"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-74539 → checkpoint-75162}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-74539 → checkpoint-75162}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-74539 → checkpoint-75162}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-74539 → checkpoint-75162}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-74539 → checkpoint-75162}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-74539 → checkpoint-75162}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-74539 → checkpoint-75162}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-74539 → checkpoint-75162}/trainer_state.json +799 -4
- model-bin/finetune/base/{checkpoint-74539 → checkpoint-75162}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629879092.38512/events.out.tfevents.1629879092.7e498afd5545.905.83 +3 -0
- model-bin/finetune/base/log/1629879755.9543498/events.out.tfevents.1629879755.7e498afd5545.905.85 +3 -0
- model-bin/finetune/base/log/1629880403.5211415/events.out.tfevents.1629880403.7e498afd5545.905.87 +3 -0
- model-bin/finetune/base/log/1629881049.788205/events.out.tfevents.1629881049.7e498afd5545.905.89 +3 -0
- model-bin/finetune/base/log/1629881697.4184577/events.out.tfevents.1629881697.7e498afd5545.905.91 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629879092.7e498afd5545.905.82 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629879755.7e498afd5545.905.84 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629880403.7e498afd5545.905.86 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629881049.7e498afd5545.905.88 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629881697.7e498afd5545.905.90 +3 -0
model-bin/finetune/base/{checkpoint-74539 → checkpoint-75162}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-74539 → checkpoint-75162}/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 sha256:54b35bdaaf4933af7fe3f9bbd9b1424d570c568987740f15bcd85cc870d83f51
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-74539 → checkpoint-75162}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-74539 → checkpoint-75162}/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:f70ffbe9923a30dbf8b87abb1106958feef550de2620a2d2b80cc91fc8a098e2
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-74539 → checkpoint-75162}/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 sha256:4be019330451dcdea1470053a969250ec2ac7591e9e422df7bf9734de62a6607
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-74539 → checkpoint-75162}/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:e3d8eee4042019854a091a06285230ead9805a530d0a1e8630cf2695e05d1d3b
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-74539 → checkpoint-75162}/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 sha256:8aae4596cc6445369fecda3c5539d240321af736664343a1c7e72481ab8edb80
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-74539 → checkpoint-75162}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.18412114350410416,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-69565",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -194199,11 +194199,806 @@
|
|
| 194199 |
"eval_steps_per_second": 0.666,
|
| 194200 |
"eval_wer": 0.18913761467889909,
|
| 194201 |
"step": 74539
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 194202 |
}
|
| 194203 |
],
|
| 194204 |
-
"max_steps":
|
| 194205 |
"num_train_epochs": 5000,
|
| 194206 |
-
"total_flos": 2.
|
| 194207 |
"trial_name": null,
|
| 194208 |
"trial_params": null
|
| 194209 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.18412114350410416,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-69565",
|
| 4 |
+
"epoch": 601.0,
|
| 5 |
+
"global_step": 75162,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 194199 |
"eval_steps_per_second": 0.666,
|
| 194200 |
"eval_wer": 0.18913761467889909,
