"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-148340 → checkpoint-148960}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-148340 → checkpoint-148960}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-148340 → checkpoint-148960}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-148340 → checkpoint-148960}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-148340 → checkpoint-148960}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-148340 → checkpoint-148960}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-148340 → checkpoint-148960}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-148340 → checkpoint-148960}/trainer_state.json +793 -4
- model-bin/finetune/base/{checkpoint-148340 → checkpoint-148960}/training_args.bin +0 -0
- model-bin/finetune/base/log/1630237299.1184735/events.out.tfevents.1630237299.cc93b136ebf5.1086.149 +3 -0
- model-bin/finetune/base/log/1630237737.6447396/events.out.tfevents.1630237737.cc93b136ebf5.1086.151 +3 -0
- model-bin/finetune/base/log/1630238176.5216558/events.out.tfevents.1630238176.cc93b136ebf5.1086.153 +3 -0
- model-bin/finetune/base/log/1630238608.102783/events.out.tfevents.1630238608.cc93b136ebf5.1086.155 +3 -0
- model-bin/finetune/base/log/1630239044.210316/events.out.tfevents.1630239044.cc93b136ebf5.1086.157 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630237299.cc93b136ebf5.1086.148 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630237737.cc93b136ebf5.1086.150 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630238176.cc93b136ebf5.1086.152 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630238608.cc93b136ebf5.1086.154 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630239044.cc93b136ebf5.1086.156 +3 -0
model-bin/finetune/base/{checkpoint-148340 → checkpoint-148960}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-148340 → checkpoint-148960}/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:f93d5bc8ae226e1c1aee31035ac894051b19c4bbe441659018307488de04cd85
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-148340 → checkpoint-148960}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-148340 → checkpoint-148960}/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 sha256:93a42bd5cb554963f46e2895d26148d57b1cf04e1e671d3cc0e6a201be5624e6
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-148340 → checkpoint-148960}/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 sha256:4057531845a0717c33153a0d43d59bbf8df52909b80221bb96290592a2c80c25
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-148340 → checkpoint-148960}/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 sha256:61520e9eb8eb51b6c2259c437be709f2dfc17a4076a25563394233f7ae24e463
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-148340 → checkpoint-148960}/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 sha256:6cdb9e88c57bc6cc6a5791bc7ab637007af04603717e93fe539b17454fd741ec
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-148340 → checkpoint-148960}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1668011442822563,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-146596",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -288102,11 +288102,800 @@
|
|
| 288102 |
"eval_steps_per_second": 0.633,
|
| 288103 |
"eval_wer": 0.1786735277301315,
|
| 288104 |
"step": 148340
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 288105 |
}
|
| 288106 |
],
|
| 288107 |
-
"max_steps":
|
| 288108 |
"num_train_epochs": 5000,
|
| 288109 |
-
"total_flos": 4.
|
| 288110 |
"trial_name": null,
|
| 288111 |
"trial_params": null
|
| 288112 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1668011442822563,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-146596",
|
| 4 |
+
"epoch": 1191.0,
|
| 5 |
+
"global_step": 148960,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 288102 |
"eval_steps_per_second": 0.633,
|
