"learning_rate": 9.298990384615386e-06,
|
| 156310 |
+
"loss": 0.356,
|
| 156311 |
+
"step": 44760
|
| 156312 |
+
},
|
| 156313 |
+
{
|
| 156314 |
+
"epoch": 357.72,
|
| 156315 |
+
"learning_rate": 9.298910256410257e-06,
|
| 156316 |
+
"loss": 0.4229,
|
| 156317 |
+
"step": 44765
|
| 156318 |
+
},
|
| 156319 |
+
{
|
| 156320 |
+
"epoch": 357.76,
|
| 156321 |
+
"learning_rate": 9.298830128205128e-06,
|
| 156322 |
+
"loss": 0.4602,
|
| 156323 |
+
"step": 44770
|
| 156324 |
+
},
|
| 156325 |
+
{
|
| 156326 |
+
"epoch": 357.8,
|
| 156327 |
+
"learning_rate": 9.298750000000001e-06,
|
| 156328 |
+
"loss": 1.2586,
|
| 156329 |
+
"step": 44775
|
| 156330 |
+
},
|
| 156331 |
+
{
|
| 156332 |
+
"epoch": 357.84,
|
| 156333 |
+
"learning_rate": 9.298669871794873e-06,
|
| 156334 |
+
"loss": 0.4194,
|
| 156335 |
+
"step": 44780
|
| 156336 |
+
},
|
| 156337 |
+
{
|
| 156338 |
+
"epoch": 357.88,
|
| 156339 |
+
"learning_rate": 9.298589743589744e-06,
|
| 156340 |
+
"loss": 0.3401,
|
| 156341 |
+
"step": 44785
|
| 156342 |
+
},
|
| 156343 |
+
{
|
| 156344 |
+
"epoch": 357.92,
|
| 156345 |
+
"learning_rate": 9.298509615384617e-06,
|
| 156346 |
+
"loss": 0.4347,
|
| 156347 |
+
"step": 44790
|
| 156348 |
+
},
|
| 156349 |
+
{
|
| 156350 |
+
"epoch": 357.96,
|
| 156351 |
+
"learning_rate": 9.298429487179489e-06,
|
| 156352 |
+
"loss": 0.5679,
|
| 156353 |
+
"step": 44795
|
| 156354 |
+
},
|
| 156355 |
+
{
|
| 156356 |
+
"epoch": 358.0,
|
| 156357 |
+
"learning_rate": 9.29834935897436e-06,
|
| 156358 |
+
"loss": 1.4365,
|
| 156359 |
+
"step": 44800
|
| 156360 |
+
},
|
| 156361 |
+
{
|
| 156362 |
+
"epoch": 358.0,
|
| 156363 |
+
"eval_loss": 0.4820411205291748,
|
| 156364 |
+
"eval_runtime": 40.4662,
|
| 156365 |
+
"eval_samples_per_second": 20.857,
|
| 156366 |
+
"eval_steps_per_second": 0.667,
|
| 156367 |
+
"eval_wer": 0.21496092296241162,
|
| 156368 |
+
"step": 44800
|
| 156369 |
+
},
|
| 156370 |
+
{
|
| 156371 |
+
"epoch": 358.04,
|
| 156372 |
+
"learning_rate": 9.298269230769231e-06,
|
| 156373 |
+
"loss": 0.3813,
|
| 156374 |
+
"step": 44805
|
| 156375 |
+
},
|
| 156376 |
+
{
|
| 156377 |
+
"epoch": 358.08,
|
| 156378 |
+
"learning_rate": 9.298189102564104e-06,
|
| 156379 |
+
"loss": 0.3869,
|
| 156380 |
+
"step": 44810
|
| 156381 |
+
},
|
| 156382 |
+
{
|
| 156383 |
+
"epoch": 358.12,
|
| 156384 |
+
"learning_rate": 9.298108974358974e-06,
|
| 156385 |
+
"loss": 0.5295,
|
| 156386 |
+
"step": 44815
|
| 156387 |
+
},
|
| 156388 |
+
{
|
| 156389 |
+
"epoch": 358.16,
|
| 156390 |
+
"learning_rate": 9.298028846153847e-06,
|
| 156391 |
+
"loss": 0.5543,
|
| 156392 |
+
"step": 44820
|
| 156393 |
+
},
|
| 156394 |
+
{
|
| 156395 |
+
"epoch": 358.2,
|
| 156396 |
+
