"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-140253 → checkpoint-140875}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-140253 → checkpoint-140875}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-140253 → checkpoint-140875}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-140253 → checkpoint-140875}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-140253 → checkpoint-140875}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-140253 → checkpoint-140875}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-140253 → checkpoint-140875}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-140253 → checkpoint-140875}/trainer_state.json +798 -3
- model-bin/finetune/base/{checkpoint-140253 → checkpoint-140875}/training_args.bin +0 -0
- model-bin/finetune/base/log/1630207944.812348/events.out.tfevents.1630207944.cc93b136ebf5.1086.19 +3 -0
- model-bin/finetune/base/log/1630208395.7358444/events.out.tfevents.1630208395.cc93b136ebf5.1086.21 +3 -0
- model-bin/finetune/base/log/1630208836.6765761/events.out.tfevents.1630208836.cc93b136ebf5.1086.23 +3 -0
- model-bin/finetune/base/log/1630209282.4121106/events.out.tfevents.1630209282.cc93b136ebf5.1086.25 +3 -0
- model-bin/finetune/base/log/1630209722.6420572/events.out.tfevents.1630209722.cc93b136ebf5.1086.27 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630207944.cc93b136ebf5.1086.18 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630208395.cc93b136ebf5.1086.20 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630208836.cc93b136ebf5.1086.22 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630209282.cc93b136ebf5.1086.24 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630209722.cc93b136ebf5.1086.26 +3 -0
model-bin/finetune/base/{checkpoint-140253 → checkpoint-140875}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-140253 → checkpoint-140875}/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:e08e769d2b4499f254b9598b9aa9f90436b21b662d97ff4c54306076791ab6da
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-140253 → checkpoint-140875}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-140253 → checkpoint-140875}/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:a1642b2f9043c2b03cc43559d6c392050e50d6e3596d4970748944f46ad265d5
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-140253 → checkpoint-140875}/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 sha256:3bc6ec187c27ef8b21c444031f6e65fbb66be1aad54c9a3dadb805aed677d27f
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-140253 → checkpoint-140875}/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 sha256:333d0050b2093ed0d375d014ab7c3da5623c8777ffbbd01b1dbb2a1f199188ea
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-140253 → checkpoint-140875}/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:c5acbf4bc880126a565d60482c3807a217efcdc2508209c3c24d996c962ca20a
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-140253 → checkpoint-140875}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1689111747851003,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-132910",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -277809,11 +277809,806 @@
|
|
| 277809 |
"eval_steps_per_second": 0.653,
|
| 277810 |
"eval_wer": 0.17799043062200956,
|
| 277811 |
"step": 140253
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 277812 |
}
|
| 277813 |
],
|
| 277814 |
"max_steps": 625000,
|
| 277815 |
"num_train_epochs": 5000,
|
| 277816 |
-
"total_flos": 3.
|
| 277817 |
"trial_name": null,
|
| 277818 |
"trial_params": null
|
| 277819 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1689111747851003,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-132910",
|
| 4 |
+
"epoch": 1127.0,
|
| 5 |
+
"global_step": 140875,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 277809 |
"eval_steps_per_second": 0.653,
|
| 277810 |
"eval_wer": 0.17799043062200956,
