"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-129674 → checkpoint-131541}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-129674 → checkpoint-131541}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-129674 → checkpoint-131541}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-131042 → checkpoint-131541}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-131042 → checkpoint-131541}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-131042 → checkpoint-131541}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-129674 → checkpoint-131541}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-131042 → checkpoint-131541}/trainer_state.json +639 -3
- model-bin/finetune/base/{checkpoint-129674 → checkpoint-131541}/training_args.bin +0 -0
- model-bin/finetune/base/{checkpoint-131042 → checkpoint-131665}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-131042 → checkpoint-131665}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-131042 → checkpoint-131665}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-129674 → checkpoint-131665}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-129674 → checkpoint-131665}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-129674 → checkpoint-131665}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-131042 → checkpoint-131665}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-129674 → checkpoint-131665}/trainer_state.json +2544 -6
- model-bin/finetune/base/{checkpoint-131042 → checkpoint-131665}/training_args.bin +0 -0
- model-bin/finetune/base/log/1630165348.6013007/events.out.tfevents.1630165348.86bb0ddabf9b.4092.211 +3 -0
- model-bin/finetune/base/log/1630165745.8357656/events.out.tfevents.1630165745.86bb0ddabf9b.4092.213 +3 -0
- model-bin/finetune/base/log/1630166142.3384323/events.out.tfevents.1630166142.86bb0ddabf9b.4092.215 +3 -0
- model-bin/finetune/base/log/1630166536.6632655/events.out.tfevents.1630166536.86bb0ddabf9b.4092.217 +3 -0
- model-bin/finetune/base/log/1630166932.4687943/events.out.tfevents.1630166932.86bb0ddabf9b.4092.219 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630165348.86bb0ddabf9b.4092.210 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630165745.86bb0ddabf9b.4092.212 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630166142.86bb0ddabf9b.4092.214 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630166536.86bb0ddabf9b.4092.216 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630166932.86bb0ddabf9b.4092.218 +3 -0
model-bin/finetune/base/{checkpoint-129674 → checkpoint-131541}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-129674 → checkpoint-131541}/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 sha256:9cf0c9d385ed21543d9c41ae50c467369420619cf68687d5e7849331918a6f11
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-129674 → checkpoint-131541}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-131042 → checkpoint-131541}/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 sha256:2200abb9a829680ff23eb433096c3388f76ffb988a1bb4d98d5b4ec3096384f0
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-131042 → checkpoint-131541}/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:ff377add17a9a4386f8bec7acd26f62d35dc8ea8471cdd283685a9058c579ed7
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-131042 → checkpoint-131541}/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 sha256:9998597ac536ec5ae933cea30d036692a73916ca500c01c04cd2b5d575a1a9a4
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-129674 → checkpoint-131541}/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 sha256:3bc58d6750d9cb17c5164124e392911ee493c05b9cb85bf63378d08f3a7f630a
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-131042 → checkpoint-131541}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.17105075053609722,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-129674",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -266091,11 +266091,647 @@
|
|
| 266091 |
"eval_steps_per_second": 0.78,
|
| 266092 |
"eval_wer": 0.1811192429954905,
|
| 266093 |
"step": 131042
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 266094 |
}
|
| 266095 |
],
|
| 266096 |
"max_steps": 625000,
|
| 266097 |
"num_train_epochs": 5000,
|
| 266098 |
-
"total_flos": 3.
|
| 266099 |
"trial_name": null,
|
| 266100 |
"trial_params": null
|
| 266101 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.17105075053609722,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-129674",
|
| 4 |
+
"epoch": 1052.0,
|
| 5 |
+
"global_step": 131541,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 266091 |
"eval_steps_per_second": 0.78,
|
| 266092 |
"eval_wer": 0.1811192429954905,
|
| 266093 |
"step": 131042
|
| 266094 |
+
},
|
| 266095 |
+
{
|
| 266096 |
+
"epoch": 1048.02,
|
| 266097 |
+
"learning_rate": 7.916762820512821e-06,
|
| 266098 |
+
"loss": 0.3444,
|
| 266099 |
+
"step": 131045
|
| 266100 |
+
},
|
| 266101 |
+
{
|
| 266102 |
+
"epoch": 1048.06,
|
| 266103 |
+
"learning_rate": 7.916682692307693e-06,
|
| 266104 |
+
"loss": 0.2465,
|
| 266105 |
+
"step": 131050
|
| 266106 |
+
},
|
| 266107 |
+
{
|
| 266108 |
+
"epoch": 1048.1,
|
| 266109 |
+
"learning_rate": 7.916602564102564e-06,
|
| 266110 |
+
"loss": 0.2852,
|
| 266111 |
+
"step": 131055
|
| 266112 |
+
},
|
| 266113 |
+
{
|
| 266114 |
+
"epoch": 1048.14,
|
| 266115 |
+
"learning_rate": 7.916522435897437e-06,
|
| 266116 |
+
"loss": 0.3522,
|
| 266117 |
+
"step": 131060
|
| 266118 |
+
},
|
| 266119 |
+
{
|
| 266120 |
+
"epoch": 1048.18,
|
| 266121 |
+
"learning_rate": 7.916442307692308e-06,
|
| 266122 |
+
"loss": 0.7194,
|
| 266123 |
+
"step": 131065
|
| 266124 |
+
},
|
| 266125 |
+
{
|
| 266126 |
+
"epoch": 1048.22,
|
| 266127 |
+
"learning_rate": 7.91636217948718e-06,
|
| 266128 |
+
"loss": 0.9185,
|
| 266129 |
+
"step": 131070
|
| 266130 |
+
},
|
| 266131 |
+
{
|
| 266132 |
+
"epoch": 1048.26,
|
| 266133 |
+
"learning_rate": 7.916282051282051e-06,
|
| 266134 |
+
"loss": 0.2673,
|
| 266135 |
+
"step": 131075
|
| 266136 |
+
},
|
| 266137 |
+
{
|
| 266138 |
+
"epoch": 1048.3,
|
| 266139 |
+
"learning_rate": 7.916201923076924e-06,
|
| 266140 |
+
"loss": 0.3111,
|
| 266141 |
+
"step": 131080
|
| 266142 |
+
},
|
| 266143 |
+
{
|
| 266144 |
+
"epoch": 1048.34,
|
| 266145 |
+
"learning_rate": 7.916121794871796e-06,
|
| 266146 |
+
"loss": 0.3996,
|
| 266147 |
+
"step": 131085
|
| 266148 |
+
},
|
| 266149 |
+
{
|
| 266150 |
+
"epoch": 1048.38,
|
| 266151 |
+
"learning_rate": 7.916041666666667e-06,
|
| 266152 |
+
"loss": 0.781,
|
| 266153 |
+
"step": 131090
|
| 266154 |
+
},
|
| 266155 |
+
{
|
| 266156 |
+
"epoch": 1048.42,
|
| 266157 |
+
"learning_rate": 7.91596153846154e-06,
|
| 266158 |
+
"loss": 0.814,
|
| 266159 |
+
"step": 131095
|
| 266160 |
+
},
|
| 266161 |
+
{
|
| 266162 |
+
"epoch": 1048.46,
|
| 266163 |
+
"learning_rate": 7.915881410256411e-06,
|
| 266164 |
+
"loss": 0.2743,
|
| 266165 |
+
"step": 131100
|
| 266166 |
+
},
|
| 266167 |
+
{
|
| 266168 |
+
"epoch": 1048.5,
|
| 266169 |
+
"learning_rate": 7.915801282051283e-06,
|
| 266170 |
+
"loss": 0.2987,
|
| 266171 |
+
"step": 131105
|
| 266172 |
+
},
|
| 266173 |
+
{
|
| 266174 |
+
"epoch": 1048.54,
|
| 266175 |
+
"learning_rate": 7.915721153846154e-06,
|
| 266176 |
+
"loss": 0.3916,
|
| 266177 |
+
"step": 131110
|
| 266178 |
+
},
|
| 266179 |
+
{
|
| 266180 |
+
"epoch": 1048.58,
|
| 266181 |
+
"learning_rate": 7.915641025641027e-06,
|
| 266182 |
+
"loss": 0.6682,
|
| 266183 |
+
"step": 131115
|
| 266184 |
+
},
|
| 266185 |
+
{
|
| 266186 |
+
"epoch": 1048.62,
|
| 266187 |
+
"learning_rate": 7.915560897435898e-06,
|
| 266188 |
+
"loss": 0.7889,
|
| 266189 |
+
"step": 131120
|
| 266190 |
+
},
|
| 266191 |
+
{
|
| 266192 |
+
"epoch": 1048.66,
|
| 266193 |
+
"learning_rate": 7.91548076923077e-06,
|
| 266194 |
+
"loss": 0.4369,
|
| 266195 |
+
"step": 131125
|
| 266196 |
+
},
|
| 266197 |
+
{
|
| 266198 |
+
"epoch": 1048.7,
|
| 266199 |
+
"learning_rate": 7.915400641025641e-06,
|
| 266200 |
+
"loss": 0.2953,
|
| 266201 |
+
"step": 131130
|
| 266202 |
+
},
|
| 266203 |
+
{
|
| 266204 |
+
"epoch": 1048.74,
|
| 266205 |
+
"learning_rate": 7.915320512820514e-06,
|
| 266206 |
+
"loss": 0.3176,
|
| 266207 |
+
"step": 131135
|
| 266208 |
+
},
|
| 266209 |
+
{
|
| 266210 |
+
"epoch": 1048.78,
|
| 266211 |
+
"learning_rate": 7.915240384615386e-06,
|
| 266212 |
+
"loss": 0.7356,
|
| 266213 |
+
"step": 131140
|
| 266214 |
+
},
|
| 266215 |
+
{
|
| 266216 |
+
"epoch": 1048.82,
|
| 266217 |
+
"learning_rate": 7.915160256410257e-06,
|
| 266218 |
+
"loss": 1.0768,
|
| 266219 |
+
"step": 131145
|
| 266220 |
+
},
|
| 266221 |
+
{
|
| 266222 |
+
"epoch": 1048.86,
|
| 266223 |
+
"learning_rate": 7.91508012820513e-06,
|
| 266224 |
+
"loss": 0.2829,
|
| 266225 |
+
"step": 131150
|
| 266226 |
+
},
|
| 266227 |
+
{
|
| 266228 |
+
"epoch": 1048.9,
|
| 266229 |
+
"learning_rate": 7.915000000000001e-06,
|
| 266230 |
+
"loss": 0.2821,
|
| 266231 |
+
"step": 131155
|
| 266232 |
+
},
|
| 266233 |
+
{
|
| 266234 |
+
"epoch": 1048.94,
|
| 266235 |
+
"learning_rate": 7.914919871794873e-06,
|
| 266236 |
+
"loss": 0.3997,
|
| 266237 |
+
"step": 131160
|
| 266238 |
+
},
|
| 266239 |
+
{
|
| 266240 |
+
"epoch": 1048.98,
|
| 266241 |
+
"learning_rate": 7.914839743589744e-06,
|
| 266242 |
+
"loss": 0.866,
|
| 266243 |
+
"step": 131165
|
| 266244 |
+
},
|
| 266245 |
+
{
|
| 266246 |
+
"epoch": 1049.0,
|
| 266247 |
+
"eval_loss": 0.4070914089679718,
|
| 266248 |
+
"eval_runtime": 37.1506,
|
| 266249 |
+
"eval_samples_per_second": 22.503,
|
| 266250 |
+
"eval_steps_per_second": 0.727,
|
| 266251 |
+
"eval_wer": 0.18080486392588305,
|
| 266252 |
+
"step": 131167
|
| 266253 |
+
},
|
| 266254 |
+
{
|
| 266255 |
+
"epoch": 1049.02,
|
| 266256 |
+
"learning_rate": 7.914759615384617e-06,
|
| 266257 |
+
"loss": 0.3212,
|
| 266258 |
+
"step": 131170
|
| 266259 |
+
},
|
| 266260 |
+
{
|
| 266261 |
+
"epoch": 1049.06,
|
| 266262 |
+
"learning_rate": 7.914679487179487e-06,
|
| 266263 |
+
"loss": 0.266,
|
| 266264 |
+
"step": 131175
|
| 266265 |
+
},
|
| 266266 |
+
{
|
| 266267 |
+
"epoch": 1049.1,
|
| 266268 |
+
"learning_rate": 7.91459935897436e-06,
|
| 266269 |
+
"loss": 0.293,
|
| 266270 |
+
"step": 131180
|
| 266271 |
+
},
|
| 266272 |
+
{
|
| 266273 |
+
"epoch": 1049.14,
|
| 266274 |
+
"learning_rate": 7.914519230769231e-06,
|
| 266275 |
+
"loss": 0.336,
|
| 266276 |
+
"step": 131185
|
| 266277 |
+
},
|
| 266278 |
+
{
|
| 266279 |
+
"epoch": 1049.18,
|
| 266280 |
+
"learning_rate": 7.914439102564103e-06,
|
| 266281 |
+
"loss": 0.7208,
|
| 266282 |
+
"step": 131190
|
| 266283 |
+
},
|
| 266284 |
+
{
|
| 266285 |
+
"epoch": 1049.22,
|
| 266286 |
+
"learning_rate": 7.914358974358976e-06,
|
| 266287 |
+
"loss": 0.8003,
|
| 266288 |
+
"step": 131195
|
| 266289 |
+
},
|
| 266290 |
+
{
|
| 266291 |
+
"epoch": 1049.26,
|
| 266292 |
+
"learning_rate": 7.914278846153847e-06,
|
| 266293 |
+
"loss": 0.3555,
|
| 266294 |
+
"step": 131200
|
| 266295 |
+
},
|
| 266296 |
+
{
|
| 266297 |
+
"epoch": 1049.3,
|
| 266298 |
+
"learning_rate": 7.914198717948718e-06,
|
| 266299 |
+
"loss": 0.2921,
|
| 266300 |
+
"step": 131205
|
| 266301 |
+
},
|
| 266302 |
+
{
|
| 266303 |
+
"epoch": 1049.34,
|
| 266304 |
+
"learning_rate": 7.91411858974359e-06,
|
| 266305 |
+
"loss": 0.351,
|
| 266306 |
+
"step": 131210
|
| 266307 |
+
},
|
| 266308 |
+
{
|
| 266309 |
+
"epoch": 1049.38,
|
| 266310 |
+
"learning_rate": 7.914038461538463e-06,
|
| 266311 |
+
"loss": 0.7633,
|
| 266312 |
+
"step": 131215
|
| 266313 |
+
},
|
| 266314 |
+
{
|
| 266315 |
+
"epoch": 1049.42,
|
| 266316 |
+
"learning_rate": 7.913958333333334e-06,
|
| 266317 |
+
"loss": 0.7887,
|
| 266318 |
+
"step": 131220
|
| 266319 |
+
},
|
| 266320 |
+
{
|
| 266321 |
+
"epoch": 1049.46,
|
| 266322 |
+
"learning_rate": 7.913878205128205e-06,
|
| 266323 |
+
"loss": 0.3027,
|
| 266324 |
+
"step": 131225
|
| 266325 |
+
},
|
| 266326 |
+
{
|
| 266327 |
+
"epoch": 1049.5,
|
| 266328 |
+
"learning_rate": 7.913798076923077e-06,
|
| 266329 |
+
"loss": 0.3407,
|
| 266330 |
+
"step": 131230
|
| 266331 |
+
},
|
| 266332 |
+
{
|
| 266333 |
+
"epoch": 1049.54,
|
| 266334 |
+
"learning_rate": 7.91371794871795e-06,
|
| 266335 |
+
"loss": 0.3005,
|
| 266336 |
+
"step": 131235
|
| 266337 |
+
},
|
| 266338 |
+
{
|
| 266339 |
+
"epoch": 1049.58,
|
| 266340 |
+
"learning_rate": 7.913637820512821e-06,
|
| 266341 |
+
"loss": 0.6735,
|
| 266342 |
+
"step": 131240
|
| 266343 |
+
},
|
| 266344 |
+
{
|
| 266345 |
+
"epoch": 1049.62,
|
| 266346 |
+
"learning_rate": 7.913557692307693e-06,
|
| 266347 |
+
"loss": 0.9273,
|
| 266348 |
+
"step": 131245
|
| 266349 |
+
},
|
| 266350 |
+
{
|
| 266351 |
+
"epoch": 1049.66,
|
| 266352 |
+
"learning_rate": 7.913477564102566e-06,
|
| 266353 |
+
"loss": 0.3269,
|
| 266354 |
+
"step": 131250
|
| 266355 |
+
},
|
| 266356 |
+
{
|
| 266357 |
+
"epoch": 1049.7,
|
| 266358 |
+
"learning_rate": 7.913397435897437e-06,
|
| 266359 |
+
"loss": 0.2882,
|
| 266360 |
+
"step": 131255
|
| 266361 |
+
},
|
| 266362 |
+
{
|
| 266363 |
+
"epoch": 1049.74,
|
| 266364 |
+
"learning_rate": 7.913317307692308e-06,
|
| 266365 |
+
"loss": 0.3379,
|
| 266366 |
+
"step": 131260
|
| 266367 |
+
},
|
| 266368 |
+
{
|
| 266369 |
+
"epoch": 1049.78,
|
| 266370 |
+
"learning_rate": 7.91323717948718e-06,