| 288103 |
"eval_wer": 0.1786735277301315,
|
| 288104 |
"step": 148340
|
| 288105 |
+
},
|
| 288106 |
+
{
|
| 288107 |
+
"epoch": 1196.04,
|
| 288108 |
+
"learning_rate": 7.620581583198708e-06,
|
| 288109 |
+
"loss": 0.2904,
|
| 288110 |
+
"step": 148345
|
| 288111 |
+
},
|
| 288112 |
+
{
|
| 288113 |
+
"epoch": 1196.08,
|
| 288114 |
+
"learning_rate": 7.620500807754443e-06,
|
| 288115 |
+
"loss": 0.2934,
|
| 288116 |
+
"step": 148350
|
| 288117 |
+
},
|
| 288118 |
+
{
|
| 288119 |
+
"epoch": 1196.12,
|
| 288120 |
+
"learning_rate": 7.620420032310178e-06,
|
| 288121 |
+
"loss": 0.2976,
|
| 288122 |
+
"step": 148355
|
| 288123 |
+
},
|
| 288124 |
+
{
|
| 288125 |
+
"epoch": 1196.16,
|
| 288126 |
+
"learning_rate": 7.620339256865913e-06,
|
| 288127 |
+
"loss": 0.4439,
|
| 288128 |
+
"step": 148360
|
| 288129 |
+
},
|
| 288130 |
+
{
|
| 288131 |
+
"epoch": 1196.2,
|
| 288132 |
+
"learning_rate": 7.620258481421648e-06,
|
| 288133 |
+
"loss": 1.1461,
|
| 288134 |
+
"step": 148365
|
| 288135 |
+
},
|
| 288136 |
+
{
|
| 288137 |
+
"epoch": 1196.24,
|
| 288138 |
+
"learning_rate": 7.620177705977384e-06,
|
| 288139 |
+
"loss": 0.3897,
|
| 288140 |
+
"step": 148370
|
| 288141 |
+
},
|
| 288142 |
+
{
|
| 288143 |
+
"epoch": 1196.28,
|
| 288144 |
+
"learning_rate": 7.620096930533119e-06,
|
| 288145 |
+
"loss": 0.2371,
|
| 288146 |
+
"step": 148375
|
| 288147 |
+
},
|
| 288148 |
+
{
|
| 288149 |
+
"epoch": 1196.32,
|
| 288150 |
+
"learning_rate": 7.620016155088854e-06,
|
| 288151 |
+
"loss": 0.2599,
|
| 288152 |
+
"step": 148380
|
| 288153 |
+
},
|
| 288154 |
+
{
|
| 288155 |
+
"epoch": 1196.36,
|
| 288156 |
+
"learning_rate": 7.619935379644589e-06,
|
| 288157 |
+
"loss": 0.4827,
|
| 288158 |
+
"step": 148385
|
| 288159 |
+
},
|
| 288160 |
+
{
|
| 288161 |
+
"epoch": 1196.4,
|
| 288162 |
+
"learning_rate": 7.619854604200324e-06,
|
| 288163 |
+
"loss": 1.2766,
|
| 288164 |
+
"step": 148390
|
| 288165 |
+
},
|
| 288166 |
+
{
|
| 288167 |
+
"epoch": 1196.44,
|
| 288168 |
+
"learning_rate": 7.619773828756059e-06,
|
| 288169 |
+
"loss": 0.3409,
|
| 288170 |
+
"step": 148395
|
| 288171 |
+
},
|
| 288172 |
+
{
|
| 288173 |
+
"epoch": 1196.48,
|
| 288174 |
+
"learning_rate": 7.619693053311794e-06,
|
| 288175 |
+
"loss": 0.2787,
|
| 288176 |
+
"step": 148400
|
| 288177 |
+
},
|
| 288178 |
+
{
|
| 288179 |
+
"epoch": 1196.52,
|
| 288180 |
+
"learning_rate": 7.619612277867529e-06,
|
| 288181 |
+
"loss": 0.2924,
|
| 288182 |
+
"step": 148405
|
| 288183 |
+
},
|
| 288184 |
+
{
|
| 288185 |
+
"epoch": 1196.56,
|
| 288186 |
+
"learning_rate": 7.619531502423264e-06,
|
| 288187 |
+
"loss": 0.396,
|
| 288188 |
+
"step": 148410
|
| 288189 |
+
},
|
| 288190 |
+
{
|
| 288191 |
+
"epoch": 1196.6,
|
| 288192 |
+
"learning_rate": 7.619450726978999e-06,
|
| 288193 |
+
"loss": 1.151,
|
| 288194 |
+
"step": 148415
|
| 288195 |
+
},
|
| 288196 |
+
{
|
| 288197 |
+
"epoch": 1196.64,
|
| 288198 |
+
"learning_rate": 7.619369951534734e-06,
|
| 288199 |
+
"loss": 0.2854,
|
| 288200 |
+
"step": 148420
|
| 288201 |
+
},
|
| 288202 |
+
{
|
| 288203 |
+
"epoch": 1196.68,
|
| 288204 |
+
"learning_rate": 7.619289176090469e-06,
|
| 288205 |
+
"loss": 0.269,
|
| 288206 |
+
"step": 148425
|
| 288207 |
+
},
|
| 288208 |
+
{
|
| 288209 |
+
"epoch": 1196.72,
|
| 288210 |
+
"learning_rate": 7.619208400646204e-06,
|
| 288211 |
+
"loss": 0.2801,
|
| 288212 |
+
"step": 148430
|
| 288213 |
+
},
|
| 288214 |
+
{
|
| 288215 |
+
"epoch": 1196.76,
|
| 288216 |
+
"learning_rate": 7.61912762520194e-06,
|
| 288217 |
+
"loss": 0.4201,
|
| 288218 |
+
"step": 148435
|
| 288219 |
+
},
|
| 288220 |
+
{
|
| 288221 |
+
"epoch": 1196.8,
|
| 288222 |
+
"learning_rate": 7.619046849757674e-06,
|
| 288223 |
+
"loss": 1.1717,
|
| 288224 |
+
"step": 148440
|
| 288225 |
+
},
|
| 288226 |
+
{
|
| 288227 |
+
"epoch": 1196.84,
|
| 288228 |
+
"learning_rate": 7.61896607431341e-06,
|
| 288229 |
