"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-139632 → checkpoint-140253}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-139632 → checkpoint-140253}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-139632 → checkpoint-140253}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-139632 → checkpoint-140253}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-139632 → checkpoint-140253}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-139632 → checkpoint-140253}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-139632 → checkpoint-140253}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-139632 → checkpoint-140253}/trainer_state.json +793 -4
- model-bin/finetune/base/{checkpoint-139632 → checkpoint-140253}/training_args.bin +0 -0
- model-bin/finetune/base/log/1630205681.0762417/events.out.tfevents.1630205681.cc93b136ebf5.1086.9 +3 -0
- model-bin/finetune/base/log/1630206127.5566013/events.out.tfevents.1630206127.cc93b136ebf5.1086.11 +3 -0
- model-bin/finetune/base/log/1630206572.6054518/events.out.tfevents.1630206572.cc93b136ebf5.1086.13 +3 -0
- model-bin/finetune/base/log/1630207019.0214832/events.out.tfevents.1630207019.cc93b136ebf5.1086.15 +3 -0
- model-bin/finetune/base/log/1630207465.424582/events.out.tfevents.1630207465.cc93b136ebf5.1086.17 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630205681.cc93b136ebf5.1086.8 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630206127.cc93b136ebf5.1086.10 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630206572.cc93b136ebf5.1086.12 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630207019.cc93b136ebf5.1086.14 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630207465.cc93b136ebf5.1086.16 +3 -0
model-bin/finetune/base/{checkpoint-139632 → checkpoint-140253}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-139632 → checkpoint-140253}/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 sha256:5bcbf3bd6bb9606b722f4044ad7b4cca358c096cb7125d78a03b09b6a667d368
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-139632 → checkpoint-140253}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-139632 → checkpoint-140253}/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:be04146c0a1f72988e88be69bb4cc9e3d8462c91e14da392807e285584dd369f
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-139632 → checkpoint-140253}/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 sha256:92ab12c7cad96755623e050a4522d6a39c67d2ae9f4b918efb3b0355d090143b
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-139632 → checkpoint-140253}/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:c04d92b6e9668a995aa67c0fd2cccd57c4af61acd1f14d12f1653f5c4a66c8e5
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-139632 → checkpoint-140253}/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:b3d4bf3c7862da618bb07ad83d6025576f9502002aca4a3defa7e8eea12e9000
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-139632 → checkpoint-140253}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1689111747851003,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-132910",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -277020,11 +277020,800 @@
|
|
| 277020 |
"eval_steps_per_second": 0.64,
|
| 277021 |
"eval_wer": 0.18331400405914758,
|
| 277022 |
"step": 139632
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 277023 |
}
|
| 277024 |
],
|
| 277025 |
-
"max_steps":
|
| 277026 |
"num_train_epochs": 5000,
|
| 277027 |
-
"total_flos": 3.
|
| 277028 |
"trial_name": null,
|
| 277029 |
"trial_params": null
|
| 277030 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1689111747851003,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-132910",
|
| 4 |
+
"epoch": 1122.0,
|
| 5 |
+
"global_step": 140253,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 277020 |
"eval_steps_per_second": 0.64,
|
| 277021 |
