"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-42062 → checkpoint-42683}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-42062 → checkpoint-42683}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-42062 → checkpoint-42683}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-42062 → checkpoint-42683}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-42062 → checkpoint-42683}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-42062 → checkpoint-42683}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-42062 → checkpoint-42683}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-42062 → checkpoint-42683}/trainer_state.json +793 -4
- model-bin/finetune/base/{checkpoint-42062 → checkpoint-42683}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629699841.8169625/events.out.tfevents.1629699841.74272264b15c.932.63 +3 -0
- model-bin/finetune/base/log/1629700484.8985372/events.out.tfevents.1629700484.74272264b15c.932.65 +3 -0
- model-bin/finetune/base/log/1629701132.1284912/events.out.tfevents.1629701132.74272264b15c.932.67 +3 -0
- model-bin/finetune/base/log/1629701784.030124/events.out.tfevents.1629701784.74272264b15c.932.69 +3 -0
- model-bin/finetune/base/log/1629702442.3360698/events.out.tfevents.1629702442.74272264b15c.932.71 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629699841.74272264b15c.932.62 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629700484.74272264b15c.932.64 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629701132.74272264b15c.932.66 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629701784.74272264b15c.932.68 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629702442.74272264b15c.932.70 +3 -0
model-bin/finetune/base/{checkpoint-42062 → checkpoint-42683}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-42062 → checkpoint-42683}/optimizer.pt
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 722165009
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:7b1052e9ba8c44adf305b728a8d3eb2a39e7937130f2f26e6b0873900f224420
|
| 3 |
size 722165009
|
model-bin/finetune/base/{checkpoint-42062 → checkpoint-42683}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-42062 → checkpoint-42683}/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:ed5297f8da34aae72091fbed7e63d6a3872859570da6b5b81ca7d654fb57fc9e
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-42062 → checkpoint-42683}/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 sha256:21e17c01a2edda57c0955930579407ad5c86c50aef6562f7d4d320e6eec35d05
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-42062 → checkpoint-42683}/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 sha256:71e4ed352775c3fc23aad60491c42f0b0b2708e2318db50a0c3bf3cdb2897ce5
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-42062 → checkpoint-42683}/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 sha256:8e8b1bad55b9b72bda909488260a59214dda02960ec0a78049bc8b16ea1e468f
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-42062 → checkpoint-42683}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.20178649875530824,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-41813",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -152880,11 +152880,800 @@
|
|
| 152880 |
"eval_steps_per_second": 0.69,
|
| 152881 |
"eval_wer": 0.21088185623026653,
|
| 152882 |
"step": 42062
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 152883 |
}
|
| 152884 |
],
|
| 152885 |
-
"max_steps":
|
| 152886 |
"num_train_epochs": 5000,
|
| 152887 |
-
"total_flos": 1.
|
| 152888 |
"trial_name": null,
|
| 152889 |
"trial_params": null
|
| 152890 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.20178649875530824,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-41813",
|
| 4 |
+
"epoch": 344.0,
|
| 5 |
+
"global_step": 42683,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 152880 |