+
"loss": 0.299,
|
| 288230 |
+
"step": 148445
|
| 288231 |
+
},
|
| 288232 |
+
{
|
| 288233 |
+
"epoch": 1196.88,
|
| 288234 |
+
"learning_rate": 7.618885298869145e-06,
|
| 288235 |
+
"loss": 0.2666,
|
| 288236 |
+
"step": 148450
|
| 288237 |
+
},
|
| 288238 |
+
{
|
| 288239 |
+
"epoch": 1196.92,
|
| 288240 |
+
"learning_rate": 7.6188045234248796e-06,
|
| 288241 |
+
"loss": 0.2868,
|
| 288242 |
+
"step": 148455
|
| 288243 |
+
},
|
| 288244 |
+
{
|
| 288245 |
+
"epoch": 1196.96,
|
| 288246 |
+
"learning_rate": 7.6187237479806146e-06,
|
| 288247 |
+
"loss": 0.4595,
|
| 288248 |
+
"step": 148460
|
| 288249 |
+
},
|
| 288250 |
+
{
|
| 288251 |
+
"epoch": 1197.0,
|
| 288252 |
+
"eval_loss": 0.3382641077041626,
|
| 288253 |
+
"eval_runtime": 42.1842,
|
| 288254 |
+
"eval_samples_per_second": 19.936,
|
| 288255 |
+
"eval_steps_per_second": 0.64,
|
| 288256 |
+
"eval_wer": 0.17046594982078853,
|
| 288257 |
+
"step": 148464
|
| 288258 |
+
},
|
| 288259 |
+
{
|
| 288260 |
+
"epoch": 1197.01,
|
| 288261 |
+
"learning_rate": 7.6186429725363495e-06,
|
| 288262 |
+
"loss": 0.4864,
|
| 288263 |
+
"step": 148465
|
| 288264 |
+
},
|
| 288265 |
+
{
|
| 288266 |
+
"epoch": 1197.05,
|
| 288267 |
+
"learning_rate": 7.6185621970920845e-06,
|
| 288268 |
+
"loss": 0.2608,
|
| 288269 |
+
"step": 148470
|
| 288270 |
+
},
|
| 288271 |
+
{
|
| 288272 |
+
"epoch": 1197.09,
|
| 288273 |
+
"learning_rate": 7.6184814216478195e-06,
|
| 288274 |
+
"loss": 0.2822,
|
| 288275 |
+
"step": 148475
|
| 288276 |
+
},
|
| 288277 |
+
{
|
| 288278 |
+
"epoch": 1197.13,
|
| 288279 |
+
"learning_rate": 7.6184006462035545e-06,
|
| 288280 |
+
"loss": 0.3627,
|
| 288281 |
+
"step": 148480
|
| 288282 |
+
},
|
| 288283 |
+
{
|
| 288284 |
+
"epoch": 1197.17,
|
| 288285 |
+
"learning_rate": 7.6183198707592895e-06,
|
| 288286 |
+
"loss": 0.5249,
|
| 288287 |
+
"step": 148485
|
| 288288 |
+
},
|
| 288289 |
+
{
|
| 288290 |
+
"epoch": 1197.21,
|
| 288291 |
+
"learning_rate": 7.618239095315025e-06,
|
| 288292 |
+
"loss": 1.042,
|
| 288293 |
+
"step": 148490
|
| 288294 |
+
},
|
| 288295 |
+
{
|
| 288296 |
+
"epoch": 1197.25,
|
| 288297 |
+
"learning_rate": 7.6181583198707595e-06,
|
| 288298 |
+
"loss": 0.2996,
|
| 288299 |
+
"step": 148495
|
| 288300 |
+
},
|
| 288301 |
+
{
|
| 288302 |
+
"epoch": 1197.29,
|
| 288303 |
+
"learning_rate": 7.618077544426495e-06,
|
| 288304 |
+
"loss": 0.2382,
|
| 288305 |
+
"step": 148500
|
| 288306 |
+
},
|
| 288307 |
+
{
|
| 288308 |
+
"epoch": 1197.33,
|
| 288309 |
+
"learning_rate": 7.6179967689822294e-06,
|
| 288310 |
+
"loss": 0.3243,
|
| 288311 |
+
"step": 148505
|
| 288312 |
+
},
|
| 288313 |
+
{
|
| 288314 |
+
"epoch": 1197.37,
|
| 288315 |
+
"learning_rate": 7.617915993537965e-06,
|
| 288316 |
+
"loss": 0.4499,
|
| 288317 |
+
"step": 148510
|
| 288318 |
+
},
|
| 288319 |
+
{
|
| 288320 |
+
"epoch": 1197.41,
|
| 288321 |
+
"learning_rate": 7.6178352180937e-06,
|
| 288322 |
+
"loss": 1.0114,
|
| 288323 |
+
"step": 148515
|
| 288324 |
+
},
|
| 288325 |
+
{
|
| 288326 |
+
"epoch": 1197.45,
|
| 288327 |
+
"learning_rate": 7.617754442649435e-06,
|
| 288328 |
+
"loss": 0.2706,
|
| 288329 |
+
"step": 148520
|
| 288330 |
+
},
|
| 288331 |
+
{
|
| 288332 |
+
"epoch": 1197.49,
|
| 288333 |
+
"learning_rate": 7.61767366720517e-06,
|
| 288334 |
+
"loss": 0.2452,
|
| 288335 |
+
"step": 148525
|
| 288336 |
+
},
|
| 288337 |
+
{
|
| 288338 |
+
"epoch": 1197.53,
|
| 288339 |
+
"learning_rate": 7.617592891760905e-06,
|
| 288340 |
+
"loss": 0.325,
|
| 288341 |
+
"step": 148530
|
| 288342 |
+
},
|
| 288343 |
+
{
|
| 288344 |
+
"epoch": 1197.57,
|
| 288345 |
+
"learning_rate": 7.61751211631664e-06,
|
| 288346 |
+
"loss": 0.5252,
|
| 288347 |
+
"step": 148535
|
| 288348 |
+
},
|
| 288349 |
+
{
|
| 288350 |
+
"epoch": 1197.61,
|
| 288351 |
+
"learning_rate": 7.617431340872375e-06,
|
| 288352 |
+
"loss": 1.137,
|
| 288353 |
+
