"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-132786 → checkpoint-133530}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-132786 → checkpoint-133530}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-132786 → checkpoint-133530}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-132786 → checkpoint-133530}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-132786 → checkpoint-133530}/rng_state.pth +2 -2
- model-bin/finetune/base/{checkpoint-132786 → checkpoint-133530}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-132786 → checkpoint-133530}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-132786 → checkpoint-133530}/trainer_state.json +954 -6
- model-bin/finetune/base/{checkpoint-132786 → checkpoint-133530}/training_args.bin +0 -0
- model-bin/finetune/base/log/1630171494.1221683/events.out.tfevents.1630171494.86bb0ddabf9b.4092.241 +3 -0
- model-bin/finetune/base/log/1630171878.5782204/events.out.tfevents.1630171878.86bb0ddabf9b.4092.243 +3 -0
- model-bin/finetune/base/log/1630172268.8488228/events.out.tfevents.1630172268.86bb0ddabf9b.4092.245 +3 -0
- model-bin/finetune/base/log/1630172658.9374063/events.out.tfevents.1630172658.86bb0ddabf9b.4092.247 +3 -0
- model-bin/finetune/base/log/1630173043.8917527/events.out.tfevents.1630173043.86bb0ddabf9b.4092.249 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630171494.86bb0ddabf9b.4092.240 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630171878.86bb0ddabf9b.4092.242 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630172268.86bb0ddabf9b.4092.244 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630172658.86bb0ddabf9b.4092.246 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630173043.86bb0ddabf9b.4092.248 +3 -0
model-bin/finetune/base/{checkpoint-132786 → checkpoint-133530}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-132786 → checkpoint-133530}/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 sha256:0131aea205c7b47df894a41ad960996f174a622cdb2a0e466e6c420dddd66eb3
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-132786 → checkpoint-133530}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-132786 → checkpoint-133530}/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 sha256:984900319b1571a20bc0eff8f0132123a9e2552f902a51b84b84e449b1d6e8ad
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-132786 → checkpoint-133530}/rng_state.pth
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:24e5394362e1ac456498e0ab0081d483b216a02a39ead2f59423571ab3011946
|
| 3 |
+
size 14503
|
model-bin/finetune/base/{checkpoint-132786 → checkpoint-133530}/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 sha256:137fd28074ff7a16078afbfd1e5c0ac3943fefd3026a1f7f05234a0ac2c9e1f9
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-132786 → checkpoint-133530}/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 sha256:8ec230c409ad1e6eb7cefef319d6d09881c43c2e19a2e87131bedb122fa28933
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-132786 → checkpoint-133530}/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,
|
|
@@ -268311,11 +268311,959 @@
|
|
| 268311 |
"eval_steps_per_second": 0.765,
|
| 268312 |
"eval_wer": 0.18351754320896077,
|
| 268313 |
"step": 132786
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 268314 |
}
|
| 268315 |
],
|
| 268316 |
-
"max_steps":
|
| 268317 |
"num_train_epochs": 5000,
|
| 268318 |
-
"total_flos": 3.
|
| 268319 |
"trial_name": null,
|
| 268320 |
"trial_params": null
|
| 268321 |
}
|
|
|
|
| 1 |
{
|
| 2 |
+
"best_metric": 0.1689111747851003,
|
| 3 |
+
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-132910",
|
| 4 |
+
"epoch": 1075.995983935743,
|
| 5 |
+
"global_step": 133530,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 268311 |
"eval_steps_per_second": 0.765,
|
| 268312 |
"eval_wer": 0.18351754320896077,
|
| 268313 |
"step": 132786
|
| 268314 |
+
},
|
| 268315 |
+
{
|
| 268316 |
+
"epoch": 1070.03,
|
| 268317 |
+
"learning_rate": 7.88883012820513e-06,
|
| 268318 |
+
"loss": 0.3289,
|
| 268319 |
+
"step": 132790
|
| 268320 |
+
},
|
| 268321 |
+
{
|
| 268322 |
+
"epoch": 1070.07,
|
| 268323 |
+
"learning_rate": 7.888750000000001e-06,
|
| 268324 |
+
"loss": 0.3066,
|
| 268325 |
+
"step": 132795
|
| 268326 |
+
},
|
| 268327 |
+
{
|
| 268328 |
+
"epoch": 1070.11,
|
| 268329 |
+
"learning_rate": 7.888669871794873e-06,
|
| 268330 |
+
"loss": 0.3182,
|
| 268331 |
+
"step": 132800
|
| 268332 |
+
},
|
| 268333 |
+
{
|
| 268334 |
+
"epoch": 1070.15,