"learning_rate": 9.297948717948718e-06,
|
| 156397 |
+
"loss": 1.3686,
|
| 156398 |
+
"step": 44825
|
| 156399 |
+
},
|
| 156400 |
+
{
|
| 156401 |
+
"epoch": 358.24,
|
| 156402 |
+
"learning_rate": 9.29786858974359e-06,
|
| 156403 |
+
"loss": 0.4286,
|
| 156404 |
+
"step": 44830
|
| 156405 |
+
},
|
| 156406 |
+
{
|
| 156407 |
+
"epoch": 358.28,
|
| 156408 |
+
"learning_rate": 9.297788461538463e-06,
|
| 156409 |
+
"loss": 0.3347,
|
| 156410 |
+
"step": 44835
|
| 156411 |
+
},
|
| 156412 |
+
{
|
| 156413 |
+
"epoch": 358.32,
|
| 156414 |
+
"learning_rate": 9.297708333333334e-06,
|
| 156415 |
+
"loss": 0.362,
|
| 156416 |
+
"step": 44840
|
| 156417 |
+
},
|
| 156418 |
+
{
|
| 156419 |
+
"epoch": 358.36,
|
| 156420 |
+
"learning_rate": 9.297628205128205e-06,
|
| 156421 |
+
"loss": 0.5454,
|
| 156422 |
+
"step": 44845
|
| 156423 |
+
},
|
| 156424 |
+
{
|
| 156425 |
+
"epoch": 358.4,
|
| 156426 |
+
"learning_rate": 9.297548076923077e-06,
|
| 156427 |
+
"loss": 1.3438,
|
| 156428 |
+
"step": 44850
|
| 156429 |
+
},
|
| 156430 |
+
{
|
| 156431 |
+
"epoch": 358.44,
|
| 156432 |
+
"learning_rate": 9.29746794871795e-06,
|
| 156433 |
+
"loss": 0.3965,
|
| 156434 |
+
"step": 44855
|
| 156435 |
+
},
|
| 156436 |
+
{
|
| 156437 |
+
"epoch": 358.48,
|
| 156438 |
+
"learning_rate": 9.297387820512821e-06,
|
| 156439 |
+
"loss": 0.3505,
|
| 156440 |
+
"step": 44860
|
| 156441 |
+
},
|
| 156442 |
+
{
|
| 156443 |
+
"epoch": 358.52,
|
| 156444 |
+
"learning_rate": 9.297307692307693e-06,
|
| 156445 |
+
"loss": 0.3767,
|
| 156446 |
+
"step": 44865
|
| 156447 |
+
},
|
| 156448 |
+
{
|
| 156449 |
+
"epoch": 358.56,
|
| 156450 |
+
"learning_rate": 9.297227564102564e-06,
|
| 156451 |
+
"loss": 0.6327,
|
| 156452 |
+
"step": 44870
|
| 156453 |
+
},
|
| 156454 |
+
{
|
| 156455 |
+
"epoch": 358.6,
|
| 156456 |
+
"learning_rate": 9.297147435897437e-06,
|
| 156457 |
+
"loss": 1.2411,
|
| 156458 |
+
"step": 44875
|
| 156459 |
+
},
|
| 156460 |
+
{
|
| 156461 |
+
"epoch": 358.64,
|
| 156462 |
+
"learning_rate": 9.297067307692308e-06,
|
| 156463 |
+
"loss": 0.3577,
|
| 156464 |
+
"step": 44880
|
| 156465 |
+
},
|
| 156466 |
+
{
|
| 156467 |
+
"epoch": 358.68,
|
| 156468 |
+
"learning_rate": 9.29698717948718e-06,
|
| 156469 |
+
"loss": 0.3982,
|
| 156470 |
+
"step": 44885
|
| 156471 |
+
},
|
| 156472 |
+
{
|
| 156473 |
+
"epoch": 358.72,
|
| 156474 |
+
"learning_rate": 9.296907051282053e-06,
|
| 156475 |
+
"loss": 0.3611,
|
| 156476 |
+
"step": 44890
|
| 156477 |
+
},
|
| 156478 |
+
{
|
| 156479 |
+
"epoch": 358.76,
|
| 156480 |
+
"learning_rate": 9.296826923076924e-06,
|
| 156481 |
+
"loss": 0.4977,
|
| 156482 |
+
"step": 44895
|
| 156483 |
+
},
|
| 156484 |
+
{
|
| 156485 |
+