|
| 277811 |
"step": 140253
|
| 277812 |
+
},
|
| 277813 |
+
{
|
| 277814 |
+
"epoch": 1122.02,
|
| 277815 |
+
"learning_rate": 7.751211631663975e-06,
|
| 277816 |
+
"loss": 0.3597,
|
| 277817 |
+
"step": 140255
|
| 277818 |
+
},
|
| 277819 |
+
{
|
| 277820 |
+
"epoch": 1122.06,
|
| 277821 |
+
"learning_rate": 7.751130856219709e-06,
|
| 277822 |
+
"loss": 0.2973,
|
| 277823 |
+
"step": 140260
|
| 277824 |
+
},
|
| 277825 |
+
{
|
| 277826 |
+
"epoch": 1122.1,
|
| 277827 |
+
"learning_rate": 7.751050080775445e-06,
|
| 277828 |
+
"loss": 0.2684,
|
| 277829 |
+
"step": 140265
|
| 277830 |
+
},
|
| 277831 |
+
{
|
| 277832 |
+
"epoch": 1122.14,
|
| 277833 |
+
"learning_rate": 7.750969305331179e-06,
|
| 277834 |
+
"loss": 0.3176,
|
| 277835 |
+
"step": 140270
|
| 277836 |
+
},
|
| 277837 |
+
{
|
| 277838 |
+
"epoch": 1122.18,
|
| 277839 |
+
"learning_rate": 7.750888529886915e-06,
|
| 277840 |
+
"loss": 0.5087,
|
| 277841 |
+
"step": 140275
|
| 277842 |
+
},
|
| 277843 |
+
{
|
| 277844 |
+
"epoch": 1122.22,
|
| 277845 |
+
"learning_rate": 7.75080775444265e-06,
|
| 277846 |
+
"loss": 1.0063,
|
| 277847 |
+
"step": 140280
|
| 277848 |
+
},
|
| 277849 |
+
{
|
| 277850 |
+
"epoch": 1122.25,
|
| 277851 |
+
"learning_rate": 7.750726978998385e-06,
|
| 277852 |
+
"loss": 0.2615,
|
| 277853 |
+
"step": 140285
|
| 277854 |
+
},
|
| 277855 |
+
{
|
| 277856 |
+
"epoch": 1122.29,
|
| 277857 |
+
"learning_rate": 7.75064620355412e-06,
|
| 277858 |
+
"loss": 0.2673,
|
| 277859 |
+
"step": 140290
|
| 277860 |
+
},
|
| 277861 |
+
{
|
| 277862 |
+
"epoch": 1122.33,
|
| 277863 |
+
"learning_rate": 7.750565428109855e-06,
|
| 277864 |
+
"loss": 0.38,
|
| 277865 |
+
"step": 140295
|
| 277866 |
+
},
|
| 277867 |
+
{
|
| 277868 |
+
"epoch": 1122.37,
|
| 277869 |
+
"learning_rate": 7.75048465266559e-06,
|
| 277870 |
+
"loss": 0.6035,
|
| 277871 |
+
"step": 140300
|
| 277872 |
+
},
|
| 277873 |
+
{
|
| 277874 |
+
"epoch": 1122.41,
|
| 277875 |
+
"learning_rate": 7.750403877221325e-06,
|
| 277876 |
+
"loss": 0.9412,
|
| 277877 |
+
"step": 140305
|
| 277878 |
+
},
|
| 277879 |
+
{
|
| 277880 |
+
"epoch": 1122.45,
|
| 277881 |
+
"learning_rate": 7.75032310177706e-06,
|
| 277882 |
+
"loss": 0.2823,
|
| 277883 |
+
"step": 140310
|
| 277884 |
+
},
|
| 277885 |
+
{
|
| 277886 |
+
"epoch": 1122.49,
|
| 277887 |
+
"learning_rate": 7.750242326332795e-06,
|
| 277888 |
+
"loss": 0.2554,
|
| 277889 |
+
"step": 140315
|
| 277890 |
+
},
|
| 277891 |
+
{
|
| 277892 |
+
"epoch": 1122.53,
|
| 277893 |
+
"learning_rate": 7.75016155088853e-06,
|
| 277894 |
+
"loss": 0.3656,
|
| 277895 |
+
"step": 140320
|
| 277896 |
+
},
|
| 277897 |
+
{
|
| 277898 |
+
"epoch": 1122.57,
|
| 277899 |
+
"learning_rate": 7.750080775444265e-06,
|
| 277900 |
+
"loss": 0.5878,
|
| 277901 |
+
"step": 140325
|
| 277902 |
+
},
|
| 277903 |
+
{
|
| 277904 |
+
"epoch": 1122.61,
|
| 277905 |
+
"learning_rate": 7.75e-06,
|
| 277906 |
+
"loss": 1.0295,
|
| 277907 |
+
"step": 140330
|
| 277908 |
+
},
|
| 277909 |
+
{
|
| 277910 |
+
"epoch": 1122.65,
|
| 277911 |
+
"learning_rate": 7.749919224555734e-06,
|
| 277912 |
+
"loss": 0.2567,
|
| 277913 |
+
"step": 140335
|
| 277914 |
+
},
|
| 277915 |
+
{
|
| 277916 |
+
"epoch": 1122.69,
|
| 277917 |
+
"learning_rate": 7.74983844911147e-06,
|
| 277918 |
+
"loss": 0.252,
|
| 277919 |
+
"step": 140340
|
| 277920 |
+
},
|
| 277921 |
+
{
|
| 277922 |
+
"epoch": 1122.73,
|
| 277923 |
+
"learning_rate": 7.749757673667206e-06,
|
| 277924 |
+
"loss": 0.3751,
|
| 277925 |
+
"step": 140345
|
| 277926 |
+
},
|
| 277927 |
+
{
|
| 277928 |
+
"epoch": 1122.77,
|
| 277929 |
+
"learning_rate": 7.74967689822294e-06,
|
| 277930 |
+
"loss": 0.5757,
|
| 277931 |
+
"step": 140350
|
| 277932 |
+
},
|
| 277933 |
+
{
|
| 277934 |
+
"epoch": 1122.81,
|
| 277935 |
+
"learning_rate": 7.749596122778676e-06,
|
| 277936 |
+
"loss": 1.067,
|
| 277937 |
+
"step": 140355