|
| 268335 |
+
"learning_rate": 7.888589743589744e-06,
|
| 268336 |
+
"loss": 0.3897,
|
| 268337 |
+
"step": 132805
|
| 268338 |
+
},
|
| 268339 |
+
{
|
| 268340 |
+
"epoch": 1070.19,
|
| 268341 |
+
"learning_rate": 7.888509615384617e-06,
|
| 268342 |
+
"loss": 0.9667,
|
| 268343 |
+
"step": 132810
|
| 268344 |
+
},
|
| 268345 |
+
{
|
| 268346 |
+
"epoch": 1070.23,
|
| 268347 |
+
"learning_rate": 7.888429487179487e-06,
|
| 268348 |
+
"loss": 0.651,
|
| 268349 |
+
"step": 132815
|
| 268350 |
+
},
|
| 268351 |
+
{
|
| 268352 |
+
"epoch": 1070.27,
|
| 268353 |
+
"learning_rate": 7.88834935897436e-06,
|
| 268354 |
+
"loss": 0.2818,
|
| 268355 |
+
"step": 132820
|
| 268356 |
+
},
|
| 268357 |
+
{
|
| 268358 |
+
"epoch": 1070.31,
|
| 268359 |
+
"learning_rate": 7.888269230769231e-06,
|
| 268360 |
+
"loss": 0.2908,
|
| 268361 |
+
"step": 132825
|
| 268362 |
+
},
|
| 268363 |
+
{
|
| 268364 |
+
"epoch": 1070.35,
|
| 268365 |
+
"learning_rate": 7.888189102564103e-06,
|
| 268366 |
+
"loss": 0.4134,
|
| 268367 |
+
"step": 132830
|
| 268368 |
+
},
|
| 268369 |
+
{
|
| 268370 |
+
"epoch": 1070.4,
|
| 268371 |
+
"learning_rate": 7.888108974358974e-06,
|
| 268372 |
+
"loss": 1.0368,
|
| 268373 |
+
"step": 132835
|
| 268374 |
+
},
|
| 268375 |
+
{
|
| 268376 |
+
"epoch": 1070.44,
|
| 268377 |
+
"learning_rate": 7.888028846153847e-06,
|
| 268378 |
+
"loss": 0.6221,
|
| 268379 |
+
"step": 132840
|
| 268380 |
+
},
|
| 268381 |
+
{
|
| 268382 |
+
"epoch": 1070.48,
|
| 268383 |
+
"learning_rate": 7.887948717948718e-06,
|
| 268384 |
+
"loss": 0.2551,
|
| 268385 |
+
"step": 132845
|
| 268386 |
+
},
|
| 268387 |
+
{
|
| 268388 |
+
"epoch": 1070.52,
|
| 268389 |
+
"learning_rate": 7.88786858974359e-06,
|
| 268390 |
+
"loss": 0.3402,
|
| 268391 |
+
"step": 132850
|
| 268392 |
+
},
|
| 268393 |
+
{
|
| 268394 |
+
"epoch": 1070.56,
|
| 268395 |
+
"learning_rate": 7.887788461538463e-06,
|
| 268396 |
+
"loss": 0.377,
|
| 268397 |
+
"step": 132855
|
| 268398 |
+
},
|
| 268399 |
+
{
|
| 268400 |
+
"epoch": 1070.6,
|
| 268401 |
+
"learning_rate": 7.887708333333334e-06,
|
| 268402 |
+
"loss": 0.9088,
|
| 268403 |
+
"step": 132860
|
| 268404 |
+
},
|
| 268405 |
+
{
|
| 268406 |
+
"epoch": 1070.64,
|
| 268407 |
+
"learning_rate": 7.887628205128205e-06,
|
| 268408 |
+
"loss": 0.5489,
|
| 268409 |
+
"step": 132865
|
| 268410 |
+
},
|
| 268411 |
+
{
|
| 268412 |
+
"epoch": 1070.68,
|
| 268413 |
+
"learning_rate": 7.887548076923077e-06,
|
| 268414 |
+
"loss": 0.3574,
|
| 268415 |
+
"step": 132870
|
| 268416 |
+
},
|
| 268417 |
+
{
|
| 268418 |
+
"epoch": 1070.72,
|
| 268419 |
+
"learning_rate": 7.88746794871795e-06,
|
| 268420 |
+
"loss": 0.2859,
|
| 268421 |
+
"step": 132875
|
| 268422 |
+
},
|
| 268423 |
+
{
|
| 268424 |
+
"epoch": 1070.76,
|
| 268425 |
+
"learning_rate": 7.887387820512821e-06,
|
| 268426 |
+
"loss": 0.3736,
|
| 268427 |
+
"step": 132880
|
| 268428 |
+
},
|
| 268429 |
+
{
|
| 268430 |
+
"epoch": 1070.8,
|
| 268431 |
+
"learning_rate": 7.887307692307693e-06,
|
| 268432 |
+
"loss": 0.9012,
|
| 268433 |
+
"step": 132885
|
| 268434 |
+
},
|
| 268435 |
+
{
|
| 268436 |
+
"epoch": 1070.84,
|
| 268437 |
+
"learning_rate": 7.887227564102566e-06,
|
| 268438 |
+
"loss": 0.6423,
|
| 268439 |
+
"step": 132890
|
| 268440 |
+
},
|
| 268441 |
+
{
|
| 268442 |
+
"epoch": 1070.88,
|
| 268443 |
+
"learning_rate": 7.887147435897437e-06,
|
| 268444 |
+
"loss": 0.2914,
|
| 268445 |
+
"step": 132895
|
| 268446 |
+
},
|
| 268447 |
+
{
|
| 268448 |
+
"epoch": 1070.92,
|
| 268449 |
+
"learning_rate": 7.887067307692308e-06,
|
| 268450 |
+
"loss": 0.3114,
|
| 268451 |
+
"step": 132900
|
| 268452 |
+
},
|
| 268453 |
+
{
|
| 268454 |
+
"epoch": 1070.96,
|
| 268455 |
+
"learning_rate": 7.88698717948718e-06,
|
| 268456 |
+
"loss": 0.4223,
|
| 268457 |
+
"step": 132905
|
| 268458 |
+
},
|
| 268459 |
+
{
|
| 268460 |
+
"epoch": 1071.0,
|
| 268461 |
+
"learning_rate": 7.886907051282053e-06,
|
| 268462 |
+
"loss": 1.1008,
|
| 268463 |
+
"step": 132910
|
| 268464 |
+
},
|
| 268465 |
+
{
|
| 268466 |
+
"epoch": 1071.0,
|
| 268467 |
+
"eval_loss": 0.39602500200271606,
|
| 268468 |
+
"eval_runtime": 36.0699,
|
| 268469 |
+
"eval_samples_per_second": 23.15,
|
| 268470 |
+
"eval_steps_per_second": 0.749,
|
| 268471 |
+
"eval_wer": 0.1689111747851003,
|
| 268472 |
+
"step": 132910
|
| 268473 |
+
},
|
| 268474 |
+
{
|
| 268475 |
+
"epoch": 1071.04,
|
| 268476 |
+
"learning_rate": 7.886826923076924e-06,
|
| 268477 |
+
"loss": 0.2871,
|
| 268478 |