"epoch": 358.8,
|
| 156486 |
+
"learning_rate": 9.296746794871796e-06,
|
| 156487 |
+
"loss": 1.3198,
|
| 156488 |
+
"step": 44900
|
| 156489 |
+
},
|
| 156490 |
+
{
|
| 156491 |
+
"epoch": 358.84,
|
| 156492 |
+
"learning_rate": 9.296666666666667e-06,
|
| 156493 |
+
"loss": 0.4275,
|
| 156494 |
+
"step": 44905
|
| 156495 |
+
},
|
| 156496 |
+
{
|
| 156497 |
+
"epoch": 358.88,
|
| 156498 |
+
"learning_rate": 9.29658653846154e-06,
|
| 156499 |
+
"loss": 0.3143,
|
| 156500 |
+
"step": 44910
|
| 156501 |
+
},
|
| 156502 |
+
{
|
| 156503 |
+
"epoch": 358.92,
|
| 156504 |
+
"learning_rate": 9.296506410256411e-06,
|
| 156505 |
+
"loss": 0.3518,
|
| 156506 |
+
"step": 44915
|
| 156507 |
+
},
|
| 156508 |
+
{
|
| 156509 |
+
"epoch": 358.96,
|
| 156510 |
+
"learning_rate": 9.296426282051283e-06,
|
| 156511 |
+
"loss": 0.5841,
|
| 156512 |
+
"step": 44920
|
| 156513 |
+
},
|
| 156514 |
+
{
|
| 156515 |
+
"epoch": 359.0,
|
| 156516 |
+
"learning_rate": 9.296346153846154e-06,
|
| 156517 |
+
"loss": 1.355,
|
| 156518 |
+
"step": 44925
|
| 156519 |
+
},
|
| 156520 |
+
{
|
| 156521 |
+
"epoch": 359.0,
|
| 156522 |
+
"eval_loss": 0.42383745312690735,
|
| 156523 |
+
"eval_runtime": 40.6859,
|
| 156524 |
+
"eval_samples_per_second": 20.744,
|
| 156525 |
+
"eval_steps_per_second": 0.664,
|
| 156526 |
+
"eval_wer": 0.21129056156467624,
|
| 156527 |
+
"step": 44925
|
| 156528 |
+
},
|
| 156529 |
+
{
|
| 156530 |
+
"epoch": 362.04,
|
| 156531 |
+
"learning_rate": 9.296266025641027e-06,
|
| 156532 |
+
"loss": 0.403,
|
| 156533 |
+
"step": 44930
|
| 156534 |
+
},
|
| 156535 |
+
{
|
| 156536 |
+
"epoch": 362.08,
|
| 156537 |
+
"learning_rate": 9.296185897435898e-06,
|
| 156538 |
+
"loss": 0.3936,
|
| 156539 |
+
"step": 44935
|
| 156540 |
+
},
|
| 156541 |
+
{
|
| 156542 |
+
"epoch": 362.12,
|
| 156543 |
+
"learning_rate": 9.29610576923077e-06,
|
| 156544 |
+
"loss": 0.3293,
|
| 156545 |
+
"step": 44940
|
| 156546 |
+
},
|
| 156547 |
+
{
|
| 156548 |
+
"epoch": 362.16,
|
| 156549 |
+
"learning_rate": 9.296025641025643e-06,
|
| 156550 |
+
"loss": 0.4635,
|
| 156551 |
+
"step": 44945
|
| 156552 |
+
},
|
| 156553 |
+
{
|
| 156554 |
+
"epoch": 362.2,
|
| 156555 |
+
"learning_rate": 9.295945512820514e-06,
|
| 156556 |
+
"loss": 1.4165,
|
| 156557 |
+
"step": 44950
|
| 156558 |
+
},
|
| 156559 |
+
{
|
| 156560 |
+
"epoch": 362.24,
|
| 156561 |
+
"learning_rate": 9.295865384615386e-06,
|
| 156562 |
+
"loss": 0.3775,
|
| 156563 |
+
"step": 44955
|
| 156564 |
+
},
|
| 156565 |
+
{
|
| 156566 |
+
"epoch": 362.28,
|
| 156567 |
+
"learning_rate": 9.295785256410257e-06,
|
| 156568 |
+
"loss": 0.3374,
|
| 156569 |
+
"step": 44960
|
| 156570 |
+
},
|
| 156571 |
+
{
|
| 156572 |
+
"epoch": 362.32,