|
| 194201 |
"step": 74539
|
| 194202 |
+
},
|
| 194203 |
+
{
|
| 194204 |
+
"epoch": 596.01,
|
| 194205 |
+
"learning_rate": 8.821939102564104e-06,
|
| 194206 |
+
"loss": 0.4295,
|
| 194207 |
+
"step": 74540
|
| 194208 |
+
},
|
| 194209 |
+
{
|
| 194210 |
+
"epoch": 596.05,
|
| 194211 |
+
"learning_rate": 8.821858974358975e-06,
|
| 194212 |
+
"loss": 0.3124,
|
| 194213 |
+
"step": 74545
|
| 194214 |
+
},
|
| 194215 |
+
{
|
| 194216 |
+
"epoch": 596.09,
|
| 194217 |
+
"learning_rate": 8.821778846153846e-06,
|
| 194218 |
+
"loss": 0.3017,
|
| 194219 |
+
"step": 74550
|
| 194220 |
+
},
|
| 194221 |
+
{
|
| 194222 |
+
"epoch": 596.13,
|
| 194223 |
+
"learning_rate": 8.82169871794872e-06,
|
| 194224 |
+
"loss": 0.3852,
|
| 194225 |
+
"step": 74555
|
| 194226 |
+
},
|
| 194227 |
+
{
|
| 194228 |
+
"epoch": 596.17,
|
| 194229 |
+
"learning_rate": 8.82161858974359e-06,
|
| 194230 |
+
"loss": 0.5428,
|
| 194231 |
+
"step": 74560
|
| 194232 |
+
},
|
| 194233 |
+
{
|
| 194234 |
+
"epoch": 596.21,
|
| 194235 |
+
"learning_rate": 8.821538461538462e-06,
|
| 194236 |
+
"loss": 1.2103,
|
| 194237 |
+
"step": 74565
|
| 194238 |
+
},
|
| 194239 |
+
{
|
| 194240 |
+
"epoch": 596.25,
|
| 194241 |
+
"learning_rate": 8.821458333333333e-06,
|
| 194242 |
+
"loss": 0.3342,
|
| 194243 |
+
"step": 74570
|
| 194244 |
+
},
|
| 194245 |
+
{
|
| 194246 |
+
"epoch": 596.29,
|
| 194247 |
+
"learning_rate": 8.821378205128206e-06,
|
| 194248 |
+
"loss": 0.279,
|
| 194249 |
+
"step": 74575
|
| 194250 |
+
},
|
| 194251 |
+
{
|
| 194252 |
+
"epoch": 596.33,
|
| 194253 |
+
"learning_rate": 8.821298076923078e-06,
|
| 194254 |
+
"loss": 0.3352,
|
| 194255 |
+
"step": 74580
|
| 194256 |
+
},
|
| 194257 |
+
{
|
| 194258 |
+
"epoch": 596.37,
|
| 194259 |
+
"learning_rate": 8.82121794871795e-06,
|
| 194260 |
+
"loss": 0.6326,
|
| 194261 |
+
"step": 74585
|
| 194262 |
+
},
|
| 194263 |
+
{
|
| 194264 |
+
"epoch": 596.41,
|
| 194265 |
+
"learning_rate": 8.821137820512822e-06,
|
| 194266 |
+
"loss": 1.2738,
|
| 194267 |
+
"step": 74590
|
| 194268 |
+
},
|
| 194269 |
+
{
|
| 194270 |
+
"epoch": 596.45,
|
| 194271 |
+
"learning_rate": 8.821057692307694e-06,
|
| 194272 |
+
"loss": 0.2898,
|
| 194273 |
+
"step": 74595
|
| 194274 |
+
},
|
| 194275 |
+
{
|
| 194276 |
+
"epoch": 596.49,
|
| 194277 |
+
"learning_rate": 8.820977564102565e-06,
|
| 194278 |
+
"loss": 0.3831,
|
| 194279 |
+
"step": 74600
|
| 194280 |
+
},
|
| 194281 |
+
{
|
| 194282 |
+
"epoch": 596.53,
|
| 194283 |
+
"learning_rate": 8.820897435897436e-06,
|
| 194284 |
+
"loss": 0.3918,
|
| 194285 |
+
"step": 74605
|
| 194286 |
+
},
|
| 194287 |
+
{
|
| 194288 |
+
"epoch": 596.57,
|
| 194289 |
+
"learning_rate": 8.82081730769231e-06,
|
| 194290 |
+
"loss": 0.4952,
|
| 194291 |
+
"step": 74610
|
| 194292 |
+
},
|
| 194293 |
+
{
|
| 194294 |
+
"epoch": 596.61,
|
| 194295 |
+
"learning_rate": 8.82073717948718e-06,
|
| 194296 |
+
"loss": 1.0689,
|
| 194297 |
+
"step": 74615
|
| 194298 |
+
},
|
| 194299 |
+
{
|
| 194300 |
+
"epoch": 596.65,
|
| 194301 |
+
"learning_rate": 8.820657051282052e-06,
|
| 194302 |
+
"loss": 0.3247,
|
| 194303 |
+
"step": 74620
|
| 194304 |
+
},
|
| 194305 |
+
{
|
| 194306 |
+
"epoch": 596.69,
|
| 194307 |
+
"learning_rate": 8.820576923076923e-06,
|
| 194308 |
+
"loss": 0.3352,
|
| 194309 |
+
"step": 74625
|
| 194310 |
+
},
|
| 194311 |
+
{
|
| 194312 |
+
"epoch": 596.73,
|
| 194313 |
+
"learning_rate": 8.820496794871796e-06,
|
| 194314 |
+
"loss": 0.3389,
|
| 194315 |
+
"step": 74630
|
| 194316 |
+
},
|
| 194317 |
+
{
|
| 194318 |
+
"epoch": 596.76,
|
| 194319 |
+
"learning_rate": 8.820416666666666e-06,
|
| 194320 |
+
"loss": 0.6059,
|
| 194321 |
+
"step": 74635
|
| 194322 |
+
},
|
| 194323 |
+
{
|
| 194324 |
+
"epoch": 596.8,
|
| 194325 |
+
"learning_rate": 8.82033653846154e-06,
|
| 194326 |
+
"loss": 1.2196,
|
| 194327 |
+