|
| 266371 |
+
"loss": 0.7038,
|
| 266372 |
+
"step": 131265
|
| 266373 |
+
},
|
| 266374 |
+
{
|
| 266375 |
+
"epoch": 1049.82,
|
| 266376 |
+
"learning_rate": 7.913157051282053e-06,
|
| 266377 |
+
"loss": 0.8137,
|
| 266378 |
+
"step": 131270
|
| 266379 |
+
},
|
| 266380 |
+
{
|
| 266381 |
+
"epoch": 1049.86,
|
| 266382 |
+
"learning_rate": 7.913076923076924e-06,
|
| 266383 |
+
"loss": 0.284,
|
| 266384 |
+
"step": 131275
|
| 266385 |
+
},
|
| 266386 |
+
{
|
| 266387 |
+
"epoch": 1049.9,
|
| 266388 |
+
"learning_rate": 7.912996794871795e-06,
|
| 266389 |
+
"loss": 0.2658,
|
| 266390 |
+
"step": 131280
|
| 266391 |
+
},
|
| 266392 |
+
{
|
| 266393 |
+
"epoch": 1049.94,
|
| 266394 |
+
"learning_rate": 7.912916666666667e-06,
|
| 266395 |
+
"loss": 0.3415,
|
| 266396 |
+
"step": 131285
|
| 266397 |
+
},
|
| 266398 |
+
{
|
| 266399 |
+
"epoch": 1049.98,
|
| 266400 |
+
"learning_rate": 7.91283653846154e-06,
|
| 266401 |
+
"loss": 0.9049,
|
| 266402 |
+
"step": 131290
|
| 266403 |
+
},
|
| 266404 |
+
{
|
| 266405 |
+
"epoch": 1050.0,
|
| 266406 |
+
"eval_loss": 0.451054185628891,
|
| 266407 |
+
"eval_runtime": 36.5574,
|
| 266408 |
+
"eval_samples_per_second": 22.868,
|
| 266409 |
+
"eval_steps_per_second": 0.739,
|
| 266410 |
+
"eval_wer": 0.1775343674189517,
|
| 266411 |
+
"step": 131292
|
| 266412 |
+
},
|
| 266413 |
+
{
|
| 266414 |
+
"epoch": 1058.02,
|
| 266415 |
+
"learning_rate": 7.912756410256411e-06,
|
| 266416 |
+
"loss": 0.3144,
|
| 266417 |
+
"step": 131295
|
| 266418 |
+
},
|
| 266419 |
+
{
|
| 266420 |
+
"epoch": 1058.06,
|
| 266421 |
+
"learning_rate": 7.912676282051283e-06,
|
| 266422 |
+
"loss": 0.3105,
|
| 266423 |
+
"step": 131300
|
| 266424 |
+
},
|
| 266425 |
+
{
|
| 266426 |
+
"epoch": 1058.1,
|
| 266427 |
+
"learning_rate": 7.912596153846156e-06,
|
| 266428 |
+
"loss": 0.3349,
|
| 266429 |
+
"step": 131305
|
| 266430 |
+
},
|
| 266431 |
+
{
|
| 266432 |
+
"epoch": 1058.15,
|
| 266433 |
+
"learning_rate": 7.912516025641025e-06,
|
| 266434 |
+
"loss": 0.3023,
|
| 266435 |
+
"step": 131310
|
| 266436 |
+
},
|
| 266437 |
+
{
|
| 266438 |
+
"epoch": 1058.19,
|
| 266439 |
+
"learning_rate": 7.912435897435898e-06,
|
| 266440 |
+
"loss": 0.6984,
|
| 266441 |
+
"step": 131315
|
| 266442 |
+
},
|
| 266443 |
+
{
|
| 266444 |
+
"epoch": 1058.23,
|
| 266445 |
+
"learning_rate": 7.91235576923077e-06,
|
| 266446 |
+
"loss": 1.0068,
|
| 266447 |
+
"step": 131320
|
| 266448 |
+
},
|
| 266449 |
+
{
|
| 266450 |
+
"epoch": 1058.27,
|
| 266451 |
+
"learning_rate": 7.912275641025641e-06,
|
| 266452 |
+
"loss": 0.293,
|
| 266453 |
+
"step": 131325
|
| 266454 |
+
},
|
| 266455 |
+
{
|
| 266456 |
+
"epoch": 1058.31,
|
| 266457 |
+
"learning_rate": 7.912195512820512e-06,
|
| 266458 |
+
"loss": 0.3049,
|
| 266459 |
+
"step": 131330
|
| 266460 |
+
},
|
| 266461 |
+
{
|
| 266462 |
+
"epoch": 1058.35,
|
| 266463 |
+
"learning_rate": 7.912115384615385e-06,
|
| 266464 |
+
"loss": 0.3829,
|
| 266465 |
+
"step": 131335
|
| 266466 |
+
},
|
| 266467 |
+
{
|
| 266468 |
+
"epoch": 1058.39,
|
| 266469 |
+
"learning_rate": 7.912035256410257e-06,
|
| 266470 |
+
"loss": 0.6978,
|
| 266471 |
+
"step": 131340
|
| 266472 |
+
},
|
| 266473 |
+
{
|
| 266474 |
+
"epoch": 1058.43,
|
| 266475 |
+
"learning_rate": 7.911955128205128e-06,
|
| 266476 |
+
"loss": 0.7856,
|
| 266477 |
+
"step": 131345
|
| 266478 |
+
},
|
| 266479 |
+
{
|
| 266480 |
+
"epoch": 1058.47,
|
| 266481 |
+
"learning_rate": 7.911875000000001e-06,
|
| 266482 |
+
"loss": 0.2529,
|
| 266483 |
+
"step": 131350
|
| 266484 |
+
},
|
| 266485 |
+
{
|
| 266486 |
+
"epoch": 1058.51,
|
| 266487 |
+
"learning_rate": 7.911794871794873e-06,
|
| 266488 |
+
"loss": 0.364,
|
| 266489 |
+
"step": 131355
|
| 266490 |
+
},
|
| 266491 |
+
{
|
| 266492 |
+
"epoch": 1058.55,
|
| 266493 |
+
"learning_rate": 7.911714743589744e-06,
|
| 266494 |
+
"loss": 0.4424,
|
| 266495 |
+
"step": 131360
|
| 266496 |
+
},
|
| 266497 |
+
{
|
| 266498 |
+
"epoch": 1058.59,
|
| 266499 |
+
"learning_rate": 7.911634615384615e-06,
|
| 266500 |
+
"loss": 0.8688,
|
| 266501 |
+
"step": 131365
|
| 266502 |
+
},
|
| 266503 |
+
{
|
| 266504 |
+
"epoch": 1058.63,
|
| 266505 |
+
"learning_rate": 7.911554487179488e-06,
|
| 266506 |
+
"loss": 0.8466,
|
| 266507 |
+
"step": 131370
|
| 266508 |
+
},
|
| 266509 |
+
{
|
| 266510 |
+
"epoch": 1058.67,
|
| 266511 |
+
"learning_rate": 7.91147435897436e-06,
|
| 266512 |
+
"loss": 0.2929,
|
| 266513 |
+
"step": 131375
|
| 266514 |
+
},
|
| 266515 |
+
{
|
| 266516 |
+
"epoch": 1058.71,
|
| 266517 |
+
"learning_rate": 7.911394230769231e-06,
|
| 266518 |
+
"loss": 0.2828,
|
| 266519 |
+
"step": 131380
|
| 266520 |
+
},
|
| 266521 |
+
{
|
| 266522 |
+
"epoch": 1058.75,
|
| 266523 |
+
"learning_rate": 7.911314102564102e-06,
|
| 266524 |
+
"loss": 0.3502,
|
| 266525 |
+
"step": 131385
|
| 266526 |
+
},
|
| 266527 |
+
{
|
| 266528 |
+
"epoch": 1058.79,
|
| 266529 |
+
"learning_rate": 7.911233974358975e-06,
|
| 266530 |
+
"loss": 0.6644,
|
| 266531 |
+
"step": 131390
|
| 266532 |
+
},
|
| 266533 |
+
{
|
| 266534 |
+
"epoch": 1058.83,
|
| 266535 |
+
"learning_rate": 7.911153846153847e-06,
|
| 266536 |
+
"loss": 0.9036,
|
| 266537 |
+
"step": 131395
|
| 266538 |
+
},
|
| 266539 |
+
{
|
| 266540 |
+
"epoch": 1058.87,
|
| 266541 |
+
"learning_rate": 7.911073717948718e-06,
|
| 266542 |
+
"loss": 0.2584,
|
| 266543 |
+
"step": 131400
|
| 266544 |
+
},
|
| 266545 |
+
{
|
| 266546 |
+
"epoch": 1058.91,
|
| 266547 |
+
"learning_rate": 7.910993589743591e-06,
|
| 266548 |
+
"loss": 0.2873,
|
| 266549 |
+
"step": 131405
|
| 266550 |
+
},
|
| 266551 |
+
{
|
| 266552 |
+
"epoch": 1058.95,
|
| 266553 |
+
"learning_rate": 7.910913461538463e-06,
|
| 266554 |
+
"loss": 0.3913,
|
| 266555 |
+
"step": 131410
|
| 266556 |
+
},
|
| 266557 |
+
{
|
| 266558 |
+
"epoch": 1058.99,
|
| 266559 |
+
"learning_rate": 7.910833333333334e-06,
|
| 266560 |
+
"loss": 0.8848,
|
| 266561 |
+
"step": 131415
|
| 266562 |
+
},
|
| 266563 |
+
{
|
| 266564 |
+
"epoch": 1059.0,
|
| 266565 |
+
"eval_loss": 0.34740835428237915,
|
| 266566 |
+
"eval_runtime": 36.7055,
|
| 266567 |
+
"eval_samples_per_second": 22.776,
|
| 266568 |
+
"eval_steps_per_second": 0.736,
|
| 266569 |
+
"eval_wer": 0.18187225816045802,
|
| 266570 |
+
"step": 131416
|
| 266571 |
+
},
|
| 266572 |
+
{
|
| 266573 |
+
"epoch": 1051.03,
|
| 266574 |
+
"learning_rate": 7.910753205128205e-06,
|
| 266575 |
+
"loss": 0.2529,
|
| 266576 |
+
"step": 131420
|
| 266577 |
+
},
|
| 266578 |
+
{
|
| 266579 |
+
"epoch": 1051.07,
|
| 266580 |
+
"learning_rate": 7.910673076923078e-06,
|
| 266581 |
+
"loss": 0.311,
|
| 266582 |
+
"step": 131425
|
| 266583 |
+
},
|
| 266584 |
+
{
|
| 266585 |
+
"epoch": 1051.11,
|
| 266586 |
+
"learning_rate": 7.91059294871795e-06,
|
| 266587 |
+
"loss": 0.3139,
|
| 266588 |
+
"step": 131430
|
| 266589 |
+
},
|
| 266590 |
+
{
|
| 266591 |
+
"epoch": 1051.15,
|
| 266592 |
+
"learning_rate": 7.910512820512821e-06,
|
| 266593 |
+
"loss": 0.5022,
|
| 266594 |
+
"step": 131435
|
| 266595 |
+
},
|
| 266596 |
+
{
|
| 266597 |
+
"epoch": 1051.19,
|
| 266598 |
+
"learning_rate": 7.910432692307694e-06,
|
| 266599 |
+
"loss": 0.8994,
|
| 266600 |
+
"step": 131440
|
| 266601 |
+
},
|
| 266602 |
+
{
|
| 266603 |
+
"epoch": 1051.23,
|
| 266604 |
+
"learning_rate": 7.910352564102566e-06,
|
| 266605 |
+
"loss": 0.6591,
|
| 266606 |
+
"step": 131445
|
| 266607 |
+
},
|
| 266608 |
+
{
|
| 266609 |
+
"epoch": 1051.27,
|
| 266610 |
+
"learning_rate": 7.910272435897437e-06,
|
| 266611 |
+
"loss": 0.2897,
|
| 266612 |
+
"step": 131450
|
| 266613 |
+
},
|
| 266614 |
+
{
|
| 266615 |
+
"epoch": 1051.31,
|
| 266616 |
+
"learning_rate": 7.910192307692308e-06,
|
| 266617 |
+
"loss": 0.344,
|
| 266618 |
+
"step": 131455
|
| 266619 |
+
},
|
| 266620 |
+
{
|
| 266621 |
+
"epoch": 1051.35,
|
| 266622 |
+
"learning_rate": 7.910112179487181e-06,
|
| 266623 |
+
"loss": 0.4109,
|
| 266624 |
+
"step": 131460
|
| 266625 |
+
},
|
| 266626 |
+
{
|
| 266627 |
+
"epoch": 1051.39,
|
| 266628 |
+
"learning_rate": 7.910032051282051e-06,
|
| 266629 |
+
"loss": 0.8273,
|
| 266630 |
+
"step": 131465
|
| 266631 |
+
},
|
| 266632 |
+
{
|
| 266633 |
+
"epoch": 1051.43,
|
| 266634 |
+
"learning_rate": 7.909951923076924e-06,
|
| 266635 |
+
"loss": 0.6015,
|
| 266636 |
+
"step": 131470
|
| 266637 |
+
},
|
| 266638 |
+
{
|
| 266639 |
+
"epoch": 1051.47,
|
| 266640 |
+
"learning_rate": 7.909871794871795e-06,
|
| 266641 |
+
"loss": 0.2959,
|
| 266642 |
+
"step": 131475
|
| 266643 |
+
},
|
| 266644 |
+
{
|
| 266645 |
+
"epoch": 1051.51,
|
| 266646 |
+
"learning_rate": 7.909791666666667e-06,
|
| 266647 |
+
"loss": 0.2969,
|
| 266648 |
+
"step": 131480
|
| 266649 |
+
},
|
| 266650 |
+
{
|
| 266651 |
+
"epoch": 1051.55,
|
| 266652 |
+
"learning_rate": 7.909711538461538e-06,
|
| 266653 |
+
"loss": 0.3375,
|
| 266654 |
+
"step": 131485
|
| 266655 |
+
},
|
| 266656 |
+
{
|
| 266657 |
+
"epoch": 1051.59,
|
| 266658 |
+
"learning_rate": 7.909631410256411e-06,
|
| 266659 |
+
"loss": 1.0552,
|
| 266660 |
+
"step": 131490
|
| 266661 |
+
},
|
| 266662 |
+
{
|
| 266663 |
+
"epoch": 1051.63,
|
| 266664 |
+
"learning_rate": 7.909551282051282e-06,
|
| 266665 |
+
"loss": 0.6516,
|
| 266666 |
+
"step": 131495
|
| 266667 |
+
},
|
| 266668 |
+
{
|
| 266669 |
+
"epoch": 1051.67,
|
| 266670 |
+
"learning_rate": 7.909471153846154e-06,
|
| 266671 |
+
"loss": 0.2981,
|
| 266672 |
+
"step": 131500
|
| 266673 |
+
},
|
| 266674 |
+
{
|
| 266675 |
+
"epoch": 1051.71,
|
| 266676 |
+
"learning_rate": 7.909391025641027e-06,
|
| 266677 |
+
"loss": 0.3007,
|
| 266678 |
+
"step": 131505
|
| 266679 |
+
},
|
| 266680 |
+
{
|
| 266681 |
+
"epoch": 1051.75,
|
| 266682 |
+
"learning_rate": 7.909310897435898e-06,
|
| 266683 |
+
"loss": 0.3721,
|
| 266684 |
+
"step": 131510
|
| 266685 |
+
},
|
| 266686 |
+
{
|
| 266687 |
+
"epoch": 1051.79,
|
| 266688 |
+
"learning_rate": 7.90923076923077e-06,
|
| 266689 |
+
"loss": 1.0284,
|
| 266690 |
+
"step": 131515
|
| 266691 |
+
},
|
| 266692 |
+
{
|
| 266693 |
+
"epoch": 1051.83,
|
| 266694 |
+
"learning_rate": 7.909150641025641e-06,
|
| 266695 |
+
"loss": 0.667,
|
| 266696 |
+
"step": 131520
|
| 266697 |
+
},
|
| 266698 |
+
{
|
| 266699 |
+
"epoch": 1051.87,
|
| 266700 |
+
"learning_rate": 7.909070512820514e-06,
|
| 266701 |
+
"loss": 0.2979,
|
| 266702 |
+
"step": 131525
|
| 266703 |
+
},
|
| 266704 |
+
{
|
| 266705 |
+
"epoch": 1051.91,
|
| 266706 |
+
"learning_rate": 7.908990384615385e-06,
|
| 266707 |
+
"loss": 0.293,
|
| 266708 |
+
"step": 131530
|
| 266709 |
+
},
|
| 266710 |
+
{
|
| 266711 |
+
"epoch": 1051.95,
|
| 266712 |
+
"learning_rate": 7.908910256410257e-06,
|
| 266713 |
+
"loss": 0.363,
|
| 266714 |
+
"step": 131535
|
| 266715 |
+
},
|
| 266716 |
+
{
|
| 266717 |
+
"epoch": 1051.99,
|
| 266718 |
+
"learning_rate": 7.908830128205128e-06,
|
| 266719 |
+
"loss": 0.9717,
|
| 266720 |
+
"step": 131540
|
| 266721 |
+
},
|
| 266722 |
+
{
|
| 266723 |
+
"epoch": 1052.0,
|
| 266724 |
+
"eval_loss": 0.3263888657093048,
|
| 266725 |
+
"eval_runtime": 36.9741,
|
| 266726 |
+
"eval_samples_per_second": 22.61,
|
| 266727 |
+
"eval_steps_per_second": 0.73,
|
| 266728 |
+
"eval_wer": 0.17733196661297407,
|
| 266729 |
+
"step": 131541
|
| 266730 |
}
|
| 266731 |
],
|
| 266732 |
"max_steps": 625000,
|
| 266733 |
"num_train_epochs": 5000,
|
| 266734 |
+
"total_flos": 3.701572850885168e+20,
|
| 266735 |
"trial_name": null,
|
| 266736 |
"trial_params": null
|
| 266737 |
}
|
model-bin/finetune/base/{checkpoint-129674 → checkpoint-131541}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-131042 → checkpoint-131665}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-131042 → checkpoint-131665}/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 sha256:563c6b73d77b4aca81b4d4ab9309ae49505a604a4bda74246659459959465b5d
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-131042 → checkpoint-131665}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-129674 → checkpoint-131665}/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 sha256:6fc661a86b40a48188de5ed83ea26c887125bf64091bbc59fb1a809f0c8f77d7
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-129674 → checkpoint-131665}/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 sha256:57d7c93039daa7ea47bac7b869a806b25ddf62153b3388d4c8cfdc851239a1bf
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-129674 → checkpoint-131665}/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 sha256:1ae34bb3b632ce84a747aa987dc0f72facb05243dbdbd4095b0555d0a6525518
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-131042 → checkpoint-131665}/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 sha256:854d93fca71c81e7e5a185a2fb60f5c42ed71e34882055e693595551e545e63a
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-129674 → checkpoint-131665}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
-
"best_metric": 0.
|
| 3 |
-
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -264348,11 +264348,2549 @@
|
|
| 264348 |
"eval_steps_per_second": 0.726,
|
| 264349 |
"eval_wer": 0.17105075053609722,
|
| 264350 |
"step": 129674
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 264351 |
}
|
| 264352 |
],
|
| 264353 |
-
"max_steps":
|
| 264354 |
"num_train_epochs": 5000,
|
| 264355 |
-
"total_flos": 3.