"eval_wer": 0.18331400405914758,
|
| 277022 |
"step": 139632
|
| 277023 |
+
},
|
| 277024 |
+
{
|
| 277025 |
+
"epoch": 1126.02,
|
| 277026 |
+
"learning_rate": 7.761227786752828e-06,
|
| 277027 |
+
"loss": 0.3216,
|
| 277028 |
+
"step": 139635
|
| 277029 |
+
},
|
| 277030 |
+
{
|
| 277031 |
+
"epoch": 1126.06,
|
| 277032 |
+
"learning_rate": 7.761147011308564e-06,
|
| 277033 |
+
"loss": 0.2688,
|
| 277034 |
+
"step": 139640
|
| 277035 |
+
},
|
| 277036 |
+
{
|
| 277037 |
+
"epoch": 1126.1,
|
| 277038 |
+
"learning_rate": 7.761066235864298e-06,
|
| 277039 |
+
"loss": 0.2752,
|
| 277040 |
+
"step": 139645
|
| 277041 |
+
},
|
| 277042 |
+
{
|
| 277043 |
+
"epoch": 1126.14,
|
| 277044 |
+
"learning_rate": 7.760985460420034e-06,
|
| 277045 |
+
"loss": 0.3519,
|
| 277046 |
+
"step": 139650
|
| 277047 |
+
},
|
| 277048 |
+
{
|
| 277049 |
+
"epoch": 1126.18,
|
| 277050 |
+
"learning_rate": 7.760904684975768e-06,
|
| 277051 |
+
"loss": 0.6544,
|
| 277052 |
+
"step": 139655
|
| 277053 |
+
},
|
| 277054 |
+
{
|
| 277055 |
+
"epoch": 1126.22,
|
| 277056 |
+
"learning_rate": 7.760823909531504e-06,
|
| 277057 |
+
"loss": 0.7993,
|
| 277058 |
+
"step": 139660
|
| 277059 |
+
},
|
| 277060 |
+
{
|
| 277061 |
+
"epoch": 1126.27,
|
| 277062 |
+
"learning_rate": 7.76074313408724e-06,
|
| 277063 |
+
"loss": 0.3324,
|
| 277064 |
+
"step": 139665
|
| 277065 |
+
},
|
| 277066 |
+
{
|
| 277067 |
+
"epoch": 1126.31,
|
| 277068 |
+
"learning_rate": 7.760662358642974e-06,
|
| 277069 |
+
"loss": 0.2886,
|
| 277070 |
+
"step": 139670
|
| 277071 |
+
},
|
| 277072 |
+
{
|
| 277073 |
+
"epoch": 1126.35,
|
| 277074 |
+
"learning_rate": 7.76058158319871e-06,
|
| 277075 |
+
"loss": 0.4051,
|
| 277076 |
+
"step": 139675
|
| 277077 |
+
},
|
| 277078 |
+
{
|
| 277079 |
+
"epoch": 1126.39,
|
| 277080 |
+
"learning_rate": 7.760500807754443e-06,
|
| 277081 |
+
"loss": 0.7113,
|
| 277082 |
+
"step": 139680
|
| 277083 |
+
},
|
| 277084 |
+
{
|
| 277085 |
+
"epoch": 1126.43,
|
| 277086 |
+
"learning_rate": 7.76042003231018e-06,
|
| 277087 |
+
"loss": 0.9837,
|
| 277088 |
+
"step": 139685
|
| 277089 |
+
},
|
| 277090 |
+
{
|
| 277091 |
+
"epoch": 1126.47,
|
| 277092 |
+
"learning_rate": 7.760339256865913e-06,
|
| 277093 |
+
"loss": 0.2824,
|
| 277094 |
+
"step": 139690
|
| 277095 |
+
},
|
| 277096 |
+
{
|
| 277097 |
+
"epoch": 1126.51,
|
| 277098 |
+
"learning_rate": 7.76025848142165e-06,
|
| 277099 |
+
"loss": 0.3018,
|
| 277100 |
+
"step": 139695
|
| 277101 |
+
},
|
| 277102 |
+
{
|
| 277103 |
+
"epoch": 1126.55,
|
| 277104 |
+
"learning_rate": 7.760177705977383e-06,
|
| 277105 |
+
"loss": 0.3242,
|
| 277106 |
+
"step": 139700
|
| 277107 |
+
},
|
| 277108 |
+
{
|
| 277109 |
+
"epoch": 1126.59,
|
| 277110 |
+
"learning_rate": 7.76009693053312e-06,
|
| 277111 |
+
"loss": 0.7377,
|
| 277112 |
+
"step": 139705
|
| 277113 |
+
},
|
| 277114 |
+
{
|
| 277115 |
+
"epoch": 1126.63,
|
| 277116 |
+
"learning_rate": 7.760016155088853e-06,
|
| 277117 |
+
"loss": 0.8722,
|
| 277118 |
+
"step": 139710
|
| 277119 |
+
},
|
| 277120 |
+
{
|
| 277121 |
+
"epoch": 1126.67,
|
| 277122 |
+
"learning_rate": 7.75993537964459e-06,
|
| 277123 |
+
"loss": 0.2306,
|
| 277124 |
+
"step": 139715
|
| 277125 |
+
},
|
| 277126 |
+
{
|
| 277127 |
+
"epoch": 1126.71,
|
| 277128 |
+
"learning_rate": 7.759854604200323e-06,
|
| 277129 |
+
"loss": 0.2928,
|
| 277130 |
+
"step": 139720
|
| 277131 |
+
},
|
| 277132 |
+
{
|
| 277133 |
+
"epoch": 1126.75,
|
| 277134 |
+
"learning_rate": 7.75977382875606e-06,
|
| 277135 |
+
"loss": 0.3686,
|
| 277136 |
+
"step": 139725
|
| 277137 |
+
},
|
| 277138 |
+
{
|
| 277139 |
+
"epoch": 1126.79,
|
| 277140 |
+
"learning_rate": 7.759693053311795e-06,
|
| 277141 |
+
"loss": 0.6808,
|
| 277142 |
+
"step": 139730
|
| 277143 |
+
},
|
| 277144 |
+
{
|
| 277145 |
+
"epoch": 1126.83,
|
| 277146 |
+
"learning_rate": 7.75961227786753e-06,
|
| 277147 |
+