"eval_steps_per_second": 0.69,
|
| 152881 |
"eval_wer": 0.21088185623026653,
|
| 152882 |
"step": 42062
|
| 152883 |
+
},
|
| 152884 |
+
{
|
| 152885 |
+
"epoch": 339.02,
|
| 152886 |
+
"learning_rate": 9.342163461538462e-06,
|
| 152887 |
+
"loss": 0.3784,
|
| 152888 |
+
"step": 42065
|
| 152889 |
+
},
|
| 152890 |
+
{
|
| 152891 |
+
"epoch": 339.06,
|
| 152892 |
+
"learning_rate": 9.342083333333333e-06,
|
| 152893 |
+
"loss": 0.3695,
|
| 152894 |
+
"step": 42070
|
| 152895 |
+
},
|
| 152896 |
+
{
|
| 152897 |
+
"epoch": 339.1,
|
| 152898 |
+
"learning_rate": 9.342003205128206e-06,
|
| 152899 |
+
"loss": 0.3803,
|
| 152900 |
+
"step": 42075
|
| 152901 |
+
},
|
| 152902 |
+
{
|
| 152903 |
+
"epoch": 339.14,
|
| 152904 |
+
"learning_rate": 9.341923076923078e-06,
|
| 152905 |
+
"loss": 0.4627,
|
| 152906 |
+
"step": 42080
|
| 152907 |
+
},
|
| 152908 |
+
{
|
| 152909 |
+
"epoch": 339.18,
|
| 152910 |
+
"learning_rate": 9.341842948717949e-06,
|
| 152911 |
+
"loss": 0.9156,
|
| 152912 |
+
"step": 42085
|
| 152913 |
+
},
|
| 152914 |
+
{
|
| 152915 |
+
"epoch": 339.22,
|
| 152916 |
+
"learning_rate": 9.34176282051282e-06,
|
| 152917 |
+
"loss": 1.5289,
|
| 152918 |
+
"step": 42090
|
| 152919 |
+
},
|
| 152920 |
+
{
|
| 152921 |
+
"epoch": 339.27,
|
| 152922 |
+
"learning_rate": 9.341682692307693e-06,
|
| 152923 |
+
"loss": 0.3324,
|
| 152924 |
+
"step": 42095
|
| 152925 |
+
},
|
| 152926 |
+
{
|
| 152927 |
+
"epoch": 339.31,
|
| 152928 |
+
"learning_rate": 9.341602564102565e-06,
|
| 152929 |
+
"loss": 0.4285,
|
| 152930 |
+
"step": 42100
|
| 152931 |
+
},
|
| 152932 |
+
{
|
| 152933 |
+
"epoch": 339.35,
|
| 152934 |
+
"learning_rate": 9.341522435897436e-06,
|
| 152935 |
+
"loss": 0.4528,
|
| 152936 |
+
"step": 42105
|
| 152937 |
+
},
|
| 152938 |
+
{
|
| 152939 |
+
"epoch": 339.39,
|
| 152940 |
+
"learning_rate": 9.341442307692309e-06,
|
| 152941 |
+
"loss": 0.8129,
|
| 152942 |
+
"step": 42110
|
| 152943 |
+
},
|
| 152944 |
+
{
|
| 152945 |
+
"epoch": 339.43,
|
| 152946 |
+
"learning_rate": 9.34136217948718e-06,
|
| 152947 |
+
"loss": 0.9654,
|
| 152948 |
+
"step": 42115
|
| 152949 |
+
},
|
| 152950 |
+
{
|
| 152951 |
+
"epoch": 339.47,
|
| 152952 |
+
"learning_rate": 9.341282051282052e-06,
|
| 152953 |
+
"loss": 0.3666,
|
| 152954 |
+
"step": 42120
|
| 152955 |
+
},
|
| 152956 |
+
{
|
| 152957 |
+
"epoch": 339.51,
|
| 152958 |
+
"learning_rate": 9.341201923076923e-06,
|
| 152959 |
+
"loss": 0.361,
|
| 152960 |
+
"step": 42125
|
| 152961 |
+
},
|
| 152962 |
+
{
|
| 152963 |
+
"epoch": 339.55,
|
| 152964 |
+
"learning_rate": 9.341121794871796e-06,
|
| 152965 |
+
"loss": 0.426,
|
| 152966 |
+
"step": 42130
|
| 152967 |
+
},
|
| 152968 |
+
{
|
| 152969 |
+
"epoch": 339.59,
|
| 152970 |
+
"learning_rate": 9.341041666666668e-06,
|
| 152971 |
+
"loss": 0.9028,
|
| 152972 |
+
"step": 42135
|
| 152973 |
+
},
|
| 152974 |
+
{
|
| 152975 |
+
"epoch": 339.63,
|
| 152976 |
+
"learning_rate": 9.340961538461539e-06,
|
| 152977 |
+
"loss": 1.02,
|
| 152978 |
+
"step": 42140
|
| 152979 |
+
},
|
| 152980 |
+
{
|
| 152981 |
+
"epoch": 339.67,
|
| 152982 |
+
"learning_rate": 9.34088141025641e-06,
|
| 152983 |
+
"loss": 0.3656,
|
| 152984 |
+
"step": 42145
|
| 152985 |
+
},
|
| 152986 |
+
{
|
| 152987 |
+
"epoch": 339.71,
|
| 152988 |
+
"learning_rate": 9.340801282051283e-06,
|
| 152989 |
+
"loss": 0.3432,
|
| 152990 |
+
"step": 42150
|
| 152991 |
+
},
|
| 152992 |
+
{
|
| 152993 |
+
"epoch": 339.75,
|
| 152994 |
+
"learning_rate": 9.340721153846155e-06,
|
| 152995 |
+
"loss": 0.46,
|
| 152996 |
+
"step": 42155
|
| 152997 |
+
},
|
| 152998 |
+
{
|
| 152999 |
+
"epoch": 339.79,
|
| 153000 |
+
"learning_rate": 9.340641025641026e-06,
|
| 153001 |
+
"loss": 0.7978,
|
| 153002 |
+
"step": 42160
|
| 153003 |
+
},
|
| 153004 |
+
{
|
| 153005 |
+
"epoch": 339.83,
|