|
| 264356 |
"trial_name": null,
|
| 264357 |
"trial_params": null
|
| 264358 |
}
|
|
|
|
| 1 |
{
|
| 2 |
+
"best_metric": 0.17062187276626162,
|
| 3 |
+
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-131665",
|
| 4 |
+
"epoch": 1060.995983935743,
|
| 5 |
+
"global_step": 131665,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 264348 |
"eval_steps_per_second": 0.726,
|
| 264349 |
"eval_wer": 0.17105075053609722,
|
| 264350 |
"step": 129674
|
| 264351 |
+
},
|
| 264352 |
+
{
|
| 264353 |
+
"epoch": 1037.01,
|
| 264354 |
+
"learning_rate": 7.938717948717949e-06,
|
| 264355 |
+
"loss": 0.4702,
|
| 264356 |
+
"step": 129675
|
| 264357 |
+
},
|
| 264358 |
+
{
|
| 264359 |
+
"epoch": 1037.05,
|
| 264360 |
+
"learning_rate": 7.938637820512822e-06,
|
| 264361 |
+
"loss": 0.2881,
|
| 264362 |
+
"step": 129680
|
| 264363 |
+
},
|
| 264364 |
+
{
|
| 264365 |
+
"epoch": 1037.09,
|
| 264366 |
+
"learning_rate": 7.938557692307693e-06,
|
| 264367 |
+
"loss": 0.3175,
|
| 264368 |
+
"step": 129685
|
| 264369 |
+
},
|
| 264370 |
+
{
|
| 264371 |
+
"epoch": 1037.13,
|
| 264372 |
+
"learning_rate": 7.938477564102564e-06,
|
| 264373 |
+
"loss": 0.3449,
|
| 264374 |
+
"step": 129690
|
| 264375 |
+
},
|
| 264376 |
+
{
|
| 264377 |
+
"epoch": 1037.17,
|
| 264378 |
+
"learning_rate": 7.938397435897438e-06,
|
| 264379 |
+
"loss": 0.5055,
|
| 264380 |
+
"step": 129695
|
| 264381 |
+
},
|
| 264382 |
+
{
|
| 264383 |
+
"epoch": 1037.21,
|
| 264384 |
+
"learning_rate": 7.938317307692309e-06,
|
| 264385 |
+
"loss": 1.1228,
|
| 264386 |
+
"step": 129700
|
| 264387 |
+
},
|
| 264388 |
+
{
|
| 264389 |
+
"epoch": 1037.25,
|
| 264390 |
+
"learning_rate": 7.93823717948718e-06,
|
| 264391 |
+
"loss": 0.3251,
|
| 264392 |
+
"step": 129705
|
| 264393 |
+
},
|
| 264394 |
+
{
|
| 264395 |
+
"epoch": 1037.29,
|
| 264396 |
+
"learning_rate": 7.938157051282052e-06,
|
| 264397 |
+
"loss": 0.2911,
|
| 264398 |
+
"step": 129710
|
| 264399 |
+
},
|
| 264400 |
+
{
|
| 264401 |
+
"epoch": 1037.33,
|
| 264402 |
+
"learning_rate": 7.938076923076925e-06,
|
| 264403 |
+
"loss": 0.2688,
|
| 264404 |
+
"step": 129715
|
| 264405 |
+
},
|
| 264406 |
+
{
|
| 264407 |
+
"epoch": 1037.37,
|
| 264408 |
+
"learning_rate": 7.937996794871794e-06,
|
| 264409 |
+
"loss": 0.5002,
|
| 264410 |
+
"step": 129720
|
| 264411 |
+
},
|
| 264412 |
+
{
|
| 264413 |
+
"epoch": 1037.41,
|
| 264414 |
+
"learning_rate": 7.937916666666667e-06,
|
| 264415 |
+
"loss": 1.0417,
|
| 264416 |
+
"step": 129725
|
| 264417 |
+
},
|
| 264418 |
+
{
|
| 264419 |
+
"epoch": 1037.45,
|
| 264420 |
+
"learning_rate": 7.937836538461539e-06,
|
| 264421 |
+
"loss": 0.3195,
|
| 264422 |
+
"step": 129730
|
| 264423 |
+
},
|
| 264424 |
+
{
|
| 264425 |
+
"epoch": 1037.49,
|
| 264426 |
+
"learning_rate": 7.93775641025641e-06,
|
| 264427 |
+
"loss": 0.3281,
|
| 264428 |
+
"step": 129735
|
| 264429 |
+
},
|
| 264430 |
+
{
|
| 264431 |
+
"epoch": 1037.53,
|
| 264432 |
+
"learning_rate": 7.937676282051283e-06,
|
| 264433 |
+
"loss": 0.3051,
|
| 264434 |
+
"step": 129740
|
| 264435 |
+
},
|
| 264436 |
+
{
|
| 264437 |
+
"epoch": 1037.57,
|
| 264438 |
+
"learning_rate": 7.937596153846155e-06,
|
| 264439 |
+
"loss": 0.558,
|
| 264440 |
+
"step": 129745
|
| 264441 |
+
},
|
| 264442 |
+
{
|
| 264443 |
+
"epoch": 1037.61,
|
| 264444 |
+
"learning_rate": 7.937516025641026e-06,
|
| 264445 |
+
"loss": 1.0366,
|
| 264446 |
+
"step": 129750
|
| 264447 |
+
},
|
| 264448 |
+
{
|
| 264449 |
+
"epoch": 1037.65,
|
| 264450 |
+
"learning_rate": 7.937435897435897e-06,
|
| 264451 |
+
"loss": 0.268,
|
| 264452 |
+
"step": 129755
|
| 264453 |
+
},
|
| 264454 |
+
{
|
| 264455 |
+
"epoch": 1037.69,
|
| 264456 |
+
"learning_rate": 7.93735576923077e-06,
|
| 264457 |
+
"loss": 0.3148,
|
| 264458 |
+
"step": 129760
|
| 264459 |
+
},
|
| 264460 |
+
{
|
| 264461 |
+
"epoch": 1037.73,
|
| 264462 |
+
"learning_rate": 7.937275641025642e-06,
|
| 264463 |
+
"loss": 0.3041,
|
| 264464 |
+
"step": 129765
|
| 264465 |
+
},
|
| 264466 |
+
{
|
| 264467 |
+
"epoch": 1037.77,
|
| 264468 |
+
"learning_rate": 7.937195512820513e-06,
|
| 264469 |
+
"loss": 0.4579,
|
| 264470 |
+
"step": 129770
|
| 264471 |
+
},
|
| 264472 |
+
{
|
| 264473 |
+
"epoch": 1037.81,
|
| 264474 |
+
"learning_rate": 7.937115384615384e-06,
|
| 264475 |
+
"loss": 1.4033,
|
| 264476 |
+
"step": 129775
|
| 264477 |
+
},
|
| 264478 |
+
{
|
| 264479 |
+
"epoch": 1037.85,
|
| 264480 |
+
"learning_rate": 7.937035256410257e-06,
|
| 264481 |
+
"loss": 0.3017,
|
| 264482 |
+
"step": 129780
|
| 264483 |
+
},
|
| 264484 |
+
{
|
| 264485 |
+
"epoch": 1037.89,
|
| 264486 |
+
"learning_rate": 7.936955128205129e-06,
|
| 264487 |
+
"loss": 0.278,
|
| 264488 |
+
"step": 129785
|
| 264489 |
+
},
|
| 264490 |
+
{
|
| 264491 |
+
"epoch": 1037.93,
|
| 264492 |
+
"learning_rate": 7.936875e-06,
|
| 264493 |
+
"loss": 0.2942,
|
| 264494 |
+
"step": 129790
|
| 264495 |
+
},
|
| 264496 |
+
{
|
| 264497 |
+
"epoch": 1037.97,
|
| 264498 |
+
"learning_rate": 7.936794871794873e-06,
|
| 264499 |
+
"loss": 0.5666,
|
| 264500 |
+
"step": 129795
|
| 264501 |
+
},
|
| 264502 |
+
{
|
| 264503 |
+
"epoch": 1038.0,
|
| 264504 |
+
"eval_loss": 0.3556685149669647,
|
| 264505 |
+
"eval_runtime": 33.9934,
|
| 264506 |
+
"eval_samples_per_second": 24.593,
|
| 264507 |
+
"eval_steps_per_second": 0.794,
|
| 264508 |
+
"eval_wer": 0.18174261316405954,
|
| 264509 |
+
"step": 129799
|
| 264510 |
+
},
|
| 264511 |
+
{
|
| 264512 |
+
"epoch": 1046.01,
|
| 264513 |
+
"learning_rate": 7.936714743589745e-06,
|
| 264514 |
+
"loss": 0.4129,
|
| 264515 |
+
"step": 129800
|
| 264516 |
+
},
|
| 264517 |
+
{
|
| 264518 |
+
"epoch": 1046.05,
|
| 264519 |
+
"learning_rate": 7.936634615384616e-06,
|
| 264520 |
+
"loss": 0.2893,
|
| 264521 |
+
"step": 129805
|
| 264522 |
+
},
|
| 264523 |
+
{
|
| 264524 |
+
"epoch": 1046.09,
|
| 264525 |
+
"learning_rate": 7.936554487179487e-06,
|
| 264526 |
+
"loss": 0.2299,
|
| 264527 |
+
"step": 129810
|
| 264528 |
+
},
|
| 264529 |
+
{
|
| 264530 |
+
"epoch": 1046.13,
|
| 264531 |
+
"learning_rate": 7.93647435897436e-06,
|
| 264532 |
+
"loss": 0.2756,
|
| 264533 |
+
"step": 129815
|
| 264534 |
+
},
|
| 264535 |
+
{
|
| 264536 |
+
"epoch": 1046.17,
|
| 264537 |
+
"learning_rate": 7.936394230769232e-06,
|
| 264538 |
+
"loss": 0.4417,
|
| 264539 |
+
"step": 129820
|
| 264540 |
+
},
|
| 264541 |
+
{
|
| 264542 |
+
"epoch": 1046.21,
|
| 264543 |
+
"learning_rate": 7.936314102564103e-06,
|
| 264544 |
+
"loss": 1.0294,
|
| 264545 |
+
"step": 129825
|
| 264546 |
+
},
|
| 264547 |
+
{
|
| 264548 |
+
"epoch": 1046.25,
|
| 264549 |
+
"learning_rate": 7.936233974358974e-06,
|
| 264550 |
+
"loss": 0.2665,
|
| 264551 |
+
"step": 129830
|
| 264552 |
+
},
|
| 264553 |
+
{
|
| 264554 |
+
"epoch": 1046.29,
|
| 264555 |
+
"learning_rate": 7.936153846153847e-06,
|
| 264556 |
+
"loss": 0.3326,
|
| 264557 |
+
"step": 129835
|
| 264558 |
+
},
|
| 264559 |
+
{
|
| 264560 |
+
"epoch": 1046.33,
|
| 264561 |
+
"learning_rate": 7.936073717948719e-06,
|
| 264562 |
+
"loss": 0.3502,
|
| 264563 |
+
"step": 129840
|
| 264564 |
+
},
|
| 264565 |
+
{
|
| 264566 |
+
"epoch": 1046.37,
|
| 264567 |
+
"learning_rate": 7.93599358974359e-06,
|
| 264568 |
+
"loss": 0.5617,
|
| 264569 |
+
"step": 129845
|
| 264570 |
+
},
|
| 264571 |
+
{
|
| 264572 |
+
"epoch": 1046.41,
|
| 264573 |
+
"learning_rate": 7.935913461538463e-06,
|
| 264574 |
+
"loss": 1.1194,
|
| 264575 |
+
"step": 129850
|
| 264576 |
+
},
|
| 264577 |
+
{
|
| 264578 |
+
"epoch": 1046.45,
|
| 264579 |
+
"learning_rate": 7.935833333333333e-06,
|
| 264580 |
+
"loss": 0.3628,
|
| 264581 |
+
"step": 129855
|
| 264582 |
+
},
|
| 264583 |
+
{
|
| 264584 |
+
"epoch": 1046.49,
|
| 264585 |
+
"learning_rate": 7.935753205128206e-06,
|
| 264586 |
+
"loss": 0.2528,
|
| 264587 |
+
"step": 129860
|
| 264588 |
+
},
|
| 264589 |
+
{
|
| 264590 |
+
"epoch": 1046.53,
|
| 264591 |
+
"learning_rate": 7.935673076923077e-06,
|
| 264592 |
+
"loss": 0.3784,
|
| 264593 |
+
"step": 129865
|
| 264594 |
+
},
|
| 264595 |
+
{
|
| 264596 |
+
"epoch": 1046.57,
|
| 264597 |
+
"learning_rate": 7.935592948717949e-06,
|
| 264598 |
+
"loss": 0.478,
|
| 264599 |
+
"step": 129870
|
| 264600 |
+
},
|
| 264601 |
+
{
|
| 264602 |
+
"epoch": 1046.61,
|
| 264603 |
+
"learning_rate": 7.93551282051282e-06,
|
| 264604 |
+
"loss": 1.186,
|
| 264605 |
+
"step": 129875
|
| 264606 |
+
},
|
| 264607 |
+
{
|
| 264608 |
+
"epoch": 1046.65,
|
| 264609 |
+
"learning_rate": 7.935432692307693e-06,
|
| 264610 |
+
"loss": 0.3365,
|
| 264611 |
+
"step": 129880
|
| 264612 |
+
},
|
| 264613 |
+
{
|
| 264614 |
+
"epoch": 1046.69,
|
| 264615 |
+
"learning_rate": 7.935352564102566e-06,
|
| 264616 |
+
"loss": 0.3178,
|
| 264617 |
+
"step": 129885
|
| 264618 |
+
},
|
| 264619 |
+
{
|
| 264620 |
+
"epoch": 1046.73,
|
| 264621 |
+
"learning_rate": 7.935272435897436e-06,
|
| 264622 |
+
"loss": 0.4071,
|
| 264623 |
+
"step": 129890
|
| 264624 |
+
},
|
| 264625 |
+
{
|
| 264626 |
+
"epoch": 1046.77,
|
| 264627 |
+
"learning_rate": 7.935192307692309e-06,
|
| 264628 |
+
"loss": 0.4907,
|
| 264629 |
+
"step": 129895
|
| 264630 |
+
},
|
| 264631 |
+
{
|
| 264632 |
+
"epoch": 1046.81,
|
| 264633 |
+
"learning_rate": 7.93511217948718e-06,
|
| 264634 |
+
"loss": 0.9865,
|
| 264635 |
+
"step": 129900
|
| 264636 |
+
},
|
| 264637 |
+
{
|
| 264638 |
+
"epoch": 1046.85,
|
| 264639 |
+
"learning_rate": 7.935032051282052e-06,
|
| 264640 |
+
"loss": 0.2768,
|
| 264641 |
+
"step": 129905
|
| 264642 |
+
},
|
| 264643 |
+
{
|
| 264644 |
+
"epoch": 1046.89,
|
| 264645 |
+
"learning_rate": 7.934951923076923e-06,
|
| 264646 |
+
"loss": 0.2493,
|
| 264647 |
+
"step": 129910
|
| 264648 |
+
},
|
| 264649 |
+
{
|
| 264650 |
+
"epoch": 1046.93,
|
| 264651 |
+
"learning_rate": 7.934871794871796e-06,
|
| 264652 |
+
"loss": 0.2988,
|
| 264653 |
+
"step": 129915
|
| 264654 |
+
},
|
| 264655 |
+
{
|
| 264656 |
+
"epoch": 1046.97,
|
| 264657 |
+
"learning_rate": 7.934791666666667e-06,
|
| 264658 |
+
"loss": 0.5019,
|
| 264659 |
+
"step": 129920
|
| 264660 |
+
},
|
| 264661 |
+
{
|
| 264662 |
+
"epoch": 1047.0,
|
| 264663 |
+
"eval_loss": 0.38806846737861633,
|
| 264664 |
+
"eval_runtime": 35.8177,
|
| 264665 |
+
"eval_samples_per_second": 23.313,
|
| 264666 |
+
"eval_steps_per_second": 0.754,
|
| 264667 |
+
"eval_wer": 0.17985185185185185,
|
| 264668 |
+
"step": 129923
|
| 264669 |
+
},
|
| 264670 |
+
{
|
| 264671 |
+
"epoch": 1047.02,
|
| 264672 |
+
"learning_rate": 7.934711538461539e-06,
|
| 264673 |
+
"loss": 0.3428,
|
| 264674 |
+
"step": 129925
|
| 264675 |
+
},
|
| 264676 |
+
{
|
| 264677 |
+
"epoch": 1047.06,
|
| 264678 |
+
"learning_rate": 7.93463141025641e-06,
|
| 264679 |
+
"loss": 0.3148,
|
| 264680 |
+
"step": 129930
|
| 264681 |
+
},
|
| 264682 |
+
{
|
| 264683 |
+
"epoch": 1047.1,
|
| 264684 |
+
"learning_rate": 7.934551282051283e-06,
|
| 264685 |
+
"loss": 0.341,
|
| 264686 |
+
"step": 129935
|
| 264687 |
+
},
|
| 264688 |
+
{
|
| 264689 |
+
"epoch": 1047.14,
|
| 264690 |
+
"learning_rate": 7.934471153846154e-06,
|
| 264691 |
+
"loss": 0.2991,
|
| 264692 |
+
"step": 129940
|
| 264693 |
+
},
|
| 264694 |
+
{
|
| 264695 |
+
"epoch": 1047.18,
|
| 264696 |
+
"learning_rate": 7.934391025641026e-06,
|
| 264697 |
+
"loss": 0.7212,
|
| 264698 |
+
"step": 129945
|
| 264699 |
+
},
|
| 264700 |
+
{
|
| 264701 |
+
"epoch": 1047.22,
|
| 264702 |
+
"learning_rate": 7.934310897435899e-06,
|
| 264703 |
+
"loss": 1.0056,
|
| 264704 |
+
"step": 129950
|
| 264705 |
+
},
|
| 264706 |
+
{
|
| 264707 |
+
"epoch": 1047.26,
|
| 264708 |
+
"learning_rate": 7.93423076923077e-06,
|
| 264709 |
+
"loss": 0.2669,
|
| 264710 |
+
"step": 129955
|
| 264711 |
+
},
|
| 264712 |
+
{
|
| 264713 |
+
"epoch": 1047.3,
|
| 264714 |
+
"learning_rate": 7.934150641025642e-06,
|
| 264715 |
+
"loss": 0.2889,
|
| 264716 |
+
"step": 129960
|
| 264717 |
+
},
|
| 264718 |
+
{
|
| 264719 |
+
"epoch": 1047.34,
|
| 264720 |
+
"learning_rate": 7.934070512820513e-06,
|
| 264721 |
+
"loss": 0.3207,
|
| 264722 |
+
"step": 129965
|
| 264723 |
+
},
|
| 264724 |
+
{
|
| 264725 |
+
"epoch": 1047.38,
|
| 264726 |
+
"learning_rate": 7.933990384615386e-06,
|
| 264727 |
+
"loss": 0.7011,
|
| 264728 |
+
"step": 129970
|
| 264729 |
+
},
|
| 264730 |
+
{
|
| 264731 |
+
"epoch": 1047.42,
|
| 264732 |
+
"learning_rate": 7.933910256410257e-06,
|
| 264733 |
+
"loss": 0.9856,
|
| 264734 |
+
"step": 129975
|
| 264735 |
+
},
|
| 264736 |
+
{
|
| 264737 |
+
"epoch": 1047.46,
|
| 264738 |
+
"learning_rate": 7.933830128205129e-06,
|
| 264739 |
+
"loss": 0.2456,
|
| 264740 |
+
"step": 129980
|
| 264741 |
+
},
|
| 264742 |
+
{
|
| 264743 |
+
"epoch": 1047.5,
|
| 264744 |
+
"learning_rate": 7.933750000000002e-06,
|
| 264745 |
+
"loss": 0.2615,
|
| 264746 |
+
"step": 129985
|
| 264747 |
+
},
|
| 264748 |
+
{
|
| 264749 |
+
"epoch": 1047.54,
|
| 264750 |
+
"learning_rate": 7.933669871794873e-06,
|
| 264751 |
+
"loss": 0.4054,
|
| 264752 |
+
"step": 129990
|
| 264753 |
+
},
|
| 264754 |
+
{
|
| 264755 |
+
"epoch": 1047.58,
|
| 264756 |
+
"learning_rate": 7.933589743589744e-06,
|
| 264757 |
+
"loss": 0.5305,
|
| 264758 |
+
"step": 129995
|
| 264759 |
+
},
|
| 264760 |
+
{
|
| 264761 |
+
"epoch": 1047.62,
|
| 264762 |
+
"learning_rate": 7.933509615384616e-06,
|
| 264763 |
+
"loss": 0.8806,
|
| 264764 |
+
"step": 130000
|
| 264765 |
+
},
|
| 264766 |
+
{
|
| 264767 |
+
"epoch": 1047.66,
|
| 264768 |
+
"learning_rate": 7.933429487179489e-06,
|
| 264769 |
+
"loss": 0.2519,
|
| 264770 |
+
"step": 130005
|
| 264771 |
+
},
|
| 264772 |
+
{
|
| 264773 |
+
"epoch": 1047.7,
|
| 264774 |
+
"learning_rate": 7.933349358974359e-06,
|
| 264775 |
+
"loss": 0.2915,
|
| 264776 |
+