"loss": 0.9933,
|
| 277148 |
+
"step": 139735
|
| 277149 |
+
},
|
| 277150 |
+
{
|
| 277151 |
+
"epoch": 1126.87,
|
| 277152 |
+
"learning_rate": 7.759531502423265e-06,
|
| 277153 |
+
"loss": 0.3629,
|
| 277154 |
+
"step": 139740
|
| 277155 |
+
},
|
| 277156 |
+
{
|
| 277157 |
+
"epoch": 1126.91,
|
| 277158 |
+
"learning_rate": 7.759450726979e-06,
|
| 277159 |
+
"loss": 0.2996,
|
| 277160 |
+
"step": 139745
|
| 277161 |
+
},
|
| 277162 |
+
{
|
| 277163 |
+
"epoch": 1126.95,
|
| 277164 |
+
"learning_rate": 7.759369951534735e-06,
|
| 277165 |
+
"loss": 0.3513,
|
| 277166 |
+
"step": 139750
|
| 277167 |
+
},
|
| 277168 |
+
{
|
| 277169 |
+
"epoch": 1126.99,
|
| 277170 |
+
"learning_rate": 7.759289176090469e-06,
|
| 277171 |
+
"loss": 0.7597,
|
| 277172 |
+
"step": 139755
|
| 277173 |
+
},
|
| 277174 |
+
{
|
| 277175 |
+
"epoch": 1127.0,
|
| 277176 |
+
"eval_loss": 0.3172752857208252,
|
| 277177 |
+
"eval_runtime": 42.8383,
|
| 277178 |
+
"eval_samples_per_second": 19.492,
|
| 277179 |
+
"eval_steps_per_second": 0.63,
|
| 277180 |
+
"eval_wer": 0.17072453107760205,
|
| 277181 |
+
"step": 139756
|
| 277182 |
+
},
|
| 277183 |
+
{
|
| 277184 |
+
"epoch": 1127.03,
|
| 277185 |
+
"learning_rate": 7.759208400646205e-06,
|
| 277186 |
+
"loss": 0.3127,
|
| 277187 |
+
"step": 139760
|
| 277188 |
+
},
|
| 277189 |
+
{
|
| 277190 |
+
"epoch": 1127.07,
|
| 277191 |
+
"learning_rate": 7.759127625201939e-06,
|
| 277192 |
+
"loss": 0.2753,
|
| 277193 |
+
"step": 139765
|
| 277194 |
+
},
|
| 277195 |
+
{
|
| 277196 |
+
"epoch": 1127.11,
|
| 277197 |
+
"learning_rate": 7.759046849757675e-06,
|
| 277198 |
+
"loss": 0.2649,
|
| 277199 |
+
"step": 139770
|
| 277200 |
+
},
|
| 277201 |
+
{
|
| 277202 |
+
"epoch": 1127.15,
|
| 277203 |
+
"learning_rate": 7.758966074313409e-06,
|
| 277204 |
+
"loss": 0.4116,
|
| 277205 |
+
"step": 139775
|
| 277206 |
+
},
|
| 277207 |
+
{
|
| 277208 |
+
"epoch": 1127.19,
|
| 277209 |
+
"learning_rate": 7.758885298869145e-06,
|
| 277210 |
+
"loss": 0.9346,
|
| 277211 |
+
"step": 139780
|
| 277212 |
+
},
|
| 277213 |
+
{
|
| 277214 |
+
"epoch": 1127.23,
|
| 277215 |
+
"learning_rate": 7.758804523424879e-06,
|
| 277216 |
+
"loss": 0.5724,
|
| 277217 |
+
"step": 139785
|
| 277218 |
+
},
|
| 277219 |
+
{
|
| 277220 |
+
"epoch": 1127.27,
|
| 277221 |
+
"learning_rate": 7.758723747980615e-06,
|
| 277222 |
+
"loss": 0.2966,
|
| 277223 |
+
"step": 139790
|
| 277224 |
+
},
|
| 277225 |
+
{
|
| 277226 |
+
"epoch": 1127.31,
|
| 277227 |
+
"learning_rate": 7.75864297253635e-06,
|
| 277228 |
+
"loss": 0.2577,
|
| 277229 |
+
"step": 139795
|
| 277230 |
+
},
|
| 277231 |
+
{
|
| 277232 |
+
"epoch": 1127.35,
|
| 277233 |
+
"learning_rate": 7.758562197092085e-06,
|
| 277234 |
+
"loss": 0.4031,
|
| 277235 |
+
"step": 139800
|
| 277236 |
+
},
|
| 277237 |
+
{
|
| 277238 |
+
"epoch": 1127.39,
|
| 277239 |
+
"learning_rate": 7.75848142164782e-06,
|
| 277240 |
+
"loss": 0.8522,
|
| 277241 |
+
"step": 139805
|
| 277242 |
+
},
|
| 277243 |
+
{
|
| 277244 |
+
"epoch": 1127.43,
|
| 277245 |
+
"learning_rate": 7.758400646203555e-06,
|
| 277246 |
+
"loss": 0.6157,
|
| 277247 |
+
"step": 139810
|
| 277248 |
+
},
|
| 277249 |
+
{
|
| 277250 |
+
"epoch": 1127.47,
|
| 277251 |
+
"learning_rate": 7.75831987075929e-06,
|
| 277252 |
+
"loss": 0.2694,
|
| 277253 |
+
"step": 139815
|
| 277254 |
+
},
|
| 277255 |
+
{
|
| 277256 |
+
"epoch": 1127.51,
|
| 277257 |
+
"learning_rate": 7.758239095315025e-06,
|
| 277258 |
+
"loss": 0.28,
|
| 277259 |
+
"step": 139820
|
| 277260 |
+
},
|
| 277261 |
+
{
|
| 277262 |
+
"epoch": 1127.55,
|
| 277263 |
+
"learning_rate": 7.75815831987076e-06,
|
| 277264 |
+
"loss": 0.4234,
|
| 277265 |
+
"step": 139825
|
| 277266 |
+
},
|
| 277267 |
+
{
|
| 277268 |
+
"epoch": 1127.59,
|
| 277269 |
+
"learning_rate": 7.758077544426495e-06,
|
| 277270 |
+
"loss": 0.8767,
|
| 277271 |
+
"step": 139830
|
