"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-81513 → checkpoint-82134}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-81513 → checkpoint-82134}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-81513 → checkpoint-82134}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-81513 → checkpoint-82134}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-81513 → checkpoint-82134}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-81513 → checkpoint-82134}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-81513 → checkpoint-82134}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-81513 → checkpoint-82134}/trainer_state.json +792 -3
- model-bin/finetune/base/{checkpoint-81513 → checkpoint-82134}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629908936.4564571/events.out.tfevents.1629908936.7e498afd5545.7645.85 +3 -0
- model-bin/finetune/base/log/1629909414.6742997/events.out.tfevents.1629909414.7e498afd5545.7645.87 +3 -0
- model-bin/finetune/base/log/1629909897.4578655/events.out.tfevents.1629909897.7e498afd5545.7645.89 +3 -0
- model-bin/finetune/base/log/1629910378.9335363/events.out.tfevents.1629910378.7e498afd5545.7645.91 +3 -0
- model-bin/finetune/base/log/1629910855.6983657/events.out.tfevents.1629910855.7e498afd5545.7645.93 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629908936.7e498afd5545.7645.84 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629909414.7e498afd5545.7645.86 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629909897.7e498afd5545.7645.88 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629910378.7e498afd5545.7645.90 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629910855.7e498afd5545.7645.92 +3 -0
model-bin/finetune/base/{checkpoint-81513 → checkpoint-82134}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-81513 → checkpoint-82134}/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 sha256:24cad9f28dc11fe9279978fcf3a0bc408a45425a697994c5b4c4738f375d14e2
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-81513 → checkpoint-82134}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-81513 → checkpoint-82134}/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 sha256:6dbf8329e6b28204af9dc8b010f5c4abbdaea31770e2d6efdbf98e30b385f207
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-81513 → checkpoint-82134}/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 sha256:3185208803e8bbdf570955ae54957beb26a219cefb74153cdf5bac8630a1e1e9
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-81513 → checkpoint-82134}/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 sha256:8a54e5d0967fa297fe8e8cf400b549017a28943c4e450107074b7aa711703e84
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-81513 → checkpoint-82134}/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 sha256:441d6ca4b8122d743fb64fa9296f5c7a8f4ce79ace5f3fd8fc1317da9d4f1aab
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-81513 → checkpoint-82134}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1822989349981638,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-79148",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -203073,11 +203073,800 @@
|
|
| 203073 |
"eval_steps_per_second": 0.594,
|
| 203074 |
"eval_wer": 0.1944121915820029,
|
| 203075 |
"step": 81513
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 203076 |
}
|
| 203077 |
],
|
| 203078 |
"max_steps": 620000,
|
| 203079 |
"num_train_epochs": 5000,
|
| 203080 |
-
"total_flos": 2.
|
| 203081 |
"trial_name": null,
|
| 203082 |
"trial_params": null
|
| 203083 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1822989349981638,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-79148",
|
| 4 |
+
"epoch": 661.995983935743,
|
| 5 |
+
"global_step": 82134,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 203073 |
"eval_steps_per_second": 0.594,