"step": 130010
|
| 264777 |
+
},
|
| 264778 |
+
{
|
| 264779 |
+
"epoch": 1047.74,
|
| 264780 |
+
"learning_rate": 7.933269230769232e-06,
|
| 264781 |
+
"loss": 0.2979,
|
| 264782 |
+
"step": 130015
|
| 264783 |
+
},
|
| 264784 |
+
{
|
| 264785 |
+
"epoch": 1047.78,
|
| 264786 |
+
"learning_rate": 7.933189102564103e-06,
|
| 264787 |
+
"loss": 0.569,
|
| 264788 |
+
"step": 130020
|
| 264789 |
+
},
|
| 264790 |
+
{
|
| 264791 |
+
"epoch": 1047.82,
|
| 264792 |
+
"learning_rate": 7.933108974358974e-06,
|
| 264793 |
+
"loss": 0.9575,
|
| 264794 |
+
"step": 130025
|
| 264795 |
+
},
|
| 264796 |
+
{
|
| 264797 |
+
"epoch": 1047.86,
|
| 264798 |
+
"learning_rate": 7.933028846153846e-06,
|
| 264799 |
+
"loss": 0.281,
|
| 264800 |
+
"step": 130030
|
| 264801 |
+
},
|
| 264802 |
+
{
|
| 264803 |
+
"epoch": 1047.9,
|
| 264804 |
+
"learning_rate": 7.932948717948719e-06,
|
| 264805 |
+
"loss": 0.2734,
|
| 264806 |
+
"step": 130035
|
| 264807 |
+
},
|
| 264808 |
+
{
|
| 264809 |
+
"epoch": 1047.94,
|
| 264810 |
+
"learning_rate": 7.93286858974359e-06,
|
| 264811 |
+
"loss": 0.3178,
|
| 264812 |
+
"step": 130040
|
| 264813 |
+
},
|
| 264814 |
+
{
|
| 264815 |
+
"epoch": 1047.98,
|
| 264816 |
+
"learning_rate": 7.932788461538461e-06,
|
| 264817 |
+
"loss": 0.768,
|
| 264818 |
+
"step": 130045
|
| 264819 |
+
},
|
| 264820 |
+
{
|
| 264821 |
+
"epoch": 1048.0,
|
| 264822 |
+
"eval_loss": 0.3541869521141052,
|
| 264823 |
+
"eval_runtime": 34.4713,
|
| 264824 |
+
"eval_samples_per_second": 24.223,
|
| 264825 |
+
"eval_steps_per_second": 0.783,
|
| 264826 |
+
"eval_wer": 0.17978857100613588,
|
| 264827 |
+
"step": 130047
|
| 264828 |
+
},
|
| 264829 |
+
{
|
| 264830 |
+
"epoch": 1048.02,
|
| 264831 |
+
"learning_rate": 7.932708333333334e-06,
|
| 264832 |
+
"loss": 0.3823,
|
| 264833 |
+
"step": 130050
|
| 264834 |
+
},
|
| 264835 |
+
{
|
| 264836 |
+
"epoch": 1048.06,
|
| 264837 |
+
"learning_rate": 7.932628205128206e-06,
|
| 264838 |
+
"loss": 0.2531,
|
| 264839 |
+
"step": 130055
|
| 264840 |
+
},
|
| 264841 |
+
{
|
| 264842 |
+
"epoch": 1048.1,
|
| 264843 |
+
"learning_rate": 7.932548076923077e-06,
|
| 264844 |
+
"loss": 0.3504,
|
| 264845 |
+
"step": 130060
|
| 264846 |
+
},
|
| 264847 |
+
{
|
| 264848 |
+
"epoch": 1048.14,
|
| 264849 |
+
"learning_rate": 7.932467948717949e-06,
|
| 264850 |
+
"loss": 0.3598,
|
| 264851 |
+
"step": 130065
|
| 264852 |
+
},
|
| 264853 |
+
{
|
| 264854 |
+
"epoch": 1048.18,
|
| 264855 |
+
"learning_rate": 7.932387820512822e-06,
|
| 264856 |
+
"loss": 0.7025,
|
| 264857 |
+
"step": 130070
|
| 264858 |
+
},
|
| 264859 |
+
{
|
| 264860 |
+
"epoch": 1048.22,
|
| 264861 |
+
"learning_rate": 7.932307692307693e-06,
|
| 264862 |
+
"loss": 0.8204,
|
| 264863 |
+
"step": 130075
|
| 264864 |
+
},
|
| 264865 |
+
{
|
| 264866 |
+
"epoch": 1048.27,
|
| 264867 |
+
"learning_rate": 7.932227564102564e-06,
|
| 264868 |
+
"loss": 0.3144,
|
| 264869 |
+
"step": 130080
|
| 264870 |
+
},
|
| 264871 |
+
{
|
| 264872 |
+
"epoch": 1048.31,
|
| 264873 |
+
"learning_rate": 7.932147435897437e-06,
|
| 264874 |
+
"loss": 0.2883,
|
| 264875 |
+
"step": 130085
|
| 264876 |
+
},
|
| 264877 |
+
{
|
| 264878 |
+
"epoch": 1048.35,
|
| 264879 |
+
"learning_rate": 7.932067307692309e-06,
|
| 264880 |
+
"loss": 0.3835,
|
| 264881 |
+
"step": 130090
|
| 264882 |
+
},
|
| 264883 |
+
{
|
| 264884 |
+
"epoch": 1048.39,
|
| 264885 |
+
"learning_rate": 7.93198717948718e-06,
|
| 264886 |
+
"loss": 0.8257,
|
| 264887 |
+
"step": 130095
|
| 264888 |
+
},
|
| 264889 |
+
{
|
| 264890 |
+
"epoch": 1048.43,
|
| 264891 |
+
"learning_rate": 7.931907051282051e-06,
|
| 264892 |
+
"loss": 0.8124,
|
| 264893 |
+
"step": 130100
|
| 264894 |
+
},
|
| 264895 |
+
{
|
| 264896 |
+
"epoch": 1048.47,
|
| 264897 |
+
"learning_rate": 7.931826923076925e-06,
|
| 264898 |
+
"loss": 0.2594,
|
| 264899 |
+
"step": 130105
|
| 264900 |
+
},
|
| 264901 |
+
{
|
| 264902 |
+
"epoch": 1048.51,
|
| 264903 |
+
"learning_rate": 7.931746794871796e-06,
|
| 264904 |
+
"loss": 0.2822,
|
| 264905 |
+
"step": 130110
|
| 264906 |
+
},
|
| 264907 |
+
{
|
| 264908 |
+
"epoch": 1048.55,
|
| 264909 |
+
"learning_rate": 7.931666666666667e-06,
|
| 264910 |
+
"loss": 0.3836,
|
| 264911 |
+
"step": 130115
|
| 264912 |
+
},
|
| 264913 |
+
{
|
| 264914 |
+
"epoch": 1048.59,
|
| 264915 |
+
"learning_rate": 7.931586538461539e-06,
|
| 264916 |
+
"loss": 0.7467,
|
| 264917 |
+
"step": 130120
|
| 264918 |
+
},
|
| 264919 |
+
{
|
| 264920 |
+
"epoch": 1048.63,
|
| 264921 |
+
"learning_rate": 7.931506410256412e-06,
|
| 264922 |
+
"loss": 0.8817,
|
| 264923 |
+
"step": 130125
|
| 264924 |
+
},
|
| 264925 |
+
{
|
| 264926 |
+
"epoch": 1048.67,
|
| 264927 |
+
"learning_rate": 7.931426282051283e-06,
|
| 264928 |
+
"loss": 0.3068,
|
| 264929 |
+
"step": 130130
|
| 264930 |
+
},
|
| 264931 |
+
{
|
| 264932 |
+
"epoch": 1048.71,
|
| 264933 |
+
"learning_rate": 7.931346153846154e-06,
|
| 264934 |
+
"loss": 0.2516,
|
| 264935 |
+
"step": 130135
|
| 264936 |
+
},
|
| 264937 |
+
{
|
| 264938 |
+
"epoch": 1048.75,
|
| 264939 |
+
"learning_rate": 7.931266025641027e-06,
|
| 264940 |
+
"loss": 0.3328,
|
| 264941 |
+
"step": 130140
|
| 264942 |
+
},
|
| 264943 |
+
{
|
| 264944 |
+
"epoch": 1048.79,
|
| 264945 |
+
"learning_rate": 7.931185897435899e-06,
|
| 264946 |
+
"loss": 0.871,
|
| 264947 |
+
"step": 130145
|
| 264948 |
+
},
|
| 264949 |
+
{
|
| 264950 |
+
"epoch": 1048.83,
|
| 264951 |
+
"learning_rate": 7.93110576923077e-06,
|
| 264952 |
+
"loss": 0.9983,
|
| 264953 |
+
"step": 130150
|
| 264954 |
+
},
|
| 264955 |
+
{
|
| 264956 |
+
"epoch": 1048.87,
|
| 264957 |
+
"learning_rate": 7.931025641025641e-06,
|
| 264958 |
+
"loss": 0.2637,
|
| 264959 |
+
"step": 130155
|
| 264960 |
+
},
|
| 264961 |
+
{
|
| 264962 |
+
"epoch": 1048.91,
|
| 264963 |
+
"learning_rate": 7.930945512820515e-06,
|
| 264964 |
+
"loss": 0.3195,
|
| 264965 |
+
"step": 130160
|
| 264966 |
+
},
|
| 264967 |
+
{
|
| 264968 |
+
"epoch": 1048.95,
|
| 264969 |
+
"learning_rate": 7.930865384615384e-06,
|
| 264970 |
+
"loss": 0.3649,
|
| 264971 |
+
"step": 130165
|
| 264972 |
+
},
|
| 264973 |
+
{
|
| 264974 |
+
"epoch": 1048.99,
|
| 264975 |
+
"learning_rate": 7.930785256410257e-06,
|
| 264976 |
+
"loss": 0.7069,
|
| 264977 |
+
"step": 130170
|
| 264978 |
+
},
|
| 264979 |
+
{
|
| 264980 |
+
"epoch": 1049.0,
|
| 264981 |
+
"eval_loss": 0.4282522201538086,
|
| 264982 |
+
"eval_runtime": 37.3121,
|
| 264983 |
+
"eval_samples_per_second": 22.406,
|
| 264984 |
+
"eval_steps_per_second": 0.724,
|
| 264985 |
+
"eval_wer": 0.17979154603358424,
|
| 264986 |
+
"step": 130171
|
| 264987 |
+
},
|
| 264988 |
+
{
|
| 264989 |
+
"epoch": 1041.03,
|
| 264990 |
+
"learning_rate": 7.930705128205129e-06,
|
| 264991 |
+
"loss": 0.3971,
|
| 264992 |
+
"step": 130175
|
| 264993 |
+
},
|
| 264994 |
+
{
|
| 264995 |
+
"epoch": 1041.07,
|
| 264996 |
+
"learning_rate": 7.930625e-06,
|
| 264997 |
+
"loss": 0.2518,
|
| 264998 |
+
"step": 130180
|
| 264999 |
+
},
|
| 265000 |
+
{
|
| 265001 |
+
"epoch": 1041.11,
|
| 265002 |
+
"learning_rate": 7.930544871794873e-06,
|
| 265003 |
+
"loss": 0.3056,
|
| 265004 |
+
"step": 130185
|
| 265005 |
+
},
|
| 265006 |
+
{
|
| 265007 |
+
"epoch": 1041.15,
|
| 265008 |
+
"learning_rate": 7.930464743589744e-06,
|
| 265009 |
+
"loss": 0.3826,
|
| 265010 |
+
"step": 130190
|
| 265011 |
+
},
|
| 265012 |
+
{
|
| 265013 |
+
"epoch": 1041.19,
|
| 265014 |
+
"learning_rate": 7.930384615384616e-06,
|
| 265015 |
+
"loss": 0.9553,
|
| 265016 |
+
"step": 130195
|
| 265017 |
+
},
|
| 265018 |
+
{
|
| 265019 |
+
"epoch": 1041.23,
|
| 265020 |
+
"learning_rate": 7.930304487179487e-06,
|
| 265021 |
+
"loss": 0.6101,
|
| 265022 |
+
"step": 130200
|
| 265023 |
+
},
|
| 265024 |
+
{
|
| 265025 |
+
"epoch": 1041.27,
|
| 265026 |
+
"learning_rate": 7.93022435897436e-06,
|
| 265027 |
+
"loss": 0.2316,
|
| 265028 |
+
"step": 130205
|
| 265029 |
+
},
|
| 265030 |
+
{
|
| 265031 |
+
"epoch": 1041.31,
|
| 265032 |
+
"learning_rate": 7.930144230769232e-06,
|
| 265033 |
+
"loss": 0.3121,
|
| 265034 |
+
"step": 130210
|
| 265035 |
+
},
|
| 265036 |
+
{
|
| 265037 |
+
"epoch": 1041.35,
|
| 265038 |
+
"learning_rate": 7.930064102564103e-06,
|
| 265039 |
+
"loss": 0.4402,
|
| 265040 |
+
"step": 130215
|
| 265041 |
+
},
|
| 265042 |
+
{
|
| 265043 |
+
"epoch": 1041.39,
|
| 265044 |
+
"learning_rate": 7.929983974358974e-06,
|
| 265045 |
+
"loss": 0.9923,
|
| 265046 |
+
"step": 130220
|
| 265047 |
+
},
|
| 265048 |
+
{
|
| 265049 |
+
"epoch": 1041.43,
|
| 265050 |
+
"learning_rate": 7.929903846153847e-06,
|
| 265051 |
+
"loss": 0.607,
|
| 265052 |
+
"step": 130225
|
| 265053 |
+
},
|
| 265054 |
+
{
|
| 265055 |
+
"epoch": 1041.47,
|
| 265056 |
+
"learning_rate": 7.929823717948719e-06,
|
| 265057 |
+
"loss": 0.2848,
|
| 265058 |
+
"step": 130230
|
| 265059 |
+
},
|
| 265060 |
+
{
|
| 265061 |
+
"epoch": 1041.51,
|
| 265062 |
+
"learning_rate": 7.92974358974359e-06,
|
| 265063 |
+
"loss": 0.2927,
|
| 265064 |
+
"step": 130235
|
| 265065 |
+
},
|
| 265066 |
+
{
|
| 265067 |
+
"epoch": 1041.55,
|
| 265068 |
+
"learning_rate": 7.929663461538463e-06,
|
| 265069 |
+
"loss": 0.4192,
|
| 265070 |
+
"step": 130240
|
| 265071 |
+
},
|
| 265072 |
+
{
|
| 265073 |
+
"epoch": 1041.59,
|
| 265074 |
+
"learning_rate": 7.929583333333334e-06,
|
| 265075 |
+
"loss": 0.825,
|
| 265076 |
+
"step": 130245
|
| 265077 |
+
},
|
| 265078 |
+
{
|
| 265079 |
+
"epoch": 1041.63,
|
| 265080 |
+
"learning_rate": 7.929503205128206e-06,
|
| 265081 |
+
"loss": 0.736,
|
| 265082 |
+
"step": 130250
|
| 265083 |
+
},
|
| 265084 |
+
{
|
| 265085 |
+
"epoch": 1041.67,
|
| 265086 |
+
"learning_rate": 7.929423076923077e-06,
|
| 265087 |
+
"loss": 0.3488,
|
| 265088 |
+
"step": 130255
|
| 265089 |
+
},
|
| 265090 |
+
{
|
| 265091 |
+
"epoch": 1041.71,
|
| 265092 |
+
"learning_rate": 7.92934294871795e-06,
|
| 265093 |
+
"loss": 0.315,
|
| 265094 |
+
"step": 130260
|
| 265095 |
+
},
|
| 265096 |
+
{
|
| 265097 |
+
"epoch": 1041.75,
|
| 265098 |
+
"learning_rate": 7.929262820512822e-06,
|
| 265099 |
+
"loss": 0.4015,
|
| 265100 |
+
"step": 130265
|
| 265101 |
+
},
|
| 265102 |
+
{
|
| 265103 |
+
"epoch": 1041.79,
|
| 265104 |
+
"learning_rate": 7.929182692307693e-06,
|
| 265105 |
+
"loss": 0.7939,
|
| 265106 |
+
"step": 130270
|
| 265107 |
+
},
|
| 265108 |
+
{
|
| 265109 |
+
"epoch": 1041.83,
|
| 265110 |
+
"learning_rate": 7.929102564102564e-06,
|
| 265111 |
+
"loss": 0.7487,
|
| 265112 |
+
"step": 130275
|
| 265113 |
+
},
|
| 265114 |
+
{
|
| 265115 |
+
"epoch": 1041.87,
|
| 265116 |
+
"learning_rate": 7.929022435897437e-06,
|
| 265117 |
+
"loss": 0.3225,
|
| 265118 |
+
"step": 130280
|
| 265119 |
+
},
|
| 265120 |
+
{
|
| 265121 |
+
"epoch": 1041.91,
|
| 265122 |
+
"learning_rate": 7.928942307692309e-06,
|
| 265123 |
+
"loss": 0.2599,
|
| 265124 |
+
"step": 130285
|
| 265125 |
+
},
|
| 265126 |
+
{
|
| 265127 |
+
"epoch": 1041.95,
|
| 265128 |
+
"learning_rate": 7.92886217948718e-06,
|
| 265129 |
+
"loss": 0.6579,
|
| 265130 |
+
"step": 130290
|
| 265131 |
+
},
|
| 265132 |
+
{
|
| 265133 |
+
"epoch": 1041.99,
|
| 265134 |
+
"learning_rate": 7.928782051282053e-06,
|
| 265135 |
+
"loss": 0.8982,
|
| 265136 |
+
"step": 130295
|
| 265137 |
+
},
|
| 265138 |
+
{
|
| 265139 |
+
"epoch": 1042.0,
|
| 265140 |
+
"eval_loss": 0.4521143138408661,
|
| 265141 |
+
"eval_runtime": 36.4023,
|
| 265142 |
+
"eval_samples_per_second": 22.966,
|
| 265143 |
+
"eval_steps_per_second": 0.742,
|
| 265144 |
+
"eval_wer": 0.17621112989781665,
|
| 265145 |
+
"step": 130296
|
| 265146 |
+
},
|
| 265147 |
+
{
|
| 265148 |
+
"epoch": 1050.03,
|
| 265149 |
+
"learning_rate": 7.928701923076923e-06,
|
| 265150 |
+
"loss": 0.2942,
|
| 265151 |
+
"step": 130300
|
| 265152 |
+
},
|
| 265153 |
+
{
|
| 265154 |
+
"epoch": 1050.07,
|
| 265155 |
+
"learning_rate": 7.928621794871796e-06,
|
| 265156 |
+
"loss": 0.3089,
|
| 265157 |
+
"step": 130305
|
| 265158 |
+
},
|
| 265159 |
+
{
|
| 265160 |
+
"epoch": 1050.11,
|
| 265161 |
+
"learning_rate": 7.928541666666667e-06,
|
| 265162 |
+
"loss": 0.28,
|
| 265163 |
+
"step": 130310
|
| 265164 |
+
},
|
| 265165 |
+
{
|
| 265166 |
+
"epoch": 1050.15,
|
| 265167 |
+
"learning_rate": 7.928461538461539e-06,
|
| 265168 |
+
"loss": 0.3548,
|
| 265169 |
+
"step": 130315
|
| 265170 |
+
},
|
| 265171 |
+
{
|
| 265172 |
+
"epoch": 1050.19,
|
| 265173 |
+
"learning_rate": 7.92838141025641e-06,
|
| 265174 |
+
"loss": 0.8331,
|
| 265175 |
+
"step": 130320
|
| 265176 |
+
},
|
| 265177 |
+
{
|
| 265178 |
+
"epoch": 1050.23,
|
| 265179 |
+
"learning_rate": 7.928301282051283e-06,
|
| 265180 |
+
"loss": 0.6048,
|
| 265181 |
+
"step": 130325
|
| 265182 |
+
},
|
| 265183 |
+
{
|
| 265184 |
+
"epoch": 1050.27,
|
| 265185 |
+
"learning_rate": 7.928221153846154e-06,
|
| 265186 |
+
"loss": 0.2592,
|
| 265187 |
+
"step": 130330
|
| 265188 |
+
},
|
| 265189 |
+
{
|
| 265190 |
+
"epoch": 1050.31,
|
| 265191 |
+
"learning_rate": 7.928141025641026e-06,
|
| 265192 |
+
"loss": 0.282,
|
| 265193 |
+
"step": 130335
|
| 265194 |
+
},
|
| 265195 |
+
{
|
| 265196 |
+
"epoch": 1050.35,
|
| 265197 |
+
"learning_rate": 7.928060897435899e-06,
|
| 265198 |
+
"loss": 0.3886,
|
| 265199 |
+
"step": 130340
|
| 265200 |
+
},
|
| 265201 |
+
{
|
| 265202 |
+
"epoch": 1050.39,
|
| 265203 |
+
"learning_rate": 7.92798076923077e-06,
|
| 265204 |
+
"loss": 0.7989,
|
| 265205 |
+
"step": 130345
|
| 265206 |
+
},
|
| 265207 |
+
{
|
| 265208 |
+
"epoch": 1050.43,
|
| 265209 |
+
"learning_rate": 7.927900641025641e-06,
|
| 265210 |
+
"loss": 0.5287,
|
| 265211 |
+
"step": 130350
|
| 265212 |
+
},
|
| 265213 |
+
{
|
| 265214 |
+
"epoch": 1050.47,
|
| 265215 |
+
"learning_rate": 7.927820512820513e-06,