"step": 74640
|
| 194328 |
+
},
|
| 194329 |
+
{
|
| 194330 |
+
"epoch": 596.84,
|
| 194331 |
+
"learning_rate": 8.820256410256412e-06,
|
| 194332 |
+
"loss": 0.3291,
|
| 194333 |
+
"step": 74645
|
| 194334 |
+
},
|
| 194335 |
+
{
|
| 194336 |
+
"epoch": 596.88,
|
| 194337 |
+
"learning_rate": 8.820176282051282e-06,
|
| 194338 |
+
"loss": 0.3028,
|
| 194339 |
+
"step": 74650
|
| 194340 |
+
},
|
| 194341 |
+
{
|
| 194342 |
+
"epoch": 596.92,
|
| 194343 |
+
"learning_rate": 8.820096153846155e-06,
|
| 194344 |
+
"loss": 0.3241,
|
| 194345 |
+
"step": 74655
|
| 194346 |
+
},
|
| 194347 |
+
{
|
| 194348 |
+
"epoch": 596.96,
|
| 194349 |
+
"learning_rate": 8.820016025641026e-06,
|
| 194350 |
+
"loss": 0.5501,
|
| 194351 |
+
"step": 74660
|
| 194352 |
+
},
|
| 194353 |
+
{
|
| 194354 |
+
"epoch": 597.0,
|
| 194355 |
+
"eval_loss": 0.3461139500141144,
|
| 194356 |
+
"eval_runtime": 40.6764,
|
| 194357 |
+
"eval_samples_per_second": 20.651,
|
| 194358 |
+
"eval_steps_per_second": 0.664,
|
| 194359 |
+
"eval_wer": 0.18884180790960453,
|
| 194360 |
+
"step": 74664
|
| 194361 |
+
},
|
| 194362 |
+
{
|
| 194363 |
+
"epoch": 602.01,
|
| 194364 |
+
"learning_rate": 8.819935897435898e-06,
|
| 194365 |
+
"loss": 0.4039,
|
| 194366 |
+
"step": 74665
|
| 194367 |
+
},
|
| 194368 |
+
{
|
| 194369 |
+
"epoch": 602.05,
|
| 194370 |
+
"learning_rate": 8.819855769230769e-06,
|
| 194371 |
+
"loss": 0.3075,
|
| 194372 |
+
"step": 74670
|
| 194373 |
+
},
|
| 194374 |
+
{
|
| 194375 |
+
"epoch": 602.09,
|
| 194376 |
+
"learning_rate": 8.819775641025642e-06,
|
| 194377 |
+
"loss": 0.2638,
|
| 194378 |
+
"step": 74675
|
| 194379 |
+
},
|
| 194380 |
+
{
|
| 194381 |
+
"epoch": 602.13,
|
| 194382 |
+
"learning_rate": 8.819695512820513e-06,
|
| 194383 |
+
"loss": 0.346,
|
| 194384 |
+
"step": 74680
|
| 194385 |
+
},
|
| 194386 |
+
{
|
| 194387 |
+
"epoch": 602.17,
|
| 194388 |
+
"learning_rate": 8.819615384615385e-06,
|
| 194389 |
+
"loss": 0.6652,
|
| 194390 |
+
"step": 74685
|
| 194391 |
+
},
|
| 194392 |
+
{
|
| 194393 |
+
"epoch": 602.21,
|
| 194394 |
+
"learning_rate": 8.819535256410258e-06,
|
| 194395 |
+
"loss": 1.2922,
|
| 194396 |
+
"step": 74690
|
| 194397 |
+
},
|
| 194398 |
+
{
|
| 194399 |
+
"epoch": 602.25,
|
| 194400 |
+
"learning_rate": 8.81945512820513e-06,
|
| 194401 |
+
"loss": 0.307,
|
| 194402 |
+
"step": 74695
|
| 194403 |
+
},
|
| 194404 |
+
{
|
| 194405 |
+
"epoch": 602.29,
|
| 194406 |
+
"learning_rate": 8.819375e-06,
|
| 194407 |
+
"loss": 0.2766,
|
| 194408 |
+
"step": 74700
|
| 194409 |
+
},
|
| 194410 |
+
{
|
| 194411 |
+
"epoch": 602.33,
|
| 194412 |
+
"learning_rate": 8.819294871794872e-06,
|
| 194413 |
+
"loss": 0.3401,
|
| 194414 |
+
"step": 74705
|
| 194415 |
+
},
|
| 194416 |
+
{
|
| 194417 |
+
"epoch": 602.37,
|
| 194418 |
+
"learning_rate": 8.819214743589745e-06,
|
| 194419 |
+
"loss": 0.5566,
|
| 194420 |
+
"step": 74710
|
| 194421 |
+
},
|
| 194422 |
+
{
|
| 194423 |
+
"epoch": 602.41,
|
| 194424 |
+
"learning_rate": 8.819134615384616e-06,
|
| 194425 |
+
"loss": 1.1703,
|
| 194426 |
+
"step": 74715
|
| 194427 |
+
},
|
| 194428 |
+
{
|
| 194429 |
+
"epoch": 602.45,
|
| 194430 |
+
"learning_rate": 8.819054487179488e-06,
|
| 194431 |
+
"loss": 0.3373,
|
| 194432 |
+
"step": 74720
|
| 194433 |
+
},
|
| 194434 |
+
{
|
| 194435 |
+
"epoch": 602.49,
|
| 194436 |
+
"learning_rate": 8.818974358974359e-06,
|
| 194437 |
+
"loss": 0.3019,
|
| 194438 |
+
"step": 74725
|
| 194439 |
+
},
|
| 194440 |
+
{
|
| 194441 |
+
"epoch": 602.53,
|
| 194442 |
+
"learning_rate": 8.818894230769232e-06,
|
| 194443 |
+
"loss": 0.351,
|
| 194444 |
+
"step": 74730
|
| 194445 |
+
},
|
| 194446 |
+
{
|
| 194447 |
+
"epoch": 602.57,
|
| 194448 |
+
"learning_rate": 8.818814102564103e-06,
|
| 194449 |
+
"loss": 0.5587,
|
| 194450 |
+
"step": 74735
|
| 194451 |
+
},
|
| 194452 |
+
{
|
| 194453 |
+