|
| 203074 |
"eval_wer": 0.1944121915820029,
|
| 203075 |
"step": 81513
|
| 203076 |
+
},
|
| 203077 |
+
{
|
| 203078 |
+
"epoch": 657.02,
|
| 203079 |
+
"learning_rate": 8.699773828756059e-06,
|
| 203080 |
+
"loss": 0.3692,
|
| 203081 |
+
"step": 81515
|
| 203082 |
+
},
|
| 203083 |
+
{
|
| 203084 |
+
"epoch": 657.06,
|
| 203085 |
+
"learning_rate": 8.699693053311795e-06,
|
| 203086 |
+
"loss": 0.3053,
|
| 203087 |
+
"step": 81520
|
| 203088 |
+
},
|
| 203089 |
+
{
|
| 203090 |
+
"epoch": 657.1,
|
| 203091 |
+
"learning_rate": 8.699612277867529e-06,
|
| 203092 |
+
"loss": 0.3155,
|
| 203093 |
+
"step": 81525
|
| 203094 |
+
},
|
| 203095 |
+
{
|
| 203096 |
+
"epoch": 657.14,
|
| 203097 |
+
"learning_rate": 8.699531502423265e-06,
|
| 203098 |
+
"loss": 0.4045,
|
| 203099 |
+
"step": 81530
|
| 203100 |
+
},
|
| 203101 |
+
{
|
| 203102 |
+
"epoch": 657.18,
|
| 203103 |
+
"learning_rate": 8.699450726979e-06,
|
| 203104 |
+
"loss": 0.6462,
|
| 203105 |
+
"step": 81535
|
| 203106 |
+
},
|
| 203107 |
+
{
|
| 203108 |
+
"epoch": 657.22,
|
| 203109 |
+
"learning_rate": 8.699369951534734e-06,
|
| 203110 |
+
"loss": 0.9745,
|
| 203111 |
+
"step": 81540
|
| 203112 |
+
},
|
| 203113 |
+
{
|
| 203114 |
+
"epoch": 657.26,
|
| 203115 |
+
"learning_rate": 8.69928917609047e-06,
|
| 203116 |
+
"loss": 0.2742,
|
| 203117 |
+
"step": 81545
|
| 203118 |
+
},
|
| 203119 |
+
{
|
| 203120 |
+
"epoch": 657.3,
|
| 203121 |
+
"learning_rate": 8.699208400646204e-06,
|
| 203122 |
+
"loss": 0.2758,
|
| 203123 |
+
"step": 81550
|
| 203124 |
+
},
|
| 203125 |
+
{
|
| 203126 |
+
"epoch": 657.34,
|
| 203127 |
+
"learning_rate": 8.69912762520194e-06,
|
| 203128 |
+
"loss": 0.3884,
|
| 203129 |
+
"step": 81555
|
| 203130 |
+
},
|
| 203131 |
+
{
|
| 203132 |
+
"epoch": 657.38,
|
| 203133 |
+
"learning_rate": 8.699046849757674e-06,
|
| 203134 |
+
"loss": 0.6769,
|
| 203135 |
+
"step": 81560
|
| 203136 |
+
},
|
| 203137 |
+
{
|
| 203138 |
+
"epoch": 657.42,
|
| 203139 |
+
"learning_rate": 8.69896607431341e-06,
|
| 203140 |
+
"loss": 1.1057,
|
| 203141 |
+
"step": 81565
|
| 203142 |
+
},
|
| 203143 |
+
{
|
| 203144 |
+
"epoch": 657.46,
|
| 203145 |
+
"learning_rate": 8.698885298869144e-06,
|
| 203146 |
+
"loss": 0.3067,
|
| 203147 |
+
"step": 81570
|
| 203148 |
+
},
|
| 203149 |
+
{
|
| 203150 |
+
"epoch": 657.5,
|
| 203151 |
+
"learning_rate": 8.69880452342488e-06,
|
| 203152 |
+
"loss": 0.3199,
|
| 203153 |
+
"step": 81575
|
| 203154 |
+
},
|
| 203155 |
+
{
|
| 203156 |
+
"epoch": 657.54,
|
| 203157 |
+
"learning_rate": 8.698723747980614e-06,
|
| 203158 |
+
"loss": 0.3696,
|
| 203159 |
+
"step": 81580
|
| 203160 |
+
},
|
| 203161 |
+
{
|
| 203162 |
+
"epoch": 657.58,
|
| 203163 |
+
"learning_rate": 8.69864297253635e-06,
|
| 203164 |
+
"loss": 0.6531,
|
| 203165 |
+
"step": 81585
|
| 203166 |
+
},
|
| 203167 |
+
{
|
| 203168 |
+
"epoch": 657.62,
|
| 203169 |
+
"learning_rate": 8.698562197092086e-06,
|
| 203170 |
+
"loss": 1.0473,
|
| 203171 |
+
"step": 81590
|
| 203172 |
+
},
|
| 203173 |
+
{
|
| 203174 |
+
"epoch": 657.66,
|
| 203175 |
+
"learning_rate": 8.69848142164782e-06,
|
| 203176 |
+
"loss": 0.321,
|
| 203177 |
+
"step": 81595
|
| 203178 |
+
},
|
| 203179 |
+
{
|
| 203180 |
+
"epoch": 657.7,
|
| 203181 |
+
"learning_rate": 8.698400646203556e-06,
|
| 203182 |
+
"loss": 0.3643,
|
| 203183 |
+
"step": 81600
|
| 203184 |
+
},
|
| 203185 |
+
{
|
| 203186 |
+
"epoch": 657.74,
|
| 203187 |
+
"learning_rate": 8.69831987075929e-06,
|
| 203188 |
+
"loss": 0.3277,
|
| 203189 |
+
"step": 81605
|
| 203190 |
+
},
|
| 203191 |
+
{
|
| 203192 |
+
"epoch": 657.78,
|
| 203193 |
+
"learning_rate": 8.698239095315026e-06,
|
| 203194 |
+
"loss": 0.6669,
|
| 203195 |
+
"step": 81610
|
| 203196 |
+
},
|
| 203197 |
+
{
|
| 203198 |
+
"epoch": 657.82,
|
| 203199 |
+
"learning_rate": 8.69815831987076e-06,