"step": 148540
|
| 288354 |
+
},
|
| 288355 |
+
{
|
| 288356 |
+
"epoch": 1197.65,
|
| 288357 |
+
"learning_rate": 7.617350565428111e-06,
|
| 288358 |
+
"loss": 0.3361,
|
| 288359 |
+
"step": 148545
|
| 288360 |
+
},
|
| 288361 |
+
{
|
| 288362 |
+
"epoch": 1197.69,
|
| 288363 |
+
"learning_rate": 7.617269789983845e-06,
|
| 288364 |
+
"loss": 0.3604,
|
| 288365 |
+
"step": 148550
|
| 288366 |
+
},
|
| 288367 |
+
{
|
| 288368 |
+
"epoch": 1197.73,
|
| 288369 |
+
"learning_rate": 7.617189014539581e-06,
|
| 288370 |
+
"loss": 0.3392,
|
| 288371 |
+
"step": 148555
|
| 288372 |
+
},
|
| 288373 |
+
{
|
| 288374 |
+
"epoch": 1197.77,
|
| 288375 |
+
"learning_rate": 7.617108239095315e-06,
|
| 288376 |
+
"loss": 0.5506,
|
| 288377 |
+
"step": 148560
|
| 288378 |
+
},
|
| 288379 |
+
{
|
| 288380 |
+
"epoch": 1197.81,
|
| 288381 |
+
"learning_rate": 7.617027463651051e-06,
|
| 288382 |
+
"loss": 1.1279,
|
| 288383 |
+
"step": 148565
|
| 288384 |
+
},
|
| 288385 |
+
{
|
| 288386 |
+
"epoch": 1197.85,
|
| 288387 |
+
"learning_rate": 7.616946688206785e-06,
|
| 288388 |
+
"loss": 0.3246,
|
| 288389 |
+
"step": 148570
|
| 288390 |
+
},
|
| 288391 |
+
{
|
| 288392 |
+
"epoch": 1197.89,
|
| 288393 |
+
"learning_rate": 7.616865912762521e-06,
|
| 288394 |
+
"loss": 0.2687,
|
| 288395 |
+
"step": 148575
|
| 288396 |
+
},
|
| 288397 |
+
{
|
| 288398 |
+
"epoch": 1197.93,
|
| 288399 |
+
"learning_rate": 7.616785137318256e-06,
|
| 288400 |
+
"loss": 0.2504,
|
| 288401 |
+
"step": 148580
|
| 288402 |
+
},
|
| 288403 |
+
{
|
| 288404 |
+
"epoch": 1197.97,
|
| 288405 |
+
"learning_rate": 7.616704361873991e-06,
|
| 288406 |
+
"loss": 0.6615,
|
| 288407 |
+
"step": 148585
|
| 288408 |
+
},
|
| 288409 |
+
{
|
| 288410 |
+
"epoch": 1198.0,
|
| 288411 |
+
"eval_loss": 0.3838162422180176,
|
| 288412 |
+
"eval_runtime": 40.8275,
|
| 288413 |
+
"eval_samples_per_second": 20.574,
|
| 288414 |
+
"eval_steps_per_second": 0.661,
|
| 288415 |
+
"eval_wer": 0.17510748378634408,
|
| 288416 |
+
"step": 148588
|
| 288417 |
+
},
|
| 288418 |
+
{
|
| 288419 |
+
"epoch": 1208.02,
|
| 288420 |
+
"learning_rate": 7.616623586429726e-06,
|
| 288421 |
+
"loss": 0.3853,
|
| 288422 |
+
"step": 148590
|
| 288423 |
+
},
|
| 288424 |
+
{
|
| 288425 |
+
"epoch": 1208.06,
|
| 288426 |
+
"learning_rate": 7.616542810985461e-06,
|
| 288427 |
+
"loss": 0.2698,
|
| 288428 |
+
"step": 148595
|
| 288429 |
+
},
|
| 288430 |
+
{
|
| 288431 |
+
"epoch": 1208.1,
|
| 288432 |
+
"learning_rate": 7.616462035541196e-06,
|
| 288433 |
+
"loss": 0.2669,
|
| 288434 |
+
"step": 148600
|
| 288435 |
+
},
|
| 288436 |
+
{
|
| 288437 |
+
"epoch": 1208.14,
|
| 288438 |
+
"learning_rate": 7.616381260096931e-06,
|
| 288439 |
+
"loss": 0.3389,
|
| 288440 |
+
"step": 148605
|
| 288441 |
+
},
|
| 288442 |
+
{
|
| 288443 |
+
"epoch": 1208.18,
|
| 288444 |
+
"learning_rate": 7.616300484652667e-06,
|
| 288445 |
+
"loss": 0.5707,
|
| 288446 |
+
"step": 148610
|
| 288447 |
+
},
|
| 288448 |
+
{
|
| 288449 |
+
"epoch": 1208.22,
|
| 288450 |
+
"learning_rate": 7.616219709208401e-06,
|
| 288451 |
+
"loss": 0.8938,
|
| 288452 |
+
"step": 148615
|
| 288453 |
+
},
|
| 288454 |
+
{
|
| 288455 |
+
"epoch": 1208.26,
|
| 288456 |
+
"learning_rate": 7.616138933764137e-06,
|
| 288457 |
+
"loss": 0.2657,
|
| 288458 |
+
"step": 148620
|
| 288459 |
+
},
|
| 288460 |
+
{
|
| 288461 |
+
"epoch": 1208.3,
|
| 288462 |
+
"learning_rate": 7.616058158319871e-06,
|
| 288463 |
+
"loss": 0.2442,
|
| 288464 |
+
"step": 148625
|
| 288465 |
+
},
|
| 288466 |
+
{
|
| 288467 |
+
"epoch": 1208.34,
|
| 288468 |
+
"learning_rate": 7.615977382875607e-06,
|
| 288469 |
+
"loss": 0.2851,
|
| 288470 |
+
"step": 148630
|
| 288471 |
+
},
|
| 288472 |
+
{
|
| 288473 |
+
"epoch": 1208.38,
|
| 288474 |
+
"learning_rate": 7.615896607431341e-06,
|
| 288475 |
+
"loss": 0.535,
|
| 288476 |
+
"step": 148635
|
| 288477 |
+
},
|
| 288478 |
+
{
|