|
| 277938 |
+
},
|
| 277939 |
+
{
|
| 277940 |
+
"epoch": 1122.85,
|
| 277941 |
+
"learning_rate": 7.74951534733441e-06,
|
| 277942 |
+
"loss": 0.2723,
|
| 277943 |
+
"step": 140360
|
| 277944 |
+
},
|
| 277945 |
+
{
|
| 277946 |
+
"epoch": 1122.89,
|
| 277947 |
+
"learning_rate": 7.749434571890146e-06,
|
| 277948 |
+
"loss": 0.2861,
|
| 277949 |
+
"step": 140365
|
| 277950 |
+
},
|
| 277951 |
+
{
|
| 277952 |
+
"epoch": 1122.93,
|
| 277953 |
+
"learning_rate": 7.74935379644588e-06,
|
| 277954 |
+
"loss": 0.3805,
|
| 277955 |
+
"step": 140370
|
| 277956 |
+
},
|
| 277957 |
+
{
|
| 277958 |
+
"epoch": 1122.97,
|
| 277959 |
+
"learning_rate": 7.749273021001616e-06,
|
| 277960 |
+
"loss": 0.5846,
|
| 277961 |
+
"step": 140375
|
| 277962 |
+
},
|
| 277963 |
+
{
|
| 277964 |
+
"epoch": 1123.0,
|
| 277965 |
+
"eval_loss": 0.3891332149505615,
|
| 277966 |
+
"eval_runtime": 41.7147,
|
| 277967 |
+
"eval_samples_per_second": 20.041,
|
| 277968 |
+
"eval_steps_per_second": 0.647,
|
| 277969 |
+
"eval_wer": 0.1826673585884795,
|
| 277970 |
+
"step": 140378
|
| 277971 |
+
},
|
| 277972 |
+
{
|
| 277973 |
+
"epoch": 1132.02,
|
| 277974 |
+
"learning_rate": 7.74919224555735e-06,
|
| 277975 |
+
"loss": 0.3525,
|
| 277976 |
+
"step": 140380
|
| 277977 |
+
},
|
| 277978 |
+
{
|
| 277979 |
+
"epoch": 1132.06,
|
| 277980 |
+
"learning_rate": 7.749111470113086e-06,
|
| 277981 |
+
"loss": 0.284,
|
| 277982 |
+
"step": 140385
|
| 277983 |
+
},
|
| 277984 |
+
{
|
| 277985 |
+
"epoch": 1132.1,
|
| 277986 |
+
"learning_rate": 7.74903069466882e-06,
|
| 277987 |
+
"loss": 0.2877,
|
| 277988 |
+
"step": 140390
|
| 277989 |
+
},
|
| 277990 |
+
{
|
| 277991 |
+
"epoch": 1132.14,
|
| 277992 |
+
"learning_rate": 7.748949919224556e-06,
|
| 277993 |
+
"loss": 0.3197,
|
| 277994 |
+
"step": 140395
|
| 277995 |
+
},
|
| 277996 |
+
{
|
| 277997 |
+
"epoch": 1132.18,
|
| 277998 |
+
"learning_rate": 7.74886914378029e-06,
|
| 277999 |
+
"loss": 0.5683,
|
| 278000 |
+
"step": 140400
|
| 278001 |
+
},
|
| 278002 |
+
{
|
| 278003 |
+
"epoch": 1132.22,
|
| 278004 |
+
"learning_rate": 7.748788368336026e-06,
|
| 278005 |
+
"loss": 1.1453,
|
| 278006 |
+
"step": 140405
|
| 278007 |
+
},
|
| 278008 |
+
{
|
| 278009 |
+
"epoch": 1132.26,
|
| 278010 |
+
"learning_rate": 7.748707592891762e-06,
|
| 278011 |
+
"loss": 0.3112,
|
| 278012 |
+
"step": 140410
|
| 278013 |
+
},
|
| 278014 |
+
{
|
| 278015 |
+
"epoch": 1132.3,
|
| 278016 |
+
"learning_rate": 7.748626817447496e-06,
|
| 278017 |
+
"loss": 0.2833,
|
| 278018 |
+
"step": 140415
|
| 278019 |
+
},
|
| 278020 |
+
{
|
| 278021 |
+
"epoch": 1132.34,
|
| 278022 |
+
"learning_rate": 7.748546042003232e-06,
|
| 278023 |
+
"loss": 0.2826,
|
| 278024 |
+
"step": 140420
|
| 278025 |
+
},
|
| 278026 |
+
{
|
| 278027 |
+
"epoch": 1132.38,
|
| 278028 |
+
"learning_rate": 7.748465266558966e-06,
|
| 278029 |
+
"loss": 0.5825,
|
| 278030 |
+
"step": 140425
|
| 278031 |
+
},
|
| 278032 |
+
{
|
| 278033 |
+
"epoch": 1132.42,
|
| 278034 |
+
"learning_rate": 7.748384491114702e-06,
|
| 278035 |
+
"loss": 1.0479,
|
| 278036 |
+
"step": 140430
|
| 278037 |
+
},
|
| 278038 |
+
{
|
| 278039 |
+
"epoch": 1132.46,
|
| 278040 |
+
"learning_rate": 7.748303715670436e-06,
|
| 278041 |
+
"loss": 0.2943,
|
| 278042 |
+
"step": 140435
|
| 278043 |
+
},
|
| 278044 |
+
{
|
| 278045 |
+
"epoch": 1132.5,
|
| 278046 |
+
"learning_rate": 7.748222940226172e-06,
|
| 278047 |
+
"loss": 0.2851,
|
| 278048 |
+
"step": 140440
|
| 278049 |
+
},
|
| 278050 |
+
{
|
| 278051 |
+
"epoch": 1132.54,
|
| 278052 |
+
"learning_rate": 7.748142164781906e-06,
|
| 278053 |
+
"loss": 0.3373,
|
| 278054 |
+
"step": 140445
|
| 278055 |
+
},
|
| 278056 |
+
{
|
| 278057 |
+
"epoch": 1132.58,
|
| 278058 |
+
"learning_rate": 7.748061389337642e-06,
|
| 278059 |
+
"loss": 0.5474,
|
| 278060 |
+
"step": 140450
|
| 278061 |
+
},
|
| 278062 |
+
{
|
| 278063 |
+
"epoch": 1132.62,