|
| 203200 |
+
"loss": 0.9607,
|
| 203201 |
+
"step": 81615
|
| 203202 |
+
},
|
| 203203 |
+
{
|
| 203204 |
+
"epoch": 657.86,
|
| 203205 |
+
"learning_rate": 8.698077544426496e-06,
|
| 203206 |
+
"loss": 0.2718,
|
| 203207 |
+
"step": 81620
|
| 203208 |
+
},
|
| 203209 |
+
{
|
| 203210 |
+
"epoch": 657.9,
|
| 203211 |
+
"learning_rate": 8.69799676898223e-06,
|
| 203212 |
+
"loss": 0.2763,
|
| 203213 |
+
"step": 81625
|
| 203214 |
+
},
|
| 203215 |
+
{
|
| 203216 |
+
"epoch": 657.94,
|
| 203217 |
+
"learning_rate": 8.697915993537966e-06,
|
| 203218 |
+
"loss": 0.3533,
|
| 203219 |
+
"step": 81630
|
| 203220 |
+
},
|
| 203221 |
+
{
|
| 203222 |
+
"epoch": 657.98,
|
| 203223 |
+
"learning_rate": 8.6978352180937e-06,
|
| 203224 |
+
"loss": 0.6378,
|
| 203225 |
+
"step": 81635
|
| 203226 |
+
},
|
| 203227 |
+
{
|
| 203228 |
+
"epoch": 658.0,
|
| 203229 |
+
"eval_loss": 0.45627066493034363,
|
| 203230 |
+
"eval_runtime": 41.9319,
|
| 203231 |
+
"eval_samples_per_second": 20.056,
|
| 203232 |
+
"eval_steps_per_second": 0.644,
|
| 203233 |
+
"eval_wer": 0.19326638248681455,
|
| 203234 |
+
"step": 81637
|
| 203235 |
+
},
|
| 203236 |
+
{
|
| 203237 |
+
"epoch": 658.02,
|
| 203238 |
+
"learning_rate": 8.697754442649436e-06,
|
| 203239 |
+
"loss": 0.4026,
|
| 203240 |
+
"step": 81640
|
| 203241 |
+
},
|
| 203242 |
+
{
|
| 203243 |
+
"epoch": 658.06,
|
| 203244 |
+
"learning_rate": 8.69767366720517e-06,
|
| 203245 |
+
"loss": 0.2801,
|
| 203246 |
+
"step": 81645
|
| 203247 |
+
},
|
| 203248 |
+
{
|
| 203249 |
+
"epoch": 658.1,
|
| 203250 |
+
"learning_rate": 8.697592891760906e-06,
|
| 203251 |
+
"loss": 0.2926,
|
| 203252 |
+
"step": 81650
|
| 203253 |
+
},
|
| 203254 |
+
{
|
| 203255 |
+
"epoch": 658.14,
|
| 203256 |
+
"learning_rate": 8.697512116316642e-06,
|
| 203257 |
+
"loss": 0.3782,
|
| 203258 |
+
"step": 81655
|
| 203259 |
+
},
|
| 203260 |
+
{
|
| 203261 |
+
"epoch": 658.18,
|
| 203262 |
+
"learning_rate": 8.697431340872376e-06,
|
| 203263 |
+
"loss": 0.7188,
|
| 203264 |
+
"step": 81660
|
| 203265 |
+
},
|
| 203266 |
+
{
|
| 203267 |
+
"epoch": 658.22,
|
| 203268 |
+
"learning_rate": 8.697350565428112e-06,
|
| 203269 |
+
"loss": 0.9318,
|
| 203270 |
+
"step": 81665
|
| 203271 |
+
},
|
| 203272 |
+
{
|
| 203273 |
+
"epoch": 658.27,
|
| 203274 |
+
"learning_rate": 8.697269789983846e-06,
|
| 203275 |
+
"loss": 0.3419,
|
| 203276 |
+
"step": 81670
|
| 203277 |
+
},
|
| 203278 |
+
{
|
| 203279 |
+
"epoch": 658.31,
|
| 203280 |
+
"learning_rate": 8.697189014539582e-06,
|
| 203281 |
+
"loss": 0.2777,
|
| 203282 |
+
"step": 81675
|
| 203283 |
+
},
|
| 203284 |
+
{
|
| 203285 |
+
"epoch": 658.35,
|
| 203286 |
+
"learning_rate": 8.697108239095316e-06,
|
| 203287 |
+
"loss": 0.3644,
|
| 203288 |
+
"step": 81680
|
| 203289 |
+
},
|
| 203290 |
+
{
|
| 203291 |
+
"epoch": 658.39,
|
| 203292 |
+
"learning_rate": 8.697027463651052e-06,
|
| 203293 |
+
"loss": 0.8372,
|
| 203294 |
+
"step": 81685
|
| 203295 |
+
},
|
| 203296 |
+
{
|
| 203297 |
+
"epoch": 658.43,
|
| 203298 |
+
"learning_rate": 8.696946688206786e-06,
|
| 203299 |
+
"loss": 0.9454,
|
| 203300 |
+
"step": 81690
|
| 203301 |
+
},
|
| 203302 |
+
{
|
| 203303 |
+
"epoch": 658.47,
|
| 203304 |
+
"learning_rate": 8.696865912762522e-06,
|
| 203305 |
+
"loss": 0.4416,
|
| 203306 |
+
"step": 81695
|
| 203307 |
+
},
|
| 203308 |
+
{
|
| 203309 |
+
"epoch": 658.51,
|
| 203310 |
+
"learning_rate": 8.696785137318256e-06,
|
| 203311 |
+
"loss": 0.2996,
|
| 203312 |
+
"step": 81700
|
| 203313 |
+
},
|
| 203314 |
+
{
|
| 203315 |
+
"epoch": 658.55,
|
| 203316 |
+
"learning_rate": 8.696704361873992e-06,
|
| 203317 |
+
"loss": 0.3234,
|
| 203318 |
+
"step": 81705
|
| 203319 |
+
},
|
| 203320 |
+
{
|
| 203321 |
+
"epoch": 658.59,
|
| 203322 |
+
"learning_rate": 8.696623586429726e-06,
|
| 203323 |
+
"loss": 0.7701,
|
| 203324 |
