"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-40819 → checkpoint-41441}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-40819 → checkpoint-41441}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-40819 → checkpoint-41441}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-40819 → checkpoint-41441}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-40819 → checkpoint-41441}/rng_state.pth +2 -2
- model-bin/finetune/base/{checkpoint-40819 → checkpoint-41441}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-40819 → checkpoint-41441}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-40819 → checkpoint-41441}/trainer_state.json +798 -3
- model-bin/finetune/base/{checkpoint-40819 → checkpoint-41441}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629693366.2436578/events.out.tfevents.1629693366.74272264b15c.932.43 +3 -0
- model-bin/finetune/base/log/1629694012.331258/events.out.tfevents.1629694012.74272264b15c.932.45 +3 -0
- model-bin/finetune/base/log/1629694653.2687066/events.out.tfevents.1629694653.74272264b15c.932.47 +3 -0
- model-bin/finetune/base/log/1629695278.107289/events.out.tfevents.1629695278.74272264b15c.932.49 +3 -0
- model-bin/finetune/base/log/1629695916.3520517/events.out.tfevents.1629695916.74272264b15c.932.51 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629693366.74272264b15c.932.42 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629694012.74272264b15c.932.44 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629694653.74272264b15c.932.46 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629695278.74272264b15c.932.48 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629695916.74272264b15c.932.50 +3 -0
model-bin/finetune/base/{checkpoint-40819 → checkpoint-41441}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-40819 → checkpoint-41441}/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 sha256:8496ff852a4eee2692a1ecb937ff5428f8c8aedcb54434c9d77d2f0aa8ae9e8b
|
| 3 |
size 722165009
|
model-bin/finetune/base/{checkpoint-40819 → checkpoint-41441}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-40819 → checkpoint-41441}/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 sha256:27fcb82ce2589b043f60b3e8ae108cd2e54bc6a4d37c76caef9d323961950097
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-40819 → checkpoint-41441}/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 sha256:6ff43d888f977e4ac4a033b8a9571e3e5d2810a4b89fbfd97d80e79e8b4b3e24
|
| 3 |
+
size 14503
|
model-bin/finetune/base/{checkpoint-40819 → checkpoint-41441}/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:eac27864a7ffc1e1d62479a5cbb3630c999046024edaba46006f14469947e2c5
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-40819 → checkpoint-41441}/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:ae2d3e6883eeddb1e27105dc1e1a8275ad0a680fe3dd8f8a12c5243a705581f4
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-40819 → checkpoint-41441}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.2028974903051145,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-39323",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -151296,11 +151296,806 @@
|
|
| 151296 |
"eval_steps_per_second": 0.689,
|
| 151297 |
"eval_wer": 0.2033972763215698,
|
| 151298 |
"step": 40819
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 151299 |
}
|
| 151300 |
],
|
| 151301 |
"max_steps": 625000,
|
| 151302 |
"num_train_epochs": 5000,
|
| 151303 |
-
"total_flos": 1.
|
| 151304 |
"trial_name": null,
|
| 151305 |
"trial_params": null
|
| 151306 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.2028974903051145,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-39323",
|
| 4 |
+
"epoch": 330.996015936255,
|
| 5 |
+
"global_step": 41441,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 151296 |