| 153006 |
+
"learning_rate": 9.3405608974359e-06,
|
| 153007 |
+
"loss": 0.9557,
|
| 153008 |
+
"step": 42165
|
| 153009 |
+
},
|
| 153010 |
+
{
|
| 153011 |
+
"epoch": 339.87,
|
| 153012 |
+
"learning_rate": 9.34048076923077e-06,
|
| 153013 |
+
"loss": 0.3946,
|
| 153014 |
+
"step": 42170
|
| 153015 |
+
},
|
| 153016 |
+
{
|
| 153017 |
+
"epoch": 339.91,
|
| 153018 |
+
"learning_rate": 9.340400641025642e-06,
|
| 153019 |
+
"loss": 0.4077,
|
| 153020 |
+
"step": 42175
|
| 153021 |
+
},
|
| 153022 |
+
{
|
| 153023 |
+
"epoch": 339.95,
|
| 153024 |
+
"learning_rate": 9.340320512820513e-06,
|
| 153025 |
+
"loss": 0.4335,
|
| 153026 |
+
"step": 42180
|
| 153027 |
+
},
|
| 153028 |
+
{
|
| 153029 |
+
"epoch": 339.99,
|
| 153030 |
+
"learning_rate": 9.340240384615386e-06,
|
| 153031 |
+
"loss": 0.9906,
|
| 153032 |
+
"step": 42185
|
| 153033 |
+
},
|
| 153034 |
+
{
|
| 153035 |
+
"epoch": 340.0,
|
| 153036 |
+
"eval_loss": 0.4212704598903656,
|
| 153037 |
+
"eval_runtime": 40.0853,
|
| 153038 |
+
"eval_samples_per_second": 20.98,
|
| 153039 |
+
"eval_steps_per_second": 0.674,
|
| 153040 |
+
"eval_wer": 0.21656574717799207,
|
| 153041 |
+
"step": 42186
|
| 153042 |
+
},
|
| 153043 |
+
{
|
| 153044 |
+
"epoch": 340.03,
|
| 153045 |
+
"learning_rate": 9.340160256410256e-06,
|
| 153046 |
+
"loss": 0.3891,
|
| 153047 |
+
"step": 42190
|
| 153048 |
+
},
|
| 153049 |
+
{
|
| 153050 |
+
"epoch": 340.07,
|
| 153051 |
+
"learning_rate": 9.340080128205129e-06,
|
| 153052 |
+
"loss": 0.3385,
|
| 153053 |
+
"step": 42195
|
| 153054 |
+
},
|
| 153055 |
+
{
|
| 153056 |
+
"epoch": 340.11,
|
| 153057 |
+
"learning_rate": 9.340000000000002e-06,
|
| 153058 |
+
"loss": 0.3938,
|
| 153059 |
+
"step": 42200
|
| 153060 |
+
},
|
| 153061 |
+
{
|
| 153062 |
+
"epoch": 340.15,
|
| 153063 |
+
"learning_rate": 9.339919871794872e-06,
|
| 153064 |
+
"loss": 0.4808,
|
| 153065 |
+
"step": 42205
|
| 153066 |
+
},
|
| 153067 |
+
{
|
| 153068 |
+
"epoch": 340.19,
|
| 153069 |
+
"learning_rate": 9.339839743589745e-06,
|
| 153070 |
+
"loss": 1.2282,
|
| 153071 |
+
"step": 42210
|
| 153072 |
+
},
|
| 153073 |
+
{
|
| 153074 |
+
"epoch": 340.23,
|
| 153075 |
+
"learning_rate": 9.339759615384616e-06,
|
| 153076 |
+
"loss": 0.8409,
|
| 153077 |
+
"step": 42215
|
| 153078 |
+
},
|
| 153079 |
+
{
|
| 153080 |
+
"epoch": 340.27,
|
| 153081 |
+
"learning_rate": 9.339679487179487e-06,
|
| 153082 |
+
"loss": 0.416,
|
| 153083 |
+
"step": 42220
|
| 153084 |
+
},
|
| 153085 |
+
{
|
| 153086 |
+
"epoch": 340.31,
|
| 153087 |
+
"learning_rate": 9.339599358974359e-06,
|
| 153088 |
+
"loss": 0.4111,
|
| 153089 |
+
"step": 42225
|
| 153090 |
+
},
|
| 153091 |
+
{
|
| 153092 |
+
"epoch": 340.35,
|
| 153093 |
+
"learning_rate": 9.339519230769232e-06,
|
| 153094 |
+
"loss": 0.512,
|
| 153095 |
+
"step": 42230
|
| 153096 |
+
},
|
| 153097 |
+
{
|
| 153098 |
+
"epoch": 340.39,
|
| 153099 |
+
"learning_rate": 9.339439102564103e-06,
|
| 153100 |
+
"loss": 1.0156,
|
| 153101 |
+
"step": 42235
|
| 153102 |
+
},
|
| 153103 |
+
{
|
| 153104 |
+
"epoch": 340.43,
|
| 153105 |
+
"learning_rate": 9.339358974358975e-06,
|
| 153106 |
+
"loss": 0.7596,
|
| 153107 |
+
"step": 42240
|
| 153108 |
+
},
|
| 153109 |
+
{
|
| 153110 |
+
"epoch": 340.47,
|
| 153111 |
+
"learning_rate": 9.339278846153846e-06,
|
| 153112 |
+
"loss": 0.3551,
|
| 153113 |
+
"step": 42245
|
| 153114 |
+
},
|
| 153115 |
+
{
|
| 153116 |
+
"epoch": 340.51,
|
| 153117 |
+
"learning_rate": 9.339198717948719e-06,
|
| 153118 |
+
"loss": 0.3959,
|
| 153119 |
+
"step": 42250
|
| 153120 |
+
},
|
| 153121 |
+
{
|
| 153122 |
+
"epoch": 340.55,
|
| 153123 |
+
"learning_rate": 9.33911858974359e-06,
|
| 153124 |
+
"loss": 0.5266,
|
| 153125 |
+
"step": 42255
|
| 153126 |
+
},
|
| 153127 |
+
{
|
| 153128 |
+
"epoch": 340.59,
|
| 153129 |
+
"learning_rate": 9.339038461538462e-06,
|