| 288479 |
+
"epoch": 1208.42,
|
| 288480 |
+
"learning_rate": 7.615815831987077e-06,
|
| 288481 |
+
"loss": 0.8672,
|
| 288482 |
+
"step": 148640
|
| 288483 |
+
},
|
| 288484 |
+
{
|
| 288485 |
+
"epoch": 1208.46,
|
| 288486 |
+
"learning_rate": 7.615735056542811e-06,
|
| 288487 |
+
"loss": 0.2416,
|
| 288488 |
+
"step": 148645
|
| 288489 |
+
},
|
| 288490 |
+
{
|
| 288491 |
+
"epoch": 1208.5,
|
| 288492 |
+
"learning_rate": 7.615654281098547e-06,
|
| 288493 |
+
"loss": 0.3156,
|
| 288494 |
+
"step": 148650
|
| 288495 |
+
},
|
| 288496 |
+
{
|
| 288497 |
+
"epoch": 1208.54,
|
| 288498 |
+
"learning_rate": 7.615573505654282e-06,
|
| 288499 |
+
"loss": 0.3208,
|
| 288500 |
+
"step": 148655
|
| 288501 |
+
},
|
| 288502 |
+
{
|
| 288503 |
+
"epoch": 1208.58,
|
| 288504 |
+
"learning_rate": 7.615492730210017e-06,
|
| 288505 |
+
"loss": 0.73,
|
| 288506 |
+
"step": 148660
|
| 288507 |
+
},
|
| 288508 |
+
{
|
| 288509 |
+
"epoch": 1208.62,
|
| 288510 |
+
"learning_rate": 7.6154119547657524e-06,
|
| 288511 |
+
"loss": 1.0101,
|
| 288512 |
+
"step": 148665
|
| 288513 |
+
},
|
| 288514 |
+
{
|
| 288515 |
+
"epoch": 1208.66,
|
| 288516 |
+
"learning_rate": 7.6153311793214866e-06,
|
| 288517 |
+
"loss": 0.3275,
|
| 288518 |
+
"step": 148670
|
| 288519 |
+
},
|
| 288520 |
+
{
|
| 288521 |
+
"epoch": 1208.7,
|
| 288522 |
+
"learning_rate": 7.615250403877222e-06,
|
| 288523 |
+
"loss": 0.2837,
|
| 288524 |
+
"step": 148675
|
| 288525 |
+
},
|
| 288526 |
+
{
|
| 288527 |
+
"epoch": 1208.74,
|
| 288528 |
+
"learning_rate": 7.6151696284329565e-06,
|
| 288529 |
+
"loss": 0.3495,
|
| 288530 |
+
"step": 148680
|
| 288531 |
+
},
|
| 288532 |
+
{
|
| 288533 |
+
"epoch": 1208.79,
|
| 288534 |
+
"learning_rate": 7.615088852988692e-06,
|
| 288535 |
+
"loss": 0.5637,
|
| 288536 |
+
"step": 148685
|
| 288537 |
+
},
|
| 288538 |
+
{
|
| 288539 |
+
"epoch": 1208.83,
|
| 288540 |
+
"learning_rate": 7.6150080775444265e-06,
|
| 288541 |
+
"loss": 0.9843,
|
| 288542 |
+
"step": 148690
|
| 288543 |
+
},
|
| 288544 |
+
{
|
| 288545 |
+
"epoch": 1208.87,
|
| 288546 |
+
"learning_rate": 7.614927302100162e-06,
|
| 288547 |
+
"loss": 0.2692,
|
| 288548 |
+
"step": 148695
|
| 288549 |
+
},
|
| 288550 |
+
{
|
| 288551 |
+
"epoch": 1208.91,
|
| 288552 |
+
"learning_rate": 7.6148465266558965e-06,
|
| 288553 |
+
"loss": 0.2533,
|
| 288554 |
+
"step": 148700
|
| 288555 |
+
},
|
| 288556 |
+
{
|
| 288557 |
+
"epoch": 1208.95,
|
| 288558 |
+
"learning_rate": 7.614765751211632e-06,
|
| 288559 |
+
"loss": 0.3418,
|
| 288560 |
+
"step": 148705
|
| 288561 |
+
},
|
| 288562 |
+
{
|
| 288563 |
+
"epoch": 1208.99,
|
| 288564 |
+
"learning_rate": 7.6146849757673665e-06,
|
| 288565 |
+
"loss": 0.7517,
|
| 288566 |
+
"step": 148710
|
| 288567 |
+
},
|
| 288568 |
+
{
|
| 288569 |
+
"epoch": 1209.0,
|
| 288570 |
+
"eval_loss": 0.3921215534210205,
|
| 288571 |
+
"eval_runtime": 42.6819,
|
| 288572 |
+
"eval_samples_per_second": 19.68,
|
| 288573 |
+
"eval_steps_per_second": 0.633,
|
| 288574 |
+
"eval_wer": 0.1728014106237602,
|
| 288575 |
+
"step": 148711
|
| 288576 |
+
},
|
| 288577 |
+
{
|
| 288578 |
+
"epoch": 1199.03,
|
| 288579 |
+
"learning_rate": 7.614604200323102e-06,
|
| 288580 |
+
"loss": 0.2649,
|
| 288581 |
+
"step": 148715
|
| 288582 |
+
},
|
| 288583 |
+
{
|
| 288584 |
+
"epoch": 1199.07,
|
| 288585 |
+
"learning_rate": 7.614523424878838e-06,
|
| 288586 |
+
"loss": 0.3106,
|
| 288587 |
+
"step": 148720
|
| 288588 |
+
},
|
| 288589 |
+
{
|
| 288590 |
+
"epoch": 1199.11,
|
| 288591 |
+
"learning_rate": 7.614442649434572e-06,
|
| 288592 |
+
"loss": 0.2858,
|
| 288593 |
+
"step": 148725
|
| 288594 |
+
},
|
| 288595 |
+
{
|
| 288596 |
+
"epoch": 1199.15,
|
| 288597 |
+
"learning_rate": 7.614361873990308e-06,
|
| 288598 |
+
"loss": 0.3628,
|
| 288599 |
+
"step": 148730
|
| 288600 |
+
},
|
| 288601 |
+
{
|
| 288602 |
+
"epoch": 1199.19,
|
| 288603 |