+
"step": 132915
|
| 268479 |
+
},
|
| 268480 |
+
{
|
| 268481 |
+
"epoch": 1071.08,
|
| 268482 |
+
"learning_rate": 7.886746794871795e-06,
|
| 268483 |
+
"loss": 0.2477,
|
| 268484 |
+
"step": 132920
|
| 268485 |
+
},
|
| 268486 |
+
{
|
| 268487 |
+
"epoch": 1071.12,
|
| 268488 |
+
"learning_rate": 7.886666666666667e-06,
|
| 268489 |
+
"loss": 0.3024,
|
| 268490 |
+
"step": 132925
|
| 268491 |
+
},
|
| 268492 |
+
{
|
| 268493 |
+
"epoch": 1071.16,
|
| 268494 |
+
"learning_rate": 7.88658653846154e-06,
|
| 268495 |
+
"loss": 0.4438,
|
| 268496 |
+
"step": 132930
|
| 268497 |
+
},
|
| 268498 |
+
{
|
| 268499 |
+
"epoch": 1071.2,
|
| 268500 |
+
"learning_rate": 7.886506410256411e-06,
|
| 268501 |
+
"loss": 1.2271,
|
| 268502 |
+
"step": 132935
|
| 268503 |
+
},
|
| 268504 |
+
{
|
| 268505 |
+
"epoch": 1071.24,
|
| 268506 |
+
"learning_rate": 7.886426282051283e-06,
|
| 268507 |
+
"loss": 0.314,
|
| 268508 |
+
"step": 132940
|
| 268509 |
+
},
|
| 268510 |
+
{
|
| 268511 |
+
"epoch": 1071.28,
|
| 268512 |
+
"learning_rate": 7.886346153846156e-06,
|
| 268513 |
+
"loss": 0.2911,
|
| 268514 |
+
"step": 132945
|
| 268515 |
+
},
|
| 268516 |
+
{
|
| 268517 |
+
"epoch": 1071.32,
|
| 268518 |
+
"learning_rate": 7.886266025641027e-06,
|
| 268519 |
+
"loss": 0.3144,
|
| 268520 |
+
"step": 132950
|
| 268521 |
+
},
|
| 268522 |
+
{
|
| 268523 |
+
"epoch": 1071.36,
|
| 268524 |
+
"learning_rate": 7.886185897435898e-06,
|
| 268525 |
+
"loss": 0.4844,
|
| 268526 |
+
"step": 132955
|
| 268527 |
+
},
|
| 268528 |
+
{
|
| 268529 |
+
"epoch": 1071.4,
|
| 268530 |
+
"learning_rate": 7.88610576923077e-06,
|
| 268531 |
+
"loss": 1.0457,
|
| 268532 |
+
"step": 132960
|
| 268533 |
+
},
|
| 268534 |
+
{
|
| 268535 |
+
"epoch": 1071.44,
|
| 268536 |
+
"learning_rate": 7.886025641025643e-06,
|
| 268537 |
+
"loss": 0.3056,
|
| 268538 |
+
"step": 132965
|
| 268539 |
+
},
|
| 268540 |
+
{
|
| 268541 |
+
"epoch": 1071.48,
|
| 268542 |
+
"learning_rate": 7.885945512820512e-06,
|
| 268543 |
+
"loss": 0.2709,
|
| 268544 |
+
"step": 132970
|
| 268545 |
+
},
|
| 268546 |
+
{
|
| 268547 |
+
"epoch": 1071.52,
|
| 268548 |
+
"learning_rate": 7.885865384615385e-06,
|
| 268549 |
+
"loss": 0.3422,
|
| 268550 |
+
"step": 132975
|
| 268551 |
+
},
|
| 268552 |
+
{
|
| 268553 |
+
"epoch": 1071.56,
|
| 268554 |
+
"learning_rate": 7.885785256410257e-06,
|
| 268555 |
+
"loss": 0.4163,
|
| 268556 |
+
"step": 132980
|
| 268557 |
+
},
|
| 268558 |
+
{
|
| 268559 |
+
"epoch": 1071.6,
|
| 268560 |
+
"learning_rate": 7.885705128205128e-06,
|
| 268561 |
+
"loss": 1.0418,
|
| 268562 |
+
"step": 132985
|
| 268563 |
+
},
|
| 268564 |
+
{
|
| 268565 |
+
"epoch": 1071.65,
|
| 268566 |
+
"learning_rate": 7.885625000000001e-06,
|
| 268567 |
+
"loss": 0.352,
|
| 268568 |
+
"step": 132990
|
| 268569 |
+
},
|
| 268570 |
+
{
|
| 268571 |
+
"epoch": 1071.69,
|
| 268572 |
+
"learning_rate": 7.885544871794873e-06,
|
| 268573 |
+
"loss": 0.2445,
|
| 268574 |
+
"step": 132995
|
| 268575 |
+
},
|
| 268576 |
+
{
|
| 268577 |
+
"epoch": 1071.73,
|
| 268578 |
+
"learning_rate": 7.885464743589744e-06,
|
| 268579 |
+
"loss": 0.3016,
|
| 268580 |
+
"step": 133000
|
| 268581 |
+
},
|
| 268582 |
+
{
|
| 268583 |
+
"epoch": 1071.77,
|
| 268584 |
+
"learning_rate": 7.885384615384615e-06,
|
| 268585 |
+
"loss": 0.4437,
|
| 268586 |
+
"step": 133005
|
| 268587 |
+
},
|
| 268588 |
+
{
|
| 268589 |
+
"epoch": 1071.81,
|
| 268590 |
+
"learning_rate": 7.885304487179488e-06,
|
| 268591 |
+
"loss": 1.064,
|
| 268592 |
+
"step": 133010
|
| 268593 |
+
},
|
| 268594 |
+
{
|
| 268595 |
+
"epoch": 1071.85,
|
| 268596 |
+
"learning_rate": 7.88522435897436e-06,
|
| 268597 |
+
"loss": 0.343,
|
| 268598 |
+
"step": 133015
|
| 268599 |
+
},
|
| 268600 |
+
{
|
| 268601 |
+
"epoch": 1071.89,
|
| 268602 |
+
"learning_rate": 7.885144230769231e-06,
|
| 268603 |
+
"loss": 0.2595,
|
| 268604 |
+
"step": 133020
|
| 268605 |
+
},
|
| 268606 |
+
{
|
| 268607 |
+
"epoch": 1071.93,
|
| 268608 |
+
"learning_rate": 7.885064102564102e-06,
|
| 268609 |
+
"loss": 0.2774,
|
| 268610 |
+
"step": 133025
|
| 268611 |
+
},
|
| 268612 |
+
{
|
| 268613 |
+
"epoch": 1071.97,
|
| 268614 |
+
"learning_rate": 7.884983974358976e-06,
|
| 268615 |
+
"loss": 0.4459,
|
| 268616 |
+
"step": 133030
|
| 268617 |
+
},
|
| 268618 |
+
{
|
| 268619 |
+
"epoch": 1072.0,
|
| 268620 |
+
"eval_loss": 0.3425444960594177,
|
| 268621 |
+
"eval_runtime": 35.8136,
|
| 268622 |
+
"eval_samples_per_second": 23.315,
|
| 268623 |
+