|
| 265216 |
+
"loss": 0.2833,
|
| 265217 |
+
"step": 130355
|
| 265218 |
+
},
|
| 265219 |
+
{
|
| 265220 |
+
"epoch": 1050.51,
|
| 265221 |
+
"learning_rate": 7.927740384615386e-06,
|
| 265222 |
+
"loss": 0.2952,
|
| 265223 |
+
"step": 130360
|
| 265224 |
+
},
|
| 265225 |
+
{
|
| 265226 |
+
"epoch": 1050.55,
|
| 265227 |
+
"learning_rate": 7.927660256410257e-06,
|
| 265228 |
+
"loss": 0.458,
|
| 265229 |
+
"step": 130365
|
| 265230 |
+
},
|
| 265231 |
+
{
|
| 265232 |
+
"epoch": 1050.59,
|
| 265233 |
+
"learning_rate": 7.927580128205129e-06,
|
| 265234 |
+
"loss": 0.8351,
|
| 265235 |
+
"step": 130370
|
| 265236 |
+
},
|
| 265237 |
+
{
|
| 265238 |
+
"epoch": 1050.63,
|
| 265239 |
+
"learning_rate": 7.9275e-06,
|
| 265240 |
+
"loss": 0.6399,
|
| 265241 |
+
"step": 130375
|
| 265242 |
+
},
|
| 265243 |
+
{
|
| 265244 |
+
"epoch": 1050.67,
|
| 265245 |
+
"learning_rate": 7.927419871794873e-06,
|
| 265246 |
+
"loss": 0.2709,
|
| 265247 |
+
"step": 130380
|
| 265248 |
+
},
|
| 265249 |
+
{
|
| 265250 |
+
"epoch": 1050.71,
|
| 265251 |
+
"learning_rate": 7.927339743589744e-06,
|
| 265252 |
+
"loss": 0.3261,
|
| 265253 |
+
"step": 130385
|
| 265254 |
+
},
|
| 265255 |
+
{
|
| 265256 |
+
"epoch": 1050.76,
|
| 265257 |
+
"learning_rate": 7.927259615384616e-06,
|
| 265258 |
+
"loss": 0.3979,
|
| 265259 |
+
"step": 130390
|
| 265260 |
+
},
|
| 265261 |
+
{
|
| 265262 |
+
"epoch": 1050.8,
|
| 265263 |
+
"learning_rate": 7.927179487179489e-06,
|
| 265264 |
+
"loss": 0.8701,
|
| 265265 |
+
"step": 130395
|
| 265266 |
+
},
|
| 265267 |
+
{
|
| 265268 |
+
"epoch": 1050.84,
|
| 265269 |
+
"learning_rate": 7.92709935897436e-06,
|
| 265270 |
+
"loss": 0.5942,
|
| 265271 |
+
"step": 130400
|
| 265272 |
+
},
|
| 265273 |
+
{
|
| 265274 |
+
"epoch": 1050.88,
|
| 265275 |
+
"learning_rate": 7.927019230769231e-06,
|
| 265276 |
+
"loss": 0.2871,
|
| 265277 |
+
"step": 130405
|
| 265278 |
+
},
|
| 265279 |
+
{
|
| 265280 |
+
"epoch": 1050.92,
|
| 265281 |
+
"learning_rate": 7.926939102564103e-06,
|
| 265282 |
+
"loss": 0.3117,
|
| 265283 |
+
"step": 130410
|
| 265284 |
+
},
|
| 265285 |
+
{
|
| 265286 |
+
"epoch": 1050.96,
|
| 265287 |
+
"learning_rate": 7.926858974358976e-06,
|
| 265288 |
+
"loss": 0.3888,
|
| 265289 |
+
"step": 130415
|
| 265290 |
+
},
|
| 265291 |
+
{
|
| 265292 |
+
"epoch": 1051.0,
|
| 265293 |
+
"learning_rate": 7.926778846153847e-06,
|
| 265294 |
+
"loss": 0.962,
|
| 265295 |
+
"step": 130420
|
| 265296 |
+
},
|
| 265297 |
+
{
|
| 265298 |
+
"epoch": 1051.0,
|
| 265299 |
+
"eval_loss": 0.36090824007987976,
|
| 265300 |
+
"eval_runtime": 36.3757,
|
| 265301 |
+
"eval_samples_per_second": 22.982,
|
| 265302 |
+
"eval_steps_per_second": 0.742,
|
| 265303 |
+
"eval_wer": 0.18075693360101122,
|
| 265304 |
+
"step": 130420
|
| 265305 |
+
},
|
| 265306 |
+
{
|
| 265307 |
+
"epoch": 1043.04,
|
| 265308 |
+
"learning_rate": 7.926698717948719e-06,
|
| 265309 |
+
"loss": 0.2924,
|
| 265310 |
+
"step": 130425
|
| 265311 |
+
},
|
| 265312 |
+
{
|
| 265313 |
+
"epoch": 1043.08,
|
| 265314 |
+
"learning_rate": 7.92661858974359e-06,
|
| 265315 |
+
"loss": 0.2821,
|
| 265316 |
+
"step": 130430
|
| 265317 |
+
},
|
| 265318 |
+
{
|
| 265319 |
+
"epoch": 1043.12,
|
| 265320 |
+
"learning_rate": 7.926538461538463e-06,
|
| 265321 |
+
"loss": 0.2655,
|
| 265322 |
+
"step": 130435
|
| 265323 |
+
},
|
| 265324 |
+
{
|
| 265325 |
+
"epoch": 1043.16,
|
| 265326 |
+
"learning_rate": 7.926458333333334e-06,
|
| 265327 |
+
"loss": 0.4237,
|
| 265328 |
+
"step": 130440
|
| 265329 |
+
},
|
| 265330 |
+
{
|
| 265331 |
+
"epoch": 1043.2,
|
| 265332 |
+
"learning_rate": 7.926378205128206e-06,
|
| 265333 |
+
"loss": 1.2634,
|
| 265334 |
+
"step": 130445
|
| 265335 |
+
},
|
| 265336 |
+
{
|
| 265337 |
+
"epoch": 1043.24,
|
| 265338 |
+
"learning_rate": 7.926298076923079e-06,
|
| 265339 |
+
"loss": 0.286,
|
| 265340 |
+
"step": 130450
|
| 265341 |
+
},
|
| 265342 |
+
{
|
| 265343 |
+
"epoch": 1043.28,
|
| 265344 |
+
"learning_rate": 7.926217948717948e-06,
|
| 265345 |
+
"loss": 0.3288,
|
| 265346 |
+
"step": 130455
|
| 265347 |
+
},
|
| 265348 |
+
{
|
| 265349 |
+
"epoch": 1043.32,
|
| 265350 |
+
"learning_rate": 7.926137820512821e-06,
|
| 265351 |
+
"loss": 0.2873,
|
| 265352 |
+
"step": 130460
|
| 265353 |
+
},
|
| 265354 |
+
{
|
| 265355 |
+
"epoch": 1043.36,
|
| 265356 |
+
"learning_rate": 7.926057692307693e-06,
|
| 265357 |
+
"loss": 0.4448,
|
| 265358 |
+
"step": 130465
|
| 265359 |
+
},
|
| 265360 |
+
{
|
| 265361 |
+
"epoch": 1043.4,
|
| 265362 |
+
"learning_rate": 7.925977564102564e-06,
|
| 265363 |
+
"loss": 1.2558,
|
| 265364 |
+
"step": 130470
|
| 265365 |
+
},
|
| 265366 |
+
{
|
| 265367 |
+
"epoch": 1043.44,
|
| 265368 |
+
"learning_rate": 7.925897435897436e-06,
|
| 265369 |
+
"loss": 0.3083,
|
| 265370 |
+
"step": 130475
|
| 265371 |
+
},
|
| 265372 |
+
{
|
| 265373 |
+
"epoch": 1043.48,
|
| 265374 |
+
"learning_rate": 7.925817307692309e-06,
|
| 265375 |
+
"loss": 0.3517,
|
| 265376 |
+
"step": 130480
|
| 265377 |
+
},
|
| 265378 |
+
{
|
| 265379 |
+
"epoch": 1043.52,
|
| 265380 |
+
"learning_rate": 7.92573717948718e-06,
|
| 265381 |
+
"loss": 0.3243,
|
| 265382 |
+
"step": 130485
|
| 265383 |
+
},
|
| 265384 |
+
{
|
| 265385 |
+
"epoch": 1043.56,
|
| 265386 |
+
"learning_rate": 7.925657051282051e-06,
|
| 265387 |
+
"loss": 0.4209,
|
| 265388 |
+
"step": 130490
|
| 265389 |
+
},
|
| 265390 |
+
{
|
| 265391 |
+
"epoch": 1043.6,
|
| 265392 |
+
"learning_rate": 7.925576923076924e-06,
|
| 265393 |
+
"loss": 1.0973,
|
| 265394 |
+
"step": 130495
|
| 265395 |
+
},
|
| 265396 |
+
{
|
| 265397 |
+
"epoch": 1043.64,
|
| 265398 |
+
"learning_rate": 7.925496794871796e-06,
|
| 265399 |
+
"loss": 0.311,
|
| 265400 |
+
"step": 130500
|
| 265401 |
+
},
|
| 265402 |
+
{
|
| 265403 |
+
"epoch": 1043.68,
|
| 265404 |
+
"learning_rate": 7.925416666666667e-06,
|
| 265405 |
+
"loss": 0.3225,
|
| 265406 |
+
"step": 130505
|
| 265407 |
+
},
|
| 265408 |
+
{
|
| 265409 |
+
"epoch": 1043.72,
|
| 265410 |
+
"learning_rate": 7.925336538461538e-06,
|
| 265411 |
+
"loss": 0.3059,
|
| 265412 |
+
"step": 130510
|
| 265413 |
+
},
|
| 265414 |
+
{
|
| 265415 |
+
"epoch": 1043.76,
|
| 265416 |
+
"learning_rate": 7.925256410256412e-06,
|
| 265417 |
+
"loss": 0.4018,
|
| 265418 |
+
"step": 130515
|
| 265419 |
+
},
|
| 265420 |
+
{
|
| 265421 |
+
"epoch": 1043.8,
|
| 265422 |
+
"learning_rate": 7.925176282051283e-06,
|
| 265423 |
+
"loss": 1.1179,
|
| 265424 |
+
"step": 130520
|
| 265425 |
+
},
|
| 265426 |
+
{
|
| 265427 |
+
"epoch": 1043.84,
|
| 265428 |
+
"learning_rate": 7.925096153846154e-06,
|
| 265429 |
+
"loss": 0.2868,
|
| 265430 |
+
"step": 130525
|
| 265431 |
+
},
|
| 265432 |
+
{
|
| 265433 |
+
"epoch": 1043.88,
|
| 265434 |
+
"learning_rate": 7.925016025641026e-06,
|
| 265435 |
+
"loss": 0.2716,
|
| 265436 |
+
"step": 130530
|
| 265437 |
+
},
|
| 265438 |
+
{
|
| 265439 |
+
"epoch": 1043.92,
|
| 265440 |
+
"learning_rate": 7.924935897435899e-06,
|
| 265441 |
+
"loss": 0.3303,
|
| 265442 |
+
"step": 130535
|
| 265443 |
+
},
|
| 265444 |
+
{
|
| 265445 |
+
"epoch": 1043.96,
|
| 265446 |
+
"learning_rate": 7.92485576923077e-06,
|
| 265447 |
+
"loss": 0.4517,
|
| 265448 |
+
"step": 130540
|
| 265449 |
+
},
|
| 265450 |
+
{
|
| 265451 |
+
"epoch": 1044.0,
|
| 265452 |
+
"learning_rate": 7.924775641025641e-06,
|
| 265453 |
+
"loss": 1.1773,
|
| 265454 |
+
"step": 130545
|
| 265455 |
+
},
|
| 265456 |
+
{
|
| 265457 |
+
"epoch": 1044.0,
|
| 265458 |
+
"eval_loss": 0.3330843150615692,
|
| 265459 |
+
"eval_runtime": 36.8417,
|
| 265460 |
+
"eval_samples_per_second": 22.692,
|
| 265461 |
+
"eval_steps_per_second": 0.733,
|
| 265462 |
+
"eval_wer": 0.1764533606677405,
|
| 265463 |
+
"step": 130545
|
| 265464 |
+
},
|
| 265465 |
+
{
|
| 265466 |
+
"epoch": 1052.04,
|
| 265467 |
+
"learning_rate": 7.924695512820514e-06,
|
| 265468 |
+
"loss": 0.3384,
|
| 265469 |
+
"step": 130550
|
| 265470 |
+
},
|
| 265471 |
+
{
|
| 265472 |
+
"epoch": 1052.08,
|
| 265473 |
+
"learning_rate": 7.924615384615386e-06,
|
| 265474 |
+
"loss": 0.2406,
|
| 265475 |
+
"step": 130555
|
| 265476 |
+
},
|
| 265477 |
+
{
|
| 265478 |
+
"epoch": 1052.12,
|
| 265479 |
+
"learning_rate": 7.924535256410257e-06,
|
| 265480 |
+
"loss": 0.3477,
|
| 265481 |
+
"step": 130560
|
| 265482 |
+
},
|
| 265483 |
+
{
|
| 265484 |
+
"epoch": 1052.16,
|
| 265485 |
+
"learning_rate": 7.924455128205128e-06,
|
| 265486 |
+
"loss": 0.4641,
|
| 265487 |
+
"step": 130565
|
| 265488 |
+
},
|
| 265489 |
+
{
|
| 265490 |
+
"epoch": 1052.2,
|
| 265491 |
+
"learning_rate": 7.924375000000002e-06,
|
| 265492 |
+
"loss": 1.1962,
|
| 265493 |
+
"step": 130570
|
| 265494 |
+
},
|
| 265495 |
+
{
|
| 265496 |
+
"epoch": 1052.24,
|
| 265497 |
+
"learning_rate": 7.924294871794871e-06,
|
| 265498 |
+
"loss": 0.3233,
|
| 265499 |
+
"step": 130575
|
| 265500 |
+
},
|
| 265501 |
+
{
|
| 265502 |
+
"epoch": 1052.28,
|
| 265503 |
+
"learning_rate": 7.924214743589744e-06,
|
| 265504 |
+
"loss": 0.2914,
|
| 265505 |
+
"step": 130580
|
| 265506 |
+
},
|
| 265507 |
+
{
|
| 265508 |
+
"epoch": 1052.32,
|
| 265509 |
+
"learning_rate": 7.924134615384616e-06,
|
| 265510 |
+
"loss": 0.2446,
|
| 265511 |
+
"step": 130585
|
| 265512 |
+
},
|
| 265513 |
+
{
|
| 265514 |
+
"epoch": 1052.36,
|
| 265515 |
+
"learning_rate": 7.924054487179487e-06,
|
| 265516 |
+
"loss": 0.3771,
|
| 265517 |
+
"step": 130590
|
| 265518 |
+
},
|
| 265519 |
+
{
|
| 265520 |
+
"epoch": 1052.4,
|
| 265521 |
+
"learning_rate": 7.92397435897436e-06,
|
| 265522 |
+
"loss": 1.2242,
|
| 265523 |
+
"step": 130595
|
| 265524 |
+
},
|
| 265525 |
+
{
|
| 265526 |
+
"epoch": 1052.44,
|
| 265527 |
+
"learning_rate": 7.923894230769231e-06,
|
| 265528 |
+
"loss": 0.2807,
|
| 265529 |
+
"step": 130600
|
| 265530 |
+
},
|
| 265531 |
+
{
|
| 265532 |
+
"epoch": 1052.48,
|
| 265533 |
+
"learning_rate": 7.923814102564103e-06,
|
| 265534 |
+
"loss": 0.3239,
|
| 265535 |
+
"step": 130605
|
| 265536 |
+
},
|
| 265537 |
+
{
|
| 265538 |
+
"epoch": 1052.52,
|
| 265539 |
+
"learning_rate": 7.923733974358974e-06,
|
| 265540 |
+
"loss": 0.3105,
|
| 265541 |
+
"step": 130610
|
| 265542 |
+
},
|
| 265543 |
+
{
|
| 265544 |
+
"epoch": 1052.56,
|
| 265545 |
+
"learning_rate": 7.923653846153847e-06,
|
| 265546 |
+
"loss": 0.5188,
|
| 265547 |
+
"step": 130615
|
| 265548 |
+
},
|
| 265549 |
+
{
|
| 265550 |
+
"epoch": 1052.6,
|
| 265551 |
+
"learning_rate": 7.923573717948719e-06,
|
| 265552 |
+
"loss": 1.2514,
|
| 265553 |
+
"step": 130620
|
| 265554 |
+
},
|
| 265555 |
+
{
|
| 265556 |
+
"epoch": 1052.64,
|
| 265557 |
+
"learning_rate": 7.92349358974359e-06,
|
| 265558 |
+
"loss": 0.3024,
|
| 265559 |
+
"step": 130625
|
| 265560 |
+
},
|
| 265561 |
+
{
|
| 265562 |
+
"epoch": 1052.68,
|
| 265563 |
+
"learning_rate": 7.923413461538461e-06,
|
| 265564 |
+
"loss": 0.2717,
|
| 265565 |
+
"step": 130630
|
| 265566 |
+
},
|
| 265567 |
+
{
|
| 265568 |
+
"epoch": 1052.72,
|
| 265569 |
+
"learning_rate": 7.923333333333334e-06,
|
| 265570 |
+
"loss": 0.3497,
|
| 265571 |
+
"step": 130635
|
| 265572 |
+
},
|
| 265573 |
+
{
|
| 265574 |
+
"epoch": 1052.76,
|
| 265575 |
+
"learning_rate": 7.923253205128206e-06,
|
| 265576 |
+
"loss": 0.4023,
|
| 265577 |
+
"step": 130640
|
| 265578 |
+
},
|
| 265579 |
+
{
|
| 265580 |
+
"epoch": 1052.8,
|
| 265581 |
+
"learning_rate": 7.923173076923077e-06,
|
| 265582 |
+
"loss": 1.178,
|
| 265583 |
+
"step": 130645
|
| 265584 |
+
},
|
| 265585 |
+
{
|
| 265586 |
+
"epoch": 1052.84,
|
| 265587 |
+
"learning_rate": 7.92309294871795e-06,
|
| 265588 |
+
"loss": 0.3058,
|
| 265589 |
+
"step": 130650
|
| 265590 |
+
},
|
| 265591 |
+
{
|
| 265592 |
+
"epoch": 1052.88,
|
| 265593 |
+
"learning_rate": 7.923012820512821e-06,
|
| 265594 |
+
"loss": 0.3344,
|
| 265595 |
+
"step": 130655
|
| 265596 |
+
},
|
| 265597 |
+
{
|
| 265598 |
+
"epoch": 1052.92,
|
| 265599 |
+
"learning_rate": 7.922932692307693e-06,
|
| 265600 |
+
"loss": 0.3237,
|
| 265601 |
+
"step": 130660
|
| 265602 |
+
},
|
| 265603 |
+
{
|
| 265604 |
+
"epoch": 1052.96,
|
| 265605 |
+
"learning_rate": 7.922852564102564e-06,
|
| 265606 |
+
"loss": 0.5455,
|
| 265607 |
+
"step": 130665
|
| 265608 |
+
},
|
| 265609 |
+
{
|
| 265610 |
+
"epoch": 1053.0,
|
| 265611 |
+
"eval_loss": 0.37590697407722473,
|
| 265612 |
+
"eval_runtime": 37.2613,
|
| 265613 |
+
"eval_samples_per_second": 22.436,
|
| 265614 |
+
"eval_steps_per_second": 0.725,
|
| 265615 |
+
"eval_wer": 0.17155110793423875,
|
| 265616 |
+
"step": 130669
|
| 265617 |
+
},
|
| 265618 |
+
{
|
| 265619 |
+
"epoch": 1053.01,
|
| 265620 |
+
"learning_rate": 7.922772435897437e-06,
|
| 265621 |
+
"loss": 0.5027,
|
| 265622 |
+
"step": 130670
|
| 265623 |
+
},
|
| 265624 |
+
{
|
| 265625 |
+
"epoch": 1053.05,
|
| 265626 |
+
"learning_rate": 7.922692307692309e-06,
|
| 265627 |
+
"loss": 0.2415,
|
| 265628 |
+
"step": 130675
|
| 265629 |
+
},
|
| 265630 |
+
{
|
| 265631 |
+
"epoch": 1053.09,
|
| 265632 |
+
"learning_rate": 7.92261217948718e-06,
|
| 265633 |
+
"loss": 0.2721,
|
| 265634 |
+
"step": 130680
|
| 265635 |
+
},
|
| 265636 |
+
{
|
| 265637 |
+
"epoch": 1053.13,
|
| 265638 |
+
"learning_rate": 7.922532051282051e-06,
|
| 265639 |
+
"loss": 0.3606,
|
| 265640 |
+
"step": 130685
|
| 265641 |
+
},
|
| 265642 |
+
{
|
| 265643 |
+
"epoch": 1053.17,
|
| 265644 |
+
"learning_rate": 7.922451923076924e-06,
|
| 265645 |
+
"loss": 0.5634,
|
| 265646 |
+
"step": 130690
|
| 265647 |
+
},
|
| 265648 |
+
{
|
| 265649 |
+
"epoch": 1053.21,
|
| 265650 |
+
"learning_rate": 7.922371794871796e-06,
|
| 265651 |
+
"loss": 1.0906,
|
| 265652 |
+
"step": 130695
|
| 265653 |
+
},
|
| 265654 |
+
{
|