+
"learning_rate": 7.614281098546042e-06,
|
| 288604 |
+
"loss": 0.8461,
|
| 288605 |
+
"step": 148735
|
| 288606 |
+
},
|
| 288607 |
+
{
|
| 288608 |
+
"epoch": 1199.23,
|
| 288609 |
+
"learning_rate": 7.614200323101778e-06,
|
| 288610 |
+
"loss": 0.5783,
|
| 288611 |
+
"step": 148740
|
| 288612 |
+
},
|
| 288613 |
+
{
|
| 288614 |
+
"epoch": 1199.27,
|
| 288615 |
+
"learning_rate": 7.614119547657512e-06,
|
| 288616 |
+
"loss": 0.5042,
|
| 288617 |
+
"step": 148745
|
| 288618 |
+
},
|
| 288619 |
+
{
|
| 288620 |
+
"epoch": 1199.31,
|
| 288621 |
+
"learning_rate": 7.614038772213248e-06,
|
| 288622 |
+
"loss": 0.2598,
|
| 288623 |
+
"step": 148750
|
| 288624 |
+
},
|
| 288625 |
+
{
|
| 288626 |
+
"epoch": 1199.35,
|
| 288627 |
+
"learning_rate": 7.613957996768982e-06,
|
| 288628 |
+
"loss": 0.3811,
|
| 288629 |
+
"step": 148755
|
| 288630 |
+
},
|
| 288631 |
+
{
|
| 288632 |
+
"epoch": 1199.39,
|
| 288633 |
+
"learning_rate": 7.613877221324718e-06,
|
| 288634 |
+
"loss": 0.9186,
|
| 288635 |
+
"step": 148760
|
| 288636 |
+
},
|
| 288637 |
+
{
|
| 288638 |
+
"epoch": 1199.43,
|
| 288639 |
+
"learning_rate": 7.613796445880452e-06,
|
| 288640 |
+
"loss": 0.6704,
|
| 288641 |
+
"step": 148765
|
| 288642 |
+
},
|
| 288643 |
+
{
|
| 288644 |
+
"epoch": 1199.47,
|
| 288645 |
+
"learning_rate": 7.613715670436188e-06,
|
| 288646 |
+
"loss": 0.2681,
|
| 288647 |
+
"step": 148770
|
| 288648 |
+
},
|
| 288649 |
+
{
|
| 288650 |
+
"epoch": 1199.51,
|
| 288651 |
+
"learning_rate": 7.613634894991922e-06,
|
| 288652 |
+
"loss": 0.2969,
|
| 288653 |
+
"step": 148775
|
| 288654 |
+
},
|
| 288655 |
+
{
|
| 288656 |
+
"epoch": 1199.55,
|
| 288657 |
+
"learning_rate": 7.613554119547658e-06,
|
| 288658 |
+
"loss": 0.3976,
|
| 288659 |
+
"step": 148780
|
| 288660 |
+
},
|
| 288661 |
+
{
|
| 288662 |
+
"epoch": 1199.59,
|
| 288663 |
+
"learning_rate": 7.613473344103394e-06,
|
| 288664 |
+
"loss": 0.969,
|
| 288665 |
+
"step": 148785
|
| 288666 |
+
},
|
| 288667 |
+
{
|
| 288668 |
+
"epoch": 1199.63,
|
| 288669 |
+
"learning_rate": 7.613392568659128e-06,
|
| 288670 |
+
"loss": 0.6605,
|
| 288671 |
+
"step": 148790
|
| 288672 |
+
},
|
| 288673 |
+
{
|
| 288674 |
+
"epoch": 1199.67,
|
| 288675 |
+
"learning_rate": 7.613311793214864e-06,
|
| 288676 |
+
"loss": 0.3244,
|
| 288677 |
+
"step": 148795
|
| 288678 |
+
},
|
| 288679 |
+
{
|
| 288680 |
+
"epoch": 1199.71,
|
| 288681 |
+
"learning_rate": 7.613231017770598e-06,
|
| 288682 |
+
"loss": 0.2454,
|
| 288683 |
+
"step": 148800
|
| 288684 |
+
},
|
| 288685 |
+
{
|
| 288686 |
+
"epoch": 1199.76,
|
| 288687 |
+
"learning_rate": 7.613150242326334e-06,
|
| 288688 |
+
"loss": 0.3354,
|
| 288689 |
+
"step": 148805
|
| 288690 |
+
},
|
| 288691 |
+
{
|
| 288692 |
+
"epoch": 1199.8,
|
| 288693 |
+
"learning_rate": 7.613069466882068e-06,
|
| 288694 |
+
"loss": 0.9782,
|
| 288695 |
+
"step": 148810
|
| 288696 |
+
},
|
| 288697 |
+
{
|
| 288698 |
+
"epoch": 1199.84,
|
| 288699 |
+
"learning_rate": 7.612988691437804e-06,
|
| 288700 |
+
"loss": 0.6457,
|
| 288701 |
+
"step": 148815
|
| 288702 |
+
},
|
| 288703 |
+
{
|
| 288704 |
+
"epoch": 1199.88,
|
| 288705 |
+
"learning_rate": 7.612907915993538e-06,
|
| 288706 |
+
"loss": 0.3441,
|
| 288707 |
+
"step": 148820
|
| 288708 |
+
},
|
| 288709 |
+
{
|
| 288710 |
+
"epoch": 1199.92,
|
| 288711 |
+
"learning_rate": 7.612827140549274e-06,
|
| 288712 |
+
"loss": 0.2688,
|
| 288713 |
+
"step": 148825
|
| 288714 |
+
},
|
| 288715 |
+
{
|
| 288716 |
+
"epoch": 1199.96,
|
| 288717 |
+
"learning_rate": 7.612746365105008e-06,
|
| 288718 |
+
"loss": 0.3786,
|
| 288719 |
+
"step": 148830
|
| 288720 |
+
},
|
| 288721 |
+
{
|
| 288722 |
+
"epoch": 1200.0,
|
| 288723 |
+
"learning_rate": 7.612665589660744e-06,
|
| 288724 |
+
"loss": 0.9727,
|
| 288725 |
+
"step": 148835
|
| 288726 |
+
},
|
| 288727 |
+
{
|
| 288728 |
+
"epoch": 1200.0,
|
| 288729 |
+