"epoch": 602.61,
|
| 194454 |
+
"learning_rate": 8.818733974358975e-06,
|
| 194455 |
+
"loss": 1.1618,
|
| 194456 |
+
"step": 74740
|
| 194457 |
+
},
|
| 194458 |
+
{
|
| 194459 |
+
"epoch": 602.65,
|
| 194460 |
+
"learning_rate": 8.818653846153848e-06,
|
| 194461 |
+
"loss": 0.327,
|
| 194462 |
+
"step": 74745
|
| 194463 |
+
},
|
| 194464 |
+
{
|
| 194465 |
+
"epoch": 602.69,
|
| 194466 |
+
"learning_rate": 8.81857371794872e-06,
|
| 194467 |
+
"loss": 0.3309,
|
| 194468 |
+
"step": 74750
|
| 194469 |
+
},
|
| 194470 |
+
{
|
| 194471 |
+
"epoch": 602.73,
|
| 194472 |
+
"learning_rate": 8.81849358974359e-06,
|
| 194473 |
+
"loss": 0.4235,
|
| 194474 |
+
"step": 74755
|
| 194475 |
+
},
|
| 194476 |
+
{
|
| 194477 |
+
"epoch": 602.77,
|
| 194478 |
+
"learning_rate": 8.818413461538462e-06,
|
| 194479 |
+
"loss": 0.5577,
|
| 194480 |
+
"step": 74760
|
| 194481 |
+
},
|
| 194482 |
+
{
|
| 194483 |
+
"epoch": 602.81,
|
| 194484 |
+
"learning_rate": 8.818333333333335e-06,
|
| 194485 |
+
"loss": 1.1446,
|
| 194486 |
+
"step": 74765
|
| 194487 |
+
},
|
| 194488 |
+
{
|
| 194489 |
+
"epoch": 602.85,
|
| 194490 |
+
"learning_rate": 8.818253205128205e-06,
|
| 194491 |
+
"loss": 0.3165,
|
| 194492 |
+
"step": 74770
|
| 194493 |
+
},
|
| 194494 |
+
{
|
| 194495 |
+
"epoch": 602.9,
|
| 194496 |
+
"learning_rate": 8.818173076923078e-06,
|
| 194497 |
+
"loss": 0.2766,
|
| 194498 |
+
"step": 74775
|
| 194499 |
+
},
|
| 194500 |
+
{
|
| 194501 |
+
"epoch": 602.94,
|
| 194502 |
+
"learning_rate": 8.818092948717949e-06,
|
| 194503 |
+
"loss": 0.3203,
|
| 194504 |
+
"step": 74780
|
| 194505 |
+
},
|
| 194506 |
+
{
|
| 194507 |
+
"epoch": 602.98,
|
| 194508 |
+
"learning_rate": 8.81801282051282e-06,
|
| 194509 |
+
"loss": 0.5715,
|
| 194510 |
+
"step": 74785
|
| 194511 |
+
},
|
| 194512 |
+
{
|
| 194513 |
+
"epoch": 603.0,
|
| 194514 |
+
"eval_loss": 0.38998129963874817,
|
| 194515 |
+
"eval_runtime": 40.9192,
|
| 194516 |
+
"eval_samples_per_second": 20.528,
|
| 194517 |
+
"eval_steps_per_second": 0.66,
|
| 194518 |
+
"eval_wer": 0.1910048391688016,
|
| 194519 |
+
"step": 74788
|
| 194520 |
+
},
|
| 194521 |
+
{
|
| 194522 |
+
"epoch": 603.02,
|
| 194523 |
+
"learning_rate": 8.817932692307694e-06,
|
| 194524 |
+
"loss": 0.3641,
|
| 194525 |
+
"step": 74790
|
| 194526 |
+
},
|
| 194527 |
+
{
|
| 194528 |
+
"epoch": 603.06,
|
| 194529 |
+
"learning_rate": 8.817852564102565e-06,
|
| 194530 |
+
"loss": 0.3938,
|
| 194531 |
+
"step": 74795
|
| 194532 |
+
},
|
| 194533 |
+
{
|
| 194534 |
+
"epoch": 603.1,
|
| 194535 |
+
"learning_rate": 8.817772435897436e-06,
|
| 194536 |
+
"loss": 0.3374,
|
| 194537 |
+
"step": 74800
|
| 194538 |
+
},
|
| 194539 |
+
{
|
| 194540 |
+
"epoch": 603.14,
|
| 194541 |
+
"learning_rate": 8.817692307692308e-06,
|
| 194542 |
+
"loss": 0.36,
|
| 194543 |
+
"step": 74805
|
| 194544 |
+
},
|
| 194545 |
+
{
|
| 194546 |
+
"epoch": 603.18,
|
| 194547 |
+
"learning_rate": 8.81761217948718e-06,
|
| 194548 |
+
"loss": 0.7182,
|
| 194549 |
+
"step": 74810
|
| 194550 |
+
},
|
| 194551 |
+
{
|
| 194552 |
+
"epoch": 603.22,
|
| 194553 |
+
"learning_rate": 8.817532051282052e-06,
|
| 194554 |
+
"loss": 1.0542,
|
| 194555 |
+
"step": 74815
|
| 194556 |
+
},
|
| 194557 |
+
{
|
| 194558 |
+
"epoch": 603.26,
|
| 194559 |
+
"learning_rate": 8.817451923076923e-06,
|
| 194560 |
+
"loss": 0.3317,
|
| 194561 |
+
"step": 74820
|
| 194562 |
+
},
|
| 194563 |
+
{
|
| 194564 |
+
"epoch": 603.3,
|
| 194565 |
+
"learning_rate": 8.817371794871795e-06,
|
| 194566 |
+
"loss": 0.4227,
|
| 194567 |
+
"step": 74825
|
| 194568 |
+
},
|
| 194569 |
+
{
|
| 194570 |
+
"epoch": 603.34,
|
| 194571 |
+
"learning_rate": 8.817291666666668e-06,
|
| 194572 |
+
"loss": 0.4196,
|
| 194573 |
+
"step": 74830
|
| 194574 |
+
},
|
| 194575 |
+
{
|
| 194576 |
+
"epoch": 603.38,
|
| 194577 |
+
"learning_rate": 8.817211538461539e-06,