| 265655 |
+
"epoch": 1053.25,
|
| 265656 |
+
"learning_rate": 7.922291666666667e-06,
|
| 265657 |
+
"loss": 0.3278,
|
| 265658 |
+
"step": 130700
|
| 265659 |
+
},
|
| 265660 |
+
{
|
| 265661 |
+
"epoch": 1053.29,
|
| 265662 |
+
"learning_rate": 7.92221153846154e-06,
|
| 265663 |
+
"loss": 0.2634,
|
| 265664 |
+
"step": 130705
|
| 265665 |
+
},
|
| 265666 |
+
{
|
| 265667 |
+
"epoch": 1053.33,
|
| 265668 |
+
"learning_rate": 7.922131410256411e-06,
|
| 265669 |
+
"loss": 0.3169,
|
| 265670 |
+
"step": 130710
|
| 265671 |
+
},
|
| 265672 |
+
{
|
| 265673 |
+
"epoch": 1053.37,
|
| 265674 |
+
"learning_rate": 7.922051282051283e-06,
|
| 265675 |
+
"loss": 0.6403,
|
| 265676 |
+
"step": 130715
|
| 265677 |
+
},
|
| 265678 |
+
{
|
| 265679 |
+
"epoch": 1053.41,
|
| 265680 |
+
"learning_rate": 7.921971153846154e-06,
|
| 265681 |
+
"loss": 1.2072,
|
| 265682 |
+
"step": 130720
|
| 265683 |
+
},
|
| 265684 |
+
{
|
| 265685 |
+
"epoch": 1053.45,
|
| 265686 |
+
"learning_rate": 7.921891025641027e-06,
|
| 265687 |
+
"loss": 0.3458,
|
| 265688 |
+
"step": 130725
|
| 265689 |
+
},
|
| 265690 |
+
{
|
| 265691 |
+
"epoch": 1053.49,
|
| 265692 |
+
"learning_rate": 7.921810897435897e-06,
|
| 265693 |
+
"loss": 0.2612,
|
| 265694 |
+
"step": 130730
|
| 265695 |
+
},
|
| 265696 |
+
{
|
| 265697 |
+
"epoch": 1053.53,
|
| 265698 |
+
"learning_rate": 7.92173076923077e-06,
|
| 265699 |
+
"loss": 0.3228,
|
| 265700 |
+
"step": 130735
|
| 265701 |
+
},
|
| 265702 |
+
{
|
| 265703 |
+
"epoch": 1053.57,
|
| 265704 |
+
"learning_rate": 7.921650641025643e-06,
|
| 265705 |
+
"loss": 0.5138,
|
| 265706 |
+
"step": 130740
|
| 265707 |
+
},
|
| 265708 |
+
{
|
| 265709 |
+
"epoch": 1053.61,
|
| 265710 |
+
"learning_rate": 7.921570512820513e-06,
|
| 265711 |
+
"loss": 1.0249,
|
| 265712 |
+
"step": 130745
|
| 265713 |
+
},
|
| 265714 |
+
{
|
| 265715 |
+
"epoch": 1053.65,
|
| 265716 |
+
"learning_rate": 7.921490384615386e-06,
|
| 265717 |
+
"loss": 0.2914,
|
| 265718 |
+
"step": 130750
|
| 265719 |
+
},
|
| 265720 |
+
{
|
| 265721 |
+
"epoch": 1053.69,
|
| 265722 |
+
"learning_rate": 7.921410256410257e-06,
|
| 265723 |
+
"loss": 0.2793,
|
| 265724 |
+
"step": 130755
|
| 265725 |
+
},
|
| 265726 |
+
{
|
| 265727 |
+
"epoch": 1053.73,
|
| 265728 |
+
"learning_rate": 7.921330128205128e-06,
|
| 265729 |
+
"loss": 0.3124,
|
| 265730 |
+
"step": 130760
|
| 265731 |
+
},
|
| 265732 |
+
{
|
| 265733 |
+
"epoch": 1053.77,
|
| 265734 |
+
"learning_rate": 7.92125e-06,
|
| 265735 |
+
"loss": 0.4488,
|
| 265736 |
+
"step": 130765
|
| 265737 |
+
},
|
| 265738 |
+
{
|
| 265739 |
+
"epoch": 1053.81,
|
| 265740 |
+
"learning_rate": 7.921169871794873e-06,
|
| 265741 |
+
"loss": 0.9875,
|
| 265742 |
+
"step": 130770
|
| 265743 |
+
},
|
| 265744 |
+
{
|
| 265745 |
+
"epoch": 1053.85,
|
| 265746 |
+
"learning_rate": 7.921089743589744e-06,
|
| 265747 |
+
"loss": 0.294,
|
| 265748 |
+
"step": 130775
|
| 265749 |
+
},
|
| 265750 |
+
{
|
| 265751 |
+
"epoch": 1053.89,
|
| 265752 |
+
"learning_rate": 7.921009615384616e-06,
|
| 265753 |
+
"loss": 0.2695,
|
| 265754 |
+
"step": 130780
|
| 265755 |
+
},
|
| 265756 |
+
{
|
| 265757 |
+
"epoch": 1053.93,
|
| 265758 |
+
"learning_rate": 7.920929487179487e-06,
|
| 265759 |
+
"loss": 0.3278,
|
| 265760 |
+
"step": 130785
|
| 265761 |
+
},
|
| 265762 |
+
{
|
| 265763 |
+
"epoch": 1053.97,
|
| 265764 |
+
"learning_rate": 7.92084935897436e-06,
|
| 265765 |
+
"loss": 0.642,
|
| 265766 |
+
"step": 130790
|
| 265767 |
+
},
|
| 265768 |
+
{
|
| 265769 |
+
"epoch": 1054.0,
|
| 265770 |
+
"eval_loss": 0.3659282624721527,
|
| 265771 |
+
"eval_runtime": 34.6363,
|
| 265772 |
+
"eval_samples_per_second": 24.137,
|
| 265773 |
+
"eval_steps_per_second": 0.78,
|
| 265774 |
+
"eval_wer": 0.18181818181818182,
|
| 265775 |
+
"step": 130793
|
| 265776 |
+
},
|
| 265777 |
+
{
|
| 265778 |
+
"epoch": 1054.02,
|
| 265779 |
+
"learning_rate": 7.920769230769231e-06,
|
| 265780 |
+
"loss": 0.3544,
|
| 265781 |
+
"step": 130795
|
| 265782 |
+
},
|
| 265783 |
+
{
|
| 265784 |
+
"epoch": 1054.06,
|
| 265785 |
+
"learning_rate": 7.920689102564103e-06,
|
| 265786 |
+
"loss": 0.2746,
|
| 265787 |
+
"step": 130800
|
| 265788 |
+
},
|
| 265789 |
+
{
|
| 265790 |
+
"epoch": 1054.1,
|
| 265791 |
+
"learning_rate": 7.920608974358976e-06,
|
| 265792 |
+
"loss": 0.3017,
|
| 265793 |
+
"step": 130805
|
| 265794 |
+
},
|
| 265795 |
+
{
|
| 265796 |
+
"epoch": 1054.14,
|
| 265797 |
+
"learning_rate": 7.920528846153847e-06,
|
| 265798 |
+
"loss": 0.3507,
|
| 265799 |
+
"step": 130810
|
| 265800 |
+
},
|
| 265801 |
+
{
|
| 265802 |
+
"epoch": 1054.18,
|
| 265803 |
+
"learning_rate": 7.920448717948718e-06,
|
| 265804 |
+
"loss": 0.6455,
|
| 265805 |
+
"step": 130815
|
| 265806 |
+
},
|
| 265807 |
+
{
|
| 265808 |
+
"epoch": 1054.22,
|
| 265809 |
+
"learning_rate": 7.92036858974359e-06,
|
| 265810 |
+
"loss": 0.9787,
|
| 265811 |
+
"step": 130820
|
| 265812 |
+
},
|
| 265813 |
+
{
|
| 265814 |
+
"epoch": 1054.26,
|
| 265815 |
+
"learning_rate": 7.920288461538463e-06,
|
| 265816 |
+
"loss": 0.2413,
|
| 265817 |
+
"step": 130825
|
| 265818 |
+
},
|
| 265819 |
+
{
|
| 265820 |
+
"epoch": 1054.3,
|
| 265821 |
+
"learning_rate": 7.920208333333334e-06,
|
| 265822 |
+
"loss": 0.2911,
|
| 265823 |
+
"step": 130830
|
| 265824 |
+
},
|
| 265825 |
+
{
|
| 265826 |
+
"epoch": 1054.34,
|
| 265827 |
+
"learning_rate": 7.920128205128206e-06,
|
| 265828 |
+
"loss": 0.3419,
|
| 265829 |
+
"step": 130835
|
| 265830 |
+
},
|
| 265831 |
+
{
|
| 265832 |
+
"epoch": 1054.38,
|
| 265833 |
+
"learning_rate": 7.920048076923079e-06,
|
| 265834 |
+
"loss": 0.635,
|
| 265835 |
+
"step": 130840
|
| 265836 |
+
},
|
| 265837 |
+
{
|
| 265838 |
+
"epoch": 1054.42,
|
| 265839 |
+
"learning_rate": 7.91996794871795e-06,
|
| 265840 |
+
"loss": 1.065,
|
| 265841 |
+
"step": 130845
|
| 265842 |
+
},
|
| 265843 |
+
{
|
| 265844 |
+
"epoch": 1054.46,
|
| 265845 |
+
"learning_rate": 7.919887820512821e-06,
|
| 265846 |
+
"loss": 0.3015,
|
| 265847 |
+
"step": 130850
|
| 265848 |
+
},
|
| 265849 |
+
{
|
| 265850 |
+
"epoch": 1054.5,
|
| 265851 |
+
"learning_rate": 7.919807692307693e-06,
|
| 265852 |
+
"loss": 0.3013,
|
| 265853 |
+
"step": 130855
|
| 265854 |
+
},
|
| 265855 |
+
{
|
| 265856 |
+
"epoch": 1054.54,
|
| 265857 |
+
"learning_rate": 7.919727564102566e-06,
|
| 265858 |
+
"loss": 0.3144,
|
| 265859 |
+
"step": 130860
|
| 265860 |
+
},
|
| 265861 |
+
{
|
| 265862 |
+
"epoch": 1054.58,
|
| 265863 |
+
"learning_rate": 7.919647435897435e-06,
|
| 265864 |
+
"loss": 0.6121,
|
| 265865 |
+
"step": 130865
|
| 265866 |
+
},
|
| 265867 |
+
{
|
| 265868 |
+
"epoch": 1054.62,
|
| 265869 |
+
"learning_rate": 7.919567307692308e-06,
|
| 265870 |
+
"loss": 0.9506,
|
| 265871 |
+
"step": 130870
|
| 265872 |
+
},
|
| 265873 |
+
{
|
| 265874 |
+
"epoch": 1054.66,
|
| 265875 |
+
"learning_rate": 7.91948717948718e-06,
|
| 265876 |
+
"loss": 0.2685,
|
| 265877 |
+
"step": 130875
|
| 265878 |
+
},
|
| 265879 |
+
{
|
| 265880 |
+
"epoch": 1054.7,
|
| 265881 |
+
"learning_rate": 7.919407051282051e-06,
|
| 265882 |
+
"loss": 0.3364,
|
| 265883 |
+
"step": 130880
|
| 265884 |
+
},
|
| 265885 |
+
{
|
| 265886 |
+
"epoch": 1054.74,
|
| 265887 |
+
"learning_rate": 7.919326923076923e-06,
|
| 265888 |
+
"loss": 0.3002,
|
| 265889 |
+
"step": 130885
|
| 265890 |
+
},
|
| 265891 |
+
{
|
| 265892 |
+
"epoch": 1054.78,
|
| 265893 |
+
"learning_rate": 7.919246794871796e-06,
|
| 265894 |
+
"loss": 0.5299,
|
| 265895 |
+
"step": 130890
|
| 265896 |
+
},
|
| 265897 |
+
{
|
| 265898 |
+
"epoch": 1054.82,
|
| 265899 |
+
"learning_rate": 7.919166666666667e-06,
|
| 265900 |
+
"loss": 0.9988,
|
| 265901 |
+
"step": 130895
|
| 265902 |
+
},
|
| 265903 |
+
{
|
| 265904 |
+
"epoch": 1054.86,
|
| 265905 |
+
"learning_rate": 7.919086538461538e-06,
|
| 265906 |
+
"loss": 0.2798,
|
| 265907 |
+
"step": 130900
|
| 265908 |
+
},
|
| 265909 |
+
{
|
| 265910 |
+
"epoch": 1054.9,
|
| 265911 |
+
"learning_rate": 7.919006410256411e-06,
|
| 265912 |
+
"loss": 0.2805,
|
| 265913 |
+
"step": 130905
|
| 265914 |
+
},
|
| 265915 |
+
{
|
| 265916 |
+
"epoch": 1054.94,
|
| 265917 |
+
"learning_rate": 7.918926282051283e-06,
|
| 265918 |
+
"loss": 0.2999,
|
| 265919 |
+
"step": 130910
|
| 265920 |
+
},
|
| 265921 |
+
{
|
| 265922 |
+
"epoch": 1054.98,
|
| 265923 |
+
"learning_rate": 7.918846153846154e-06,
|
| 265924 |
+
"loss": 0.6856,
|
| 265925 |
+
"step": 130915
|
| 265926 |
+
},
|
| 265927 |
+
{
|
| 265928 |
+
"epoch": 1055.0,
|
| 265929 |
+
"eval_loss": 0.4071400761604309,
|
| 265930 |
+
"eval_runtime": 36.0409,
|
| 265931 |
+
"eval_samples_per_second": 23.168,
|
| 265932 |
+
"eval_steps_per_second": 0.749,
|
| 265933 |
+
"eval_wer": 0.17985185185185185,
|
| 265934 |
+
"step": 130917
|
| 265935 |
+
},
|
| 265936 |
+
{
|
| 265937 |
+
"epoch": 1047.02,
|
| 265938 |
+
"learning_rate": 7.918766025641025e-06,
|
| 265939 |
+
"loss": 0.3123,
|
| 265940 |
+
"step": 130920
|
| 265941 |
+
},
|
| 265942 |
+
{
|
| 265943 |
+
"epoch": 1047.06,
|
| 265944 |
+
"learning_rate": 7.918685897435898e-06,
|
| 265945 |
+
"loss": 0.2641,
|
| 265946 |
+
"step": 130925
|
| 265947 |
+
},
|
| 265948 |
+
{
|
| 265949 |
+
"epoch": 1047.1,
|
| 265950 |
+
"learning_rate": 7.91860576923077e-06,
|
| 265951 |
+
"loss": 0.3179,
|
| 265952 |
+
"step": 130930
|
| 265953 |
+
},
|
| 265954 |
+
{
|
| 265955 |
+
"epoch": 1047.14,
|
| 265956 |
+
"learning_rate": 7.918525641025641e-06,
|
| 265957 |
+
"loss": 0.3773,
|
| 265958 |
+
"step": 130935
|
| 265959 |
+
},
|
| 265960 |
+
{
|
| 265961 |
+
"epoch": 1047.18,
|
| 265962 |
+
"learning_rate": 7.918445512820514e-06,
|
| 265963 |
+
"loss": 0.6746,
|
| 265964 |
+
"step": 130940
|
| 265965 |
+
},
|
| 265966 |
+
{
|
| 265967 |
+
"epoch": 1047.22,
|
| 265968 |
+
"learning_rate": 7.918365384615386e-06,
|
| 265969 |
+
"loss": 0.8553,
|
| 265970 |
+
"step": 130945
|
| 265971 |
+
},
|
| 265972 |
+
{
|
| 265973 |
+
"epoch": 1047.26,
|
| 265974 |
+
"learning_rate": 7.918285256410257e-06,
|
| 265975 |
+
"loss": 0.2629,
|
| 265976 |
+
"step": 130950
|
| 265977 |
+
},
|
| 265978 |
+
{
|
| 265979 |
+
"epoch": 1047.3,
|
| 265980 |
+
"learning_rate": 7.918205128205128e-06,
|
| 265981 |
+
"loss": 0.2765,
|
| 265982 |
+
"step": 130955
|
| 265983 |
+
},
|
| 265984 |
+
{
|
| 265985 |
+
"epoch": 1047.34,
|
| 265986 |
+
"learning_rate": 7.918125000000001e-06,
|
| 265987 |
+
"loss": 0.344,
|
| 265988 |
+
"step": 130960
|
| 265989 |
+
},
|
| 265990 |
+
{
|
| 265991 |
+
"epoch": 1047.38,
|
| 265992 |
+
"learning_rate": 7.918044871794873e-06,
|
| 265993 |
+
"loss": 0.6662,
|
| 265994 |
+
"step": 130965
|
| 265995 |
+
},
|
| 265996 |
+
{
|
| 265997 |
+
"epoch": 1047.42,
|
| 265998 |
+
"learning_rate": 7.917964743589744e-06,
|
| 265999 |
+
"loss": 0.8144,
|
| 266000 |
+
"step": 130970
|
| 266001 |
+
},
|
| 266002 |
+
{
|
| 266003 |
+
"epoch": 1047.46,
|
| 266004 |
+
"learning_rate": 7.917884615384615e-06,
|
| 266005 |
+
"loss": 0.2599,
|
| 266006 |
+
"step": 130975
|
| 266007 |
+
},
|
| 266008 |
+
{
|
| 266009 |
+
"epoch": 1047.5,
|
| 266010 |
+
"learning_rate": 7.917804487179489e-06,
|
| 266011 |
+
"loss": 0.3034,
|
| 266012 |
+
"step": 130980
|
| 266013 |
+
},
|
| 266014 |
+
{
|
| 266015 |
+
"epoch": 1047.54,
|
| 266016 |
+
"learning_rate": 7.91772435897436e-06,
|
| 266017 |
+
"loss": 0.3557,
|
| 266018 |
+
"step": 130985
|
| 266019 |
+
},
|
| 266020 |
+
{
|
| 266021 |
+
"epoch": 1047.58,
|
| 266022 |
+
"learning_rate": 7.917644230769231e-06,
|
| 266023 |
+
"loss": 0.6239,
|
| 266024 |
+
"step": 130990
|
| 266025 |
+
},
|
| 266026 |
+
{
|
| 266027 |
+
"epoch": 1047.62,
|
| 266028 |
+
"learning_rate": 7.917564102564104e-06,
|
| 266029 |
+
"loss": 0.7923,
|
| 266030 |
+
"step": 130995
|
| 266031 |
+
},
|
| 266032 |
+
{
|
| 266033 |
+
"epoch": 1047.66,
|
| 266034 |
+
"learning_rate": 7.917483974358976e-06,
|
| 266035 |
+
"loss": 0.2728,
|
| 266036 |
+
"step": 131000
|
| 266037 |
+
},
|
| 266038 |
+
{
|
| 266039 |
+
"epoch": 1047.7,
|
| 266040 |
+
"learning_rate": 7.917403846153847e-06,
|
| 266041 |
+
"loss": 0.2946,
|
| 266042 |
+
"step": 131005
|
| 266043 |
+
},
|
| 266044 |
+
{
|
| 266045 |
+
"epoch": 1047.74,
|
| 266046 |
+
"learning_rate": 7.917323717948718e-06,
|
| 266047 |
+
"loss": 0.3642,
|
| 266048 |
+
"step": 131010
|
| 266049 |
+
},
|
| 266050 |
+
{
|
| 266051 |
+
"epoch": 1047.78,
|
| 266052 |
+
"learning_rate": 7.917243589743591e-06,
|
| 266053 |
+
"loss": 0.7808,
|
| 266054 |
+
"step": 131015
|
| 266055 |
+
},
|
| 266056 |
+
{
|
| 266057 |
+
"epoch": 1047.82,
|
| 266058 |
+
"learning_rate": 7.917163461538461e-06,
|
| 266059 |
+
"loss": 0.8567,
|
| 266060 |
+
"step": 131020
|
| 266061 |
+
},
|
| 266062 |
+
{
|
| 266063 |
+
"epoch": 1047.86,
|
| 266064 |
+
"learning_rate": 7.917083333333334e-06,
|
| 266065 |
+
"loss": 0.3216,
|
| 266066 |
+
"step": 131025
|
| 266067 |
+
},
|
| 266068 |
+
{
|
| 266069 |
+
"epoch": 1047.9,
|
| 266070 |
+
"learning_rate": 7.917003205128205e-06,
|
| 266071 |
+
"loss": 0.3074,
|
| 266072 |
+
"step": 131030
|
| 266073 |
+
},
|
| 266074 |
+
{
|
| 266075 |
+
"epoch": 1047.94,
|
| 266076 |
+
"learning_rate": 7.916923076923077e-06,
|
| 266077 |
+
"loss": 0.4341,
|
| 266078 |
+
"step": 131035
|
| 266079 |
+
},
|
| 266080 |
+
{
|
| 266081 |
+
"epoch": 1047.98,
|
| 266082 |
+
"learning_rate": 7.91684294871795e-06,
|
| 266083 |
+
"loss": 0.8587,
|
| 266084 |
+
"step": 131040
|
| 266085 |
+
},
|
| 266086 |
+
{
|
| 266087 |
+
"epoch": 1048.0,
|
| 266088 |
+
"eval_loss": 0.35665926337242126,
|
| 266089 |
+
"eval_runtime": 34.609,
|
| 266090 |
+
"eval_samples_per_second": 24.127,
|
| 266091 |
+
"eval_steps_per_second": 0.78,
|
| 266092 |
+
"eval_wer": 0.1811192429954905,
|
| 266093 |