"eval_steps_per_second": 0.754,
|
| 268624 |
+
"eval_wer": 0.17275353729905918,
|
| 268625 |
+
"step": 133034
|
| 268626 |
+
},
|
| 268627 |
+
{
|
| 268628 |
+
"epoch": 1072.01,
|
| 268629 |
+
"learning_rate": 7.884903846153847e-06,
|
| 268630 |
+
"loss": 0.4965,
|
| 268631 |
+
"step": 133035
|
| 268632 |
+
},
|
| 268633 |
+
{
|
| 268634 |
+
"epoch": 1072.05,
|
| 268635 |
+
"learning_rate": 7.884823717948718e-06,
|
| 268636 |
+
"loss": 0.2984,
|
| 268637 |
+
"step": 133040
|
| 268638 |
+
},
|
| 268639 |
+
{
|
| 268640 |
+
"epoch": 1072.09,
|
| 268641 |
+
"learning_rate": 7.884743589743591e-06,
|
| 268642 |
+
"loss": 0.2917,
|
| 268643 |
+
"step": 133045
|
| 268644 |
+
},
|
| 268645 |
+
{
|
| 268646 |
+
"epoch": 1072.13,
|
| 268647 |
+
"learning_rate": 7.884663461538463e-06,
|
| 268648 |
+
"loss": 0.2915,
|
| 268649 |
+
"step": 133050
|
| 268650 |
+
},
|
| 268651 |
+
{
|
| 268652 |
+
"epoch": 1072.17,
|
| 268653 |
+
"learning_rate": 7.884583333333334e-06,
|
| 268654 |
+
"loss": 0.5281,
|
| 268655 |
+
"step": 133055
|
| 268656 |
+
},
|
| 268657 |
+
{
|
| 268658 |
+
"epoch": 1072.21,
|
| 268659 |
+
"learning_rate": 7.884503205128205e-06,
|
| 268660 |
+
"loss": 1.0444,
|
| 268661 |
+
"step": 133060
|
| 268662 |
+
},
|
| 268663 |
+
{
|
| 268664 |
+
"epoch": 1072.25,
|
| 268665 |
+
"learning_rate": 7.884423076923078e-06,
|
| 268666 |
+
"loss": 0.2852,
|
| 268667 |
+
"step": 133065
|
| 268668 |
+
},
|
| 268669 |
+
{
|
| 268670 |
+
"epoch": 1072.29,
|
| 268671 |
+
"learning_rate": 7.88434294871795e-06,
|
| 268672 |
+
"loss": 0.3252,
|
| 268673 |
+
"step": 133070
|
| 268674 |
+
},
|
| 268675 |
+
{
|
| 268676 |
+
"epoch": 1072.33,
|
| 268677 |
+
"learning_rate": 7.884262820512821e-06,
|
| 268678 |
+
"loss": 0.2731,
|
| 268679 |
+
"step": 133075
|
| 268680 |
+
},
|
| 268681 |
+
{
|
| 268682 |
+
"epoch": 1072.37,
|
| 268683 |
+
"learning_rate": 7.884182692307692e-06,
|
| 268684 |
+
"loss": 0.4665,
|
| 268685 |
+
"step": 133080
|
| 268686 |
+
},
|
| 268687 |
+
{
|
| 268688 |
+
"epoch": 1072.41,
|
| 268689 |
+
"learning_rate": 7.884102564102566e-06,
|
| 268690 |
+
"loss": 1.0977,
|
| 268691 |
+
"step": 133085
|
| 268692 |
+
},
|
| 268693 |
+
{
|
| 268694 |
+
"epoch": 1072.45,
|
| 268695 |
+
"learning_rate": 7.884022435897435e-06,
|
| 268696 |
+
"loss": 0.3109,
|
| 268697 |
+
"step": 133090
|
| 268698 |
+
},
|
| 268699 |
+
{
|
| 268700 |
+
"epoch": 1072.49,
|
| 268701 |
+
"learning_rate": 7.883942307692308e-06,
|
| 268702 |
+
"loss": 0.2731,
|
| 268703 |
+
"step": 133095
|
| 268704 |
+
},
|
| 268705 |
+
{
|
| 268706 |
+
"epoch": 1072.53,
|
| 268707 |
+
"learning_rate": 7.883862179487181e-06,
|
| 268708 |
+
"loss": 0.3565,
|
| 268709 |
+
"step": 133100
|
| 268710 |
+
},
|
| 268711 |
+
{
|
| 268712 |
+
"epoch": 1072.57,
|
| 268713 |
+
"learning_rate": 7.883782051282051e-06,
|
| 268714 |
+
"loss": 0.5563,
|
| 268715 |
+
"step": 133105
|
| 268716 |
+
},
|
| 268717 |
+
{
|
| 268718 |
+
"epoch": 1072.61,
|
| 268719 |
+
"learning_rate": 7.883701923076924e-06,
|
| 268720 |
+
"loss": 1.0868,
|
| 268721 |
+
"step": 133110
|
| 268722 |
+
},
|
| 268723 |
+
{
|
| 268724 |
+
"epoch": 1072.65,
|
| 268725 |
+
"learning_rate": 7.883621794871795e-06,
|
| 268726 |
+
"loss": 0.2862,
|
| 268727 |
+
"step": 133115
|
| 268728 |
+
},
|
| 268729 |
+
{
|
| 268730 |
+
"epoch": 1072.69,
|
| 268731 |
+
"learning_rate": 7.883541666666667e-06,
|
| 268732 |
+
"loss": 0.3271,
|
| 268733 |
+
"step": 133120
|
| 268734 |
+
},
|
| 268735 |
+
{
|
| 268736 |
+
"epoch": 1072.73,
|
| 268737 |
+
"learning_rate": 7.883461538461538e-06,
|
| 268738 |
+
"loss": 0.3216,
|
| 268739 |
+
"step": 133125
|
| 268740 |
+
},
|
| 268741 |
+
{
|
| 268742 |
+
"epoch": 1072.77,
|
| 268743 |
+
"learning_rate": 7.883381410256411e-06,
|
| 268744 |
+
"loss": 0.4251,
|
| 268745 |
+
"step": 133130
|
| 268746 |
+
},
|
| 268747 |
+
{
|
| 268748 |
+
"epoch": 1072.81,
|
| 268749 |
+
"learning_rate": 7.883301282051283e-06,
|
| 268750 |
+
"loss": 1.0783,
|
| 268751 |
+
"step": 133135
|
| 268752 |
+
},
|
| 268753 |
+
{
|
| 268754 |
+
"epoch": 1072.85,
|
| 268755 |
+
"learning_rate": 7.883221153846154e-06,
|
| 268756 |
+
"loss": 0.3133,
|
| 268757 |
+
"step": 133140
|
| 268758 |
+
},
|
| 268759 |
+
{
|
| 268760 |
+
"epoch": 1072.89,
|
| 268761 |
+
"learning_rate": 7.883141025641027e-06,
|
| 268762 |
+
"loss": 0.3055,
|
| 268763 |
+
"step": 133145
|
| 268764 |
+
},
|
| 268765 |
+
{
|
| 268766 |
+
"epoch": 1072.93,
|
| 268767 |
+
"learning_rate": 7.883060897435898e-06,
|
| 268768 |
+