|
| 194578 |
+
"loss": 0.6258,
|
| 194579 |
+
"step": 74835
|
| 194580 |
+
},
|
| 194581 |
+
{
|
| 194582 |
+
"epoch": 603.42,
|
| 194583 |
+
"learning_rate": 8.81713141025641e-06,
|
| 194584 |
+
"loss": 1.0404,
|
| 194585 |
+
"step": 74840
|
| 194586 |
+
},
|
| 194587 |
+
{
|
| 194588 |
+
"epoch": 603.46,
|
| 194589 |
+
"learning_rate": 8.817051282051284e-06,
|
| 194590 |
+
"loss": 0.3388,
|
| 194591 |
+
"step": 74845
|
| 194592 |
+
},
|
| 194593 |
+
{
|
| 194594 |
+
"epoch": 603.5,
|
| 194595 |
+
"learning_rate": 8.816971153846155e-06,
|
| 194596 |
+
"loss": 0.4682,
|
| 194597 |
+
"step": 74850
|
| 194598 |
+
},
|
| 194599 |
+
{
|
| 194600 |
+
"epoch": 603.54,
|
| 194601 |
+
"learning_rate": 8.816891025641026e-06,
|
| 194602 |
+
"loss": 0.3918,
|
| 194603 |
+
"step": 74855
|
| 194604 |
+
},
|
| 194605 |
+
{
|
| 194606 |
+
"epoch": 603.58,
|
| 194607 |
+
"learning_rate": 8.816810897435898e-06,
|
| 194608 |
+
"loss": 0.6384,
|
| 194609 |
+
"step": 74860
|
| 194610 |
+
},
|
| 194611 |
+
{
|
| 194612 |
+
"epoch": 603.62,
|
| 194613 |
+
"learning_rate": 8.81673076923077e-06,
|
| 194614 |
+
"loss": 1.0819,
|
| 194615 |
+
"step": 74865
|
| 194616 |
+
},
|
| 194617 |
+
{
|
| 194618 |
+
"epoch": 603.66,
|
| 194619 |
+
"learning_rate": 8.816650641025642e-06,
|
| 194620 |
+
"loss": 0.2616,
|
| 194621 |
+
"step": 74870
|
| 194622 |
+
},
|
| 194623 |
+
{
|
| 194624 |
+
"epoch": 603.7,
|
| 194625 |
+
"learning_rate": 8.816570512820513e-06,
|
| 194626 |
+
"loss": 0.3247,
|
| 194627 |
+
"step": 74875
|
| 194628 |
+
},
|
| 194629 |
+
{
|
| 194630 |
+
"epoch": 603.74,
|
| 194631 |
+
"learning_rate": 8.816490384615385e-06,
|
| 194632 |
+
"loss": 0.318,
|
| 194633 |
+
"step": 74880
|
| 194634 |
+
},
|
| 194635 |
+
{
|
| 194636 |
+
"epoch": 603.78,
|
| 194637 |
+
"learning_rate": 8.816410256410258e-06,
|
| 194638 |
+
"loss": 0.6705,
|
| 194639 |
+
"step": 74885
|
| 194640 |
+
},
|
| 194641 |
+
{
|
| 194642 |
+
"epoch": 603.82,
|
| 194643 |
+
"learning_rate": 8.816330128205129e-06,
|
| 194644 |
+
"loss": 1.2115,
|
| 194645 |
+
"step": 74890
|
| 194646 |
+
},
|
| 194647 |
+
{
|
| 194648 |
+
"epoch": 603.86,
|
| 194649 |
+
"learning_rate": 8.81625e-06,
|
| 194650 |
+
"loss": 0.3773,
|
| 194651 |
+
"step": 74895
|
| 194652 |
+
},
|
| 194653 |
+
{
|
| 194654 |
+
"epoch": 603.9,
|
| 194655 |
+
"learning_rate": 8.816169871794874e-06,
|
| 194656 |
+
"loss": 0.3811,
|
| 194657 |
+
"step": 74900
|
| 194658 |
+
},
|
| 194659 |
+
{
|
| 194660 |
+
"epoch": 603.94,
|
| 194661 |
+
"learning_rate": 8.816089743589745e-06,
|
| 194662 |
+
"loss": 0.388,
|
| 194663 |
+
"step": 74905
|
| 194664 |
+
},
|
| 194665 |
+
{
|
| 194666 |
+
"epoch": 603.98,
|
| 194667 |
+
"learning_rate": 8.816009615384616e-06,
|
| 194668 |
+
"loss": 0.8204,
|
| 194669 |
+
"step": 74910
|
| 194670 |
+
},
|
| 194671 |
+
{
|
| 194672 |
+
"epoch": 604.0,
|
| 194673 |
+
"eval_loss": 0.4254598617553711,
|
| 194674 |
+
"eval_runtime": 39.7945,
|
| 194675 |
+
"eval_samples_per_second": 21.108,
|
| 194676 |
+
"eval_steps_per_second": 0.678,
|
| 194677 |
+
"eval_wer": 0.19010397463313916,
|
| 194678 |
+
"step": 74912
|
| 194679 |
+
},
|
| 194680 |
+
{
|
| 194681 |
+
"epoch": 599.02,
|
| 194682 |
+
"learning_rate": 8.815929487179488e-06,
|
| 194683 |
+
"loss": 0.3721,
|
| 194684 |
+
"step": 74915
|
| 194685 |
+
},
|
| 194686 |
+
{
|
| 194687 |
+
"epoch": 599.06,
|
| 194688 |
+
"learning_rate": 8.81584935897436e-06,
|
| 194689 |
+
"loss": 0.2658,
|
| 194690 |
+
"step": 74920
|
| 194691 |
+
},
|
| 194692 |
+
{
|
| 194693 |
+
"epoch": 599.1,
|
| 194694 |
+
"learning_rate": 8.81576923076923e-06,
|
| 194695 |
+
"loss": 0.3926,
|
| 194696 |
+
"step": 74925
|
| 194697 |
+
},
|
| 194698 |
+
{
|
| 194699 |
+
"epoch": 599.14,
|
| 194700 |
+
"learning_rate": 8.815689102564103e-06,
|
| 194701 |
+
"loss": 0.3716,
|
| 194702 |
+
"step": 74930
|
| 194703 |