| 153130 |
+
"loss": 1.1358,
|
| 153131 |
+
"step": 42260
|
| 153132 |
+
},
|
| 153133 |
+
{
|
| 153134 |
+
"epoch": 340.63,
|
| 153135 |
+
"learning_rate": 9.338958333333335e-06,
|
| 153136 |
+
"loss": 0.7483,
|
| 153137 |
+
"step": 42265
|
| 153138 |
+
},
|
| 153139 |
+
{
|
| 153140 |
+
"epoch": 340.67,
|
| 153141 |
+
"learning_rate": 9.338878205128206e-06,
|
| 153142 |
+
"loss": 0.3271,
|
| 153143 |
+
"step": 42270
|
| 153144 |
+
},
|
| 153145 |
+
{
|
| 153146 |
+
"epoch": 340.71,
|
| 153147 |
+
"learning_rate": 9.338798076923078e-06,
|
| 153148 |
+
"loss": 0.45,
|
| 153149 |
+
"step": 42275
|
| 153150 |
+
},
|
| 153151 |
+
{
|
| 153152 |
+
"epoch": 340.76,
|
| 153153 |
+
"learning_rate": 9.338717948717949e-06,
|
| 153154 |
+
"loss": 0.543,
|
| 153155 |
+
"step": 42280
|
| 153156 |
+
},
|
| 153157 |
+
{
|
| 153158 |
+
"epoch": 340.8,
|
| 153159 |
+
"learning_rate": 9.338637820512822e-06,
|
| 153160 |
+
"loss": 1.0552,
|
| 153161 |
+
"step": 42285
|
| 153162 |
+
},
|
| 153163 |
+
{
|
| 153164 |
+
"epoch": 340.84,
|
| 153165 |
+
"learning_rate": 9.338557692307693e-06,
|
| 153166 |
+
"loss": 0.7232,
|
| 153167 |
+
"step": 42290
|
| 153168 |
+
},
|
| 153169 |
+
{
|
| 153170 |
+
"epoch": 340.88,
|
| 153171 |
+
"learning_rate": 9.338477564102565e-06,
|
| 153172 |
+
"loss": 0.6572,
|
| 153173 |
+
"step": 42295
|
| 153174 |
+
},
|
| 153175 |
+
{
|
| 153176 |
+
"epoch": 340.92,
|
| 153177 |
+
"learning_rate": 9.338397435897438e-06,
|
| 153178 |
+
"loss": 0.3963,
|
| 153179 |
+
"step": 42300
|
| 153180 |
+
},
|
| 153181 |
+
{
|
| 153182 |
+
"epoch": 340.96,
|
| 153183 |
+
"learning_rate": 9.338317307692309e-06,
|
| 153184 |
+
"loss": 0.6189,
|
| 153185 |
+
"step": 42305
|
| 153186 |
+
},
|
| 153187 |
+
{
|
| 153188 |
+
"epoch": 341.0,
|
| 153189 |
+
"learning_rate": 9.33823717948718e-06,
|
| 153190 |
+
"loss": 1.1999,
|
| 153191 |
+
"step": 42310
|
| 153192 |
+
},
|
| 153193 |
+
{
|
| 153194 |
+
"epoch": 341.0,
|
| 153195 |
+
"eval_loss": 0.41303303837776184,
|
| 153196 |
+
"eval_runtime": 41.1131,
|
| 153197 |
+
"eval_samples_per_second": 20.456,
|
| 153198 |
+
"eval_steps_per_second": 0.657,
|
| 153199 |
+
"eval_wer": 0.2064,
|
| 153200 |
+
"step": 42310
|
| 153201 |
+
},
|
| 153202 |
+
{
|
| 153203 |
+
"epoch": 341.04,
|
| 153204 |
+
"learning_rate": 9.338157051282052e-06,
|
| 153205 |
+
"loss": 0.4639,
|
| 153206 |
+
"step": 42315
|
| 153207 |
+
},
|
| 153208 |
+
{
|
| 153209 |
+
"epoch": 341.08,
|
| 153210 |
+
"learning_rate": 9.338076923076925e-06,
|
| 153211 |
+
"loss": 0.3526,
|
| 153212 |
+
"step": 42320
|
| 153213 |
+
},
|
| 153214 |
+
{
|
| 153215 |
+
"epoch": 341.12,
|
| 153216 |
+
"learning_rate": 9.337996794871794e-06,
|
| 153217 |
+
"loss": 0.3283,
|
| 153218 |
+
"step": 42325
|
| 153219 |
+
},
|
| 153220 |
+
{
|
| 153221 |
+
"epoch": 341.16,
|
| 153222 |
+
"learning_rate": 9.337916666666668e-06,
|
| 153223 |
+
"loss": 0.5197,
|
| 153224 |
+
"step": 42330
|
| 153225 |
+
},
|
| 153226 |
+
{
|
| 153227 |
+
"epoch": 341.2,
|
| 153228 |
+
"learning_rate": 9.337836538461539e-06,
|
| 153229 |
+
"loss": 1.348,
|
| 153230 |
+
"step": 42335
|
| 153231 |
+
},
|
| 153232 |
+
{
|
| 153233 |
+
"epoch": 341.24,
|
| 153234 |
+
"learning_rate": 9.337756410256412e-06,
|
| 153235 |
+
"loss": 0.4031,
|
| 153236 |
+
"step": 42340
|
| 153237 |
+
},
|
| 153238 |
+
{
|
| 153239 |
+
"epoch": 341.28,
|
| 153240 |
+
"learning_rate": 9.337676282051282e-06,
|
| 153241 |
+
"loss": 0.3728,
|
| 153242 |
+
"step": 42345
|
| 153243 |
+
},
|
| 153244 |
+
{
|
| 153245 |
+
"epoch": 341.32,
|
| 153246 |
+
"learning_rate": 9.337596153846155e-06,
|
| 153247 |
+
"loss": 0.3682,
|
| 153248 |
+
"step": 42350
|
| 153249 |
+
},
|
| 153250 |
+
{
|
| 153251 |
+
"epoch": 341.36,
|
| 153252 |
+
"learning_rate": 9.337516025641028e-06,
|
| 153253 |
+
"loss": 0.5583,
|
| 153254 |
+
"step": 42355
|