"loss": 0.2986,
|
| 268769 |
+
"step": 133150
|
| 268770 |
+
},
|
| 268771 |
+
{
|
| 268772 |
+
"epoch": 1072.97,
|
| 268773 |
+
"learning_rate": 7.88298076923077e-06,
|
| 268774 |
+
"loss": 0.5821,
|
| 268775 |
+
"step": 133155
|
| 268776 |
+
},
|
| 268777 |
+
{
|
| 268778 |
+
"epoch": 1073.0,
|
| 268779 |
+
"eval_loss": 0.3429014980792999,
|
| 268780 |
+
"eval_runtime": 36.8127,
|
| 268781 |
+
"eval_samples_per_second": 22.71,
|
| 268782 |
+
"eval_steps_per_second": 0.733,
|
| 268783 |
+
"eval_wer": 0.18048328462470598,
|
| 268784 |
+
"step": 133158
|
| 268785 |
+
},
|
| 268786 |
+
{
|
| 268787 |
+
"epoch": 1073.02,
|
| 268788 |
+
"learning_rate": 7.882900641025641e-06,
|
| 268789 |
+
"loss": 0.2765,
|
| 268790 |
+
"step": 133160
|
| 268791 |
+
},
|
| 268792 |
+
{
|
| 268793 |
+
"epoch": 1073.06,
|
| 268794 |
+
"learning_rate": 7.882820512820514e-06,
|
| 268795 |
+
"loss": 0.3097,
|
| 268796 |
+
"step": 133165
|
| 268797 |
+
},
|
| 268798 |
+
{
|
| 268799 |
+
"epoch": 1073.1,
|
| 268800 |
+
"learning_rate": 7.882740384615385e-06,
|
| 268801 |
+
"loss": 0.2492,
|
| 268802 |
+
"step": 133170
|
| 268803 |
+
},
|
| 268804 |
+
{
|
| 268805 |
+
"epoch": 1073.14,
|
| 268806 |
+
"learning_rate": 7.882660256410257e-06,
|
| 268807 |
+
"loss": 0.3436,
|
| 268808 |
+
"step": 133175
|
| 268809 |
+
},
|
| 268810 |
+
{
|
| 268811 |
+
"epoch": 1073.18,
|
| 268812 |
+
"learning_rate": 7.882580128205128e-06,
|
| 268813 |
+
"loss": 0.6949,
|
| 268814 |
+
"step": 133180
|
| 268815 |
+
},
|
| 268816 |
+
{
|
| 268817 |
+
"epoch": 1073.22,
|
| 268818 |
+
"learning_rate": 7.882500000000001e-06,
|
| 268819 |
+
"loss": 1.0907,
|
| 268820 |
+
"step": 133185
|
| 268821 |
+
},
|
| 268822 |
+
{
|
| 268823 |
+
"epoch": 1073.26,
|
| 268824 |
+
"learning_rate": 7.882419871794873e-06,
|
| 268825 |
+
"loss": 0.3416,
|
| 268826 |
+
"step": 133190
|
| 268827 |
+
},
|
| 268828 |
+
{
|
| 268829 |
+
"epoch": 1073.3,
|
| 268830 |
+
"learning_rate": 7.882339743589744e-06,
|
| 268831 |
+
"loss": 0.2677,
|
| 268832 |
+
"step": 133195
|
| 268833 |
+
},
|
| 268834 |
+
{
|
| 268835 |
+
"epoch": 1073.34,
|
| 268836 |
+
"learning_rate": 7.882259615384617e-06,
|
| 268837 |
+
"loss": 0.2954,
|
| 268838 |
+
"step": 133200
|
| 268839 |
+
},
|
| 268840 |
+
{
|
| 268841 |
+
"epoch": 1073.38,
|
| 268842 |
+
"learning_rate": 7.882179487179488e-06,
|
| 268843 |
+
"loss": 0.6008,
|
| 268844 |
+
"step": 133205
|
| 268845 |
+
},
|
| 268846 |
+
{
|
| 268847 |
+
"epoch": 1073.42,
|
| 268848 |
+
"learning_rate": 7.88209935897436e-06,
|
| 268849 |
+
"loss": 1.0349,
|
| 268850 |
+
"step": 133210
|
| 268851 |
+
},
|
| 268852 |
+
{
|
| 268853 |
+
"epoch": 1073.46,
|
| 268854 |
+
"learning_rate": 7.882019230769231e-06,
|
| 268855 |
+
"loss": 0.2499,
|
| 268856 |
+
"step": 133215
|
| 268857 |
+
},
|
| 268858 |
+
{
|
| 268859 |
+
"epoch": 1073.5,
|
| 268860 |
+
"learning_rate": 7.881939102564104e-06,
|
| 268861 |
+
"loss": 0.2891,
|
| 268862 |
+
"step": 133220
|
| 268863 |
+
},
|
| 268864 |
+
{
|
| 268865 |
+
"epoch": 1073.54,
|
| 268866 |
+
"learning_rate": 7.881858974358975e-06,
|
| 268867 |
+
"loss": 0.3114,
|
| 268868 |
+
"step": 133225
|
| 268869 |
+
},
|
| 268870 |
+
{
|
| 268871 |
+
"epoch": 1073.58,
|
| 268872 |
+
"learning_rate": 7.881778846153847e-06,
|
| 268873 |
+
"loss": 0.5413,
|
| 268874 |
+
"step": 133230
|
| 268875 |
+
},
|
| 268876 |
+
{
|
| 268877 |
+
"epoch": 1073.62,
|
| 268878 |
+
"learning_rate": 7.881698717948718e-06,
|
| 268879 |
+
"loss": 0.9867,
|
| 268880 |
+
"step": 133235
|
| 268881 |
+
},
|
| 268882 |
+
{
|
| 268883 |
+
"epoch": 1073.66,
|
| 268884 |
+
"learning_rate": 7.881618589743591e-06,
|
| 268885 |
+
"loss": 0.2607,
|
| 268886 |
+
"step": 133240
|
| 268887 |
+
},
|
| 268888 |
+
{
|
| 268889 |
+
"epoch": 1073.7,
|
| 268890 |
+
"learning_rate": 7.881538461538463e-06,
|
| 268891 |
+
"loss": 0.2457,
|
| 268892 |
+
"step": 133245
|
| 268893 |
+
},
|
| 268894 |
+
{
|
| 268895 |
+
"epoch": 1073.74,
|
| 268896 |
+
"learning_rate": 7.881458333333334e-06,
|
| 268897 |
+
"loss": 0.2933,
|
| 268898 |
+
"step": 133250
|
| 268899 |
+
},
|
| 268900 |
+
{
|
| 268901 |
+
"epoch": 1073.78,
|
| 268902 |
+
"learning_rate": 7.881378205128207e-06,
|
| 268903 |
+
"loss": 0.6377,
|
| 268904 |
+
"step": 133255
|
| 268905 |
+
},
|
| 268906 |
+
{
|
| 268907 |
+
"epoch": 1073.82,
|
| 268908 |
+
"learning_rate": 7.881298076923077e-06,
|
| 268909 |
+
"loss": 1.0415,
|
| 268910 |
+
"step": 133260
|
| 268911 |
+
},
|
| 268912 |
+
{
|
| 268913 |
+
"epoch": 1073.86,