| 277272 |
+
},
|
| 277273 |
+
{
|
| 277274 |
+
"epoch": 1127.63,
|
| 277275 |
+
"learning_rate": 7.75799676898223e-06,
|
| 277276 |
+
"loss": 0.6523,
|
| 277277 |
+
"step": 139835
|
| 277278 |
+
},
|
| 277279 |
+
{
|
| 277280 |
+
"epoch": 1127.67,
|
| 277281 |
+
"learning_rate": 7.757915993537965e-06,
|
| 277282 |
+
"loss": 0.2484,
|
| 277283 |
+
"step": 139840
|
| 277284 |
+
},
|
| 277285 |
+
{
|
| 277286 |
+
"epoch": 1127.71,
|
| 277287 |
+
"learning_rate": 7.7578352180937e-06,
|
| 277288 |
+
"loss": 0.2454,
|
| 277289 |
+
"step": 139845
|
| 277290 |
+
},
|
| 277291 |
+
{
|
| 277292 |
+
"epoch": 1127.76,
|
| 277293 |
+
"learning_rate": 7.757754442649435e-06,
|
| 277294 |
+
"loss": 0.4232,
|
| 277295 |
+
"step": 139850
|
| 277296 |
+
},
|
| 277297 |
+
{
|
| 277298 |
+
"epoch": 1127.8,
|
| 277299 |
+
"learning_rate": 7.75767366720517e-06,
|
| 277300 |
+
"loss": 0.8989,
|
| 277301 |
+
"step": 139855
|
| 277302 |
+
},
|
| 277303 |
+
{
|
| 277304 |
+
"epoch": 1127.84,
|
| 277305 |
+
"learning_rate": 7.757592891760905e-06,
|
| 277306 |
+
"loss": 0.6786,
|
| 277307 |
+
"step": 139860
|
| 277308 |
+
},
|
| 277309 |
+
{
|
| 277310 |
+
"epoch": 1127.88,
|
| 277311 |
+
"learning_rate": 7.75751211631664e-06,
|
| 277312 |
+
"loss": 0.3215,
|
| 277313 |
+
"step": 139865
|
| 277314 |
+
},
|
| 277315 |
+
{
|
| 277316 |
+
"epoch": 1127.92,
|
| 277317 |
+
"learning_rate": 7.757431340872376e-06,
|
| 277318 |
+
"loss": 0.2762,
|
| 277319 |
+
"step": 139870
|
| 277320 |
+
},
|
| 277321 |
+
{
|
| 277322 |
+
"epoch": 1127.96,
|
| 277323 |
+
"learning_rate": 7.75735056542811e-06,
|
| 277324 |
+
"loss": 0.4383,
|
| 277325 |
+
"step": 139875
|
| 277326 |
+
},
|
| 277327 |
+
{
|
| 277328 |
+
"epoch": 1128.0,
|
| 277329 |
+
"learning_rate": 7.757269789983846e-06,
|
| 277330 |
+
"loss": 1.0332,
|
| 277331 |
+
"step": 139880
|
| 277332 |
+
},
|
| 277333 |
+
{
|
| 277334 |
+
"epoch": 1128.0,
|
| 277335 |
+
"eval_loss": 0.3175179064273834,
|
| 277336 |
+
"eval_runtime": 42.4105,
|
| 277337 |
+
"eval_samples_per_second": 19.689,
|
| 277338 |
+
"eval_steps_per_second": 0.637,
|
| 277339 |
+
"eval_wer": 0.17878896452884271,
|
| 277340 |
+
"step": 139880
|
| 277341 |
+
},
|
| 277342 |
+
{
|
| 277343 |
+
"epoch": 1128.04,
|
| 277344 |
+
"learning_rate": 7.75718901453958e-06,
|
| 277345 |
+
"loss": 0.3254,
|
| 277346 |
+
"step": 139885
|
| 277347 |
+
},
|
| 277348 |
+
{
|
| 277349 |
+
"epoch": 1128.08,
|
| 277350 |
+
"learning_rate": 7.757108239095316e-06,
|
| 277351 |
+
"loss": 0.2949,
|
| 277352 |
+
"step": 139890
|
| 277353 |
+
},
|
| 277354 |
+
{
|
| 277355 |
+
"epoch": 1128.12,
|
| 277356 |
+
"learning_rate": 7.75702746365105e-06,
|
| 277357 |
+
"loss": 0.3332,
|
| 277358 |
+
"step": 139895
|
| 277359 |
+
},
|
| 277360 |
+
{
|
| 277361 |
+
"epoch": 1128.16,
|
| 277362 |
+
"learning_rate": 7.756946688206786e-06,
|
| 277363 |
+
"loss": 0.4165,
|
| 277364 |
+
"step": 139900
|
| 277365 |
+
},
|
| 277366 |
+
{
|
| 277367 |
+
"epoch": 1128.2,
|
| 277368 |
+
"learning_rate": 7.75686591276252e-06,
|
| 277369 |
+
"loss": 1.1338,
|
| 277370 |
+
"step": 139905
|
| 277371 |
+
},
|
| 277372 |
+
{
|
| 277373 |
+
"epoch": 1128.24,
|
| 277374 |
+
"learning_rate": 7.756785137318256e-06,
|
| 277375 |
+
"loss": 0.3002,
|
| 277376 |
+
"step": 139910
|
| 277377 |
+
},
|
| 277378 |
+
{
|
| 277379 |
+
"epoch": 1128.28,
|
| 277380 |
+
"learning_rate": 7.75670436187399e-06,
|
| 277381 |
+
"loss": 0.2491,
|
| 277382 |
+
"step": 139915
|
| 277383 |
+
},
|
| 277384 |
+
{
|
| 277385 |
+
"epoch": 1128.32,
|
| 277386 |
+
"learning_rate": 7.756623586429726e-06,
|
| 277387 |
+
"loss": 0.3151,
|
| 277388 |
+
"step": 139920
|
| 277389 |
+
},
|
| 277390 |
+
{
|
| 277391 |
+
"epoch": 1128.36,
|
| 277392 |
+
"learning_rate": 7.75654281098546e-06,
|
| 277393 |
+
"loss": 0.4347,
|
| 277394 |
+
"step": 139925
|
| 277395 |
+
},
|
| 277396 |
+
{
|
| 277397 |
+