+
},
|
| 194704 |
+
{
|
| 194705 |
+
"epoch": 599.18,
|
| 194706 |
+
"learning_rate": 8.815608974358975e-06,
|
| 194707 |
+
"loss": 0.7772,
|
| 194708 |
+
"step": 74935
|
| 194709 |
+
},
|
| 194710 |
+
{
|
| 194711 |
+
"epoch": 599.22,
|
| 194712 |
+
"learning_rate": 8.815528846153846e-06,
|
| 194713 |
+
"loss": 0.8761,
|
| 194714 |
+
"step": 74940
|
| 194715 |
+
},
|
| 194716 |
+
{
|
| 194717 |
+
"epoch": 599.26,
|
| 194718 |
+
"learning_rate": 8.81544871794872e-06,
|
| 194719 |
+
"loss": 0.3565,
|
| 194720 |
+
"step": 74945
|
| 194721 |
+
},
|
| 194722 |
+
{
|
| 194723 |
+
"epoch": 599.3,
|
| 194724 |
+
"learning_rate": 8.81536858974359e-06,
|
| 194725 |
+
"loss": 0.363,
|
| 194726 |
+
"step": 74950
|
| 194727 |
+
},
|
| 194728 |
+
{
|
| 194729 |
+
"epoch": 599.34,
|
| 194730 |
+
"learning_rate": 8.815288461538462e-06,
|
| 194731 |
+
"loss": 0.361,
|
| 194732 |
+
"step": 74955
|
| 194733 |
+
},
|
| 194734 |
+
{
|
| 194735 |
+
"epoch": 599.38,
|
| 194736 |
+
"learning_rate": 8.815208333333333e-06,
|
| 194737 |
+
"loss": 0.7271,
|
| 194738 |
+
"step": 74960
|
| 194739 |
+
},
|
| 194740 |
+
{
|
| 194741 |
+
"epoch": 599.42,
|
| 194742 |
+
"learning_rate": 8.815128205128206e-06,
|
| 194743 |
+
"loss": 0.9404,
|
| 194744 |
+
"step": 74965
|
| 194745 |
+
},
|
| 194746 |
+
{
|
| 194747 |
+
"epoch": 599.46,
|
| 194748 |
+
"learning_rate": 8.815048076923078e-06,
|
| 194749 |
+
"loss": 0.3435,
|
| 194750 |
+
"step": 74970
|
| 194751 |
+
},
|
| 194752 |
+
{
|
| 194753 |
+
"epoch": 599.5,
|
| 194754 |
+
"learning_rate": 8.814967948717949e-06,
|
| 194755 |
+
"loss": 0.3344,
|
| 194756 |
+
"step": 74975
|
| 194757 |
+
},
|
| 194758 |
+
{
|
| 194759 |
+
"epoch": 599.54,
|
| 194760 |
+
"learning_rate": 8.81488782051282e-06,
|
| 194761 |
+
"loss": 0.4215,
|
| 194762 |
+
"step": 74980
|
| 194763 |
+
},
|
| 194764 |
+
{
|
| 194765 |
+
"epoch": 599.58,
|
| 194766 |
+
"learning_rate": 8.814807692307693e-06,
|
| 194767 |
+
"loss": 0.7642,
|
| 194768 |
+
"step": 74985
|
| 194769 |
+
},
|
| 194770 |
+
{
|
| 194771 |
+
"epoch": 599.62,
|
| 194772 |
+
"learning_rate": 8.814727564102565e-06,
|
| 194773 |
+
"loss": 0.9729,
|
| 194774 |
+
"step": 74990
|
| 194775 |
+
},
|
| 194776 |
+
{
|
| 194777 |
+
"epoch": 599.66,
|
| 194778 |
+
"learning_rate": 8.814647435897436e-06,
|
| 194779 |
+
"loss": 0.3428,
|
| 194780 |
+
"step": 74995
|
| 194781 |
+
},
|
| 194782 |
+
{
|
| 194783 |
+
"epoch": 599.7,
|
| 194784 |
+
"learning_rate": 8.81456730769231e-06,
|
| 194785 |
+
"loss": 0.2913,
|
| 194786 |
+
"step": 75000
|
| 194787 |
+
},
|
| 194788 |
+
{
|
| 194789 |
+
"epoch": 599.74,
|
| 194790 |
+
"learning_rate": 8.81448717948718e-06,
|
| 194791 |
+
"loss": 0.3823,
|
| 194792 |
+
"step": 75005
|
| 194793 |
+
},
|
| 194794 |
+
{
|
| 194795 |
+
"epoch": 599.78,
|
| 194796 |
+
"learning_rate": 8.814407051282052e-06,
|
| 194797 |
+
"loss": 0.6696,
|
| 194798 |
+
"step": 75010
|
| 194799 |
+
},
|
| 194800 |
+
{
|
| 194801 |
+
"epoch": 599.82,
|
| 194802 |
+
"learning_rate": 8.814326923076923e-06,
|
| 194803 |
+
"loss": 0.9848,
|
| 194804 |
+
"step": 75015
|
| 194805 |
+
},
|
| 194806 |
+
{
|
| 194807 |
+
"epoch": 599.86,
|
| 194808 |
+
"learning_rate": 8.814246794871796e-06,
|
| 194809 |
+
"loss": 0.2948,
|
| 194810 |
+
"step": 75020
|
| 194811 |
+
},
|
| 194812 |
+
{
|
| 194813 |
+
"epoch": 599.9,
|
| 194814 |
+
"learning_rate": 8.814166666666668e-06,
|
| 194815 |
+
"loss": 0.3141,
|
| 194816 |
+
"step": 75025
|
| 194817 |
+
},
|
| 194818 |
+
{
|
| 194819 |
+
"epoch": 599.94,
|
| 194820 |
+
"learning_rate": 8.814086538461539e-06,
|
| 194821 |
+
"loss": 0.4113,
|
| 194822 |
+
"step": 75030
|
| 194823 |
+
},
|
| 194824 |
+
{
|
| 194825 |
+
"epoch": 599.98,
|
| 194826 |
+
"learning_rate": 8.81400641025641e-06,
|
| 194827 |
+
"loss": 0.7153,
|
| 194828 |
+
"step": 75035
|
| 194829 |
+
},
|
| 194830 |
+
{
|