|
| 278064 |
+
"learning_rate": 7.747980613893376e-06,
|
| 278065 |
+
"loss": 0.8722,
|
| 278066 |
+
"step": 140455
|
| 278067 |
+
},
|
| 278068 |
+
{
|
| 278069 |
+
"epoch": 1132.66,
|
| 278070 |
+
"learning_rate": 7.747899838449112e-06,
|
| 278071 |
+
"loss": 0.2612,
|
| 278072 |
+
"step": 140460
|
| 278073 |
+
},
|
| 278074 |
+
{
|
| 278075 |
+
"epoch": 1132.7,
|
| 278076 |
+
"learning_rate": 7.747819063004846e-06,
|
| 278077 |
+
"loss": 0.2451,
|
| 278078 |
+
"step": 140465
|
| 278079 |
+
},
|
| 278080 |
+
{
|
| 278081 |
+
"epoch": 1132.74,
|
| 278082 |
+
"learning_rate": 7.747738287560582e-06,
|
| 278083 |
+
"loss": 0.4083,
|
| 278084 |
+
"step": 140470
|
| 278085 |
+
},
|
| 278086 |
+
{
|
| 278087 |
+
"epoch": 1132.78,
|
| 278088 |
+
"learning_rate": 7.747657512116318e-06,
|
| 278089 |
+
"loss": 0.5761,
|
| 278090 |
+
"step": 140475
|
| 278091 |
+
},
|
| 278092 |
+
{
|
| 278093 |
+
"epoch": 1132.82,
|
| 278094 |
+
"learning_rate": 7.747576736672052e-06,
|
| 278095 |
+
"loss": 0.963,
|
| 278096 |
+
"step": 140480
|
| 278097 |
+
},
|
| 278098 |
+
{
|
| 278099 |
+
"epoch": 1132.86,
|
| 278100 |
+
"learning_rate": 7.747495961227787e-06,
|
| 278101 |
+
"loss": 0.2942,
|
| 278102 |
+
"step": 140485
|
| 278103 |
+
},
|
| 278104 |
+
{
|
| 278105 |
+
"epoch": 1132.9,
|
| 278106 |
+
"learning_rate": 7.747415185783522e-06,
|
| 278107 |
+
"loss": 0.2917,
|
| 278108 |
+
"step": 140490
|
| 278109 |
+
},
|
| 278110 |
+
{
|
| 278111 |
+
"epoch": 1132.94,
|
| 278112 |
+
"learning_rate": 7.747334410339257e-06,
|
| 278113 |
+
"loss": 0.3014,
|
| 278114 |
+
"step": 140495
|
| 278115 |
+
},
|
| 278116 |
+
{
|
| 278117 |
+
"epoch": 1132.98,
|
| 278118 |
+
"learning_rate": 7.747253634894992e-06,
|
| 278119 |
+
"loss": 0.7486,
|
| 278120 |
+
"step": 140500
|
| 278121 |
+
},
|
| 278122 |
+
{
|
| 278123 |
+
"epoch": 1133.0,
|
| 278124 |
+
"eval_loss": 0.35742509365081787,
|
| 278125 |
+
"eval_runtime": 40.3458,
|
| 278126 |
+
"eval_samples_per_second": 20.721,
|
| 278127 |
+
"eval_steps_per_second": 0.669,
|
| 278128 |
+
"eval_wer": 0.18371763301246175,
|
| 278129 |
+
"step": 140502
|
| 278130 |
+
},
|
| 278131 |
+
{
|
| 278132 |
+
"epoch": 1133.02,
|
| 278133 |
+
"learning_rate": 7.747172859450727e-06,
|
| 278134 |
+
"loss": 0.3033,
|
| 278135 |
+
"step": 140505
|
| 278136 |
+
},
|
| 278137 |
+
{
|
| 278138 |
+
"epoch": 1133.06,
|
| 278139 |
+
"learning_rate": 7.747092084006462e-06,
|
| 278140 |
+
"loss": 0.272,
|
| 278141 |
+
"step": 140510
|
| 278142 |
+
},
|
| 278143 |
+
{
|
| 278144 |
+
"epoch": 1133.1,
|
| 278145 |
+
"learning_rate": 7.747011308562197e-06,
|
| 278146 |
+
"loss": 0.248,
|
| 278147 |
+
"step": 140515
|
| 278148 |
+
},
|
| 278149 |
+
{
|
| 278150 |
+
"epoch": 1133.14,
|
| 278151 |
+
"learning_rate": 7.746930533117932e-06,
|
| 278152 |
+
"loss": 0.4174,
|
| 278153 |
+
"step": 140520
|
| 278154 |
+
},
|
| 278155 |
+
{
|
| 278156 |
+
"epoch": 1133.18,
|
| 278157 |
+
"learning_rate": 7.746849757673667e-06,
|
| 278158 |
+
"loss": 0.7289,
|
| 278159 |
+
"step": 140525
|
| 278160 |
+
},
|
| 278161 |
+
{
|
| 278162 |
+
"epoch": 1133.22,
|
| 278163 |
+
"learning_rate": 7.746768982229403e-06,
|
| 278164 |
+
"loss": 0.8883,
|
| 278165 |
+
"step": 140530
|
| 278166 |
+
},
|
| 278167 |
+
{
|
| 278168 |
+
"epoch": 1133.27,
|
| 278169 |
+
"learning_rate": 7.746688206785137e-06,
|
| 278170 |
+
"loss": 0.2844,
|
| 278171 |
+
"step": 140535
|
| 278172 |
+
},
|
| 278173 |
+
{
|
| 278174 |
+
"epoch": 1133.31,
|
| 278175 |
+
"learning_rate": 7.746607431340873e-06,
|
| 278176 |
+
"loss": 0.3098,
|
| 278177 |
+
"step": 140540
|
| 278178 |
+
},
|
| 278179 |
+
{
|
| 278180 |
+
"epoch": 1133.35,
|
| 278181 |
+
"learning_rate": 7.746526655896607e-06,
|
| 278182 |
+
"loss": 0.3032,
|
| 278183 |
+
"step": 140545
|
| 278184 |
+
},
|
| 278185 |
+
{
|
| 278186 |
+
"epoch": 1133.39,
|
| 278187 |
+
"learning_rate": 7.746445880452343e-06,