"eval_loss": 0.3765946924686432,
|
| 288730 |
+
"eval_runtime": 41.4291,
|
| 288731 |
+
"eval_samples_per_second": 20.276,
|
| 288732 |
+
"eval_steps_per_second": 0.652,
|
| 288733 |
+
"eval_wer": 0.17325978079407708,
|
| 288734 |
+
"step": 148835
|
| 288735 |
+
},
|
| 288736 |
+
{
|
| 288737 |
+
"epoch": 1190.04,
|
| 288738 |
+
"learning_rate": 7.6125848142164795e-06,
|
| 288739 |
+
"loss": 0.3049,
|
| 288740 |
+
"step": 148840
|
| 288741 |
+
},
|
| 288742 |
+
{
|
| 288743 |
+
"epoch": 1190.08,
|
| 288744 |
+
"learning_rate": 7.612504038772214e-06,
|
| 288745 |
+
"loss": 0.244,
|
| 288746 |
+
"step": 148845
|
| 288747 |
+
},
|
| 288748 |
+
{
|
| 288749 |
+
"epoch": 1190.12,
|
| 288750 |
+
"learning_rate": 7.6124232633279495e-06,
|
| 288751 |
+
"loss": 0.2753,
|
| 288752 |
+
"step": 148850
|
| 288753 |
+
},
|
| 288754 |
+
{
|
| 288755 |
+
"epoch": 1190.16,
|
| 288756 |
+
"learning_rate": 7.612342487883684e-06,
|
| 288757 |
+
"loss": 0.3841,
|
| 288758 |
+
"step": 148855
|
| 288759 |
+
},
|
| 288760 |
+
{
|
| 288761 |
+
"epoch": 1190.2,
|
| 288762 |
+
"learning_rate": 7.6122617124394195e-06,
|
| 288763 |
+
"loss": 0.9727,
|
| 288764 |
+
"step": 148860
|
| 288765 |
+
},
|
| 288766 |
+
{
|
| 288767 |
+
"epoch": 1190.24,
|
| 288768 |
+
"learning_rate": 7.612180936995154e-06,
|
| 288769 |
+
"loss": 0.3428,
|
| 288770 |
+
"step": 148865
|
| 288771 |
+
},
|
| 288772 |
+
{
|
| 288773 |
+
"epoch": 1190.28,
|
| 288774 |
+
"learning_rate": 7.6121001615508895e-06,
|
| 288775 |
+
"loss": 0.2587,
|
| 288776 |
+
"step": 148870
|
| 288777 |
+
},
|
| 288778 |
+
{
|
| 288779 |
+
"epoch": 1190.32,
|
| 288780 |
+
"learning_rate": 7.612019386106624e-06,
|
| 288781 |
+
"loss": 0.2912,
|
| 288782 |
+
"step": 148875
|
| 288783 |
+
},
|
| 288784 |
+
{
|
| 288785 |
+
"epoch": 1190.36,
|
| 288786 |
+
"learning_rate": 7.6119386106623594e-06,
|
| 288787 |
+
"loss": 0.4598,
|
| 288788 |
+
"step": 148880
|
| 288789 |
+
},
|
| 288790 |
+
{
|
| 288791 |
+
"epoch": 1190.4,
|
| 288792 |
+
"learning_rate": 7.611857835218094e-06,
|
| 288793 |
+
"loss": 1.111,
|
| 288794 |
+
"step": 148885
|
| 288795 |
+
},
|
| 288796 |
+
{
|
| 288797 |
+
"epoch": 1190.44,
|
| 288798 |
+
"learning_rate": 7.611777059773829e-06,
|
| 288799 |
+
"loss": 0.295,
|
| 288800 |
+
"step": 148890
|
| 288801 |
+
},
|
| 288802 |
+
{
|
| 288803 |
+
"epoch": 1190.48,
|
| 288804 |
+
"learning_rate": 7.611696284329565e-06,
|
| 288805 |
+
"loss": 0.2636,
|
| 288806 |
+
"step": 148895
|
| 288807 |
+
},
|
| 288808 |
+
{
|
| 288809 |
+
"epoch": 1190.52,
|
| 288810 |
+
"learning_rate": 7.611615508885299e-06,
|
| 288811 |
+
"loss": 0.3022,
|
| 288812 |
+
"step": 148900
|
| 288813 |
+
},
|
| 288814 |
+
{
|
| 288815 |
+
"epoch": 1190.56,
|
| 288816 |
+
"learning_rate": 7.611534733441035e-06,
|
| 288817 |
+
"loss": 0.4243,
|
| 288818 |
+
"step": 148905
|
| 288819 |
+
},
|
| 288820 |
+
{
|
| 288821 |
+
"epoch": 1190.6,
|
| 288822 |
+
"learning_rate": 7.611453957996769e-06,
|
| 288823 |
+
"loss": 1.1794,
|
| 288824 |
+
"step": 148910
|
| 288825 |
+
},
|
| 288826 |
+
{
|
| 288827 |
+
"epoch": 1190.64,
|
| 288828 |
+
"learning_rate": 7.611373182552505e-06,
|
| 288829 |
+
"loss": 0.3277,
|
| 288830 |
+
"step": 148915
|
| 288831 |
+
},
|
| 288832 |
+
{
|
| 288833 |
+
"epoch": 1190.68,
|
| 288834 |
+
"learning_rate": 7.611292407108239e-06,
|
| 288835 |
+
"loss": 0.2458,
|
| 288836 |
+
"step": 148920
|
| 288837 |
+
},
|
| 288838 |
+
{
|
| 288839 |
+
"epoch": 1190.72,
|
| 288840 |
+
"learning_rate": 7.611211631663975e-06,
|
| 288841 |
+
"loss": 0.3399,
|
| 288842 |
+
"step": 148925
|
| 288843 |
+
},
|
| 288844 |
+
{
|
| 288845 |
+
"epoch": 1190.76,
|
| 288846 |
+
"learning_rate": 7.611130856219709e-06,
|
| 288847 |
+
"loss": 0.4827,
|
| 288848 |
+
"step": 148930
|
| 288849 |
+
},
|
| 288850 |
+
{
|
| 288851 |
+
"epoch": 1190.8,
|
| 288852 |
+
"learning_rate": 7.611050080775445e-06,
|
| 288853 |