"epoch": 1128.4,
|
| 277398 |
+
"learning_rate": 7.756462035541196e-06,
|
| 277399 |
+
"loss": 1.1322,
|
| 277400 |
+
"step": 139930
|
| 277401 |
+
},
|
| 277402 |
+
{
|
| 277403 |
+
"epoch": 1128.44,
|
| 277404 |
+
"learning_rate": 7.756381260096932e-06,
|
| 277405 |
+
"loss": 0.4139,
|
| 277406 |
+
"step": 139935
|
| 277407 |
+
},
|
| 277408 |
+
{
|
| 277409 |
+
"epoch": 1128.48,
|
| 277410 |
+
"learning_rate": 7.756300484652666e-06,
|
| 277411 |
+
"loss": 0.2697,
|
| 277412 |
+
"step": 139940
|
| 277413 |
+
},
|
| 277414 |
+
{
|
| 277415 |
+
"epoch": 1128.52,
|
| 277416 |
+
"learning_rate": 7.756219709208402e-06,
|
| 277417 |
+
"loss": 0.3003,
|
| 277418 |
+
"step": 139945
|
| 277419 |
+
},
|
| 277420 |
+
{
|
| 277421 |
+
"epoch": 1128.56,
|
| 277422 |
+
"learning_rate": 7.756138933764136e-06,
|
| 277423 |
+
"loss": 0.4227,
|
| 277424 |
+
"step": 139950
|
| 277425 |
+
},
|
| 277426 |
+
{
|
| 277427 |
+
"epoch": 1128.6,
|
| 277428 |
+
"learning_rate": 7.756058158319872e-06,
|
| 277429 |
+
"loss": 1.1699,
|
| 277430 |
+
"step": 139955
|
| 277431 |
+
},
|
| 277432 |
+
{
|
| 277433 |
+
"epoch": 1128.64,
|
| 277434 |
+
"learning_rate": 7.755977382875606e-06,
|
| 277435 |
+
"loss": 0.2963,
|
| 277436 |
+
"step": 139960
|
| 277437 |
+
},
|
| 277438 |
+
{
|
| 277439 |
+
"epoch": 1128.68,
|
| 277440 |
+
"learning_rate": 7.755896607431342e-06,
|
| 277441 |
+
"loss": 0.4297,
|
| 277442 |
+
"step": 139965
|
| 277443 |
+
},
|
| 277444 |
+
{
|
| 277445 |
+
"epoch": 1128.72,
|
| 277446 |
+
"learning_rate": 7.755815831987076e-06,
|
| 277447 |
+
"loss": 0.2703,
|
| 277448 |
+
"step": 139970
|
| 277449 |
+
},
|
| 277450 |
+
{
|
| 277451 |
+
"epoch": 1128.76,
|
| 277452 |
+
"learning_rate": 7.755735056542812e-06,
|
| 277453 |
+
"loss": 0.4239,
|
| 277454 |
+
"step": 139975
|
| 277455 |
+
},
|
| 277456 |
+
{
|
| 277457 |
+
"epoch": 1128.8,
|
| 277458 |
+
"learning_rate": 7.755654281098546e-06,
|
| 277459 |
+
"loss": 1.1696,
|
| 277460 |
+
"step": 139980
|
| 277461 |
+
},
|
| 277462 |
+
{
|
| 277463 |
+
"epoch": 1128.84,
|
| 277464 |
+
"learning_rate": 7.755573505654282e-06,
|
| 277465 |
+
"loss": 0.2746,
|
| 277466 |
+
"step": 139985
|
| 277467 |
+
},
|
| 277468 |
+
{
|
| 277469 |
+
"epoch": 1128.88,
|
| 277470 |
+
"learning_rate": 7.755492730210016e-06,
|
| 277471 |
+
"loss": 0.3473,
|
| 277472 |
+
"step": 139990
|
| 277473 |
+
},
|
| 277474 |
+
{
|
| 277475 |
+
"epoch": 1128.92,
|
| 277476 |
+
"learning_rate": 7.755411954765752e-06,
|
| 277477 |
+
"loss": 0.2985,
|
| 277478 |
+
"step": 139995
|
| 277479 |
+
},
|
| 277480 |
+
{
|
| 277481 |
+
"epoch": 1128.96,
|
| 277482 |
+
"learning_rate": 7.755331179321488e-06,
|
| 277483 |
+
"loss": 0.3748,
|
| 277484 |
+
"step": 140000
|
| 277485 |
+
},
|
| 277486 |
+
{
|
| 277487 |
+
"epoch": 1129.0,
|
| 277488 |
+
"eval_loss": 0.313136488199234,
|
| 277489 |
+
"eval_runtime": 41.8574,
|
| 277490 |
+
"eval_samples_per_second": 19.973,
|
| 277491 |
+
"eval_steps_per_second": 0.645,
|
| 277492 |
+
"eval_wer": 0.17273896270895842,
|
| 277493 |
+
"step": 140004
|
| 277494 |
+
},
|
| 277495 |
+
{
|
| 277496 |
+
"epoch": 1129.01,
|
| 277497 |
+
"learning_rate": 7.755250403877222e-06,
|
| 277498 |
+
"loss": 0.4476,
|
| 277499 |
+
"step": 140005
|
| 277500 |
+
},
|
| 277501 |
+
{
|
| 277502 |
+
"epoch": 1129.05,
|
| 277503 |
+
"learning_rate": 7.755169628432958e-06,
|
| 277504 |
+
"loss": 0.306,
|
| 277505 |
+
"step": 140010
|
| 277506 |
+
},
|
| 277507 |
+
{
|
| 277508 |
+
"epoch": 1129.09,
|
| 277509 |
+
"learning_rate": 7.755088852988692e-06,
|
| 277510 |
+
"loss": 0.3078,
|
| 277511 |
+
"step": 140015
|
| 277512 |
+
},
|
| 277513 |
+
{
|
| 277514 |
+
"epoch": 1129.13,
|
| 277515 |
+
"learning_rate": 7.755008077544428e-06,
|
| 277516 |
+
"loss": 0.3265,
|
| 277517 |
+
"step": 140020
|
| 277518 |
+
},
|
| 277519 |
+
{
|
| 277520 |
+
"epoch": 1129.17,
|
| 277521 |
+