|
| 278188 |
+
"loss": 0.7064,
|
| 278189 |
+
"step": 140550
|
| 278190 |
+
},
|
| 278191 |
+
{
|
| 278192 |
+
"epoch": 1133.43,
|
| 278193 |
+
"learning_rate": 7.746365105008077e-06,
|
| 278194 |
+
"loss": 0.8404,
|
| 278195 |
+
"step": 140555
|
| 278196 |
+
},
|
| 278197 |
+
{
|
| 278198 |
+
"epoch": 1133.47,
|
| 278199 |
+
"learning_rate": 7.746284329563813e-06,
|
| 278200 |
+
"loss": 0.2751,
|
| 278201 |
+
"step": 140560
|
| 278202 |
+
},
|
| 278203 |
+
{
|
| 278204 |
+
"epoch": 1133.51,
|
| 278205 |
+
"learning_rate": 7.746203554119547e-06,
|
| 278206 |
+
"loss": 0.3098,
|
| 278207 |
+
"step": 140565
|
| 278208 |
+
},
|
| 278209 |
+
{
|
| 278210 |
+
"epoch": 1133.55,
|
| 278211 |
+
"learning_rate": 7.746122778675283e-06,
|
| 278212 |
+
"loss": 0.3457,
|
| 278213 |
+
"step": 140570
|
| 278214 |
+
},
|
| 278215 |
+
{
|
| 278216 |
+
"epoch": 1133.59,
|
| 278217 |
+
"learning_rate": 7.746042003231017e-06,
|
| 278218 |
+
"loss": 0.6564,
|
| 278219 |
+
"step": 140575
|
| 278220 |
+
},
|
| 278221 |
+
{
|
| 278222 |
+
"epoch": 1133.63,
|
| 278223 |
+
"learning_rate": 7.745961227786753e-06,
|
| 278224 |
+
"loss": 0.8131,
|
| 278225 |
+
"step": 140580
|
| 278226 |
+
},
|
| 278227 |
+
{
|
| 278228 |
+
"epoch": 1133.67,
|
| 278229 |
+
"learning_rate": 7.745880452342487e-06,
|
| 278230 |
+
"loss": 0.3265,
|
| 278231 |
+
"step": 140585
|
| 278232 |
+
},
|
| 278233 |
+
{
|
| 278234 |
+
"epoch": 1133.71,
|
| 278235 |
+
"learning_rate": 7.745799676898223e-06,
|
| 278236 |
+
"loss": 0.2746,
|
| 278237 |
+
"step": 140590
|
| 278238 |
+
},
|
| 278239 |
+
{
|
| 278240 |
+
"epoch": 1133.75,
|
| 278241 |
+
"learning_rate": 7.745718901453959e-06,
|
| 278242 |
+
"loss": 0.3602,
|
| 278243 |
+
"step": 140595
|
| 278244 |
+
},
|
| 278245 |
+
{
|
| 278246 |
+
"epoch": 1133.79,
|
| 278247 |
+
"learning_rate": 7.745638126009693e-06,
|
| 278248 |
+
"loss": 0.6824,
|
| 278249 |
+
"step": 140600
|
| 278250 |
+
},
|
| 278251 |
+
{
|
| 278252 |
+
"epoch": 1133.83,
|
| 278253 |
+
"learning_rate": 7.745557350565429e-06,
|
| 278254 |
+
"loss": 0.8483,
|
| 278255 |
+
"step": 140605
|
| 278256 |
+
},
|
| 278257 |
+
{
|
| 278258 |
+
"epoch": 1133.87,
|
| 278259 |
+
"learning_rate": 7.745476575121163e-06,
|
| 278260 |
+
"loss": 0.2513,
|
| 278261 |
+
"step": 140610
|
| 278262 |
+
},
|
| 278263 |
+
{
|
| 278264 |
+
"epoch": 1133.91,
|
| 278265 |
+
"learning_rate": 7.745395799676899e-06,
|
| 278266 |
+
"loss": 0.3173,
|
| 278267 |
+
"step": 140615
|
| 278268 |
+
},
|
| 278269 |
+
{
|
| 278270 |
+
"epoch": 1133.95,
|
| 278271 |
+
"learning_rate": 7.745315024232633e-06,
|
| 278272 |
+
"loss": 0.3481,
|
| 278273 |
+
"step": 140620
|
| 278274 |
+
},
|
| 278275 |
+
{
|
| 278276 |
+
"epoch": 1133.99,
|
| 278277 |
+
"learning_rate": 7.745234248788369e-06,
|
| 278278 |
+
"loss": 0.8273,
|
| 278279 |
+
"step": 140625
|
| 278280 |
+
},
|
| 278281 |
+
{
|
| 278282 |
+
"epoch": 1134.0,
|
| 278283 |
+
"eval_loss": 0.3965514600276947,
|
| 278284 |
+
"eval_runtime": 42.6409,
|
| 278285 |
+
"eval_samples_per_second": 19.606,
|
| 278286 |
+
"eval_steps_per_second": 0.633,
|
| 278287 |
+
"eval_wer": 0.1822267323861989,
|
| 278288 |
+
"step": 140626
|
| 278289 |
+
},
|
| 278290 |
+
{
|
| 278291 |
+
"epoch": 1134.03,
|
| 278292 |
+
"learning_rate": 7.745153473344103e-06,
|
| 278293 |
+
"loss": 0.3243,
|
| 278294 |
+
"step": 140630
|
| 278295 |
+
},
|
| 278296 |
+
{
|
| 278297 |
+
"epoch": 1134.07,
|
| 278298 |
+
"learning_rate": 7.745072697899839e-06,
|
| 278299 |
+
"loss": 0.2471,
|
| 278300 |
+
"step": 140635
|
| 278301 |
+
},
|
| 278302 |
+
{
|
| 278303 |
+
"epoch": 1134.11,
|
| 278304 |
+
"learning_rate": 7.744991922455573e-06,
|
| 278305 |
+
"loss": 0.2839,
|
| 278306 |
+
"step": 140640
|
| 278307 |
+
},
|
| 278308 |
+
{
|
| 278309 |
+
"epoch": 1134.15,
|
| 278310 |
+
"learning_rate": 7.744911147011309e-06,
|
| 278311 |
+
"loss": 0.3705,
|
| 278312 |
+
"step": 140645