|
| 156573 |
+
"learning_rate": 9.29570512820513e-06,
|
| 156574 |
+
"loss": 0.4018,
|
| 156575 |
+
"step": 44965
|
| 156576 |
+
},
|
| 156577 |
+
{
|
| 156578 |
+
"epoch": 362.36,
|
| 156579 |
+
"learning_rate": 9.295625e-06,
|
| 156580 |
+
"loss": 0.6313,
|
| 156581 |
+
"step": 44970
|
| 156582 |
+
},
|
| 156583 |
+
{
|
| 156584 |
+
"epoch": 362.4,
|
| 156585 |
+
"learning_rate": 9.295544871794873e-06,
|
| 156586 |
+
"loss": 1.3116,
|
| 156587 |
+
"step": 44975
|
| 156588 |
+
},
|
| 156589 |
+
{
|
| 156590 |
+
"epoch": 362.44,
|
| 156591 |
+
"learning_rate": 9.295464743589744e-06,
|
| 156592 |
+
"loss": 0.3469,
|
| 156593 |
+
"step": 44980
|
| 156594 |
+
},
|
| 156595 |
+
{
|
| 156596 |
+
"epoch": 362.48,
|
| 156597 |
+
"learning_rate": 9.295384615384615e-06,
|
| 156598 |
+
"loss": 0.3339,
|
| 156599 |
+
"step": 44985
|
| 156600 |
+
},
|
| 156601 |
+
{
|
| 156602 |
+
"epoch": 362.52,
|
| 156603 |
+
"learning_rate": 9.295304487179488e-06,
|
| 156604 |
+
"loss": 0.3281,
|
| 156605 |
+
"step": 44990
|
| 156606 |
+
},
|
| 156607 |
+
{
|
| 156608 |
+
"epoch": 362.56,
|
| 156609 |
+
"learning_rate": 9.29522435897436e-06,
|
| 156610 |
+
"loss": 0.5257,
|
| 156611 |
+
"step": 44995
|
| 156612 |
+
},
|
| 156613 |
+
{
|
| 156614 |
+
"epoch": 362.6,
|
| 156615 |
+
"learning_rate": 9.295144230769231e-06,
|
| 156616 |
+
"loss": 1.2097,
|
| 156617 |
+
"step": 45000
|
| 156618 |
+
},
|
| 156619 |
+
{
|
| 156620 |
+
"epoch": 362.64,
|
| 156621 |
+
"learning_rate": 9.295064102564103e-06,
|
| 156622 |
+
"loss": 0.434,
|
| 156623 |
+
"step": 45005
|
| 156624 |
+
},
|
| 156625 |
+
{
|
| 156626 |
+
"epoch": 362.68,
|
| 156627 |
+
"learning_rate": 9.294983974358976e-06,
|
| 156628 |
+
"loss": 0.3269,
|
| 156629 |
+
"step": 45010
|
| 156630 |
+
},
|
| 156631 |
+
{
|
| 156632 |
+
"epoch": 362.72,
|
| 156633 |
+
"learning_rate": 9.294903846153847e-06,
|
| 156634 |
+
"loss": 0.4081,
|
| 156635 |
+
"step": 45015
|
| 156636 |
+
},
|
| 156637 |
+
{
|
| 156638 |
+
"epoch": 362.76,
|
| 156639 |
+
"learning_rate": 9.294823717948718e-06,
|
| 156640 |
+
"loss": 0.4983,
|
| 156641 |
+
"step": 45020
|
| 156642 |
+
},
|
| 156643 |
+
{
|
| 156644 |
+
"epoch": 362.8,
|
| 156645 |
+
"learning_rate": 9.29474358974359e-06,
|
| 156646 |
+
"loss": 1.309,
|
| 156647 |
+
"step": 45025
|
| 156648 |
+
},
|
| 156649 |
+
{
|
| 156650 |
+
"epoch": 362.84,
|
| 156651 |
+
"learning_rate": 9.294663461538463e-06,
|
| 156652 |
+
"loss": 0.4292,
|
| 156653 |
+
"step": 45030
|
| 156654 |
+
},
|
| 156655 |
+
{
|
| 156656 |
+
"epoch": 362.88,
|
| 156657 |
+
"learning_rate": 9.294583333333334e-06,
|
| 156658 |
+
"loss": 0.3685,
|
| 156659 |
+
"step": 45035
|
| 156660 |
+
},
|
| 156661 |
+
{
|
| 156662 |