"learning_rate": 7.754927302100162e-06,
|
| 277522 |
+
"loss": 0.5433,
|
| 277523 |
+
"step": 140025
|
| 277524 |
+
},
|
| 277525 |
+
{
|
| 277526 |
+
"epoch": 1129.21,
|
| 277527 |
+
"learning_rate": 7.754846526655898e-06,
|
| 277528 |
+
"loss": 1.1187,
|
| 277529 |
+
"step": 140030
|
| 277530 |
+
},
|
| 277531 |
+
{
|
| 277532 |
+
"epoch": 1129.25,
|
| 277533 |
+
"learning_rate": 7.754765751211632e-06,
|
| 277534 |
+
"loss": 0.2686,
|
| 277535 |
+
"step": 140035
|
| 277536 |
+
},
|
| 277537 |
+
{
|
| 277538 |
+
"epoch": 1129.29,
|
| 277539 |
+
"learning_rate": 7.754684975767368e-06,
|
| 277540 |
+
"loss": 0.3152,
|
| 277541 |
+
"step": 140040
|
| 277542 |
+
},
|
| 277543 |
+
{
|
| 277544 |
+
"epoch": 1129.33,
|
| 277545 |
+
"learning_rate": 7.754604200323102e-06,
|
| 277546 |
+
"loss": 0.3167,
|
| 277547 |
+
"step": 140045
|
| 277548 |
+
},
|
| 277549 |
+
{
|
| 277550 |
+
"epoch": 1129.37,
|
| 277551 |
+
"learning_rate": 7.754523424878838e-06,
|
| 277552 |
+
"loss": 0.4683,
|
| 277553 |
+
"step": 140050
|
| 277554 |
+
},
|
| 277555 |
+
{
|
| 277556 |
+
"epoch": 1129.41,
|
| 277557 |
+
"learning_rate": 7.754442649434572e-06,
|
| 277558 |
+
"loss": 1.0714,
|
| 277559 |
+
"step": 140055
|
| 277560 |
+
},
|
| 277561 |
+
{
|
| 277562 |
+
"epoch": 1129.45,
|
| 277563 |
+
"learning_rate": 7.754361873990308e-06,
|
| 277564 |
+
"loss": 0.2651,
|
| 277565 |
+
"step": 140060
|
| 277566 |
+
},
|
| 277567 |
+
{
|
| 277568 |
+
"epoch": 1129.49,
|
| 277569 |
+
"learning_rate": 7.754281098546042e-06,
|
| 277570 |
+
"loss": 0.2786,
|
| 277571 |
+
"step": 140065
|
| 277572 |
+
},
|
| 277573 |
+
{
|
| 277574 |
+
"epoch": 1129.53,
|
| 277575 |
+
"learning_rate": 7.754200323101778e-06,
|
| 277576 |
+
"loss": 0.2964,
|
| 277577 |
+
"step": 140070
|
| 277578 |
+
},
|
| 277579 |
+
{
|
| 277580 |
+
"epoch": 1129.57,
|
| 277581 |
+
"learning_rate": 7.754119547657513e-06,
|
| 277582 |
+
"loss": 0.5536,
|
| 277583 |
+
"step": 140075
|
| 277584 |
+
},
|
| 277585 |
+
{
|
| 277586 |
+
"epoch": 1129.61,
|
| 277587 |
+
"learning_rate": 7.754038772213248e-06,
|
| 277588 |
+
"loss": 1.1089,
|
| 277589 |
+
"step": 140080
|
| 277590 |
+
},
|
| 277591 |
+
{
|
| 277592 |
+
"epoch": 1129.65,
|
| 277593 |
+
"learning_rate": 7.753957996768983e-06,
|
| 277594 |
+
"loss": 0.3057,
|
| 277595 |
+
"step": 140085
|
| 277596 |
+
},
|
| 277597 |
+
{
|
| 277598 |
+
"epoch": 1129.69,
|
| 277599 |
+
"learning_rate": 7.753877221324718e-06,
|
| 277600 |
+
"loss": 0.2479,
|
| 277601 |
+
"step": 140090
|
| 277602 |
+
},
|
| 277603 |
+
{
|
| 277604 |
+
"epoch": 1129.73,
|
| 277605 |
+
"learning_rate": 7.753796445880453e-06,
|
| 277606 |
+
"loss": 0.3417,
|
| 277607 |
+
"step": 140095
|
| 277608 |
+
},
|
| 277609 |
+
{
|
| 277610 |
+
"epoch": 1129.77,
|
| 277611 |
+
"learning_rate": 7.753715670436188e-06,
|
| 277612 |
+
"loss": 0.5828,
|
| 277613 |
+
"step": 140100
|
| 277614 |
+
},
|
| 277615 |
+
{
|
| 277616 |
+
"epoch": 1129.81,
|
| 277617 |
+
"learning_rate": 7.753634894991923e-06,
|
| 277618 |
+
"loss": 1.0685,
|
| 277619 |
+
"step": 140105
|
| 277620 |
+
},
|
| 277621 |
+
{
|
| 277622 |
+
"epoch": 1129.85,
|
| 277623 |
+
"learning_rate": 7.753554119547657e-06,
|
| 277624 |
+
"loss": 0.2756,
|
| 277625 |
+
"step": 140110
|
| 277626 |
+
},
|
| 277627 |
+
{
|
| 277628 |
+
"epoch": 1129.89,
|
| 277629 |
+
"learning_rate": 7.753473344103393e-06,
|
| 277630 |
+
"loss": 0.2826,
|
| 277631 |
+
"step": 140115
|
| 277632 |
+
},
|
| 277633 |
+
{
|
| 277634 |
+
"epoch": 1129.93,
|
| 277635 |
+
"learning_rate": 7.753392568659127e-06,
|
| 277636 |
+
"loss": 0.3145,
|
| 277637 |
+
"step": 140120
|
| 277638 |
+
},
|
| 277639 |
+
{
|
| 277640 |
+
"epoch": 1129.97,
|
| 277641 |
+
"learning_rate": 7.753311793214863e-06,
|
| 277642 |
+
"loss": 0.5671,
|
| 277643 |
+
"step": 140125
|
| 277644 |
+
},
|
| 277645 |
+
{
|
| 277646 |
+
"epoch": 1130.0,
|
| 277647 |
+