|
| 278313 |
+
},
|
| 278314 |
+
{
|
| 278315 |
+
"epoch": 1134.19,
|
| 278316 |
+
"learning_rate": 7.744830371567045e-06,
|
| 278317 |
+
"loss": 0.8877,
|
| 278318 |
+
"step": 140650
|
| 278319 |
+
},
|
| 278320 |
+
{
|
| 278321 |
+
"epoch": 1134.23,
|
| 278322 |
+
"learning_rate": 7.744749596122779e-06,
|
| 278323 |
+
"loss": 0.6301,
|
| 278324 |
+
"step": 140655
|
| 278325 |
+
},
|
| 278326 |
+
{
|
| 278327 |
+
"epoch": 1134.27,
|
| 278328 |
+
"learning_rate": 7.744668820678515e-06,
|
| 278329 |
+
"loss": 0.2588,
|
| 278330 |
+
"step": 140660
|
| 278331 |
+
},
|
| 278332 |
+
{
|
| 278333 |
+
"epoch": 1134.31,
|
| 278334 |
+
"learning_rate": 7.744588045234249e-06,
|
| 278335 |
+
"loss": 0.2737,
|
| 278336 |
+
"step": 140665
|
| 278337 |
+
},
|
| 278338 |
+
{
|
| 278339 |
+
"epoch": 1134.35,
|
| 278340 |
+
"learning_rate": 7.744507269789985e-06,
|
| 278341 |
+
"loss": 0.3338,
|
| 278342 |
+
"step": 140670
|
| 278343 |
+
},
|
| 278344 |
+
{
|
| 278345 |
+
"epoch": 1134.4,
|
| 278346 |
+
"learning_rate": 7.744426494345719e-06,
|
| 278347 |
+
"loss": 0.916,
|
| 278348 |
+
"step": 140675
|
| 278349 |
+
},
|
| 278350 |
+
{
|
| 278351 |
+
"epoch": 1134.44,
|
| 278352 |
+
"learning_rate": 7.744345718901455e-06,
|
| 278353 |
+
"loss": 0.7044,
|
| 278354 |
+
"step": 140680
|
| 278355 |
+
},
|
| 278356 |
+
{
|
| 278357 |
+
"epoch": 1134.48,
|
| 278358 |
+
"learning_rate": 7.744264943457189e-06,
|
| 278359 |
+
"loss": 0.2781,
|
| 278360 |
+
"step": 140685
|
| 278361 |
+
},
|
| 278362 |
+
{
|
| 278363 |
+
"epoch": 1134.52,
|
| 278364 |
+
"learning_rate": 7.744184168012925e-06,
|
| 278365 |
+
"loss": 0.3119,
|
| 278366 |
+
"step": 140690
|
| 278367 |
+
},
|
| 278368 |
+
{
|
| 278369 |
+
"epoch": 1134.56,
|
| 278370 |
+
"learning_rate": 7.744103392568659e-06,
|
| 278371 |
+
"loss": 0.4419,
|
| 278372 |
+
"step": 140695
|
| 278373 |
+
},
|
| 278374 |
+
{
|
| 278375 |
+
"epoch": 1134.6,
|
| 278376 |
+
"learning_rate": 7.744022617124394e-06,
|
| 278377 |
+
"loss": 0.8221,
|
| 278378 |
+
"step": 140700
|
| 278379 |
+
},
|
| 278380 |
+
{
|
| 278381 |
+
"epoch": 1134.64,
|
| 278382 |
+
"learning_rate": 7.74394184168013e-06,
|
| 278383 |
+
"loss": 0.5401,
|
| 278384 |
+
"step": 140705
|
| 278385 |
+
},
|
| 278386 |
+
{
|
| 278387 |
+
"epoch": 1134.68,
|
| 278388 |
+
"learning_rate": 7.743861066235864e-06,
|
| 278389 |
+
"loss": 0.2905,
|
| 278390 |
+
"step": 140710
|
| 278391 |
+
},
|
| 278392 |
+
{
|
| 278393 |
+
"epoch": 1134.72,
|
| 278394 |
+
"learning_rate": 7.7437802907916e-06,
|
| 278395 |
+
"loss": 0.3626,
|
| 278396 |
+
"step": 140715
|
| 278397 |
+
},
|
| 278398 |
+
{
|
| 278399 |
+
"epoch": 1134.76,
|
| 278400 |
+
"learning_rate": 7.743699515347334e-06,
|
| 278401 |
+
"loss": 0.3865,
|
| 278402 |
+
"step": 140720
|
| 278403 |
+
},
|
| 278404 |
+
{
|
| 278405 |
+
"epoch": 1134.8,
|
| 278406 |
+
"learning_rate": 7.74361873990307e-06,
|
| 278407 |
+
"loss": 0.8632,
|
| 278408 |
+
"step": 140725
|
| 278409 |
+
},
|
| 278410 |
+
{
|
| 278411 |
+
"epoch": 1134.84,
|
| 278412 |
+
"learning_rate": 7.743537964458804e-06,
|
| 278413 |
+
"loss": 0.6607,
|
| 278414 |
+
"step": 140730
|
| 278415 |
+
},
|
| 278416 |
+
{
|
| 278417 |
+
"epoch": 1134.88,
|
| 278418 |
+
"learning_rate": 7.74345718901454e-06,
|
| 278419 |
+
"loss": 0.3065,
|
| 278420 |
+
"step": 140735
|
| 278421 |
+
},
|
| 278422 |
+
{
|
| 278423 |
+
"epoch": 1134.92,
|
| 278424 |
+
"learning_rate": 7.743376413570274e-06,
|
| 278425 |
+
"loss": 0.3455,
|
| 278426 |
+
"step": 140740
|
| 278427 |
+
},
|
| 278428 |
+
{
|
| 278429 |
+
"epoch": 1134.96,
|
| 278430 |
+
"learning_rate": 7.74329563812601e-06,
|
| 278431 |
+
"loss": 0.4063,
|
| 278432 |
+
"step": 140745
|
| 278433 |
+
},
|
| 278434 |
+
{
|
| 278435 |
+
"epoch": 1135.0,
|
| 278436 |
+
"learning_rate": 7.743214862681744e-06,
|
| 278437 |
+
"loss": 1.2105,
|
| 278438 |
+
"step": 140750
|
| 278439 |
+
},
|