+
"epoch": 362.92,
|
| 156663 |
+
"learning_rate": 9.294503205128205e-06,
|
| 156664 |
+
"loss": 0.3778,
|
| 156665 |
+
"step": 45040
|
| 156666 |
+
},
|
| 156667 |
+
{
|
| 156668 |
+
"epoch": 362.96,
|
| 156669 |
+
"learning_rate": 9.294423076923078e-06,
|
| 156670 |
+
"loss": 0.5988,
|
| 156671 |
+
"step": 45045
|
| 156672 |
+
},
|
| 156673 |
+
{
|
| 156674 |
+
"epoch": 363.0,
|
| 156675 |
+
"eval_loss": 0.42374563217163086,
|
| 156676 |
+
"eval_runtime": 39.8218,
|
| 156677 |
+
"eval_samples_per_second": 21.169,
|
| 156678 |
+
"eval_steps_per_second": 0.678,
|
| 156679 |
+
"eval_wer": 0.20569620253164558,
|
| 156680 |
+
"step": 45049
|
| 156681 |
+
},
|
| 156682 |
+
{
|
| 156683 |
+
"epoch": 363.01,
|
| 156684 |
+
"learning_rate": 9.29434294871795e-06,
|
| 156685 |
+
"loss": 0.4274,
|
| 156686 |
+
"step": 45050
|
| 156687 |
+
},
|
| 156688 |
+
{
|
| 156689 |
+
"epoch": 363.05,
|
| 156690 |
+
"learning_rate": 9.294262820512821e-06,
|
| 156691 |
+
"loss": 0.3244,
|
| 156692 |
+
"step": 45055
|
| 156693 |
+
},
|
| 156694 |
+
{
|
| 156695 |
+
"epoch": 363.09,
|
| 156696 |
+
"learning_rate": 9.294182692307693e-06,
|
| 156697 |
+
"loss": 0.3628,
|
| 156698 |
+
"step": 45060
|
| 156699 |
+
},
|
| 156700 |
+
{
|
| 156701 |
+
"epoch": 363.13,
|
| 156702 |
+
"learning_rate": 9.294102564102566e-06,
|
| 156703 |
+
"loss": 0.3902,
|
| 156704 |
+
"step": 45065
|
| 156705 |
+
},
|
| 156706 |
+
{
|
| 156707 |
+
"epoch": 363.17,
|
| 156708 |
+
"learning_rate": 9.294022435897437e-06,
|
| 156709 |
+
"loss": 0.6742,
|
| 156710 |
+
"step": 45070
|
| 156711 |
+
},
|
| 156712 |
+
{
|
| 156713 |
+
"epoch": 363.21,
|
| 156714 |
+
"learning_rate": 9.293942307692308e-06,
|
| 156715 |
+
"loss": 1.2861,
|
| 156716 |
+
"step": 45075
|
| 156717 |
+
},
|
| 156718 |
+
{
|
| 156719 |
+
"epoch": 363.25,
|
| 156720 |
+
"learning_rate": 9.29386217948718e-06,
|
| 156721 |
+
"loss": 0.3662,
|
| 156722 |
+
"step": 45080
|
| 156723 |
+
},
|
| 156724 |
+
{
|
| 156725 |
+
"epoch": 363.29,
|
| 156726 |
+
"learning_rate": 9.293782051282053e-06,
|
| 156727 |
+
"loss": 0.3126,
|
| 156728 |
+
"step": 45085
|
| 156729 |
+
},
|
| 156730 |
+
{
|
| 156731 |
+
"epoch": 363.33,
|
| 156732 |
+
"learning_rate": 9.293701923076924e-06,
|
| 156733 |
+
"loss": 0.3738,
|
| 156734 |
+
"step": 45090
|
| 156735 |
+
},
|
| 156736 |
+
{
|
| 156737 |
+
"epoch": 363.37,
|
| 156738 |
+
"learning_rate": 9.293621794871795e-06,
|
| 156739 |
+
"loss": 0.6139,
|
| 156740 |
+
"step": 45095
|
| 156741 |
+
},
|
| 156742 |
+
{
|
| 156743 |
+
"epoch": 363.41,
|
| 156744 |
+
"learning_rate": 9.293541666666668e-06,
|
| 156745 |
+
"loss": 1.1466,
|
| 156746 |
+
"step": 45100
|
| 156747 |
+
},
|
| 156748 |
+
{
|
| 156749 |