+
"step": 81710
|
| 203325 |
+
},
|
| 203326 |
+
{
|
| 203327 |
+
"epoch": 658.63,
|
| 203328 |
+
"learning_rate": 8.696542810985462e-06,
|
| 203329 |
+
"loss": 0.908,
|
| 203330 |
+
"step": 81715
|
| 203331 |
+
},
|
| 203332 |
+
{
|
| 203333 |
+
"epoch": 658.67,
|
| 203334 |
+
"learning_rate": 8.696462035541197e-06,
|
| 203335 |
+
"loss": 0.3623,
|
| 203336 |
+
"step": 81720
|
| 203337 |
+
},
|
| 203338 |
+
{
|
| 203339 |
+
"epoch": 658.71,
|
| 203340 |
+
"learning_rate": 8.696381260096932e-06,
|
| 203341 |
+
"loss": 0.2868,
|
| 203342 |
+
"step": 81725
|
| 203343 |
+
},
|
| 203344 |
+
{
|
| 203345 |
+
"epoch": 658.75,
|
| 203346 |
+
"learning_rate": 8.696300484652667e-06,
|
| 203347 |
+
"loss": 0.3455,
|
| 203348 |
+
"step": 81730
|
| 203349 |
+
},
|
| 203350 |
+
{
|
| 203351 |
+
"epoch": 658.79,
|
| 203352 |
+
"learning_rate": 8.696219709208402e-06,
|
| 203353 |
+
"loss": 0.8271,
|
| 203354 |
+
"step": 81735
|
| 203355 |
+
},
|
| 203356 |
+
{
|
| 203357 |
+
"epoch": 658.83,
|
| 203358 |
+
"learning_rate": 8.696138933764137e-06,
|
| 203359 |
+
"loss": 0.932,
|
| 203360 |
+
"step": 81740
|
| 203361 |
+
},
|
| 203362 |
+
{
|
| 203363 |
+
"epoch": 658.87,
|
| 203364 |
+
"learning_rate": 8.696058158319872e-06,
|
| 203365 |
+
"loss": 0.3335,
|
| 203366 |
+
"step": 81745
|
| 203367 |
+
},
|
| 203368 |
+
{
|
| 203369 |
+
"epoch": 658.91,
|
| 203370 |
+
"learning_rate": 8.695977382875607e-06,
|
| 203371 |
+
"loss": 0.2877,
|
| 203372 |
+
"step": 81750
|
| 203373 |
+
},
|
| 203374 |
+
{
|
| 203375 |
+
"epoch": 658.95,
|
| 203376 |
+
"learning_rate": 8.695896607431341e-06,
|
| 203377 |
+
"loss": 0.4708,
|
| 203378 |
+
"step": 81755
|
| 203379 |
+
},
|
| 203380 |
+
{
|
| 203381 |
+
"epoch": 658.99,
|
| 203382 |
+
"learning_rate": 8.695815831987077e-06,
|
| 203383 |
+
"loss": 0.7955,
|
| 203384 |
+
"step": 81760
|
| 203385 |
+
},
|
| 203386 |
+
{
|
| 203387 |
+
"epoch": 659.0,
|
| 203388 |
+
"eval_loss": 0.3713783025741577,
|
| 203389 |
+
"eval_runtime": 41.2538,
|
| 203390 |
+
"eval_samples_per_second": 20.386,
|
| 203391 |
+
"eval_steps_per_second": 0.654,
|
| 203392 |
+
"eval_wer": 0.18491199883151976,
|
| 203393 |
+
"step": 81761
|
| 203394 |
+
},
|
| 203395 |
+
{
|
| 203396 |
+
"epoch": 659.03,
|
| 203397 |
+
"learning_rate": 8.695735056542811e-06,
|
| 203398 |
+
"loss": 0.3734,
|
| 203399 |
+
"step": 81765
|
| 203400 |
+
},
|
| 203401 |
+
{
|
| 203402 |
+
"epoch": 659.07,
|
| 203403 |
+
"learning_rate": 8.695654281098547e-06,
|
| 203404 |
+
"loss": 0.3216,
|
| 203405 |
+
"step": 81770
|
| 203406 |
+
},
|
| 203407 |
+
{
|
| 203408 |
+
"epoch": 659.11,
|
| 203409 |
+
"learning_rate": 8.695573505654281e-06,
|
| 203410 |
+
"loss": 0.3628,
|
| 203411 |
+
"step": 81775
|
| 203412 |
+
},
|
| 203413 |
+
{
|
| 203414 |
+
"epoch": 659.15,
|
| 203415 |
+
"learning_rate": 8.695492730210017e-06,
|
| 203416 |
+
"loss": 0.4213,
|
| 203417 |
+
"step": 81780
|
| 203418 |
+
},
|
| 203419 |
+
{
|
| 203420 |
+
"epoch": 659.19,
|
| 203421 |
+
"learning_rate": 8.695411954765753e-06,
|
| 203422 |
+
"loss": 0.9702,
|
| 203423 |
+
"step": 81785
|
| 203424 |
+
},
|
| 203425 |
+
{
|
| 203426 |
+
"epoch": 659.23,
|
| 203427 |
+
"learning_rate": 8.695331179321487e-06,
|
| 203428 |
+
"loss": 1.3116,
|
| 203429 |
+
"step": 81790
|
| 203430 |
+
},
|
| 203431 |
+
{
|
| 203432 |
+
"epoch": 659.27,
|
| 203433 |
+
"learning_rate": 8.695250403877223e-06,
|
| 203434 |
+
"loss": 0.2895,
|
| 203435 |
+
"step": 81795
|
| 203436 |
+
},
|
| 203437 |
+
{
|
| 203438 |
+
"epoch": 659.31,
|
| 203439 |
+
"learning_rate": 8.695169628432957e-06,
|
| 203440 |
+
"loss": 0.2964,
|
| 203441 |
+
"step": 81800
|
| 203442 |
+
},
|
| 203443 |
+
{
|
| 203444 |
+
"epoch": 659.35,
|
| 203445 |
+
"learning_rate": 8.695088852988693e-06,
|
| 203446 |
+
"loss": 0.3786,
|
| 203447 |
+
"step": 81805
|
| 203448 |
+
},
|