|
| 268914 |
+
"learning_rate": 7.88121794871795e-06,
|
| 268915 |
+
"loss": 0.3253,
|
| 268916 |
+
"step": 133265
|
| 268917 |
+
},
|
| 268918 |
+
{
|
| 268919 |
+
"epoch": 1073.9,
|
| 268920 |
+
"learning_rate": 7.881137820512821e-06,
|
| 268921 |
+
"loss": 0.3159,
|
| 268922 |
+
"step": 133270
|
| 268923 |
+
},
|
| 268924 |
+
{
|
| 268925 |
+
"epoch": 1073.94,
|
| 268926 |
+
"learning_rate": 7.881057692307692e-06,
|
| 268927 |
+
"loss": 0.3446,
|
| 268928 |
+
"step": 133275
|
| 268929 |
+
},
|
| 268930 |
+
{
|
| 268931 |
+
"epoch": 1073.98,
|
| 268932 |
+
"learning_rate": 7.880977564102564e-06,
|
| 268933 |
+
"loss": 0.6176,
|
| 268934 |
+
"step": 133280
|
| 268935 |
+
},
|
| 268936 |
+
{
|
| 268937 |
+
"epoch": 1074.0,
|
| 268938 |
+
"eval_loss": 0.39306333661079407,
|
| 268939 |
+
"eval_runtime": 36.4207,
|
| 268940 |
+
"eval_samples_per_second": 22.927,
|
| 268941 |
+
"eval_steps_per_second": 0.741,
|
| 268942 |
+
"eval_wer": 0.18625053933553862,
|
| 268943 |
+
"step": 133282
|
| 268944 |
+
},
|
| 268945 |
+
{
|
| 268946 |
+
"epoch": 1074.02,
|
| 268947 |
+
"learning_rate": 7.880897435897437e-06,
|
| 268948 |
+
"loss": 0.2615,
|
| 268949 |
+
"step": 133285
|
| 268950 |
+
},
|
| 268951 |
+
{
|
| 268952 |
+
"epoch": 1074.06,
|
| 268953 |
+
"learning_rate": 7.880817307692308e-06,
|
| 268954 |
+
"loss": 0.2366,
|
| 268955 |
+
"step": 133290
|
| 268956 |
+
},
|
| 268957 |
+
{
|
| 268958 |
+
"epoch": 1074.1,
|
| 268959 |
+
"learning_rate": 7.88073717948718e-06,
|
| 268960 |
+
"loss": 0.3324,
|
| 268961 |
+
"step": 133295
|
| 268962 |
+
},
|
| 268963 |
+
{
|
| 268964 |
+
"epoch": 1074.14,
|
| 268965 |
+
"learning_rate": 7.880657051282053e-06,
|
| 268966 |
+
"loss": 0.3884,
|
| 268967 |
+
"step": 133300
|
| 268968 |
+
},
|
| 268969 |
+
{
|
| 268970 |
+
"epoch": 1074.18,
|
| 268971 |
+
"learning_rate": 7.880576923076924e-06,
|
| 268972 |
+
"loss": 0.8582,
|
| 268973 |
+
"step": 133305
|
| 268974 |
+
},
|
| 268975 |
+
{
|
| 268976 |
+
"epoch": 1074.22,
|
| 268977 |
+
"learning_rate": 7.880496794871795e-06,
|
| 268978 |
+
"loss": 0.8879,
|
| 268979 |
+
"step": 133310
|
| 268980 |
+
},
|
| 268981 |
+
{
|
| 268982 |
+
"epoch": 1074.27,
|
| 268983 |
+
"learning_rate": 7.880416666666667e-06,
|
| 268984 |
+
"loss": 0.2533,
|
| 268985 |
+
"step": 133315
|
| 268986 |
+
},
|
| 268987 |
+
{
|
| 268988 |
+
"epoch": 1074.31,
|
| 268989 |
+
"learning_rate": 7.88033653846154e-06,
|
| 268990 |
+
"loss": 0.2945,
|
| 268991 |
+
"step": 133320
|
| 268992 |
+
},
|
| 268993 |
+
{
|
| 268994 |
+
"epoch": 1074.35,
|
| 268995 |
+
"learning_rate": 7.880256410256411e-06,
|
| 268996 |
+
"loss": 0.3862,
|
| 268997 |
+
"step": 133325
|
| 268998 |
+
},
|
| 268999 |
+
{
|
| 269000 |
+
"epoch": 1074.39,
|
| 269001 |
+
"learning_rate": 7.880176282051282e-06,
|
| 269002 |
+
"loss": 0.6537,
|
| 269003 |
+
"step": 133330
|
| 269004 |
+
},
|
| 269005 |
+
{
|
| 269006 |
+
"epoch": 1074.43,
|
| 269007 |
+
"learning_rate": 7.880096153846154e-06,
|
| 269008 |
+
"loss": 0.7725,
|
| 269009 |
+
"step": 133335
|
| 269010 |
+
},
|
| 269011 |
+
{
|
| 269012 |
+
"epoch": 1074.47,
|
| 269013 |
+
"learning_rate": 7.880016025641027e-06,
|
| 269014 |
+
"loss": 0.3357,
|
| 269015 |
+
"step": 133340
|
| 269016 |
+
},
|
| 269017 |
+
{
|
| 269018 |
+
"epoch": 1074.51,
|
| 269019 |
+
"learning_rate": 7.879935897435898e-06,
|
| 269020 |
+
"loss": 0.3361,
|
| 269021 |
+
"step": 133345
|
| 269022 |
+
},
|
| 269023 |
+
{
|
| 269024 |
+
"epoch": 1074.55,
|
| 269025 |
+
"learning_rate": 7.87985576923077e-06,
|
| 269026 |
+
"loss": 0.3801,
|
| 269027 |
+
"step": 133350
|
| 269028 |
+
},
|
| 269029 |
+
{
|
| 269030 |
+
"epoch": 1074.59,
|
| 269031 |
+
"learning_rate": 7.879775641025643e-06,
|
| 269032 |
+
"loss": 0.8344,
|
| 269033 |
+
"step": 133355
|
| 269034 |
+
},
|
| 269035 |
+
{
|
| 269036 |
+
"epoch": 1074.63,
|
| 269037 |
+
"learning_rate": 7.879695512820514e-06,
|
| 269038 |
+
"loss": 0.9052,
|
| 269039 |
+
"step": 133360
|
| 269040 |
+
},
|
| 269041 |
+
{
|
| 269042 |
+
"epoch": 1074.67,
|
| 269043 |
+
"learning_rate": 7.879615384615385e-06,
|
| 269044 |
+
"loss": 0.2434,
|
| 269045 |
+
"step": 133365
|
| 269046 |
+
},
|
| 269047 |
+
{
|
| 269048 |
+
"epoch": 1074.71,
|
| 269049 |
+
"learning_rate": 7.879535256410257e-06,
|
| 269050 |
+
"loss": 0.3301,
|
| 269051 |
+
"step": 133370
|
| 269052 |
+
},
|
| 269053 |
+
{
|
| 269054 |
+
"epoch": 1074.75,
|
| 269055 |
+
"learning_rate": 7.87945512820513e-06,
|
| 269056 |
+
"loss": 0.3821,
|
| 269057 |