| 153255 |
+
},
|
| 153256 |
+
{
|
| 153257 |
+
"epoch": 341.4,
|
| 153258 |
+
"learning_rate": 9.337435897435897e-06,
|
| 153259 |
+
"loss": 1.3059,
|
| 153260 |
+
"step": 42360
|
| 153261 |
+
},
|
| 153262 |
+
{
|
| 153263 |
+
"epoch": 341.44,
|
| 153264 |
+
"learning_rate": 9.33735576923077e-06,
|
| 153265 |
+
"loss": 0.4093,
|
| 153266 |
+
"step": 42365
|
| 153267 |
+
},
|
| 153268 |
+
{
|
| 153269 |
+
"epoch": 341.48,
|
| 153270 |
+
"learning_rate": 9.337275641025642e-06,
|
| 153271 |
+
"loss": 0.3289,
|
| 153272 |
+
"step": 42370
|
| 153273 |
+
},
|
| 153274 |
+
{
|
| 153275 |
+
"epoch": 341.52,
|
| 153276 |
+
"learning_rate": 9.337195512820513e-06,
|
| 153277 |
+
"loss": 0.3548,
|
| 153278 |
+
"step": 42375
|
| 153279 |
+
},
|
| 153280 |
+
{
|
| 153281 |
+
"epoch": 341.56,
|
| 153282 |
+
"learning_rate": 9.337115384615385e-06,
|
| 153283 |
+
"loss": 0.5825,
|
| 153284 |
+
"step": 42380
|
| 153285 |
+
},
|
| 153286 |
+
{
|
| 153287 |
+
"epoch": 341.6,
|
| 153288 |
+
"learning_rate": 9.337035256410258e-06,
|
| 153289 |
+
"loss": 1.2219,
|
| 153290 |
+
"step": 42385
|
| 153291 |
+
},
|
| 153292 |
+
{
|
| 153293 |
+
"epoch": 341.64,
|
| 153294 |
+
"learning_rate": 9.336955128205129e-06,
|
| 153295 |
+
"loss": 0.4102,
|
| 153296 |
+
"step": 42390
|
| 153297 |
+
},
|
| 153298 |
+
{
|
| 153299 |
+
"epoch": 341.68,
|
| 153300 |
+
"learning_rate": 9.336875e-06,
|
| 153301 |
+
"loss": 0.3547,
|
| 153302 |
+
"step": 42395
|
| 153303 |
+
},
|
| 153304 |
+
{
|
| 153305 |
+
"epoch": 341.72,
|
| 153306 |
+
"learning_rate": 9.336794871794873e-06,
|
| 153307 |
+
"loss": 0.3619,
|
| 153308 |
+
"step": 42400
|
| 153309 |
+
},
|
| 153310 |
+
{
|
| 153311 |
+
"epoch": 341.76,
|
| 153312 |
+
"learning_rate": 9.336714743589745e-06,
|
| 153313 |
+
"loss": 0.6202,
|
| 153314 |
+
"step": 42405
|
| 153315 |
+
},
|
| 153316 |
+
{
|
| 153317 |
+
"epoch": 341.8,
|
| 153318 |
+
"learning_rate": 9.336634615384616e-06,
|
| 153319 |
+
"loss": 1.4368,
|
| 153320 |
+
"step": 42410
|
| 153321 |
+
},
|
| 153322 |
+
{
|
| 153323 |
+
"epoch": 341.84,
|
| 153324 |
+
"learning_rate": 9.336554487179487e-06,
|
| 153325 |
+
"loss": 0.3809,
|
| 153326 |
+
"step": 42415
|
| 153327 |
+
},
|
| 153328 |
+
{
|
| 153329 |
+
"epoch": 341.88,
|
| 153330 |
+
"learning_rate": 9.33647435897436e-06,
|
| 153331 |
+
"loss": 0.3271,
|
| 153332 |
+
"step": 42420
|
| 153333 |
+
},
|
| 153334 |
+
{
|
| 153335 |
+
"epoch": 341.92,
|
| 153336 |
+
"learning_rate": 9.336394230769232e-06,
|
| 153337 |
+
"loss": 0.3632,
|
| 153338 |
+
"step": 42425
|
| 153339 |
+
},
|
| 153340 |
+
{
|
| 153341 |
+
"epoch": 341.96,
|
| 153342 |
+
"learning_rate": 9.336314102564103e-06,
|
| 153343 |
+
"loss": 0.5768,
|
| 153344 |
+
"step": 42430
|
| 153345 |
+
},
|
| 153346 |
+
{
|
| 153347 |
+
"epoch": 342.0,
|
| 153348 |
+
"eval_loss": 0.4199758768081665,
|
| 153349 |
+
"eval_runtime": 37.6372,
|
| 153350 |
+
"eval_samples_per_second": 22.318,
|
| 153351 |
+
"eval_steps_per_second": 0.717,
|
| 153352 |
+
"eval_wer": 0.21552898441047375,
|
| 153353 |
+
"step": 42434
|
| 153354 |
+
},
|
| 153355 |
+
{
|
| 153356 |
+
"epoch": 339.01,
|
| 153357 |
+
"learning_rate": 9.336233974358975e-06,
|
| 153358 |
+
"loss": 0.4293,
|
| 153359 |
+
"step": 42435
|
| 153360 |
+
},
|
| 153361 |
+
{
|
| 153362 |
+
"epoch": 339.05,
|
| 153363 |
+
"learning_rate": 9.336153846153848e-06,
|
| 153364 |
+
"loss": 0.4086,
|
| 153365 |
+
"step": 42440
|
| 153366 |
+
},
|
| 153367 |
+
{
|
| 153368 |
+
"epoch": 339.09,
|
| 153369 |
+
"learning_rate": 9.336073717948719e-06,
|
| 153370 |
+
"loss": 0.317,
|
| 153371 |
+
"step": 42445
|
| 153372 |
+
},
|
| 153373 |
+
{
|
| 153374 |
+
"epoch": 339.13,
|
| 153375 |
+
"learning_rate": 9.33599358974359e-06,
|
| 153376 |
+
"loss": 0.4464,
|
| 153377 |
+
"step": 42450
|
| 153378 |
+
},
|
| 153379 |
+
{
|
| 153380 |
+
"epoch": 339.17,