| 194831 |
+
"epoch": 600.0,
|
| 194832 |
+
"eval_loss": 0.5015895962715149,
|
| 194833 |
+
"eval_runtime": 39.7322,
|
| 194834 |
+
"eval_samples_per_second": 21.142,
|
| 194835 |
+
"eval_steps_per_second": 0.68,
|
| 194836 |
+
"eval_wer": 0.20653854649878506,
|
| 194837 |
+
"step": 75037
|
| 194838 |
+
},
|
| 194839 |
+
{
|
| 194840 |
+
"epoch": 600.02,
|
| 194841 |
+
"learning_rate": 8.813926282051283e-06,
|
| 194842 |
+
"loss": 0.3683,
|
| 194843 |
+
"step": 75040
|
| 194844 |
+
},
|
| 194845 |
+
{
|
| 194846 |
+
"epoch": 600.06,
|
| 194847 |
+
"learning_rate": 8.813846153846155e-06,
|
| 194848 |
+
"loss": 0.27,
|
| 194849 |
+
"step": 75045
|
| 194850 |
+
},
|
| 194851 |
+
{
|
| 194852 |
+
"epoch": 600.1,
|
| 194853 |
+
"learning_rate": 8.813766025641026e-06,
|
| 194854 |
+
"loss": 0.3321,
|
| 194855 |
+
"step": 75050
|
| 194856 |
+
},
|
| 194857 |
+
{
|
| 194858 |
+
"epoch": 600.14,
|
| 194859 |
+
"learning_rate": 8.8136858974359e-06,
|
| 194860 |
+
"loss": 0.4144,
|
| 194861 |
+
"step": 75055
|
| 194862 |
+
},
|
| 194863 |
+
{
|
| 194864 |
+
"epoch": 600.18,
|
| 194865 |
+
"learning_rate": 8.813605769230769e-06,
|
| 194866 |
+
"loss": 0.9552,
|
| 194867 |
+
"step": 75060
|
| 194868 |
+
},
|
| 194869 |
+
{
|
| 194870 |
+
"epoch": 600.22,
|
| 194871 |
+
"learning_rate": 8.813525641025642e-06,
|
| 194872 |
+
"loss": 0.9213,
|
| 194873 |
+
"step": 75065
|
| 194874 |
+
},
|
| 194875 |
+
{
|
| 194876 |
+
"epoch": 600.26,
|
| 194877 |
+
"learning_rate": 8.813445512820513e-06,
|
| 194878 |
+
"loss": 0.3018,
|
| 194879 |
+
"step": 75070
|
| 194880 |
+
},
|
| 194881 |
+
{
|
| 194882 |
+
"epoch": 600.3,
|
| 194883 |
+
"learning_rate": 8.813365384615385e-06,
|
| 194884 |
+
"loss": 0.3287,
|
| 194885 |
+
"step": 75075
|
| 194886 |
+
},
|
| 194887 |
+
{
|
| 194888 |
+
"epoch": 600.34,
|
| 194889 |
+
"learning_rate": 8.813285256410256e-06,
|
| 194890 |
+
"loss": 0.4535,
|
| 194891 |
+
"step": 75080
|
| 194892 |
+
},
|
| 194893 |
+
{
|
| 194894 |
+
"epoch": 600.38,
|
| 194895 |
+
"learning_rate": 8.813205128205129e-06,
|
| 194896 |
+
"loss": 0.8355,
|
| 194897 |
+
"step": 75085
|
| 194898 |
+
},
|
| 194899 |
+
{
|
| 194900 |
+
"epoch": 600.42,
|
| 194901 |
+
"learning_rate": 8.813125e-06,
|
| 194902 |
+
"loss": 0.8693,
|
| 194903 |
+
"step": 75090
|
| 194904 |
+
},
|
| 194905 |
+
{
|
| 194906 |
+
"epoch": 600.46,
|
| 194907 |
+
"learning_rate": 8.813044871794872e-06,
|
| 194908 |
+
"loss": 0.2923,
|
| 194909 |
+
"step": 75095
|
| 194910 |
+
},
|
| 194911 |
+
{
|
| 194912 |
+
"epoch": 600.5,
|
| 194913 |
+
"learning_rate": 8.812964743589745e-06,
|
| 194914 |
+
"loss": 0.3067,
|
| 194915 |
+
"step": 75100
|
| 194916 |
+
},
|
| 194917 |
+
{
|
| 194918 |
+
"epoch": 600.54,
|
| 194919 |
+
"learning_rate": 8.812884615384616e-06,
|
| 194920 |
+
"loss": 0.3881,
|
| 194921 |
+
"step": 75105
|
| 194922 |
+
},
|
| 194923 |
+
{
|
| 194924 |
+
"epoch": 600.58,
|
| 194925 |
+
"learning_rate": 8.812804487179488e-06,
|
| 194926 |
+
"loss": 0.7686,
|
| 194927 |
+
"step": 75110
|
| 194928 |
+
},
|
| 194929 |
+
{
|
| 194930 |
+
"epoch": 600.62,
|
| 194931 |
+
"learning_rate": 8.812724358974359e-06,
|
| 194932 |
+
"loss": 0.9086,
|
| 194933 |
+
"step": 75115
|
| 194934 |
+
},
|
| 194935 |
+
{
|
| 194936 |
+
"epoch": 600.66,
|
| 194937 |
+
"learning_rate": 8.812644230769232e-06,
|
| 194938 |
+
"loss": 0.2994,
|
| 194939 |
+
"step": 75120
|
| 194940 |
+
},
|
| 194941 |
+
{
|
| 194942 |
+
"epoch": 600.7,
|
| 194943 |
+
"learning_rate": 8.812564102564103e-06,
|
| 194944 |
+
"loss": 0.3495,
|
| 194945 |
+
"step": 75125
|
| 194946 |
+
},
|
| 194947 |
+
{
|
| 194948 |
+
"epoch": 600.74,
|
| 194949 |
+
"learning_rate": 8.812483974358975e-06,
|
| 194950 |
+
"loss": 0.3605,
|
| 194951 |
+
"step": 75130
|
| 194952 |
+
},
|
| 194953 |
+
{
|
| 194954 |
+
"epoch": 600.78,
|
| 194955 |
+
"learning_rate": 8.812403846153846e-06,