+
"epoch": 363.45,
|
| 156750 |
+
"learning_rate": 9.293461538461538e-06,
|
| 156751 |
+
"loss": 0.4075,
|
| 156752 |
+
"step": 45105
|
| 156753 |
+
},
|
| 156754 |
+
{
|
| 156755 |
+
"epoch": 363.49,
|
| 156756 |
+
"learning_rate": 9.293381410256411e-06,
|
| 156757 |
+
"loss": 0.3647,
|
| 156758 |
+
"step": 45110
|
| 156759 |
+
},
|
| 156760 |
+
{
|
| 156761 |
+
"epoch": 363.53,
|
| 156762 |
+
"learning_rate": 9.293301282051283e-06,
|
| 156763 |
+
"loss": 0.4528,
|
| 156764 |
+
"step": 45115
|
| 156765 |
+
},
|
| 156766 |
+
{
|
| 156767 |
+
"epoch": 363.57,
|
| 156768 |
+
"learning_rate": 9.293221153846154e-06,
|
| 156769 |
+
"loss": 0.5232,
|
| 156770 |
+
"step": 45120
|
| 156771 |
+
},
|
| 156772 |
+
{
|
| 156773 |
+
"epoch": 363.61,
|
| 156774 |
+
"learning_rate": 9.293141025641025e-06,
|
| 156775 |
+
"loss": 1.3171,
|
| 156776 |
+
"step": 45125
|
| 156777 |
+
},
|
| 156778 |
+
{
|
| 156779 |
+
"epoch": 363.65,
|
| 156780 |
+
"learning_rate": 9.293060897435898e-06,
|
| 156781 |
+
"loss": 0.3799,
|
| 156782 |
+
"step": 45130
|
| 156783 |
+
},
|
| 156784 |
+
{
|
| 156785 |
+
"epoch": 363.69,
|
| 156786 |
+
"learning_rate": 9.29298076923077e-06,
|
| 156787 |
+
"loss": 0.3184,
|
| 156788 |
+
"step": 45135
|
| 156789 |
+
},
|
| 156790 |
+
{
|
| 156791 |
+
"epoch": 363.73,
|
| 156792 |
+
"learning_rate": 9.292900641025641e-06,
|
| 156793 |
+
"loss": 0.3589,
|
| 156794 |
+
"step": 45140
|
| 156795 |
+
},
|
| 156796 |
+
{
|
| 156797 |
+
"epoch": 363.77,
|
| 156798 |
+
"learning_rate": 9.292820512820514e-06,
|
| 156799 |
+
"loss": 0.5893,
|
| 156800 |
+
"step": 45145
|
| 156801 |
+
},
|
| 156802 |
+
{
|
| 156803 |
+
"epoch": 363.81,
|
| 156804 |
+
"learning_rate": 9.292740384615385e-06,
|
| 156805 |
+
"loss": 1.2559,
|
| 156806 |
+
"step": 45150
|
| 156807 |
+
},
|
| 156808 |
+
{
|
| 156809 |
+
"epoch": 363.85,
|
| 156810 |
+
"learning_rate": 9.292660256410257e-06,
|
| 156811 |
+
"loss": 0.424,
|
| 156812 |
+
"step": 45155
|
| 156813 |
+
},
|
| 156814 |
+
{
|
| 156815 |
+
"epoch": 363.89,
|
| 156816 |
+
"learning_rate": 9.292580128205128e-06,
|
| 156817 |
+
"loss": 0.3622,
|
| 156818 |
+
"step": 45160
|
| 156819 |
+
},
|
| 156820 |
+
{
|
| 156821 |
+
"epoch": 363.93,
|
| 156822 |
+
"learning_rate": 9.292500000000001e-06,
|
| 156823 |
+
"loss": 0.4648,
|
| 156824 |
+
"step": 45165
|
| 156825 |
+
},
|
| 156826 |
+
{
|
| 156827 |
+
"epoch": 363.97,
|
| 156828 |
+
"learning_rate": 9.292419871794873e-06,
|
| 156829 |
+
"loss": 0.7258,
|
| 156830 |
+
"step": 45170
|
| 156831 |
+
},
|
| 156832 |
+
{
|
| 156833 |
+
"epoch": 364.0,
|
| 156834 |
+
"eval_loss": 0.40380603075027466,
|
| 156835 |
+
"eval_runtime": 39.3667,
|
| 156836 |
+
"eval_samples_per_second": 21.439,