|
| 153381 |
+
"learning_rate": 9.335913461538463e-06,
|
| 153382 |
+
"loss": 0.6737,
|
| 153383 |
+
"step": 42455
|
| 153384 |
+
},
|
| 153385 |
+
{
|
| 153386 |
+
"epoch": 339.21,
|
| 153387 |
+
"learning_rate": 9.335833333333335e-06,
|
| 153388 |
+
"loss": 1.3228,
|
| 153389 |
+
"step": 42460
|
| 153390 |
+
},
|
| 153391 |
+
{
|
| 153392 |
+
"epoch": 339.25,
|
| 153393 |
+
"learning_rate": 9.335753205128206e-06,
|
| 153394 |
+
"loss": 0.4304,
|
| 153395 |
+
"step": 42465
|
| 153396 |
+
},
|
| 153397 |
+
{
|
| 153398 |
+
"epoch": 339.29,
|
| 153399 |
+
"learning_rate": 9.335673076923077e-06,
|
| 153400 |
+
"loss": 0.3829,
|
| 153401 |
+
"step": 42470
|
| 153402 |
+
},
|
| 153403 |
+
{
|
| 153404 |
+
"epoch": 339.33,
|
| 153405 |
+
"learning_rate": 9.33559294871795e-06,
|
| 153406 |
+
"loss": 0.5297,
|
| 153407 |
+
"step": 42475
|
| 153408 |
+
},
|
| 153409 |
+
{
|
| 153410 |
+
"epoch": 339.37,
|
| 153411 |
+
"learning_rate": 9.33551282051282e-06,
|
| 153412 |
+
"loss": 0.7097,
|
| 153413 |
+
"step": 42480
|
| 153414 |
+
},
|
| 153415 |
+
{
|
| 153416 |
+
"epoch": 339.41,
|
| 153417 |
+
"learning_rate": 9.335432692307693e-06,
|
| 153418 |
+
"loss": 1.3097,
|
| 153419 |
+
"step": 42485
|
| 153420 |
+
},
|
| 153421 |
+
{
|
| 153422 |
+
"epoch": 339.45,
|
| 153423 |
+
"learning_rate": 9.335352564102565e-06,
|
| 153424 |
+
"loss": 0.379,
|
| 153425 |
+
"step": 42490
|
| 153426 |
+
},
|
| 153427 |
+
{
|
| 153428 |
+
"epoch": 339.49,
|
| 153429 |
+
"learning_rate": 9.335272435897436e-06,
|
| 153430 |
+
"loss": 0.3751,
|
| 153431 |
+
"step": 42495
|
| 153432 |
+
},
|
| 153433 |
+
{
|
| 153434 |
+
"epoch": 339.53,
|
| 153435 |
+
"learning_rate": 9.335192307692309e-06,
|
| 153436 |
+
"loss": 0.4265,
|
| 153437 |
+
"step": 42500
|
| 153438 |
+
},
|
| 153439 |
+
{
|
| 153440 |
+
"epoch": 339.57,
|
| 153441 |
+
"learning_rate": 9.33511217948718e-06,
|
| 153442 |
+
"loss": 0.6526,
|
| 153443 |
+
"step": 42505
|
| 153444 |
+
},
|
| 153445 |
+
{
|
| 153446 |
+
"epoch": 339.61,
|
| 153447 |
+
"learning_rate": 9.335032051282052e-06,
|
| 153448 |
+
"loss": 1.1962,
|
| 153449 |
+
"step": 42510
|
| 153450 |
+
},
|
| 153451 |
+
{
|
| 153452 |
+
"epoch": 339.65,
|
| 153453 |
+
"learning_rate": 9.334951923076923e-06,
|
| 153454 |
+
"loss": 0.3464,
|
| 153455 |
+
"step": 42515
|
| 153456 |
+
},
|
| 153457 |
+
{
|
| 153458 |
+
"epoch": 339.69,
|
| 153459 |
+
"learning_rate": 9.334871794871796e-06,
|
| 153460 |
+
"loss": 0.3569,
|
| 153461 |
+
"step": 42520
|
| 153462 |
+
},
|
| 153463 |
+
{
|
| 153464 |
+
"epoch": 339.73,
|
| 153465 |
+
"learning_rate": 9.334791666666667e-06,
|
| 153466 |
+
"loss": 0.4268,
|
| 153467 |
+
"step": 42525
|
| 153468 |
+
},
|
| 153469 |
+
{
|
| 153470 |
+
"epoch": 339.76,
|
| 153471 |
+
"learning_rate": 9.334711538461539e-06,
|
| 153472 |
+
"loss": 0.6771,
|
| 153473 |
+
"step": 42530
|
| 153474 |
+
},
|
| 153475 |
+
{
|
| 153476 |
+
"epoch": 339.8,
|
| 153477 |
+
"learning_rate": 9.33463141025641e-06,
|
| 153478 |
+
"loss": 1.2774,
|
| 153479 |
+
"step": 42535
|
| 153480 |
+
},
|
| 153481 |
+
{
|
| 153482 |
+
"epoch": 339.84,
|
| 153483 |
+
"learning_rate": 9.334551282051283e-06,
|
| 153484 |
+
"loss": 0.3607,
|
| 153485 |
+
"step": 42540
|
| 153486 |
+
},
|
| 153487 |
+
{
|
| 153488 |
+
"epoch": 339.88,
|
| 153489 |
+
"learning_rate": 9.334471153846155e-06,
|
| 153490 |
+
"loss": 0.3786,
|
| 153491 |
+
"step": 42545
|
| 153492 |
+
},
|
| 153493 |
+
{
|
| 153494 |
+
"epoch": 339.92,
|
| 153495 |
+
"learning_rate": 9.334391025641026e-06,
|
| 153496 |
+
"loss": 0.356,
|
| 153497 |
+
"step": 42550
|
| 153498 |
+
},
|
| 153499 |
+
{
|
| 153500 |
+
"epoch": 339.96,
|
| 153501 |
+
"learning_rate": 9.334310897435899e-06,
|
| 153502 |
+
"loss": 0.6029,
|
| 153503 |
+
"step": 42555
|
| 153504 |
+
},
|
| 153505 |
+
{
|
| 153506 |
+
"epoch": 340.0,
|
| 153507 |