"eval_loss": 0.355304479598999,
|
| 277648 |
+
"eval_runtime": 41.1345,
|
| 277649 |
+
"eval_samples_per_second": 20.324,
|
| 277650 |
+
"eval_steps_per_second": 0.656,
|
| 277651 |
+
"eval_wer": 0.17302096545549503,
|
| 277652 |
+
"step": 140128
|
| 277653 |
+
},
|
| 277654 |
+
{
|
| 277655 |
+
"epoch": 1121.02,
|
| 277656 |
+
"learning_rate": 7.753231017770597e-06,
|
| 277657 |
+
"loss": 0.3267,
|
| 277658 |
+
"step": 140130
|
| 277659 |
+
},
|
| 277660 |
+
{
|
| 277661 |
+
"epoch": 1121.06,
|
| 277662 |
+
"learning_rate": 7.753150242326333e-06,
|
| 277663 |
+
"loss": 0.2669,
|
| 277664 |
+
"step": 140135
|
| 277665 |
+
},
|
| 277666 |
+
{
|
| 277667 |
+
"epoch": 1121.1,
|
| 277668 |
+
"learning_rate": 7.753069466882069e-06,
|
| 277669 |
+
"loss": 0.3241,
|
| 277670 |
+
"step": 140140
|
| 277671 |
+
},
|
| 277672 |
+
{
|
| 277673 |
+
"epoch": 1121.14,
|
| 277674 |
+
"learning_rate": 7.752988691437803e-06,
|
| 277675 |
+
"loss": 0.3853,
|
| 277676 |
+
"step": 140145
|
| 277677 |
+
},
|
| 277678 |
+
{
|
| 277679 |
+
"epoch": 1121.18,
|
| 277680 |
+
"learning_rate": 7.752907915993539e-06,
|
| 277681 |
+
"loss": 0.6596,
|
| 277682 |
+
"step": 140150
|
| 277683 |
+
},
|
| 277684 |
+
{
|
| 277685 |
+
"epoch": 1121.22,
|
| 277686 |
+
"learning_rate": 7.752827140549273e-06,
|
| 277687 |
+
"loss": 1.0272,
|
| 277688 |
+
"step": 140155
|
| 277689 |
+
},
|
| 277690 |
+
{
|
| 277691 |
+
"epoch": 1121.26,
|
| 277692 |
+
"learning_rate": 7.752746365105009e-06,
|
| 277693 |
+
"loss": 0.3261,
|
| 277694 |
+
"step": 140160
|
| 277695 |
+
},
|
| 277696 |
+
{
|
| 277697 |
+
"epoch": 1121.3,
|
| 277698 |
+
"learning_rate": 7.752665589660743e-06,
|
| 277699 |
+
"loss": 0.2747,
|
| 277700 |
+
"step": 140165
|
| 277701 |
+
},
|
| 277702 |
+
{
|
| 277703 |
+
"epoch": 1121.34,
|
| 277704 |
+
"learning_rate": 7.752584814216479e-06,
|
| 277705 |
+
"loss": 0.3155,
|
| 277706 |
+
"step": 140170
|
| 277707 |
+
},
|
| 277708 |
+
{
|
| 277709 |
+
"epoch": 1121.38,
|
| 277710 |
+
"learning_rate": 7.752504038772213e-06,
|
| 277711 |
+
"loss": 0.6354,
|
| 277712 |
+
"step": 140175
|
| 277713 |
+
},
|
| 277714 |
+
{
|
| 277715 |
+
"epoch": 1121.42,
|
| 277716 |
+
"learning_rate": 7.752423263327949e-06,
|
| 277717 |
+
"loss": 1.0513,
|
| 277718 |
+
"step": 140180
|
| 277719 |
+
},
|
| 277720 |
+
{
|
| 277721 |
+
"epoch": 1121.46,
|
| 277722 |
+
"learning_rate": 7.752342487883683e-06,
|
| 277723 |
+
"loss": 0.2732,
|
| 277724 |
+
"step": 140185
|
| 277725 |
+
},
|
| 277726 |
+
{
|
| 277727 |
+
"epoch": 1121.5,
|
| 277728 |
+
"learning_rate": 7.752261712439419e-06,
|
| 277729 |
+
"loss": 0.2639,
|
| 277730 |
+
"step": 140190
|
| 277731 |
+
},
|
| 277732 |
+
{
|
| 277733 |
+
"epoch": 1121.54,
|
| 277734 |
+
"learning_rate": 7.752180936995153e-06,
|
| 277735 |
+
"loss": 0.2941,
|
| 277736 |
+
"step": 140195
|
| 277737 |
+
},
|
| 277738 |
+
{
|
| 277739 |
+
"epoch": 1121.58,
|
| 277740 |
+
"learning_rate": 7.752100161550889e-06,
|
| 277741 |
+
"loss": 0.5625,
|
| 277742 |
+
"step": 140200
|
| 277743 |
+
},
|
| 277744 |
+
{
|
| 277745 |
+
"epoch": 1121.62,
|
| 277746 |
+
"learning_rate": 7.752019386106625e-06,
|
| 277747 |
+
"loss": 0.9816,
|
| 277748 |
+
"step": 140205
|
| 277749 |
+
},
|
| 277750 |
+
{
|
| 277751 |
+
"epoch": 1121.66,
|
| 277752 |
+
"learning_rate": 7.751938610662359e-06,
|
| 277753 |
+
"loss": 0.3077,
|
| 277754 |
+
"step": 140210
|
| 277755 |
+
},
|
| 277756 |
+
{
|
| 277757 |
+
"epoch": 1121.7,
|
| 277758 |
+
"learning_rate": 7.751857835218095e-06,
|
| 277759 |
+
"loss": 0.2621,
|
| 277760 |
+
"step": 140215
|
| 277761 |
+
},
|
| 277762 |
+
{
|
| 277763 |
+
"epoch": 1121.74,
|
| 277764 |
+
"learning_rate": 7.751777059773829e-06,
|
| 277765 |
+
"loss": 0.3137,
|
| 277766 |
+
"step": 140220
|
| 277767 |
+
},
|
| 277768 |
+
{
|
| 277769 |
+
"epoch": 1121.78,
|
| 277770 |
+
"learning_rate": 7.751696284329565e-06,
|
| 277771 |