| 203449 |
+
{
|
| 203450 |
+
"epoch": 659.39,
|
| 203451 |
+
"learning_rate": 8.695008077544427e-06,
|
| 203452 |
+
"loss": 0.8122,
|
| 203453 |
+
"step": 81810
|
| 203454 |
+
},
|
| 203455 |
+
{
|
| 203456 |
+
"epoch": 659.43,
|
| 203457 |
+
"learning_rate": 8.694927302100163e-06,
|
| 203458 |
+
"loss": 0.6819,
|
| 203459 |
+
"step": 81815
|
| 203460 |
+
},
|
| 203461 |
+
{
|
| 203462 |
+
"epoch": 659.47,
|
| 203463 |
+
"learning_rate": 8.694846526655897e-06,
|
| 203464 |
+
"loss": 0.3317,
|
| 203465 |
+
"step": 81820
|
| 203466 |
+
},
|
| 203467 |
+
{
|
| 203468 |
+
"epoch": 659.51,
|
| 203469 |
+
"learning_rate": 8.694765751211633e-06,
|
| 203470 |
+
"loss": 0.3594,
|
| 203471 |
+
"step": 81825
|
| 203472 |
+
},
|
| 203473 |
+
{
|
| 203474 |
+
"epoch": 659.55,
|
| 203475 |
+
"learning_rate": 8.694684975767367e-06,
|
| 203476 |
+
"loss": 0.3832,
|
| 203477 |
+
"step": 81830
|
| 203478 |
+
},
|
| 203479 |
+
{
|
| 203480 |
+
"epoch": 659.59,
|
| 203481 |
+
"learning_rate": 8.694604200323103e-06,
|
| 203482 |
+
"loss": 1.1161,
|
| 203483 |
+
"step": 81835
|
| 203484 |
+
},
|
| 203485 |
+
{
|
| 203486 |
+
"epoch": 659.63,
|
| 203487 |
+
"learning_rate": 8.694523424878837e-06,
|
| 203488 |
+
"loss": 0.6244,
|
| 203489 |
+
"step": 81840
|
| 203490 |
+
},
|
| 203491 |
+
{
|
| 203492 |
+
"epoch": 659.67,
|
| 203493 |
+
"learning_rate": 8.694442649434573e-06,
|
| 203494 |
+
"loss": 0.2728,
|
| 203495 |
+
"step": 81845
|
| 203496 |
+
},
|
| 203497 |
+
{
|
| 203498 |
+
"epoch": 659.71,
|
| 203499 |
+
"learning_rate": 8.694361873990309e-06,
|
| 203500 |
+
"loss": 0.3422,
|
| 203501 |
+
"step": 81850
|
| 203502 |
+
},
|
| 203503 |
+
{
|
| 203504 |
+
"epoch": 659.76,
|
| 203505 |
+
"learning_rate": 8.694281098546043e-06,
|
| 203506 |
+
"loss": 0.4158,
|
| 203507 |
+
"step": 81855
|
| 203508 |
+
},
|
| 203509 |
+
{
|
| 203510 |
+
"epoch": 659.8,
|
| 203511 |
+
"learning_rate": 8.694200323101779e-06,
|
| 203512 |
+
"loss": 0.9222,
|
| 203513 |
+
"step": 81860
|
| 203514 |
+
},
|
| 203515 |
+
{
|
| 203516 |
+
"epoch": 659.84,
|
| 203517 |
+
"learning_rate": 8.694119547657513e-06,
|
| 203518 |
+
"loss": 0.6799,
|
| 203519 |
+
"step": 81865
|
| 203520 |
+
},
|
| 203521 |
+
{
|
| 203522 |
+
"epoch": 659.88,
|
| 203523 |
+
"learning_rate": 8.694038772213249e-06,
|
| 203524 |
+
"loss": 0.3446,
|
| 203525 |
+
"step": 81870
|
| 203526 |
+
},
|
| 203527 |
+
{
|
| 203528 |
+
"epoch": 659.92,
|
| 203529 |
+
"learning_rate": 8.693957996768983e-06,
|
| 203530 |
+
"loss": 0.319,
|
| 203531 |
+
"step": 81875
|
| 203532 |
+
},
|
| 203533 |
+
{
|
| 203534 |
+
"epoch": 659.96,
|
| 203535 |
+
"learning_rate": 8.693877221324719e-06,
|
| 203536 |
+
"loss": 0.3603,
|
| 203537 |
+
"step": 81880
|
| 203538 |
+
},
|
| 203539 |
+
{
|
| 203540 |
+
"epoch": 660.0,
|
| 203541 |
+
"learning_rate": 8.693796445880453e-06,
|
| 203542 |
+
"loss": 1.205,
|
| 203543 |
+
"step": 81885
|
| 203544 |
+
},
|
| 203545 |
+
{
|
| 203546 |
+
"epoch": 660.0,
|
| 203547 |
+
"eval_loss": 0.4129309058189392,
|
| 203548 |
+
"eval_runtime": 44.2238,
|
| 203549 |
+
"eval_samples_per_second": 19.017,
|
| 203550 |
+
"eval_steps_per_second": 0.611,
|
| 203551 |
+
"eval_wer": 0.1928848546958495,
|
| 203552 |
+
"step": 81885
|
| 203553 |
+
},
|
| 203554 |
+
{
|
| 203555 |
+
"epoch": 655.04,
|
| 203556 |
+
"learning_rate": 8.693715670436189e-06,
|
| 203557 |
+
"loss": 0.3022,
|
| 203558 |
+
"step": 81890
|
| 203559 |
+
},
|
| 203560 |
+
{
|
| 203561 |
+
"epoch": 655.08,
|
| 203562 |
+
"learning_rate": 8.693634894991923e-06,
|
| 203563 |
+
"loss": 0.2905,
|
| 203564 |
+
"step": 81895
|
| 203565 |
+
},
|
| 203566 |
+
{
|
| 203567 |
+
"epoch": 655.12,
|
| 203568 |
+
"learning_rate": 8.693554119547659e-06,
|
| 203569 |
+
"loss": 0.3468,
|
| 203570 |
+
"step": 81900
|
| 203571 |
+
},
|
| 203572 |
+
{
|
| 203573 |
+
"epoch": 655.16,