|
| 194956 |
+
"loss": 0.8099,
|
| 194957 |
+
"step": 75135
|
| 194958 |
+
},
|
| 194959 |
+
{
|
| 194960 |
+
"epoch": 600.82,
|
| 194961 |
+
"learning_rate": 8.812323717948719e-06,
|
| 194962 |
+
"loss": 1.0115,
|
| 194963 |
+
"step": 75140
|
| 194964 |
+
},
|
| 194965 |
+
{
|
| 194966 |
+
"epoch": 600.86,
|
| 194967 |
+
"learning_rate": 8.81224358974359e-06,
|
| 194968 |
+
"loss": 0.3424,
|
| 194969 |
+
"step": 75145
|
| 194970 |
+
},
|
| 194971 |
+
{
|
| 194972 |
+
"epoch": 600.9,
|
| 194973 |
+
"learning_rate": 8.812163461538462e-06,
|
| 194974 |
+
"loss": 0.2935,
|
| 194975 |
+
"step": 75150
|
| 194976 |
+
},
|
| 194977 |
+
{
|
| 194978 |
+
"epoch": 600.94,
|
| 194979 |
+
"learning_rate": 8.812083333333335e-06,
|
| 194980 |
+
"loss": 0.3522,
|
| 194981 |
+
"step": 75155
|
| 194982 |
+
},
|
| 194983 |
+
{
|
| 194984 |
+
"epoch": 600.98,
|
| 194985 |
+
"learning_rate": 8.812003205128206e-06,
|
| 194986 |
+
"loss": 0.9099,
|
| 194987 |
+
"step": 75160
|
| 194988 |
+
},
|
| 194989 |
+
{
|
| 194990 |
+
"epoch": 601.0,
|
| 194991 |
+
"eval_loss": 0.41707369685173035,
|
| 194992 |
+
"eval_runtime": 40.6126,
|
| 194993 |
+
"eval_samples_per_second": 20.708,
|
| 194994 |
+
"eval_steps_per_second": 0.665,
|
| 194995 |
+
"eval_wer": 0.2013764427557531,
|
| 194996 |
+
"step": 75162
|
| 194997 |
}
|
| 194998 |
],
|
| 194999 |
+
"max_steps": 625000,
|
| 195000 |
"num_train_epochs": 5000,
|
| 195001 |
+
"total_flos": 2.115183129640869e+20,
|
| 195002 |
"trial_name": null,
|
| 195003 |
"trial_params": null
|
| 195004 |
}
|
model-bin/finetune/base/{checkpoint-74539 → checkpoint-75162}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629879092.38512/events.out.tfevents.1629879092.7e498afd5545.905.83
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:de16129103ae23782997a2f6e16821239d7b72f259380afebb6a9c2bafa80298
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629879755.9543498/events.out.tfevents.1629879755.7e498afd5545.905.85
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:daf01e48c1f5fdaa4932cbf36a0e8f1a360a6c9d8d78d37ea17dbc37f9889fbc
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629880403.5211415/events.out.tfevents.1629880403.7e498afd5545.905.87
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:73f895454c759316114bbc447d0826926107eacd8b1c68db68b79c05418dfdb3
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629881049.788205/events.out.tfevents.1629881049.7e498afd5545.905.89
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:5eb08ad02c8458e333fddd2e3bcab37c63f5cf5bd79fc60136fe6ab62c5abfad
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629881697.4184577/events.out.tfevents.1629881697.7e498afd5545.905.91
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a9e7fb9b395276852faf078e20ad5747928b80f779c086d8d8d6e472d95dc267
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629879092.7e498afd5545.905.82
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1f4d2778fc57e6d6438bb3f914aa609d397f723631fb70490c2bf77e9ff71107
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629879755.7e498afd5545.905.84
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:5e987a946247bf31d37e50ff0bec8308355b873abadef1ee8a0ea045987b7ecf
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629880403.7e498afd5545.905.86
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:0fca7d1a3a06e4fe4d2adbfbe329d1bff6334e8c455a0a4ef00b8cbd407248b0
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629881049.7e498afd5545.905.88
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:36541710ccff020bfe48255466aa1183422f6b931e5ec67e8803f8e2442a82a2
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629881697.7e498afd5545.905.90
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:cee8924f68cd5f67852530719b67535c0badcd4e01529498c6b85fcc616fd300
|
| 3 |
+
size 8622
|