+
"step": 131042
|
| 266094 |
+
},
|
| 266095 |
+
{
|
| 266096 |
+
"epoch": 1048.02,
|
| 266097 |
+
"learning_rate": 7.916762820512821e-06,
|
| 266098 |
+
"loss": 0.3444,
|
| 266099 |
+
"step": 131045
|
| 266100 |
+
},
|
| 266101 |
+
{
|
| 266102 |
+
"epoch": 1048.06,
|
| 266103 |
+
"learning_rate": 7.916682692307693e-06,
|
| 266104 |
+
"loss": 0.2465,
|
| 266105 |
+
"step": 131050
|
| 266106 |
+
},
|
| 266107 |
+
{
|
| 266108 |
+
"epoch": 1048.1,
|
| 266109 |
+
"learning_rate": 7.916602564102564e-06,
|
| 266110 |
+
"loss": 0.2852,
|
| 266111 |
+
"step": 131055
|
| 266112 |
+
},
|
| 266113 |
+
{
|
| 266114 |
+
"epoch": 1048.14,
|
| 266115 |
+
"learning_rate": 7.916522435897437e-06,
|
| 266116 |
+
"loss": 0.3522,
|
| 266117 |
+
"step": 131060
|
| 266118 |
+
},
|
| 266119 |
+
{
|
| 266120 |
+
"epoch": 1048.18,
|
| 266121 |
+
"learning_rate": 7.916442307692308e-06,
|
| 266122 |
+
"loss": 0.7194,
|
| 266123 |
+
"step": 131065
|
| 266124 |
+
},
|
| 266125 |
+
{
|
| 266126 |
+
"epoch": 1048.22,
|
| 266127 |
+
"learning_rate": 7.91636217948718e-06,
|
| 266128 |
+
"loss": 0.9185,
|
| 266129 |
+
"step": 131070
|
| 266130 |
+
},
|
| 266131 |
+
{
|
| 266132 |
+
"epoch": 1048.26,
|
| 266133 |
+
"learning_rate": 7.916282051282051e-06,
|
| 266134 |
+
"loss": 0.2673,
|
| 266135 |
+
"step": 131075
|
| 266136 |
+
},
|
| 266137 |
+
{
|
| 266138 |
+
"epoch": 1048.3,
|
| 266139 |
+
"learning_rate": 7.916201923076924e-06,
|
| 266140 |
+
"loss": 0.3111,
|
| 266141 |
+
"step": 131080
|
| 266142 |
+
},
|
| 266143 |
+
{
|
| 266144 |
+
"epoch": 1048.34,
|
| 266145 |
+
"learning_rate": 7.916121794871796e-06,
|
| 266146 |
+
"loss": 0.3996,
|
| 266147 |
+
"step": 131085
|
| 266148 |
+
},
|
| 266149 |
+
{
|
| 266150 |
+
"epoch": 1048.38,
|
| 266151 |
+
"learning_rate": 7.916041666666667e-06,
|
| 266152 |
+
"loss": 0.781,
|
| 266153 |
+
"step": 131090
|
| 266154 |
+
},
|
| 266155 |
+
{
|
| 266156 |
+
"epoch": 1048.42,
|
| 266157 |
+
"learning_rate": 7.91596153846154e-06,
|
| 266158 |
+
"loss": 0.814,
|
| 266159 |
+
"step": 131095
|
| 266160 |
+
},
|
| 266161 |
+
{
|
| 266162 |
+
"epoch": 1048.46,
|
| 266163 |
+
"learning_rate": 7.915881410256411e-06,
|
| 266164 |
+
"loss": 0.2743,
|
| 266165 |
+
"step": 131100
|
| 266166 |
+
},
|
| 266167 |
+
{
|
| 266168 |
+
"epoch": 1048.5,
|
| 266169 |
+
"learning_rate": 7.915801282051283e-06,
|
| 266170 |
+
"loss": 0.2987,
|
| 266171 |
+
"step": 131105
|
| 266172 |
+
},
|
| 266173 |
+
{
|
| 266174 |
+
"epoch": 1048.54,
|
| 266175 |
+
"learning_rate": 7.915721153846154e-06,
|
| 266176 |
+
"loss": 0.3916,
|
| 266177 |
+
"step": 131110
|
| 266178 |
+
},
|
| 266179 |
+
{
|
| 266180 |
+
"epoch": 1048.58,
|
| 266181 |
+
"learning_rate": 7.915641025641027e-06,
|
| 266182 |
+
"loss": 0.6682,
|
| 266183 |
+
"step": 131115
|
| 266184 |
+
},
|
| 266185 |
+
{
|
| 266186 |
+
"epoch": 1048.62,
|
| 266187 |
+
"learning_rate": 7.915560897435898e-06,
|
| 266188 |
+
"loss": 0.7889,
|
| 266189 |
+
"step": 131120
|
| 266190 |
+
},
|
| 266191 |
+
{
|
| 266192 |
+
"epoch": 1048.66,
|
| 266193 |
+
"learning_rate": 7.91548076923077e-06,
|
| 266194 |
+
"loss": 0.4369,
|
| 266195 |
+
"step": 131125
|
| 266196 |
+
},
|
| 266197 |
+
{
|
| 266198 |
+
"epoch": 1048.7,
|
| 266199 |
+
"learning_rate": 7.915400641025641e-06,
|
| 266200 |
+
"loss": 0.2953,
|
| 266201 |
+
"step": 131130
|
| 266202 |
+
},
|
| 266203 |
+
{
|
| 266204 |
+
"epoch": 1048.74,
|
| 266205 |
+
"learning_rate": 7.915320512820514e-06,
|
| 266206 |
+
"loss": 0.3176,
|
| 266207 |
+
"step": 131135
|
| 266208 |
+
},
|
| 266209 |
+
{
|
| 266210 |
+
"epoch": 1048.78,
|
| 266211 |
+
"learning_rate": 7.915240384615386e-06,
|
| 266212 |
+
"loss": 0.7356,
|
| 266213 |
+
"step": 131140
|
| 266214 |
+
},
|
| 266215 |
+
{
|
| 266216 |
+
"epoch": 1048.82,
|
| 266217 |
+
"learning_rate": 7.915160256410257e-06,
|
| 266218 |
+
"loss": 1.0768,
|
| 266219 |
+
"step": 131145
|
| 266220 |
+
},
|
| 266221 |
+
{
|
| 266222 |
+
"epoch": 1048.86,
|
| 266223 |
+
"learning_rate": 7.91508012820513e-06,
|
| 266224 |
+
"loss": 0.2829,
|
| 266225 |
+
"step": 131150
|
| 266226 |
+
},
|
| 266227 |
+
{
|
| 266228 |
+
"epoch": 1048.9,
|
| 266229 |
+
"learning_rate": 7.915000000000001e-06,
|
| 266230 |
+
"loss": 0.2821,
|
| 266231 |
+
"step": 131155
|
| 266232 |
+
},
|
| 266233 |
+
{
|
| 266234 |
+
"epoch": 1048.94,
|
| 266235 |
+
"learning_rate": 7.914919871794873e-06,
|
| 266236 |
+
"loss": 0.3997,
|
| 266237 |
+
"step": 131160
|
| 266238 |
+
},
|
| 266239 |
+
{
|
| 266240 |
+
"epoch": 1048.98,
|
| 266241 |
+
"learning_rate": 7.914839743589744e-06,
|
| 266242 |
+
"loss": 0.866,
|
| 266243 |
+
"step": 131165
|
| 266244 |
+
},
|
| 266245 |
+
{
|
| 266246 |
+
"epoch": 1049.0,
|
| 266247 |
+
"eval_loss": 0.4070914089679718,
|
| 266248 |
+
"eval_runtime": 37.1506,
|
| 266249 |
+
"eval_samples_per_second": 22.503,
|
| 266250 |
+
"eval_steps_per_second": 0.727,
|
| 266251 |
+
"eval_wer": 0.18080486392588305,
|
| 266252 |
+
"step": 131167
|
| 266253 |
+
},
|
| 266254 |
+
{
|
| 266255 |
+
"epoch": 1049.02,
|
| 266256 |
+
"learning_rate": 7.914759615384617e-06,
|
| 266257 |
+
"loss": 0.3212,
|
| 266258 |
+
"step": 131170
|
| 266259 |
+
},
|
| 266260 |
+
{
|
| 266261 |
+
"epoch": 1049.06,
|
| 266262 |
+
"learning_rate": 7.914679487179487e-06,
|
| 266263 |
+
"loss": 0.266,
|
| 266264 |
+
"step": 131175
|
| 266265 |
+
},
|
| 266266 |
+
{
|
| 266267 |
+
"epoch": 1049.1,
|
| 266268 |
+
"learning_rate": 7.91459935897436e-06,
|
| 266269 |
+
"loss": 0.293,
|
| 266270 |
+
"step": 131180
|
| 266271 |
+
},
|
| 266272 |
+
{
|
| 266273 |
+
"epoch": 1049.14,
|
| 266274 |
+
"learning_rate": 7.914519230769231e-06,
|
| 266275 |
+
"loss": 0.336,
|
| 266276 |
+
"step": 131185
|
| 266277 |
+
},
|
| 266278 |
+
{
|
| 266279 |
+
"epoch": 1049.18,
|
| 266280 |
+
"learning_rate": 7.914439102564103e-06,
|
| 266281 |
+
"loss": 0.7208,
|
| 266282 |
+
"step": 131190
|
| 266283 |
+
},
|
| 266284 |
+
{
|
| 266285 |
+
"epoch": 1049.22,
|
| 266286 |
+
"learning_rate": 7.914358974358976e-06,
|
| 266287 |
+
"loss": 0.8003,
|
| 266288 |
+
"step": 131195
|
| 266289 |
+
},
|
| 266290 |
+
{
|
| 266291 |
+
"epoch": 1049.26,
|
| 266292 |
+
"learning_rate": 7.914278846153847e-06,
|
| 266293 |
+
"loss": 0.3555,
|
| 266294 |
+
"step": 131200
|
| 266295 |
+
},
|
| 266296 |
+
{
|
| 266297 |
+
"epoch": 1049.3,
|
| 266298 |
+
"learning_rate": 7.914198717948718e-06,
|
| 266299 |
+
"loss": 0.2921,
|
| 266300 |
+
"step": 131205
|
| 266301 |
+
},
|
| 266302 |
+
{
|
| 266303 |
+
"epoch": 1049.34,
|
| 266304 |
+
"learning_rate": 7.91411858974359e-06,
|
| 266305 |
+
"loss": 0.351,
|
| 266306 |
+
"step": 131210
|
| 266307 |
+
},
|
| 266308 |
+
{
|
| 266309 |
+
"epoch": 1049.38,
|
| 266310 |
+
"learning_rate": 7.914038461538463e-06,
|
| 266311 |
+
"loss": 0.7633,
|
| 266312 |
+
"step": 131215
|
| 266313 |
+
},
|
| 266314 |
+
{
|
| 266315 |
+
"epoch": 1049.42,
|
| 266316 |
+
"learning_rate": 7.913958333333334e-06,
|
| 266317 |
+
"loss": 0.7887,
|
| 266318 |
+
"step": 131220
|
| 266319 |
+
},
|
| 266320 |
+
{
|
| 266321 |
+
"epoch": 1049.46,
|
| 266322 |
+
"learning_rate": 7.913878205128205e-06,
|
| 266323 |
+
"loss": 0.3027,
|
| 266324 |
+
"step": 131225
|
| 266325 |
+
},
|
| 266326 |
+
{
|
| 266327 |
+
"epoch": 1049.5,
|
| 266328 |
+
"learning_rate": 7.913798076923077e-06,
|
| 266329 |
+
"loss": 0.3407,
|
| 266330 |
+
"step": 131230
|
| 266331 |
+
},
|
| 266332 |
+
{
|
| 266333 |
+
"epoch": 1049.54,
|
| 266334 |
+
"learning_rate": 7.91371794871795e-06,
|
| 266335 |
+
"loss": 0.3005,
|
| 266336 |
+
"step": 131235
|
| 266337 |
+
},
|
| 266338 |
+
{
|
| 266339 |
+
"epoch": 1049.58,
|
| 266340 |
+
"learning_rate": 7.913637820512821e-06,
|
| 266341 |
+
"loss": 0.6735,
|
| 266342 |
+
"step": 131240
|
| 266343 |
+
},
|
| 266344 |
+
{
|
| 266345 |
+
"epoch": 1049.62,
|
| 266346 |
+
"learning_rate": 7.913557692307693e-06,
|
| 266347 |
+
"loss": 0.9273,
|
| 266348 |
+
"step": 131245
|
| 266349 |
+
},
|
| 266350 |
+
{
|
| 266351 |
+
"epoch": 1049.66,
|
| 266352 |
+
"learning_rate": 7.913477564102566e-06,
|
| 266353 |
+
"loss": 0.3269,
|
| 266354 |
+
"step": 131250
|
| 266355 |
+
},
|
| 266356 |
+
{
|
| 266357 |
+
"epoch": 1049.7,
|
| 266358 |
+
"learning_rate": 7.913397435897437e-06,
|
| 266359 |
+
"loss": 0.2882,
|
| 266360 |
+
"step": 131255
|
| 266361 |
+
},
|
| 266362 |
+
{
|
| 266363 |
+
"epoch": 1049.74,
|
| 266364 |
+
"learning_rate": 7.913317307692308e-06,
|
| 266365 |
+
"loss": 0.3379,
|
| 266366 |
+
"step": 131260
|
| 266367 |
+
},
|
| 266368 |
+
{
|
| 266369 |
+
"epoch": 1049.78,
|
| 266370 |
+
"learning_rate": 7.91323717948718e-06,
|
| 266371 |
+
"loss": 0.7038,
|
| 266372 |
+
"step": 131265
|
| 266373 |
+
},
|
| 266374 |
+
{
|
| 266375 |
+
"epoch": 1049.82,
|
| 266376 |
+
"learning_rate": 7.913157051282053e-06,
|
| 266377 |
+
"loss": 0.8137,
|
| 266378 |
+
"step": 131270
|
| 266379 |
+
},
|
| 266380 |
+
{
|
| 266381 |
+
"epoch": 1049.86,
|
| 266382 |
+
"learning_rate": 7.913076923076924e-06,
|
| 266383 |
+
"loss": 0.284,
|
| 266384 |
+
"step": 131275
|
| 266385 |
+
},
|
| 266386 |
+
{
|
| 266387 |
+
"epoch": 1049.9,
|
| 266388 |
+
"learning_rate": 7.912996794871795e-06,
|
| 266389 |
+
"loss": 0.2658,
|
| 266390 |
+
"step": 131280
|
| 266391 |
+
},
|
| 266392 |
+
{
|
| 266393 |
+
"epoch": 1049.94,
|
| 266394 |
+
"learning_rate": 7.912916666666667e-06,
|
| 266395 |
+
"loss": 0.3415,
|
| 266396 |
+
"step": 131285
|
| 266397 |
+
},
|
| 266398 |
+
{
|
| 266399 |
+
"epoch": 1049.98,
|
| 266400 |
+
"learning_rate": 7.91283653846154e-06,
|
| 266401 |
+
"loss": 0.9049,
|
| 266402 |
+
"step": 131290
|
| 266403 |
+
},
|
| 266404 |
+
{
|
| 266405 |
+
"epoch": 1050.0,
|
| 266406 |
+
"eval_loss": 0.451054185628891,
|
| 266407 |
+
"eval_runtime": 36.5574,
|
| 266408 |
+
"eval_samples_per_second": 22.868,
|
| 266409 |
+
"eval_steps_per_second": 0.739,
|
| 266410 |
+
"eval_wer": 0.1775343674189517,
|
| 266411 |
+
"step": 131292
|
| 266412 |
+
},
|
| 266413 |
+
{
|
| 266414 |
+
"epoch": 1058.02,
|
| 266415 |
+
"learning_rate": 7.912756410256411e-06,
|
| 266416 |
+
"loss": 0.3144,
|
| 266417 |
+
"step": 131295
|
| 266418 |
+
},
|
| 266419 |
+
{
|
| 266420 |
+
"epoch": 1058.06,
|
| 266421 |
+
"learning_rate": 7.912676282051283e-06,
|
| 266422 |
+
"loss": 0.3105,
|
| 266423 |
+
"step": 131300
|
| 266424 |
+
},
|
| 266425 |
+
{
|
| 266426 |
+
"epoch": 1058.1,
|
| 266427 |
+
"learning_rate": 7.912596153846156e-06,
|
| 266428 |
+
"loss": 0.3349,
|
| 266429 |
+
"step": 131305
|
| 266430 |
+
},
|
| 266431 |
+
{
|
| 266432 |
+
"epoch": 1058.15,
|
| 266433 |
+
"learning_rate": 7.912516025641025e-06,
|
| 266434 |
+
"loss": 0.3023,
|
| 266435 |
+
"step": 131310
|
| 266436 |
+
},
|
| 266437 |
+
{
|
| 266438 |
+
"epoch": 1058.19,
|
| 266439 |
+
"learning_rate": 7.912435897435898e-06,
|
| 266440 |
+
"loss": 0.6984,
|
| 266441 |
+
"step": 131315
|
| 266442 |
+
},
|
| 266443 |
+
{
|
| 266444 |
+
"epoch": 1058.23,
|
| 266445 |
+
"learning_rate": 7.91235576923077e-06,
|
| 266446 |
+
"loss": 1.0068,
|
| 266447 |
+
"step": 131320
|
| 266448 |
+
},
|
| 266449 |
+
{
|
| 266450 |
+
"epoch": 1058.27,
|
| 266451 |
+
"learning_rate": 7.912275641025641e-06,
|
| 266452 |
+
"loss": 0.293,
|
| 266453 |
+
"step": 131325
|
| 266454 |
+
},
|
| 266455 |
+
{
|
| 266456 |
+
"epoch": 1058.31,
|
| 266457 |
+
"learning_rate": 7.912195512820512e-06,
|
| 266458 |
+
"loss": 0.3049,
|
| 266459 |
+
"step": 131330
|
| 266460 |
+
},
|
| 266461 |
+
{
|
| 266462 |
+
"epoch": 1058.35,
|
| 266463 |
+
"learning_rate": 7.912115384615385e-06,
|
| 266464 |
+
"loss": 0.3829,
|
| 266465 |
+
"step": 131335
|
| 266466 |
+
},
|
| 266467 |
+
{
|
| 266468 |
+
"epoch": 1058.39,
|
| 266469 |
+
"learning_rate": 7.912035256410257e-06,
|
| 266470 |
+
"loss": 0.6978,
|
| 266471 |
+
"step": 131340
|
| 266472 |
+
},
|
| 266473 |
+
{
|
| 266474 |
+
"epoch": 1058.43,
|
| 266475 |
+
"learning_rate": 7.911955128205128e-06,
|
| 266476 |
+
"loss": 0.7856,
|
| 266477 |
+
"step": 131345
|
| 266478 |
+
},
|
| 266479 |
+
{
|
| 266480 |
+
"epoch": 1058.47,
|
| 266481 |
+
"learning_rate": 7.911875000000001e-06,
|
| 266482 |
+
"loss": 0.2529,
|
| 266483 |
+
"step": 131350
|
| 266484 |
+
},
|
| 266485 |
+
{
|
| 266486 |
+
"epoch": 1058.51,
|
| 266487 |
+
"learning_rate": 7.911794871794873e-06,
|
| 266488 |
+
"loss": 0.364,
|
| 266489 |
+
"step": 131355
|
| 266490 |
+
},
|
| 266491 |
+
{
|
| 266492 |
+
"epoch": 1058.55,
|
| 266493 |
+
"learning_rate": 7.911714743589744e-06,
|
| 266494 |
+
"loss": 0.4424,
|
| 266495 |
+
"step": 131360
|
| 266496 |
+
},
|
| 266497 |
+
{
|
| 266498 |
+
"epoch": 1058.59,
|
| 266499 |
+
"learning_rate": 7.911634615384615e-06,
|
| 266500 |
+
"loss": 0.8688,
|
| 266501 |
+
"step": 131365
|
| 266502 |
+
},
|
| 266503 |
+
{
|
| 266504 |
+
"epoch": 1058.63,
|
| 266505 |
+
"learning_rate": 7.911554487179488e-06,
|
| 266506 |
+
"loss": 0.8466,
|
| 266507 |
+
"step": 131370
|
| 266508 |
+
},
|
| 266509 |
+
{
|
| 266510 |
+
"epoch": 1058.67,
|
| 266511 |
+
"learning_rate": 7.91147435897436e-06,
|
| 266512 |
+
"loss": 0.2929,
|
| 266513 |
+
"step": 131375
|
| 266514 |
+
},
|
| 266515 |
+
{
|
| 266516 |
+
"epoch": 1058.71,
|
| 266517 |
+
"learning_rate": 7.911394230769231e-06,
|
| 266518 |
+
"loss": 0.2828,
|
| 266519 |
+
"step": 131380
|
| 266520 |
+
},
|
| 266521 |
+
{
|
| 266522 |
+
"epoch": 1058.75,
|
| 266523 |
+
"learning_rate": 7.911314102564102e-06,
|
| 266524 |
+
"loss": 0.3502,
|
| 266525 |
+
"step": 131385
|
| 266526 |
+
},
|
| 266527 |
+
{
|
| 266528 |
+
"epoch": 1058.79,
|
| 266529 |
+
"learning_rate": 7.911233974358975e-06,
|
| 266530 |
+
"loss": 0.6644,
|
| 266531 |
+
"step": 131390
|
| 266532 |
+
},
|
| 266533 |
+
{
|
| 266534 |
+
"epoch": 1058.83,