|
| 203574 |
+
"learning_rate": 8.693473344103393e-06,
|
| 203575 |
+
"loss": 0.4777,
|
| 203576 |
+
"step": 81905
|
| 203577 |
+
},
|
| 203578 |
+
{
|
| 203579 |
+
"epoch": 655.2,
|
| 203580 |
+
"learning_rate": 8.693392568659129e-06,
|
| 203581 |
+
"loss": 1.1993,
|
| 203582 |
+
"step": 81910
|
| 203583 |
+
},
|
| 203584 |
+
{
|
| 203585 |
+
"epoch": 655.24,
|
| 203586 |
+
"learning_rate": 8.693311793214864e-06,
|
| 203587 |
+
"loss": 0.287,
|
| 203588 |
+
"step": 81915
|
| 203589 |
+
},
|
| 203590 |
+
{
|
| 203591 |
+
"epoch": 655.28,
|
| 203592 |
+
"learning_rate": 8.693231017770599e-06,
|
| 203593 |
+
"loss": 0.3406,
|
| 203594 |
+
"step": 81920
|
| 203595 |
+
},
|
| 203596 |
+
{
|
| 203597 |
+
"epoch": 655.32,
|
| 203598 |
+
"learning_rate": 8.693150242326334e-06,
|
| 203599 |
+
"loss": 0.3755,
|
| 203600 |
+
"step": 81925
|
| 203601 |
+
},
|
| 203602 |
+
{
|
| 203603 |
+
"epoch": 655.36,
|
| 203604 |
+
"learning_rate": 8.693069466882069e-06,
|
| 203605 |
+
"loss": 0.4703,
|
| 203606 |
+
"step": 81930
|
| 203607 |
+
},
|
| 203608 |
+
{
|
| 203609 |
+
"epoch": 655.4,
|
| 203610 |
+
"learning_rate": 8.692988691437804e-06,
|
| 203611 |
+
"loss": 1.1881,
|
| 203612 |
+
"step": 81935
|
| 203613 |
+
},
|
| 203614 |
+
{
|
| 203615 |
+
"epoch": 655.44,
|
| 203616 |
+
"learning_rate": 8.692907915993539e-06,
|
| 203617 |
+
"loss": 0.3511,
|
| 203618 |
+
"step": 81940
|
| 203619 |
+
},
|
| 203620 |
+
{
|
| 203621 |
+
"epoch": 655.48,
|
| 203622 |
+
"learning_rate": 8.692827140549274e-06,
|
| 203623 |
+
"loss": 0.2856,
|
| 203624 |
+
"step": 81945
|
| 203625 |
+
},
|
| 203626 |
+
{
|
| 203627 |
+
"epoch": 655.52,
|
| 203628 |
+
"learning_rate": 8.692746365105009e-06,
|
| 203629 |
+
"loss": 0.2809,
|
| 203630 |
+
"step": 81950
|
| 203631 |
+
},
|
| 203632 |
+
{
|
| 203633 |
+
"epoch": 655.56,
|
| 203634 |
+
"learning_rate": 8.692665589660744e-06,
|
| 203635 |
+
"loss": 0.4301,
|
| 203636 |
+
"step": 81955
|
| 203637 |
+
},
|
| 203638 |
+
{
|
| 203639 |
+
"epoch": 655.6,
|
| 203640 |
+
"learning_rate": 8.692584814216479e-06,
|
| 203641 |
+
"loss": 1.2094,
|
| 203642 |
+
"step": 81960
|
| 203643 |
+
},
|
| 203644 |
+
{
|
| 203645 |
+
"epoch": 655.64,
|
| 203646 |
+
"learning_rate": 8.692504038772214e-06,
|
| 203647 |
+
"loss": 0.3171,
|
| 203648 |
+
"step": 81965
|
| 203649 |
+
},
|
| 203650 |
+
{
|
| 203651 |
+
"epoch": 655.68,
|
| 203652 |
+
"learning_rate": 8.692423263327948e-06,
|
| 203653 |
+
"loss": 0.3045,
|
| 203654 |
+
"step": 81970
|
| 203655 |
+
},
|
| 203656 |
+
{
|
| 203657 |
+
"epoch": 655.72,
|
| 203658 |
+
"learning_rate": 8.692342487883684e-06,
|
| 203659 |
+
"loss": 0.3573,
|
| 203660 |
+
"step": 81975
|
| 203661 |
+
},
|
| 203662 |
+
{
|
| 203663 |
+
"epoch": 655.76,
|
| 203664 |
+
"learning_rate": 8.692261712439418e-06,
|
| 203665 |
+
"loss": 0.4384,
|
| 203666 |
+
"step": 81980
|
| 203667 |
+
},
|
| 203668 |
+
{
|
| 203669 |
+
"epoch": 655.8,
|
| 203670 |
+
"learning_rate": 8.692180936995154e-06,
|
| 203671 |
+
"loss": 1.3024,
|
| 203672 |
+
"step": 81985
|
| 203673 |
+
},
|
| 203674 |
+
{
|
| 203675 |
+
"epoch": 655.84,
|
| 203676 |
+
"learning_rate": 8.69210016155089e-06,
|
| 203677 |
+
"loss": 0.3106,
|
| 203678 |
+
"step": 81990
|
| 203679 |
+
},
|
| 203680 |
+
{
|
| 203681 |
+
"epoch": 655.88,
|
| 203682 |
+
"learning_rate": 8.692019386106624e-06,
|
| 203683 |
+
"loss": 0.2804,
|
| 203684 |
+
"step": 81995
|
| 203685 |
+
},
|
| 203686 |
+
{
|
| 203687 |
+
"epoch": 655.92,
|
| 203688 |
+
"learning_rate": 8.69193861066236e-06,
|
| 203689 |
+
"loss": 0.4209,
|
| 203690 |
+
"step": 82000
|
| 203691 |
+
},
|
| 203692 |
+
{
|
| 203693 |
+
"epoch": 655.96,
|
| 203694 |
+
"learning_rate": 8.691857835218094e-06,
|
| 203695 |
+
"loss": 0.4513,
|
| 203696 |
+
"step": 82005
|
| 203697 |
+
},
|
| 203698 |
+
{
|
| 203699 |
+
"epoch": 656.0,
|
| 203700 |