+
"eval_loss": 0.40952178835868835,
|
| 153508 |
+
"eval_runtime": 39.5602,
|
| 153509 |
+
"eval_samples_per_second": 21.233,
|
| 153510 |
+
"eval_steps_per_second": 0.683,
|
| 153511 |
+
"eval_wer": 0.20599112426035504,
|
| 153512 |
+
"step": 42559
|
| 153513 |
+
},
|
| 153514 |
+
{
|
| 153515 |
+
"epoch": 343.01,
|
| 153516 |
+
"learning_rate": 9.33423076923077e-06,
|
| 153517 |
+
"loss": 0.3787,
|
| 153518 |
+
"step": 42560
|
| 153519 |
+
},
|
| 153520 |
+
{
|
| 153521 |
+
"epoch": 343.05,
|
| 153522 |
+
"learning_rate": 9.334150641025642e-06,
|
| 153523 |
+
"loss": 0.3543,
|
| 153524 |
+
"step": 42565
|
| 153525 |
+
},
|
| 153526 |
+
{
|
| 153527 |
+
"epoch": 343.09,
|
| 153528 |
+
"learning_rate": 9.334070512820513e-06,
|
| 153529 |
+
"loss": 0.4122,
|
| 153530 |
+
"step": 42570
|
| 153531 |
+
},
|
| 153532 |
+
{
|
| 153533 |
+
"epoch": 343.13,
|
| 153534 |
+
"learning_rate": 9.333990384615386e-06,
|
| 153535 |
+
"loss": 0.546,
|
| 153536 |
+
"step": 42575
|
| 153537 |
+
},
|
| 153538 |
+
{
|
| 153539 |
+
"epoch": 343.17,
|
| 153540 |
+
"learning_rate": 9.333910256410257e-06,
|
| 153541 |
+
"loss": 0.7324,
|
| 153542 |
+
"step": 42580
|
| 153543 |
+
},
|
| 153544 |
+
{
|
| 153545 |
+
"epoch": 343.21,
|
| 153546 |
+
"learning_rate": 9.333830128205129e-06,
|
| 153547 |
+
"loss": 1.267,
|
| 153548 |
+
"step": 42585
|
| 153549 |
+
},
|
| 153550 |
+
{
|
| 153551 |
+
"epoch": 343.25,
|
| 153552 |
+
"learning_rate": 9.33375e-06,
|
| 153553 |
+
"loss": 0.4008,
|
| 153554 |
+
"step": 42590
|
| 153555 |
+
},
|
| 153556 |
+
{
|
| 153557 |
+
"epoch": 343.29,
|
| 153558 |
+
"learning_rate": 9.333669871794873e-06,
|
| 153559 |
+
"loss": 0.3772,
|
| 153560 |
+
"step": 42595
|
| 153561 |
+
},
|
| 153562 |
+
{
|
| 153563 |
+
"epoch": 343.33,
|
| 153564 |
+
"learning_rate": 9.333589743589745e-06,
|
| 153565 |
+
"loss": 0.3913,
|
| 153566 |
+
"step": 42600
|
| 153567 |
+
},
|
| 153568 |
+
{
|
| 153569 |
+
"epoch": 343.37,
|
| 153570 |
+
"learning_rate": 9.333509615384616e-06,
|
| 153571 |
+
"loss": 0.6971,
|
| 153572 |
+
"step": 42605
|
| 153573 |
+
},
|
| 153574 |
+
{
|
| 153575 |
+
"epoch": 343.41,
|
| 153576 |
+
"learning_rate": 9.333429487179489e-06,
|
| 153577 |
+
"loss": 1.2577,
|
| 153578 |
+
"step": 42610
|
| 153579 |
+
},
|
| 153580 |
+
{
|
| 153581 |
+
"epoch": 343.45,
|
| 153582 |
+
"learning_rate": 9.33334935897436e-06,
|
| 153583 |
+
"loss": 0.4398,
|
| 153584 |
+
"step": 42615
|
| 153585 |
+
},
|
| 153586 |
+
{
|
| 153587 |
+
"epoch": 343.49,
|
| 153588 |
+
"learning_rate": 9.333269230769232e-06,
|
| 153589 |
+
"loss": 0.3462,
|
| 153590 |
+
"step": 42620
|
| 153591 |
+
},
|
| 153592 |
+
{
|
| 153593 |
+
"epoch": 343.53,
|
| 153594 |
+
"learning_rate": 9.333189102564103e-06,
|
| 153595 |
+
"loss": 0.392,
|
| 153596 |
+
"step": 42625
|
| 153597 |
+
},
|
| 153598 |
+
{
|
| 153599 |
+
"epoch": 343.57,
|
| 153600 |
+
"learning_rate": 9.333108974358976e-06,
|
| 153601 |
+
"loss": 0.6626,
|
| 153602 |
+
"step": 42630
|
| 153603 |
+
},
|
| 153604 |
+
{
|
| 153605 |
+
"epoch": 343.61,
|
| 153606 |
+
"learning_rate": 9.333028846153846e-06,
|
| 153607 |
+
"loss": 1.36,
|
| 153608 |
+
"step": 42635
|
| 153609 |
+
},
|
| 153610 |
+
{
|
| 153611 |
+
"epoch": 343.65,
|
| 153612 |
+
"learning_rate": 9.332948717948719e-06,
|
| 153613 |
+
"loss": 0.3631,
|
| 153614 |
+
"step": 42640
|
| 153615 |
+
},
|
| 153616 |
+
{
|
| 153617 |
+
"epoch": 343.69,
|
| 153618 |
+
"learning_rate": 9.33286858974359e-06,
|
| 153619 |
+
"loss": 0.3492,
|
| 153620 |
+
"step": 42645
|
| 153621 |
+
},
|
| 153622 |
+
{
|
| 153623 |
+
"epoch": 343.73,
|
| 153624 |
+
"learning_rate": 9.332788461538462e-06,
|
| 153625 |
+
"loss": 0.3442,
|
| 153626 |
+
"step": 42650
|
| 153627 |
+
},
|
| 153628 |
+
{
|
| 153629 |
+
"epoch": 343.77,
|
| 153630 |
+
"learning_rate": 9.332708333333335e-06,
|
| 153631 |
+