|
| 266535 |
+
"learning_rate": 7.911153846153847e-06,
|
| 266536 |
+
"loss": 0.9036,
|
| 266537 |
+
"step": 131395
|
| 266538 |
+
},
|
| 266539 |
+
{
|
| 266540 |
+
"epoch": 1058.87,
|
| 266541 |
+
"learning_rate": 7.911073717948718e-06,
|
| 266542 |
+
"loss": 0.2584,
|
| 266543 |
+
"step": 131400
|
| 266544 |
+
},
|
| 266545 |
+
{
|
| 266546 |
+
"epoch": 1058.91,
|
| 266547 |
+
"learning_rate": 7.910993589743591e-06,
|
| 266548 |
+
"loss": 0.2873,
|
| 266549 |
+
"step": 131405
|
| 266550 |
+
},
|
| 266551 |
+
{
|
| 266552 |
+
"epoch": 1058.95,
|
| 266553 |
+
"learning_rate": 7.910913461538463e-06,
|
| 266554 |
+
"loss": 0.3913,
|
| 266555 |
+
"step": 131410
|
| 266556 |
+
},
|
| 266557 |
+
{
|
| 266558 |
+
"epoch": 1058.99,
|
| 266559 |
+
"learning_rate": 7.910833333333334e-06,
|
| 266560 |
+
"loss": 0.8848,
|
| 266561 |
+
"step": 131415
|
| 266562 |
+
},
|
| 266563 |
+
{
|
| 266564 |
+
"epoch": 1059.0,
|
| 266565 |
+
"eval_loss": 0.34740835428237915,
|
| 266566 |
+
"eval_runtime": 36.7055,
|
| 266567 |
+
"eval_samples_per_second": 22.776,
|
| 266568 |
+
"eval_steps_per_second": 0.736,
|
| 266569 |
+
"eval_wer": 0.18187225816045802,
|
| 266570 |
+
"step": 131416
|
| 266571 |
+
},
|
| 266572 |
+
{
|
| 266573 |
+
"epoch": 1051.03,
|
| 266574 |
+
"learning_rate": 7.910753205128205e-06,
|
| 266575 |
+
"loss": 0.2529,
|
| 266576 |
+
"step": 131420
|
| 266577 |
+
},
|
| 266578 |
+
{
|
| 266579 |
+
"epoch": 1051.07,
|
| 266580 |
+
"learning_rate": 7.910673076923078e-06,
|
| 266581 |
+
"loss": 0.311,
|
| 266582 |
+
"step": 131425
|
| 266583 |
+
},
|
| 266584 |
+
{
|
| 266585 |
+
"epoch": 1051.11,
|
| 266586 |
+
"learning_rate": 7.91059294871795e-06,
|
| 266587 |
+
"loss": 0.3139,
|
| 266588 |
+
"step": 131430
|
| 266589 |
+
},
|
| 266590 |
+
{
|
| 266591 |
+
"epoch": 1051.15,
|
| 266592 |
+
"learning_rate": 7.910512820512821e-06,
|
| 266593 |
+
"loss": 0.5022,
|
| 266594 |
+
"step": 131435
|
| 266595 |
+
},
|
| 266596 |
+
{
|
| 266597 |
+
"epoch": 1051.19,
|
| 266598 |
+
"learning_rate": 7.910432692307694e-06,
|
| 266599 |
+
"loss": 0.8994,
|
| 266600 |
+
"step": 131440
|
| 266601 |
+
},
|
| 266602 |
+
{
|
| 266603 |
+
"epoch": 1051.23,
|
| 266604 |
+
"learning_rate": 7.910352564102566e-06,
|
| 266605 |
+
"loss": 0.6591,
|
| 266606 |
+
"step": 131445
|
| 266607 |
+
},
|
| 266608 |
+
{
|
| 266609 |
+
"epoch": 1051.27,
|
| 266610 |
+
"learning_rate": 7.910272435897437e-06,
|
| 266611 |
+
"loss": 0.2897,
|
| 266612 |
+
"step": 131450
|
| 266613 |
+
},
|
| 266614 |
+
{
|
| 266615 |
+
"epoch": 1051.31,
|
| 266616 |
+
"learning_rate": 7.910192307692308e-06,
|
| 266617 |
+
"loss": 0.344,
|
| 266618 |
+
"step": 131455
|
| 266619 |
+
},
|
| 266620 |
+
{
|
| 266621 |
+
"epoch": 1051.35,
|
| 266622 |
+
"learning_rate": 7.910112179487181e-06,
|
| 266623 |
+
"loss": 0.4109,
|
| 266624 |
+
"step": 131460
|
| 266625 |
+
},
|
| 266626 |
+
{
|
| 266627 |
+
"epoch": 1051.39,
|
| 266628 |
+
"learning_rate": 7.910032051282051e-06,
|
| 266629 |
+
"loss": 0.8273,
|
| 266630 |
+
"step": 131465
|
| 266631 |
+
},
|
| 266632 |
+
{
|
| 266633 |
+
"epoch": 1051.43,
|
| 266634 |
+
"learning_rate": 7.909951923076924e-06,
|
| 266635 |
+
"loss": 0.6015,
|
| 266636 |
+
"step": 131470
|
| 266637 |
+
},
|
| 266638 |
+
{
|
| 266639 |
+
"epoch": 1051.47,
|
| 266640 |
+
"learning_rate": 7.909871794871795e-06,
|
| 266641 |
+
"loss": 0.2959,
|
| 266642 |
+
"step": 131475
|
| 266643 |
+
},
|
| 266644 |
+
{
|
| 266645 |
+
"epoch": 1051.51,
|
| 266646 |
+
"learning_rate": 7.909791666666667e-06,
|
| 266647 |
+
"loss": 0.2969,
|
| 266648 |
+
"step": 131480
|
| 266649 |
+
},
|
| 266650 |
+
{
|
| 266651 |
+
"epoch": 1051.55,
|
| 266652 |
+
"learning_rate": 7.909711538461538e-06,
|
| 266653 |
+
"loss": 0.3375,
|
| 266654 |
+
"step": 131485
|
| 266655 |
+
},
|
| 266656 |
+
{
|
| 266657 |
+
"epoch": 1051.59,
|
| 266658 |
+
"learning_rate": 7.909631410256411e-06,
|
| 266659 |
+
"loss": 1.0552,
|
| 266660 |
+
"step": 131490
|
| 266661 |
+
},
|
| 266662 |
+
{
|
| 266663 |
+
"epoch": 1051.63,
|
| 266664 |
+
"learning_rate": 7.909551282051282e-06,
|
| 266665 |
+
"loss": 0.6516,
|
| 266666 |
+
"step": 131495
|
| 266667 |
+
},
|
| 266668 |
+
{
|
| 266669 |
+
"epoch": 1051.67,
|
| 266670 |
+
"learning_rate": 7.909471153846154e-06,
|
| 266671 |
+
"loss": 0.2981,
|
| 266672 |
+
"step": 131500
|
| 266673 |
+
},
|
| 266674 |
+
{
|
| 266675 |
+
"epoch": 1051.71,
|
| 266676 |
+
"learning_rate": 7.909391025641027e-06,
|
| 266677 |
+
"loss": 0.3007,
|
| 266678 |
+
"step": 131505
|
| 266679 |
+
},
|
| 266680 |
+
{
|
| 266681 |
+
"epoch": 1051.75,
|
| 266682 |
+
"learning_rate": 7.909310897435898e-06,
|
| 266683 |
+
"loss": 0.3721,
|
| 266684 |
+
"step": 131510
|
| 266685 |
+
},
|
| 266686 |
+
{
|
| 266687 |
+
"epoch": 1051.79,
|
| 266688 |
+
"learning_rate": 7.90923076923077e-06,
|
| 266689 |
+
"loss": 1.0284,
|
| 266690 |
+
"step": 131515
|
| 266691 |
+
},
|
| 266692 |
+
{
|
| 266693 |
+
"epoch": 1051.83,
|
| 266694 |
+
"learning_rate": 7.909150641025641e-06,
|
| 266695 |
+
"loss": 0.667,
|
| 266696 |
+
"step": 131520
|
| 266697 |
+
},
|
| 266698 |
+
{
|
| 266699 |
+
"epoch": 1051.87,
|
| 266700 |
+
"learning_rate": 7.909070512820514e-06,
|
| 266701 |
+
"loss": 0.2979,
|
| 266702 |
+
"step": 131525
|
| 266703 |
+
},
|
| 266704 |
+
{
|
| 266705 |
+
"epoch": 1051.91,
|
| 266706 |
+
"learning_rate": 7.908990384615385e-06,
|
| 266707 |
+
"loss": 0.293,
|
| 266708 |
+
"step": 131530
|
| 266709 |
+
},
|
| 266710 |
+
{
|
| 266711 |
+
"epoch": 1051.95,
|
| 266712 |
+
"learning_rate": 7.908910256410257e-06,
|
| 266713 |
+
"loss": 0.363,
|
| 266714 |
+
"step": 131535
|
| 266715 |
+
},
|
| 266716 |
+
{
|
| 266717 |
+
"epoch": 1051.99,
|
| 266718 |
+
"learning_rate": 7.908830128205128e-06,
|
| 266719 |
+
"loss": 0.9717,
|
| 266720 |
+
"step": 131540
|
| 266721 |
+
},
|
| 266722 |
+
{
|
| 266723 |
+
"epoch": 1052.0,
|
| 266724 |
+
"eval_loss": 0.3263888657093048,
|
| 266725 |
+
"eval_runtime": 36.9741,
|
| 266726 |
+
"eval_samples_per_second": 22.61,
|
| 266727 |
+
"eval_steps_per_second": 0.73,
|
| 266728 |
+
"eval_wer": 0.17733196661297407,
|
| 266729 |
+
"step": 131541
|
| 266730 |
+
},
|
| 266731 |
+
{
|
| 266732 |
+
"epoch": 1060.03,
|
| 266733 |
+
"learning_rate": 7.908750000000001e-06,
|
| 266734 |
+
"loss": 0.2501,
|
| 266735 |
+
"step": 131545
|
| 266736 |
+
},
|
| 266737 |
+
{
|
| 266738 |
+
"epoch": 1060.07,
|
| 266739 |
+
"learning_rate": 7.908669871794873e-06,
|
| 266740 |
+
"loss": 0.2856,
|
| 266741 |
+
"step": 131550
|
| 266742 |
+
},
|
| 266743 |
+
{
|
| 266744 |
+
"epoch": 1060.11,
|
| 266745 |
+
"learning_rate": 7.908589743589744e-06,
|
| 266746 |
+
"loss": 0.4052,
|
| 266747 |
+
"step": 131555
|
| 266748 |
+
},
|
| 266749 |
+
{
|
| 266750 |
+
"epoch": 1060.15,
|
| 266751 |
+
"learning_rate": 7.908509615384617e-06,
|
| 266752 |
+
"loss": 0.391,
|
| 266753 |
+
"step": 131560
|
| 266754 |
+
},
|
| 266755 |
+
{
|
| 266756 |
+
"epoch": 1060.19,
|
| 266757 |
+
"learning_rate": 7.908429487179488e-06,
|
| 266758 |
+
"loss": 1.0021,
|
| 266759 |
+
"step": 131565
|
| 266760 |
+
},
|
| 266761 |
+
{
|
| 266762 |
+
"epoch": 1060.23,
|
| 266763 |
+
"learning_rate": 7.90834935897436e-06,
|
| 266764 |
+
"loss": 0.6628,
|
| 266765 |
+
"step": 131570
|
| 266766 |
+
},
|
| 266767 |
+
{
|
| 266768 |
+
"epoch": 1060.27,
|
| 266769 |
+
"learning_rate": 7.908269230769231e-06,
|
| 266770 |
+
"loss": 0.292,
|
| 266771 |
+
"step": 131575
|
| 266772 |
+
},
|
| 266773 |
+
{
|
| 266774 |
+
"epoch": 1060.31,
|
| 266775 |
+
"learning_rate": 7.908189102564104e-06,
|
| 266776 |
+
"loss": 0.2786,
|
| 266777 |
+
"step": 131580
|
| 266778 |
+
},
|
| 266779 |
+
{
|
| 266780 |
+
"epoch": 1060.35,
|
| 266781 |
+
"learning_rate": 7.908108974358974e-06,
|
| 266782 |
+
"loss": 0.3931,
|
| 266783 |
+
"step": 131585
|
| 266784 |
+
},
|
| 266785 |
+
{
|
| 266786 |
+
"epoch": 1060.39,
|
| 266787 |
+
"learning_rate": 7.908028846153847e-06,
|
| 266788 |
+
"loss": 0.9163,
|
| 266789 |
+
"step": 131590
|
| 266790 |
+
},
|
| 266791 |
+
{
|
| 266792 |
+
"epoch": 1060.43,
|
| 266793 |
+
"learning_rate": 7.90794871794872e-06,
|
| 266794 |
+
"loss": 0.7397,
|
| 266795 |
+
"step": 131595
|
| 266796 |
+
},
|
| 266797 |
+
{
|
| 266798 |
+
"epoch": 1060.47,
|
| 266799 |
+
"learning_rate": 7.90786858974359e-06,
|
| 266800 |
+
"loss": 0.3026,
|
| 266801 |
+
"step": 131600
|
| 266802 |
+
},
|
| 266803 |
+
{
|
| 266804 |
+
"epoch": 1060.51,
|
| 266805 |
+
"learning_rate": 7.907788461538463e-06,
|
| 266806 |
+
"loss": 0.3751,
|
| 266807 |
+
"step": 131605
|
| 266808 |
+
},
|
| 266809 |
+
{
|
| 266810 |
+
"epoch": 1060.55,
|
| 266811 |
+
"learning_rate": 7.907708333333334e-06,
|
| 266812 |
+
"loss": 0.4241,
|
| 266813 |
+
"step": 131610
|
| 266814 |
+
},
|
| 266815 |
+
{
|
| 266816 |
+
"epoch": 1060.59,
|
| 266817 |
+
"learning_rate": 7.907628205128205e-06,
|
| 266818 |
+
"loss": 0.9136,
|
| 266819 |
+
"step": 131615
|
| 266820 |
+
},
|
| 266821 |
+
{
|
| 266822 |
+
"epoch": 1060.63,
|
| 266823 |
+
"learning_rate": 7.907548076923077e-06,
|
| 266824 |
+
"loss": 0.6556,
|
| 266825 |
+
"step": 131620
|
| 266826 |
+
},
|
| 266827 |
+
{
|
| 266828 |
+
"epoch": 1060.67,
|
| 266829 |
+
"learning_rate": 7.90746794871795e-06,
|
| 266830 |
+
"loss": 0.2733,
|
| 266831 |
+
"step": 131625
|
| 266832 |
+
},
|
| 266833 |
+
{
|
| 266834 |
+
"epoch": 1060.71,
|
| 266835 |
+
"learning_rate": 7.907387820512821e-06,
|
| 266836 |
+
"loss": 0.286,
|
| 266837 |
+
"step": 131630
|
| 266838 |
+
},
|
| 266839 |
+
{
|
| 266840 |
+
"epoch": 1060.76,
|
| 266841 |
+
"learning_rate": 7.907307692307692e-06,
|
| 266842 |
+
"loss": 0.3578,
|
| 266843 |
+
"step": 131635
|
| 266844 |
+
},
|
| 266845 |
+
{
|
| 266846 |
+
"epoch": 1060.8,
|
| 266847 |
+
"learning_rate": 7.907227564102564e-06,
|
| 266848 |
+
"loss": 0.8609,
|
| 266849 |
+
"step": 131640
|
| 266850 |
+
},
|
| 266851 |
+
{
|
| 266852 |
+
"epoch": 1060.84,
|
| 266853 |
+
"learning_rate": 7.907147435897437e-06,
|
| 266854 |
+
"loss": 0.6417,
|
| 266855 |
+
"step": 131645
|
| 266856 |
+
},
|
| 266857 |
+
{
|
| 266858 |
+
"epoch": 1060.88,
|
| 266859 |
+
"learning_rate": 7.907067307692308e-06,
|
| 266860 |
+
"loss": 0.2615,
|
| 266861 |
+
"step": 131650
|
| 266862 |
+
},
|
| 266863 |
+
{
|
| 266864 |
+
"epoch": 1060.92,
|
| 266865 |
+
"learning_rate": 7.90698717948718e-06,
|
| 266866 |
+
"loss": 0.3053,
|
| 266867 |
+
"step": 131655
|
| 266868 |
+
},
|
| 266869 |
+
{
|
| 266870 |
+
"epoch": 1060.96,
|
| 266871 |
+
"learning_rate": 7.906907051282053e-06,
|
| 266872 |
+
"loss": 0.4595,
|
| 266873 |
+
"step": 131660
|
| 266874 |
+
},
|
| 266875 |
+
{
|
| 266876 |
+
"epoch": 1061.0,
|
| 266877 |
+
"learning_rate": 7.906826923076924e-06,
|
| 266878 |
+
"loss": 1.1019,
|
| 266879 |
+
"step": 131665
|
| 266880 |
+
},
|
| 266881 |
+
{
|
| 266882 |
+
"epoch": 1061.0,
|
| 266883 |
+
"eval_loss": 0.37313929200172424,
|
| 266884 |
+
"eval_runtime": 37.0356,
|
| 266885 |
+
"eval_samples_per_second": 22.573,
|
| 266886 |
+
"eval_steps_per_second": 0.729,
|
| 266887 |
+
"eval_wer": 0.17062187276626162,
|
| 266888 |
+
"step": 131665
|
| 266889 |
}
|
| 266890 |
],
|
| 266891 |
+
"max_steps": 620000,
|
| 266892 |
"num_train_epochs": 5000,
|
| 266893 |
+
"total_flos": 3.7050613849030145e+20,
|
| 266894 |
"trial_name": null,
|
| 266895 |
"trial_params": null
|
| 266896 |
}
|
model-bin/finetune/base/{checkpoint-131042 → checkpoint-131665}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1630165348.6013007/events.out.tfevents.1630165348.86bb0ddabf9b.4092.211
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:5bd312c019242f10cb81ff01e09dc77c54fabd4b41c91a86b52c1710a4f8971c
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630165745.8357656/events.out.tfevents.1630165745.86bb0ddabf9b.4092.213
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:73b5c1ccd8ea65b3d53fb3803fd280a499b6184f23e2c79fb08f6a70f62a839e
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630166142.3384323/events.out.tfevents.1630166142.86bb0ddabf9b.4092.215
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e54f94394e721c3afdb49b8275059e5af8cb4878d8adf3d951cfa564600073c6
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630166536.6632655/events.out.tfevents.1630166536.86bb0ddabf9b.4092.217
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:14d419fb8201898387798479236ba196e83b9ab5cf814b2379d3722dcb22316c
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630166932.4687943/events.out.tfevents.1630166932.86bb0ddabf9b.4092.219
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:910cf0d597a73dbe5bbd8a253f27a0d67cd5b232d350ae477e730da85fd79be0
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1630165348.86bb0ddabf9b.4092.210
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e41da7bbb75929a5ba0235e4f5eef4030abe6dbe809830537144bca621d293d8
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630165745.86bb0ddabf9b.4092.212
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ce8abe787bce9b0ccc0462b74411d4a6ca30d851b4be51f40fc708988f566578
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630166142.86bb0ddabf9b.4092.214
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:347d57c4061f2783d5cd9547995170a5c7f4b1c5cca6dc4bc4eddbb010aebcf9
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630166536.86bb0ddabf9b.4092.216
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6b9d6e68331f4fe5e64d7abff3689d4d8c66ecad862e594e00e51388ac46d0fa
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630166932.86bb0ddabf9b.4092.218
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:052a04fe29f81dc13f88a3b169e5b973898abfece890298bab0b681c42ee5e31
|
| 3 |
+
size 8622
|