| 278440 |
+
{
|
| 278441 |
+
"epoch": 1135.0,
|
| 278442 |
+
"eval_loss": 0.32200777530670166,
|
| 278443 |
+
"eval_runtime": 42.577,
|
| 278444 |
+
"eval_samples_per_second": 19.612,
|
| 278445 |
+
"eval_steps_per_second": 0.634,
|
| 278446 |
+
"eval_wer": 0.1698418536226554,
|
| 278447 |
+
"step": 140750
|
| 278448 |
+
},
|
| 278449 |
+
{
|
| 278450 |
+
"epoch": 1126.04,
|
| 278451 |
+
"learning_rate": 7.74313408723748e-06,
|
| 278452 |
+
"loss": 0.8644,
|
| 278453 |
+
"step": 140755
|
| 278454 |
+
},
|
| 278455 |
+
{
|
| 278456 |
+
"epoch": 1126.08,
|
| 278457 |
+
"learning_rate": 7.743053311793214e-06,
|
| 278458 |
+
"loss": 0.2758,
|
| 278459 |
+
"step": 140760
|
| 278460 |
+
},
|
| 278461 |
+
{
|
| 278462 |
+
"epoch": 1126.12,
|
| 278463 |
+
"learning_rate": 7.74297253634895e-06,
|
| 278464 |
+
"loss": 0.3146,
|
| 278465 |
+
"step": 140765
|
| 278466 |
+
},
|
| 278467 |
+
{
|
| 278468 |
+
"epoch": 1126.16,
|
| 278469 |
+
"learning_rate": 7.742891760904686e-06,
|
| 278470 |
+
"loss": 0.5039,
|
| 278471 |
+
"step": 140770
|
| 278472 |
+
},
|
| 278473 |
+
{
|
| 278474 |
+
"epoch": 1126.2,
|
| 278475 |
+
"learning_rate": 7.74281098546042e-06,
|
| 278476 |
+
"loss": 1.2046,
|
| 278477 |
+
"step": 140775
|
| 278478 |
+
},
|
| 278479 |
+
{
|
| 278480 |
+
"epoch": 1126.24,
|
| 278481 |
+
"learning_rate": 7.742730210016156e-06,
|
| 278482 |
+
"loss": 0.3246,
|
| 278483 |
+
"step": 140780
|
| 278484 |
+
},
|
| 278485 |
+
{
|
| 278486 |
+
"epoch": 1126.28,
|
| 278487 |
+
"learning_rate": 7.74264943457189e-06,
|
| 278488 |
+
"loss": 0.2897,
|
| 278489 |
+
"step": 140785
|
| 278490 |
+
},
|
| 278491 |
+
{
|
| 278492 |
+
"epoch": 1126.32,
|
| 278493 |
+
"learning_rate": 7.742568659127626e-06,
|
| 278494 |
+
"loss": 0.2996,
|
| 278495 |
+
"step": 140790
|
| 278496 |
+
},
|
| 278497 |
+
{
|
| 278498 |
+
"epoch": 1126.36,
|
| 278499 |
+
"learning_rate": 7.74248788368336e-06,
|
| 278500 |
+
"loss": 0.4034,
|
| 278501 |
+
"step": 140795
|
| 278502 |
+
},
|
| 278503 |
+
{
|
| 278504 |
+
"epoch": 1126.4,
|
| 278505 |
+
"learning_rate": 7.742407108239096e-06,
|
| 278506 |
+
"loss": 1.2926,
|
| 278507 |
+
"step": 140800
|
| 278508 |
+
},
|
| 278509 |
+
{
|
| 278510 |
+
"epoch": 1126.44,
|
| 278511 |
+
"learning_rate": 7.74232633279483e-06,
|
| 278512 |
+
"loss": 0.3318,
|
| 278513 |
+
"step": 140805
|
| 278514 |
+
},
|
| 278515 |
+
{
|
| 278516 |
+
"epoch": 1126.48,
|
| 278517 |
+
"learning_rate": 7.742245557350566e-06,
|
| 278518 |
+
"loss": 0.2861,
|
| 278519 |
+
"step": 140810
|
| 278520 |
+
},
|
| 278521 |
+
{
|
| 278522 |
+
"epoch": 1126.52,
|
| 278523 |
+
"learning_rate": 7.7421647819063e-06,
|
| 278524 |
+
"loss": 0.2783,
|
| 278525 |
+
"step": 140815
|
| 278526 |
+
},
|
| 278527 |
+
{
|
| 278528 |
+
"epoch": 1126.56,
|
| 278529 |
+
"learning_rate": 7.742084006462036e-06,
|
| 278530 |
+
"loss": 0.4202,
|
| 278531 |
+
"step": 140820
|
| 278532 |
+
},
|
| 278533 |
+
{
|
| 278534 |
+
"epoch": 1126.6,
|
| 278535 |
+
"learning_rate": 7.742003231017772e-06,
|
| 278536 |
+
"loss": 1.0495,
|
| 278537 |
+
"step": 140825
|
| 278538 |
+
},
|
| 278539 |
+
{
|
| 278540 |
+
"epoch": 1126.64,
|
| 278541 |
+
"learning_rate": 7.741922455573506e-06,
|
| 278542 |
+
"loss": 0.2931,
|
| 278543 |
+
"step": 140830
|
| 278544 |
+
},
|
| 278545 |
+
{
|
| 278546 |
+
"epoch": 1126.68,
|
| 278547 |
+
"learning_rate": 7.741841680129242e-06,
|
| 278548 |
+
"loss": 0.2207,
|
| 278549 |
+
"step": 140835
|
| 278550 |
+
},
|
| 278551 |
+
{
|
| 278552 |
+
"epoch": 1126.72,
|
| 278553 |
+
"learning_rate": 7.741760904684976e-06,
|
| 278554 |
+
"loss": 0.281,
|
| 278555 |
+
"step": 140840
|
| 278556 |
+
},
|
| 278557 |
+
{
|
| 278558 |
+
"epoch": 1126.76,
|
| 278559 |
+
"learning_rate": 7.741680129240712e-06,
|
| 278560 |
+
"loss": 0.4278,
|
| 278561 |
+
"step": 140845
|
| 278562 |
+
},
|
| 278563 |
+
{
|
| 278564 |
+
"epoch": 1126.8,
|
| 278565 |
+