+
"loss": 1.2963,
|
| 288854 |
+
"step": 148935
|
| 288855 |
+
},
|
| 288856 |
+
{
|
| 288857 |
+
"epoch": 1190.84,
|
| 288858 |
+
"learning_rate": 7.610969305331179e-06,
|
| 288859 |
+
"loss": 0.3011,
|
| 288860 |
+
"step": 148940
|
| 288861 |
+
},
|
| 288862 |
+
{
|
| 288863 |
+
"epoch": 1190.88,
|
| 288864 |
+
"learning_rate": 7.610888529886915e-06,
|
| 288865 |
+
"loss": 0.2551,
|
| 288866 |
+
"step": 148945
|
| 288867 |
+
},
|
| 288868 |
+
{
|
| 288869 |
+
"epoch": 1190.92,
|
| 288870 |
+
"learning_rate": 7.610807754442649e-06,
|
| 288871 |
+
"loss": 0.2753,
|
| 288872 |
+
"step": 148950
|
| 288873 |
+
},
|
| 288874 |
+
{
|
| 288875 |
+
"epoch": 1190.96,
|
| 288876 |
+
"learning_rate": 7.610726978998385e-06,
|
| 288877 |
+
"loss": 0.409,
|
| 288878 |
+
"step": 148955
|
| 288879 |
+
},
|
| 288880 |
+
{
|
| 288881 |
+
"epoch": 1191.0,
|
| 288882 |
+
"learning_rate": 7.610646203554121e-06,
|
| 288883 |
+
"loss": 1.1083,
|
| 288884 |
+
"step": 148960
|
| 288885 |
+
},
|
| 288886 |
+
{
|
| 288887 |
+
"epoch": 1191.0,
|
| 288888 |
+
"eval_loss": 0.3409165143966675,
|
| 288889 |
+
"eval_runtime": 41.855,
|
| 288890 |
+
"eval_samples_per_second": 20.093,
|
| 288891 |
+
"eval_steps_per_second": 0.645,
|
| 288892 |
+
"eval_wer": 0.17429476108232586,
|
| 288893 |
+
"step": 148960
|
| 288894 |
}
|
| 288895 |
],
|
| 288896 |
+
"max_steps": 625000,
|
| 288897 |
"num_train_epochs": 5000,
|
| 288898 |
+
"total_flos": 4.1919118003071346e+20,
|
| 288899 |
"trial_name": null,
|
| 288900 |
"trial_params": null
|
| 288901 |
}
|
model-bin/finetune/base/{checkpoint-148340 → checkpoint-148960}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1630237299.1184735/events.out.tfevents.1630237299.cc93b136ebf5.1086.149
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:055281116f10c036024a1c591328c8686c68d52ebebf1c09411dbb557a2d3b20
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630237737.6447396/events.out.tfevents.1630237737.cc93b136ebf5.1086.151
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:68674c462a5a2d4abd34805035e7a648cfd8fb5ea855bb6a2530d65bbdc535cc
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630238176.5216558/events.out.tfevents.1630238176.cc93b136ebf5.1086.153
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:7c02f1a64f951ac3de63f2c2f42f5023db15ab13380ede957d260304a16f2a4a
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630238608.102783/events.out.tfevents.1630238608.cc93b136ebf5.1086.155
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9991109f76ad167908d19792327fcc959dd4093a639c032198f6aaceca8d89ff
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630239044.210316/events.out.tfevents.1630239044.cc93b136ebf5.1086.157
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ff0d81a3769965a1487bdcfab204f85e0435df7cfbbcb78b2944c162b0056976
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1630237299.cc93b136ebf5.1086.148
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2acd2d728b970db419ea96f68ccb0c0fc0538d7bd57261950be955e0c72dab20
|
| 3 |
+
size 8462
|
model-bin/finetune/base/log/events.out.tfevents.1630237737.cc93b136ebf5.1086.150
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b15ba1e3ce4911b844aadc4448753febcee107b492d9bed71fc8a1bf6f4883c4
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630238176.cc93b136ebf5.1086.152
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f4e38775e10091e47e180161166c1a47622a06e6349bacafb3e41b6e6813042e
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630238608.cc93b136ebf5.1086.154
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:4fde57f985d0dd5de8e7954a18587d2ea779ae74850ddeb7191395a3e81c971c
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630239044.cc93b136ebf5.1086.156
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:5d4f53e1b364a692f61dc000d469a055b64babd41998b702fadb88e5e19e7b50
|
| 3 |
+
size 8622
|