+
"learning_rate": 8.69177705977383e-06,
|
| 203701 |
+
"loss": 1.2045,
|
| 203702 |
+
"step": 82010
|
| 203703 |
+
},
|
| 203704 |
+
{
|
| 203705 |
+
"epoch": 656.0,
|
| 203706 |
+
"eval_loss": 0.3971728980541229,
|
| 203707 |
+
"eval_runtime": 43.3663,
|
| 203708 |
+
"eval_samples_per_second": 19.393,
|
| 203709 |
+
"eval_steps_per_second": 0.623,
|
| 203710 |
+
"eval_wer": 0.19588447653429603,
|
| 203711 |
+
"step": 82010
|
| 203712 |
+
},
|
| 203713 |
+
{
|
| 203714 |
+
"epoch": 661.04,
|
| 203715 |
+
"learning_rate": 8.691696284329564e-06,
|
| 203716 |
+
"loss": 0.3776,
|
| 203717 |
+
"step": 82015
|
| 203718 |
+
},
|
| 203719 |
+
{
|
| 203720 |
+
"epoch": 661.08,
|
| 203721 |
+
"learning_rate": 8.6916155088853e-06,
|
| 203722 |
+
"loss": 0.3147,
|
| 203723 |
+
"step": 82020
|
| 203724 |
+
},
|
| 203725 |
+
{
|
| 203726 |
+
"epoch": 661.12,
|
| 203727 |
+
"learning_rate": 8.691534733441034e-06,
|
| 203728 |
+
"loss": 0.3062,
|
| 203729 |
+
"step": 82025
|
| 203730 |
+
},
|
| 203731 |
+
{
|
| 203732 |
+
"epoch": 661.16,
|
| 203733 |
+
"learning_rate": 8.69145395799677e-06,
|
| 203734 |
+
"loss": 0.4539,
|
| 203735 |
+
"step": 82030
|
| 203736 |
+
},
|
| 203737 |
+
{
|
| 203738 |
+
"epoch": 661.2,
|
| 203739 |
+
"learning_rate": 8.691373182552504e-06,
|
| 203740 |
+
"loss": 1.04,
|
| 203741 |
+
"step": 82035
|
| 203742 |
+
},
|
| 203743 |
+
{
|
| 203744 |
+
"epoch": 661.24,
|
| 203745 |
+
"learning_rate": 8.69129240710824e-06,
|
| 203746 |
+
"loss": 0.3421,
|
| 203747 |
+
"step": 82040
|
| 203748 |
+
},
|
| 203749 |
+
{
|
| 203750 |
+
"epoch": 661.28,
|
| 203751 |
+
"learning_rate": 8.691211631663974e-06,
|
| 203752 |
+
"loss": 0.3186,
|
| 203753 |
+
"step": 82045
|
| 203754 |
+
},
|
| 203755 |
+
{
|
| 203756 |
+
"epoch": 661.32,
|
| 203757 |
+
"learning_rate": 8.69113085621971e-06,
|
| 203758 |
+
"loss": 0.3361,
|
| 203759 |
+
"step": 82050
|
| 203760 |
+
},
|
| 203761 |
+
{
|
| 203762 |
+
"epoch": 661.36,
|
| 203763 |
+
"learning_rate": 8.691050080775446e-06,
|
| 203764 |
+
"loss": 0.4341,
|
| 203765 |
+
"step": 82055
|
| 203766 |
+
},
|
| 203767 |
+
{
|
| 203768 |
+
"epoch": 661.4,
|
| 203769 |
+
"learning_rate": 8.69096930533118e-06,
|
| 203770 |
+
"loss": 1.3379,
|
| 203771 |
+
"step": 82060
|
| 203772 |
+
},
|
| 203773 |
+
{
|
| 203774 |
+
"epoch": 661.44,
|
| 203775 |
+
"learning_rate": 8.690888529886916e-06,
|
| 203776 |
+
"loss": 0.3727,
|
| 203777 |
+
"step": 82065
|
| 203778 |
+
},
|
| 203779 |
+
{
|
| 203780 |
+
"epoch": 661.48,
|
| 203781 |
+
"learning_rate": 8.69080775444265e-06,
|
| 203782 |
+
"loss": 0.3282,
|
| 203783 |
+
"step": 82070
|
| 203784 |
+
},
|
| 203785 |
+
{
|
| 203786 |
+
"epoch": 661.52,
|
| 203787 |
+
"learning_rate": 8.690726978998386e-06,
|
| 203788 |
+
"loss": 0.3196,
|
| 203789 |
+
"step": 82075
|
| 203790 |
+
},
|
| 203791 |
+
{
|
| 203792 |
+
"epoch": 661.56,
|
| 203793 |
+
"learning_rate": 8.69064620355412e-06,
|
| 203794 |
+
"loss": 0.4741,
|
| 203795 |
+
"step": 82080
|
| 203796 |
+
},
|
| 203797 |
+
{
|
| 203798 |
+
"epoch": 661.6,
|
| 203799 |
+
"learning_rate": 8.690565428109856e-06,
|
| 203800 |
+
"loss": 1.3915,
|
| 203801 |
+
"step": 82085
|
| 203802 |
+
},
|
| 203803 |
+
{
|
| 203804 |
+
"epoch": 661.64,
|
| 203805 |
+
"learning_rate": 8.69048465266559e-06,
|
| 203806 |
+
"loss": 0.3393,
|
| 203807 |
+
"step": 82090
|
| 203808 |
+
},
|
| 203809 |
+
{
|
| 203810 |
+
"epoch": 661.68,
|
| 203811 |
+
"learning_rate": 8.690403877221326e-06,
|
| 203812 |
+
"loss": 0.28,
|
| 203813 |
+
"step": 82095
|
| 203814 |
+
},
|
| 203815 |
+
{
|
| 203816 |
+
"epoch": 661.72,
|
| 203817 |
+
"learning_rate": 8.69032310177706e-06,
|
| 203818 |
+
"loss": 0.3303,
|
| 203819 |
+
"step": 82100
|
| 203820 |
+
},
|
| 203821 |
+
{
|
| 203822 |
+
"epoch": 661.76,
|
| 203823 |
+
"learning_rate": 8.690242326332796e-06,
|
| 203824 |
+