+
"step": 133375
|
| 269058 |
+
},
|
| 269059 |
+
{
|
| 269060 |
+
"epoch": 1074.79,
|
| 269061 |
+
"learning_rate": 7.879375e-06,
|
| 269062 |
+
"loss": 0.7044,
|
| 269063 |
+
"step": 133380
|
| 269064 |
+
},
|
| 269065 |
+
{
|
| 269066 |
+
"epoch": 1074.83,
|
| 269067 |
+
"learning_rate": 7.879294871794872e-06,
|
| 269068 |
+
"loss": 0.8513,
|
| 269069 |
+
"step": 133385
|
| 269070 |
+
},
|
| 269071 |
+
{
|
| 269072 |
+
"epoch": 1074.87,
|
| 269073 |
+
"learning_rate": 7.879214743589744e-06,
|
| 269074 |
+
"loss": 0.3133,
|
| 269075 |
+
"step": 133390
|
| 269076 |
+
},
|
| 269077 |
+
{
|
| 269078 |
+
"epoch": 1074.91,
|
| 269079 |
+
"learning_rate": 7.879134615384615e-06,
|
| 269080 |
+
"loss": 0.2895,
|
| 269081 |
+
"step": 133395
|
| 269082 |
+
},
|
| 269083 |
+
{
|
| 269084 |
+
"epoch": 1074.95,
|
| 269085 |
+
"learning_rate": 7.879054487179488e-06,
|
| 269086 |
+
"loss": 0.3408,
|
| 269087 |
+
"step": 133400
|
| 269088 |
+
},
|
| 269089 |
+
{
|
| 269090 |
+
"epoch": 1074.99,
|
| 269091 |
+
"learning_rate": 7.87897435897436e-06,
|
| 269092 |
+
"loss": 0.8444,
|
| 269093 |
+
"step": 133405
|
| 269094 |
+
},
|
| 269095 |
+
{
|
| 269096 |
+
"epoch": 1075.0,
|
| 269097 |
+
"eval_loss": 0.33311131596565247,
|
| 269098 |
+
"eval_runtime": 35.8766,
|
| 269099 |
+
"eval_samples_per_second": 23.274,
|
| 269100 |
+
"eval_steps_per_second": 0.753,
|
| 269101 |
+
"eval_wer": 0.17353530853196703,
|
| 269102 |
+
"step": 133406
|
| 269103 |
+
},
|
| 269104 |
+
{
|
| 269105 |
+
"epoch": 1075.03,
|
| 269106 |
+
"learning_rate": 7.878894230769231e-06,
|
| 269107 |
+
"loss": 0.3067,
|
| 269108 |
+
"step": 133410
|
| 269109 |
+
},
|
| 269110 |
+
{
|
| 269111 |
+
"epoch": 1075.07,
|
| 269112 |
+
"learning_rate": 7.878814102564102e-06,
|
| 269113 |
+
"loss": 0.2842,
|
| 269114 |
+
"step": 133415
|
| 269115 |
+
},
|
| 269116 |
+
{
|
| 269117 |
+
"epoch": 1075.11,
|
| 269118 |
+
"learning_rate": 7.878733974358975e-06,
|
| 269119 |
+
"loss": 0.287,
|
| 269120 |
+
"step": 133420
|
| 269121 |
+
},
|
| 269122 |
+
{
|
| 269123 |
+
"epoch": 1075.15,
|
| 269124 |
+
"learning_rate": 7.878653846153847e-06,
|
| 269125 |
+
"loss": 0.4559,
|
| 269126 |
+
"step": 133425
|
| 269127 |
+
},
|
| 269128 |
+
{
|
| 269129 |
+
"epoch": 1075.19,
|
| 269130 |
+
"learning_rate": 7.878573717948718e-06,
|
| 269131 |
+
"loss": 0.7729,
|
| 269132 |
+
"step": 133430
|
| 269133 |
+
},
|
| 269134 |
+
{
|
| 269135 |
+
"epoch": 1075.23,
|
| 269136 |
+
"learning_rate": 7.87849358974359e-06,
|
| 269137 |
+
"loss": 0.6402,
|
| 269138 |
+
"step": 133435
|
| 269139 |
+
},
|
| 269140 |
+
{
|
| 269141 |
+
"epoch": 1075.27,
|
| 269142 |
+
"learning_rate": 7.878413461538462e-06,
|
| 269143 |
+
"loss": 0.2491,
|
| 269144 |
+
"step": 133440
|
| 269145 |
+
},
|
| 269146 |
+
{
|
| 269147 |
+
"epoch": 1075.31,
|
| 269148 |
+
"learning_rate": 7.878333333333334e-06,
|
| 269149 |
+
"loss": 0.273,
|
| 269150 |
+
"step": 133445
|
| 269151 |
+
},
|
| 269152 |
+
{
|
| 269153 |
+
"epoch": 1075.35,
|
| 269154 |
+
"learning_rate": 7.878253205128205e-06,
|
| 269155 |
+
"loss": 0.4031,
|
| 269156 |
+
"step": 133450
|
| 269157 |
+
},
|
| 269158 |
+
{
|
| 269159 |
+
"epoch": 1075.39,
|
| 269160 |
+
"learning_rate": 7.878173076923078e-06,
|
| 269161 |
+
"loss": 0.9203,
|
| 269162 |
+
"step": 133455
|
| 269163 |
+
},
|
| 269164 |
+
{
|
| 269165 |
+
"epoch": 1075.43,
|
| 269166 |
+
"learning_rate": 7.87809294871795e-06,
|
| 269167 |
+
"loss": 0.6169,
|
| 269168 |
+
"step": 133460
|
| 269169 |
+
},
|
| 269170 |
+
{
|
| 269171 |
+
"epoch": 1075.47,
|
| 269172 |
+
"learning_rate": 7.878012820512821e-06,
|
| 269173 |
+
"loss": 0.2476,
|
| 269174 |
+
"step": 133465
|
| 269175 |
+
},
|
| 269176 |
+
{
|
| 269177 |
+
"epoch": 1075.51,
|
| 269178 |
+
"learning_rate": 7.877932692307692e-06,
|
| 269179 |
+
"loss": 0.2902,
|
| 269180 |
+
"step": 133470
|
| 269181 |
+
},
|
| 269182 |
+
{
|
| 269183 |
+
"epoch": 1075.55,
|
| 269184 |
+
"learning_rate": 7.877852564102565e-06,
|
| 269185 |
+
"loss": 0.3484,
|
| 269186 |
+
"step": 133475
|
| 269187 |
+
},
|
| 269188 |
+
{
|
| 269189 |
+
"epoch": 1075.59,
|
| 269190 |
+
"learning_rate": 7.877772435897437e-06,
|
| 269191 |
+
"loss": 0.9193,
|
| 269192 |
+
"step": 133480
|
| 269193 |
+
},
|
| 269194 |
+
{
|
| 269195 |
+
"epoch": 1075.63,
|
| 269196 |
+
"learning_rate": 7.877692307692308e-06,
|
| 269197 |
+
"loss": 0.6227,
|
| 269198 |
+
"step": 133485
|
| 269199 |
+
},
|
| 269200 |
+
{
|
| 269201 |
+
"epoch": 1075.67,
|
| 269202 |
+