"learning_rate": 7.741599353796446e-06,
|
| 278566 |
+
"loss": 1.1692,
|
| 278567 |
+
"step": 140850
|
| 278568 |
+
},
|
| 278569 |
+
{
|
| 278570 |
+
"epoch": 1126.84,
|
| 278571 |
+
"learning_rate": 7.741518578352182e-06,
|
| 278572 |
+
"loss": 0.3526,
|
| 278573 |
+
"step": 140855
|
| 278574 |
+
},
|
| 278575 |
+
{
|
| 278576 |
+
"epoch": 1126.88,
|
| 278577 |
+
"learning_rate": 7.741437802907916e-06,
|
| 278578 |
+
"loss": 0.2201,
|
| 278579 |
+
"step": 140860
|
| 278580 |
+
},
|
| 278581 |
+
{
|
| 278582 |
+
"epoch": 1126.92,
|
| 278583 |
+
"learning_rate": 7.741357027463652e-06,
|
| 278584 |
+
"loss": 0.316,
|
| 278585 |
+
"step": 140865
|
| 278586 |
+
},
|
| 278587 |
+
{
|
| 278588 |
+
"epoch": 1126.96,
|
| 278589 |
+
"learning_rate": 7.741276252019386e-06,
|
| 278590 |
+
"loss": 0.456,
|
| 278591 |
+
"step": 140870
|
| 278592 |
+
},
|
| 278593 |
+
{
|
| 278594 |
+
"epoch": 1127.0,
|
| 278595 |
+
"learning_rate": 7.741195476575122e-06,
|
| 278596 |
+
"loss": 1.2677,
|
| 278597 |
+
"step": 140875
|
| 278598 |
+
},
|
| 278599 |
+
{
|
| 278600 |
+
"epoch": 1127.0,
|
| 278601 |
+
"eval_loss": 0.3179488778114319,
|
| 278602 |
+
"eval_runtime": 43.8327,
|
| 278603 |
+
"eval_samples_per_second": 19.05,
|
| 278604 |
+
"eval_steps_per_second": 0.616,
|
| 278605 |
+
"eval_wer": 0.1768541741311358,
|
| 278606 |
+
"step": 140875
|
| 278607 |
}
|
| 278608 |
],
|
| 278609 |
"max_steps": 625000,
|
| 278610 |
"num_train_epochs": 5000,
|
| 278611 |
+
"total_flos": 3.964100039273066e+20,
|
| 278612 |
"trial_name": null,
|
| 278613 |
"trial_params": null
|
| 278614 |
}
|
model-bin/finetune/base/{checkpoint-140253 → checkpoint-140875}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1630207944.812348/events.out.tfevents.1630207944.cc93b136ebf5.1086.19
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9bb0038ce346d71b093907088ca38c2b2b16883a5e3d10add28d2c424d8327e0
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630208395.7358444/events.out.tfevents.1630208395.cc93b136ebf5.1086.21
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d4593caa2715a9a0e31b64ef2fc43b36191ba597fd60e23db8c5fae35cf2c083
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630208836.6765761/events.out.tfevents.1630208836.cc93b136ebf5.1086.23
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6340c7941a847698c58a28774ea1706aec6badab8d2709f2e2cde14a65a6bdf0
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630209282.4121106/events.out.tfevents.1630209282.cc93b136ebf5.1086.25
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e408630b5c363786025b061330ae8b793e767f9f773a981af39821763adda054
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630209722.6420572/events.out.tfevents.1630209722.cc93b136ebf5.1086.27
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a0809335fbb59faaa6c0c2e2ea9febfc2a180e09abed842d48f2846939f06fc7
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1630207944.cc93b136ebf5.1086.18
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:562cb0d0d237f091e575975de725f51cb1612dbdbadf2f74656c8c865ca05409
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630208395.cc93b136ebf5.1086.20
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f5f3a228287e12060660051d42b0e4bed33878d7690861834cb9f884ce2a5005
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630208836.cc93b136ebf5.1086.22
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1bb29e5116108ae4da557b1250f30f34cab8ace022eb0c12f842aa2351c8cc95
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630209282.cc93b136ebf5.1086.24
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1defd4275b686a10c8155d47b897160a37787a21324464d8e008a616057ef6d6
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630209722.cc93b136ebf5.1086.26
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:59f57bafb62b0b2db966571cdeee919741079ee95f25bd05e341f9afb0854e89
|
| 3 |
+
size 8622
|