+
"loss": 0.5958,
|
| 277772 |
+
"step": 140225
|
| 277773 |
+
},
|
| 277774 |
+
{
|
| 277775 |
+
"epoch": 1121.82,
|
| 277776 |
+
"learning_rate": 7.751615508885299e-06,
|
| 277777 |
+
"loss": 0.8971,
|
| 277778 |
+
"step": 140230
|
| 277779 |
+
},
|
| 277780 |
+
{
|
| 277781 |
+
"epoch": 1121.86,
|
| 277782 |
+
"learning_rate": 7.751534733441035e-06,
|
| 277783 |
+
"loss": 0.3331,
|
| 277784 |
+
"step": 140235
|
| 277785 |
+
},
|
| 277786 |
+
{
|
| 277787 |
+
"epoch": 1121.9,
|
| 277788 |
+
"learning_rate": 7.751453957996769e-06,
|
| 277789 |
+
"loss": 0.3037,
|
| 277790 |
+
"step": 140240
|
| 277791 |
+
},
|
| 277792 |
+
{
|
| 277793 |
+
"epoch": 1121.94,
|
| 277794 |
+
"learning_rate": 7.751373182552505e-06,
|
| 277795 |
+
"loss": 0.4154,
|
| 277796 |
+
"step": 140245
|
| 277797 |
+
},
|
| 277798 |
+
{
|
| 277799 |
+
"epoch": 1121.98,
|
| 277800 |
+
"learning_rate": 7.751292407108239e-06,
|
| 277801 |
+
"loss": 0.6362,
|
| 277802 |
+
"step": 140250
|
| 277803 |
+
},
|
| 277804 |
+
{
|
| 277805 |
+
"epoch": 1122.0,
|
| 277806 |
+
"eval_loss": 0.36987918615341187,
|
| 277807 |
+
"eval_runtime": 41.3421,
|
| 277808 |
+
"eval_samples_per_second": 20.222,
|
| 277809 |
+
"eval_steps_per_second": 0.653,
|
| 277810 |
+
"eval_wer": 0.17799043062200956,
|
| 277811 |
+
"step": 140253
|
| 277812 |
}
|
| 277813 |
],
|
| 277814 |
+
"max_steps": 625000,
|
| 277815 |
"num_train_epochs": 5000,
|
| 277816 |
+
"total_flos": 3.946598257885967e+20,
|
| 277817 |
"trial_name": null,
|
| 277818 |
"trial_params": null
|
| 277819 |
}
|
model-bin/finetune/base/{checkpoint-139632 → checkpoint-140253}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1630205681.0762417/events.out.tfevents.1630205681.cc93b136ebf5.1086.9
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d2b8c923a98a9b84cd701aba33019982aec28389dc8dfb3cf07b8c472a00d64c
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630206127.5566013/events.out.tfevents.1630206127.cc93b136ebf5.1086.11
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e024394a609a84328c13e3953acff7dc1602a63107e3b1d568d292af039e6fc6
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630206572.6054518/events.out.tfevents.1630206572.cc93b136ebf5.1086.13
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:feb627a0f18551df7b7f9ea6403a8623c8ecda2afe905eec825e9b9a47730ca9
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630207019.0214832/events.out.tfevents.1630207019.cc93b136ebf5.1086.15
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:7f8c11c732190d48ddfbc996a007c0cb7489f98b58995d8f1019a0964ce76499
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630207465.424582/events.out.tfevents.1630207465.cc93b136ebf5.1086.17
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d43da565dbefd479c9ed2b79bed53093da2b54a3c12a2e505a5ad8d9d46776e2
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1630205681.cc93b136ebf5.1086.8
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2b16cb140b2ec07184c32989a8bd64104839a8fe1a504fbdbd6a687999f07875
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630206127.cc93b136ebf5.1086.10
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:0da8b0e2a84658a160e7a73ed753d3d740aaafb84d952945698590efdd9170e6
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630206572.cc93b136ebf5.1086.12
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8de712b9d64ee44308764d11ee0424ef42abdef6b36316182f11c3e5dd35396f
|
| 3 |
+
size 8462
|
model-bin/finetune/base/log/events.out.tfevents.1630207019.cc93b136ebf5.1086.14
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:eb0ea75c8c4afee650eed769f9b7f2fb65492aca54035399d1efe9939dd97bf9
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630207465.cc93b136ebf5.1086.16
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:eff381fef3359b434a412f346a083e97c83fcea7a8b0ca4dc89f701106460768
|
| 3 |
+
size 8622
|