"learning_rate": 7.87761217948718e-06,
|
| 269203 |
+
"loss": 0.3372,
|
| 269204 |
+
"step": 133490
|
| 269205 |
+
},
|
| 269206 |
+
{
|
| 269207 |
+
"epoch": 1075.71,
|
| 269208 |
+
"learning_rate": 7.877532051282053e-06,
|
| 269209 |
+
"loss": 0.3326,
|
| 269210 |
+
"step": 133495
|
| 269211 |
+
},
|
| 269212 |
+
{
|
| 269213 |
+
"epoch": 1075.76,
|
| 269214 |
+
"learning_rate": 7.877451923076924e-06,
|
| 269215 |
+
"loss": 0.3769,
|
| 269216 |
+
"step": 133500
|
| 269217 |
+
},
|
| 269218 |
+
{
|
| 269219 |
+
"epoch": 1075.8,
|
| 269220 |
+
"learning_rate": 7.877371794871795e-06,
|
| 269221 |
+
"loss": 0.8798,
|
| 269222 |
+
"step": 133505
|
| 269223 |
+
},
|
| 269224 |
+
{
|
| 269225 |
+
"epoch": 1075.84,
|
| 269226 |
+
"learning_rate": 7.877291666666668e-06,
|
| 269227 |
+
"loss": 0.7368,
|
| 269228 |
+
"step": 133510
|
| 269229 |
+
},
|
| 269230 |
+
{
|
| 269231 |
+
"epoch": 1075.88,
|
| 269232 |
+
"learning_rate": 7.87721153846154e-06,
|
| 269233 |
+
"loss": 0.2917,
|
| 269234 |
+
"step": 133515
|
| 269235 |
+
},
|
| 269236 |
+
{
|
| 269237 |
+
"epoch": 1075.92,
|
| 269238 |
+
"learning_rate": 7.877131410256411e-06,
|
| 269239 |
+
"loss": 0.2877,
|
| 269240 |
+
"step": 133520
|
| 269241 |
+
},
|
| 269242 |
+
{
|
| 269243 |
+
"epoch": 1075.96,
|
| 269244 |
+
"learning_rate": 7.877051282051282e-06,
|
| 269245 |
+
"loss": 0.3613,
|
| 269246 |
+
"step": 133525
|
| 269247 |
+
},
|
| 269248 |
+
{
|
| 269249 |
+
"epoch": 1076.0,
|
| 269250 |
+
"learning_rate": 7.876971153846155e-06,
|
| 269251 |
+
"loss": 1.0196,
|
| 269252 |
+
"step": 133530
|
| 269253 |
+
},
|
| 269254 |
+
{
|
| 269255 |
+
"epoch": 1076.0,
|
| 269256 |
+
"eval_loss": 0.31553006172180176,
|
| 269257 |
+
"eval_runtime": 35.6523,
|
| 269258 |
+
"eval_samples_per_second": 23.421,
|
| 269259 |
+
"eval_steps_per_second": 0.757,
|
| 269260 |
+
"eval_wer": 0.17710974284679465,
|
| 269261 |
+
"step": 133530
|
| 269262 |
}
|
| 269263 |
],
|
| 269264 |
+
"max_steps": 620000,
|
| 269265 |
"num_train_epochs": 5000,
|
| 269266 |
+
"total_flos": 3.75745971037642e+20,
|
| 269267 |
"trial_name": null,
|
| 269268 |
"trial_params": null
|
| 269269 |
}
|
model-bin/finetune/base/{checkpoint-132786 → checkpoint-133530}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1630171494.1221683/events.out.tfevents.1630171494.86bb0ddabf9b.4092.241
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e6f4a8a36a6fac8d1c5515e5f6c66152f9d7a24029fe39c328a3fe5005ee1c41
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630171878.5782204/events.out.tfevents.1630171878.86bb0ddabf9b.4092.243
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ad950c2dc867a5bcbe8be380b03a193a2af8b9e891e2aa238b4a79ba0cf6db10
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630172268.8488228/events.out.tfevents.1630172268.86bb0ddabf9b.4092.245
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:0bbe35877b98bd4fd507a9913e60d40106990d6c5f484548a33ab4a34ac2fffd
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630172658.9374063/events.out.tfevents.1630172658.86bb0ddabf9b.4092.247
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9a8af6238765a65a0c20ae4ed4c9be60049b78788082d6587aef7318d9359256
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630173043.8917527/events.out.tfevents.1630173043.86bb0ddabf9b.4092.249
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f3f8cc7faffaccb2cf9c5ac8e10f59600b5355935522ec2bb1af7a1372edcf58
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1630171494.86bb0ddabf9b.4092.240
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:799142f0a822d1fc067b3fb2b8cd9246dd13efd2d7a688b049c94a46ce3d0a7d
|
| 3 |
+
size 8462
|
model-bin/finetune/base/log/events.out.tfevents.1630171878.86bb0ddabf9b.4092.242
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6e1ad04ae3efc429939c8d83d612e6fda4ea949b18b793d9a69a497dda7cda50
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630172268.86bb0ddabf9b.4092.244
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:4212d35653cbac345b82d995ba40d8b4f6ef71436094c4043e3611d91fa40fd5
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630172658.86bb0ddabf9b.4092.246
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:596801ef559b8a9ec04d90f8ba7581f112cfe8b6c17c3ecd10db28aa35a0cc03
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630173043.86bb0ddabf9b.4092.248
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:36ba137324e4dd3f5781b363b8870850203a88045c54447aebd880cbc2dbd5ff
|
| 3 |
+
size 8622
|