"loss": 0.6743,
|
| 203825 |
+
"step": 82105
|
| 203826 |
+
},
|
| 203827 |
+
{
|
| 203828 |
+
"epoch": 661.8,
|
| 203829 |
+
"learning_rate": 8.69016155088853e-06,
|
| 203830 |
+
"loss": 1.2618,
|
| 203831 |
+
"step": 82110
|
| 203832 |
+
},
|
| 203833 |
+
{
|
| 203834 |
+
"epoch": 661.84,
|
| 203835 |
+
"learning_rate": 8.690080775444266e-06,
|
| 203836 |
+
"loss": 0.3449,
|
| 203837 |
+
"step": 82115
|
| 203838 |
+
},
|
| 203839 |
+
{
|
| 203840 |
+
"epoch": 661.88,
|
| 203841 |
+
"learning_rate": 8.690000000000002e-06,
|
| 203842 |
+
"loss": 0.2685,
|
| 203843 |
+
"step": 82120
|
| 203844 |
+
},
|
| 203845 |
+
{
|
| 203846 |
+
"epoch": 661.92,
|
| 203847 |
+
"learning_rate": 8.689919224555736e-06,
|
| 203848 |
+
"loss": 0.3878,
|
| 203849 |
+
"step": 82125
|
| 203850 |
+
},
|
| 203851 |
+
{
|
| 203852 |
+
"epoch": 661.96,
|
| 203853 |
+
"learning_rate": 8.689838449111471e-06,
|
| 203854 |
+
"loss": 0.4573,
|
| 203855 |
+
"step": 82130
|
| 203856 |
+
},
|
| 203857 |
+
{
|
| 203858 |
+
"epoch": 662.0,
|
| 203859 |
+
"eval_loss": 0.4005543291568756,
|
| 203860 |
+
"eval_runtime": 41.7634,
|
| 203861 |
+
"eval_samples_per_second": 20.137,
|
| 203862 |
+
"eval_steps_per_second": 0.646,
|
| 203863 |
+
"eval_wer": 0.1882754117390989,
|
| 203864 |
+
"step": 82134
|
| 203865 |
}
|
| 203866 |
],
|
| 203867 |
"max_steps": 620000,
|
| 203868 |
"num_train_epochs": 5000,
|
| 203869 |
+
"total_flos": 2.311439623193885e+20,
|
| 203870 |
"trial_name": null,
|
| 203871 |
"trial_params": null
|
| 203872 |
}
|
model-bin/finetune/base/{checkpoint-81513 → checkpoint-82134}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629908936.4564571/events.out.tfevents.1629908936.7e498afd5545.7645.85
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:88043dbfc014775fe34ca52feef56a845d5c72b7f834cd2eedd1f005e7c3513d
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629909414.6742997/events.out.tfevents.1629909414.7e498afd5545.7645.87
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:539ba4c32aa9ba29fc524630bc75b00c130f9b9dc379ec813c8e151c46813bb0
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629909897.4578655/events.out.tfevents.1629909897.7e498afd5545.7645.89
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:46d079512bd7c680908d7de7849f2a7ba8c53b3461349a64dea7f4d8278b42a4
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629910378.9335363/events.out.tfevents.1629910378.7e498afd5545.7645.91
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:7360042c324e7514187a2a1547a3649d50d5a8ca62f7a386040ae23d995edb64
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629910855.6983657/events.out.tfevents.1629910855.7e498afd5545.7645.93
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:cfb2728e096f7c3f4446de6971af6f710e4ddc59901ae230c6e1442fbc4fddad
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629908936.7e498afd5545.7645.84
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:0ef0885530910541691764e60631a45e999938a1cc3a3cc4a013d99dfb3186db
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629909414.7e498afd5545.7645.86
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:30d2a2747b4fa0af5a584e24ce818500b405bbf8d7d908aee5384cf2621ac9d2
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629909897.7e498afd5545.7645.88
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e10c39d49cfbac496ce109b11fbc1ecc98c61886ba2f40ec35b9b1b12f137036
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629910378.7e498afd5545.7645.90
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:af8657c949f1bec17cdd19d42d9b4f80924851715f071810fe3ea20f2fb35ed9
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629910855.7e498afd5545.7645.92
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f4c2303e1d3c4b46ba59d1a386403959162cae895c54fd8f726e8ab1e8cb0bb8
|
| 3 |
+
size 8462
|