|
| 156837 |
+
"eval_steps_per_second": 0.686,
|
| 156838 |
+
"eval_wer": 0.2054794520547945,
|
| 156839 |
+
"step": 45173
|
| 156840 |
}
|
| 156841 |
],
|
| 156842 |
"max_steps": 620000,
|
| 156843 |
"num_train_epochs": 5000,
|
| 156844 |
+
"total_flos": 1.2710145731316258e+20,
|
| 156845 |
"trial_name": null,
|
| 156846 |
"trial_params": null
|
| 156847 |
}
|
model-bin/finetune/base/{checkpoint-44550 → checkpoint-45173}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629713095.9356778/events.out.tfevents.1629713095.74272264b15c.932.103
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:03490ee5889e4a0d18f563d4b12bf93852f7130b2f31adc574d89ad80b5b858b
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629713745.6345289/events.out.tfevents.1629713745.74272264b15c.932.105
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:3667fc4d22b83d2d0e832db0868b8699f1b618e0dbcb1d2cecdbbe2f250bb695
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629714384.1845772/events.out.tfevents.1629714384.74272264b15c.932.107
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d4f1d100883b76f213e92ab145479a1b1bc479077f5f6b396f5f15ca4eae98c8
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629715035.40795/events.out.tfevents.1629715035.74272264b15c.932.109
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f7e1de040a6598a138eb15c7d1de24800c95521ee990ec8642a876cbb3a3f169
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629715675.1128497/events.out.tfevents.1629715675.74272264b15c.932.111
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:63c5b11cd8da0403672e2d6fd661cd0c8c06d0df40666d07af0d91f9981d5025
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629713095.74272264b15c.932.102
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:fdc64f27307d916378988e4b72c756678b1633d46fe4e31792c862a74f1b2d64
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629713745.74272264b15c.932.104
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:24aa9042af2f1103ee146f260233639fc8cd98fa4612adc0ad3b46654d6469c1
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629714384.74272264b15c.932.106
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1beb1c4cacf29e0e4f818d9fcdc2af6644abee0037d4468c88b29db67494a2b3
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629715035.74272264b15c.932.108
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d3ce452acddf9a15738ca53538f4045e90fa13745d4df5d50035a765bc67f392
|
| 3 |
+
size 8462
|
model-bin/finetune/base/log/events.out.tfevents.1629715675.74272264b15c.932.110
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2fa2acf847b6398b132c94a582cb19b16a2c0b66d2975c336f203206d762cdb3
|
| 3 |
+
size 8622
|