"loss": 0.5685,
|
| 153632 |
+
"step": 42655
|
| 153633 |
+
},
|
| 153634 |
+
{
|
| 153635 |
+
"epoch": 343.81,
|
| 153636 |
+
"learning_rate": 9.332628205128206e-06,
|
| 153637 |
+
"loss": 1.1764,
|
| 153638 |
+
"step": 42660
|
| 153639 |
+
},
|
| 153640 |
+
{
|
| 153641 |
+
"epoch": 343.85,
|
| 153642 |
+
"learning_rate": 9.332548076923077e-06,
|
| 153643 |
+
"loss": 0.4174,
|
| 153644 |
+
"step": 42665
|
| 153645 |
+
},
|
| 153646 |
+
{
|
| 153647 |
+
"epoch": 343.9,
|
| 153648 |
+
"learning_rate": 9.332467948717949e-06,
|
| 153649 |
+
"loss": 0.412,
|
| 153650 |
+
"step": 42670
|
| 153651 |
+
},
|
| 153652 |
+
{
|
| 153653 |
+
"epoch": 343.94,
|
| 153654 |
+
"learning_rate": 9.332387820512822e-06,
|
| 153655 |
+
"loss": 0.4007,
|
| 153656 |
+
"step": 42675
|
| 153657 |
+
},
|
| 153658 |
+
{
|
| 153659 |
+
"epoch": 343.98,
|
| 153660 |
+
"learning_rate": 9.332307692307693e-06,
|
| 153661 |
+
"loss": 0.7013,
|
| 153662 |
+
"step": 42680
|
| 153663 |
+
},
|
| 153664 |
+
{
|
| 153665 |
+
"epoch": 344.0,
|
| 153666 |
+
"eval_loss": 0.44121620059013367,
|
| 153667 |
+
"eval_runtime": 38.9351,
|
| 153668 |
+
"eval_samples_per_second": 21.574,
|
| 153669 |
+
"eval_steps_per_second": 0.693,
|
| 153670 |
+
"eval_wer": 0.20641501200087278,
|
| 153671 |
+
"step": 42683
|
| 153672 |
}
|
| 153673 |
],
|
| 153674 |
+
"max_steps": 620000,
|
| 153675 |
"num_train_epochs": 5000,
|
| 153676 |
+
"total_flos": 1.2009602703328253e+20,
|
| 153677 |
"trial_name": null,
|
| 153678 |
"trial_params": null
|
| 153679 |
}
|
model-bin/finetune/base/{checkpoint-42062 → checkpoint-42683}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629699841.8169625/events.out.tfevents.1629699841.74272264b15c.932.63
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c6f9c54684ebdac28b12d136ce5f08857f5a0ed5c9ec624673cfdeeeb7f46bfa
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629700484.8985372/events.out.tfevents.1629700484.74272264b15c.932.65
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2be98e261e8b8f5bb274c417459fb23fa23a8fc4aae2411665019e57c1e3836f
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629701132.1284912/events.out.tfevents.1629701132.74272264b15c.932.67
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9111ee628957d7e2d885685f8d03636cd638c9e680dc19ca7118c44115fe5494
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629701784.030124/events.out.tfevents.1629701784.74272264b15c.932.69
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:cb50f8ddb29d08073821b5af13f1915bbd1f318b1c58a04aafb245b0d77fe227
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629702442.3360698/events.out.tfevents.1629702442.74272264b15c.932.71
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c932c3dddc1c7c287207befb2385dd6e746638c0c57e056958e837afde700d63
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629699841.74272264b15c.932.62
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d529fb814cd4e26a32c36ec06ef189843ff3cd915c257eec104b2a28c9e5075b
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629700484.74272264b15c.932.64
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:189a2af358fca4f39798745546f79aee91c14caad9def7f2893b48cdc57060ff
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629701132.74272264b15c.932.66
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1780edb3fb23f3269750b6e87d616b1ba16ebc199d14fd6987414b329c06f1c2
|
| 3 |
+
size 8462
|
model-bin/finetune/base/log/events.out.tfevents.1629701784.74272264b15c.932.68
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:4e80f33d3777db70c9404bea2e52547c3c3fdcfb3848f3e6783b0fdebc124005
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629702442.74272264b15c.932.70
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f29003ad9d1701e8e3f7941777b832725a9901dafa7c549a7c9bdd182a43e14a
|
| 3 |
+
size 8622
|