"eval_steps_per_second": 0.689,
|
| 151297 |
"eval_wer": 0.2033972763215698,
|
| 151298 |
"step": 40819
|
| 151299 |
+
},
|
| 151300 |
+
{
|
| 151301 |
+
"epoch": 329.01,
|
| 151302 |
+
"learning_rate": 9.362115384615385e-06,
|
| 151303 |
+
"loss": 0.4174,
|
| 151304 |
+
"step": 40820
|
| 151305 |
+
},
|
| 151306 |
+
{
|
| 151307 |
+
"epoch": 329.05,
|
| 151308 |
+
"learning_rate": 9.362035256410256e-06,
|
| 151309 |
+
"loss": 0.3265,
|
| 151310 |
+
"step": 40825
|
| 151311 |
+
},
|
| 151312 |
+
{
|
| 151313 |
+
"epoch": 329.09,
|
| 151314 |
+
"learning_rate": 9.36195512820513e-06,
|
| 151315 |
+
"loss": 0.3712,
|
| 151316 |
+
"step": 40830
|
| 151317 |
+
},
|
| 151318 |
+
{
|
| 151319 |
+
"epoch": 329.13,
|
| 151320 |
+
"learning_rate": 9.361875000000001e-06,
|
| 151321 |
+
"loss": 0.421,
|
| 151322 |
+
"step": 40835
|
| 151323 |
+
},
|
| 151324 |
+
{
|
| 151325 |
+
"epoch": 329.17,
|
| 151326 |
+
"learning_rate": 9.361794871794872e-06,
|
| 151327 |
+
"loss": 0.7053,
|
| 151328 |
+
"step": 40840
|
| 151329 |
+
},
|
| 151330 |
+
{
|
| 151331 |
+
"epoch": 329.21,
|
| 151332 |
+
"learning_rate": 9.361714743589745e-06,
|
| 151333 |
+
"loss": 1.4004,
|
| 151334 |
+
"step": 40845
|
| 151335 |
+
},
|
| 151336 |
+
{
|
| 151337 |
+
"epoch": 329.25,
|
| 151338 |
+
"learning_rate": 9.361634615384617e-06,
|
| 151339 |
+
"loss": 0.4308,
|
| 151340 |
+
"step": 40850
|
| 151341 |
+
},
|
| 151342 |
+
{
|
| 151343 |
+
"epoch": 329.29,
|
| 151344 |
+
"learning_rate": 9.361554487179488e-06,
|
| 151345 |
+
"loss": 0.3684,
|
| 151346 |
+
"step": 40855
|
| 151347 |
+
},
|
| 151348 |
+
{
|
| 151349 |
+
"epoch": 329.33,
|
| 151350 |
+
"learning_rate": 9.36147435897436e-06,
|
| 151351 |
+
"loss": 0.3801,
|
| 151352 |
+
"step": 40860
|
| 151353 |
+
},
|
| 151354 |
+
{
|
| 151355 |
+
"epoch": 329.37,
|
| 151356 |
+
"learning_rate": 9.361394230769232e-06,
|
| 151357 |
+
"loss": 0.655,
|
| 151358 |
+
"step": 40865
|
| 151359 |
+
},
|
| 151360 |
+
{
|
| 151361 |
+
"epoch": 329.41,
|
| 151362 |
+
"learning_rate": 9.361314102564104e-06,
|
| 151363 |
+
"loss": 1.3581,
|
| 151364 |
+
"step": 40870
|
| 151365 |
+
},
|
| 151366 |
+
{
|
| 151367 |
+
"epoch": 329.45,
|
| 151368 |
+
"learning_rate": 9.361233974358975e-06,
|
| 151369 |
+
"loss": 0.3331,
|
| 151370 |
+
"step": 40875
|
| 151371 |
+
},
|
| 151372 |
+
{
|
| 151373 |
+
"epoch": 329.49,
|
| 151374 |
+
"learning_rate": 9.361153846153847e-06,
|
| 151375 |
+
"loss": 0.3514,
|
| 151376 |
+
"step": 40880
|
| 151377 |
+
},
|
| 151378 |
+
{
|
| 151379 |
+
"epoch": 329.53,
|
| 151380 |
+
"learning_rate": 9.36107371794872e-06,
|
| 151381 |
+
"loss": 0.367,
|
| 151382 |
+
"step": 40885
|
| 151383 |
+
},
|
| 151384 |
+
{
|
| 151385 |
+
"epoch": 329.57,
|
| 151386 |
+
"learning_rate": 9.36099358974359e-06,
|
| 151387 |
+
"loss": 0.5829,
|
| 151388 |
+
"step": 40890
|
| 151389 |
+
},
|
| 151390 |
+
{
|
| 151391 |
+
"epoch": 329.61,
|
| 151392 |
+
"learning_rate": 9.360913461538462e-06,
|
| 151393 |
+
"loss": 1.3516,
|
| 151394 |
+
"step": 40895
|
| 151395 |
+
},
|
| 151396 |
+
{
|
| 151397 |
+
"epoch": 329.65,
|
| 151398 |
+
"learning_rate": 9.360833333333335e-06,
|
| 151399 |
+
"loss": 0.4371,
|
| 151400 |
+
"step": 40900
|
| 151401 |
+
},
|
| 151402 |
+
{
|
| 151403 |
+
"epoch": 329.69,
|
| 151404 |
+
"learning_rate": 9.360753205128205e-06,
|
| 151405 |
+
"loss": 0.3099,
|
| 151406 |
+
"step": 40905
|
| 151407 |
+
},
|
| 151408 |
+
{
|
| 151409 |
+
"epoch": 329.73,
|
| 151410 |
+
"learning_rate": 9.360673076923078e-06,
|
| 151411 |
+
"loss": 0.3453,
|
| 151412 |
+
"step": 40910
|
| 151413 |
+
},
|
| 151414 |
+
{
|
| 151415 |
+
"epoch": 329.77,
|
| 151416 |
+
"learning_rate": 9.36059294871795e-06,
|
| 151417 |
+
"loss": 0.5304,
|
| 151418 |
+
"step": 40915
|
| 151419 |
+
},
|
| 151420 |
+
{
|
| 151421 |
+
"epoch": 329.81,
|
| 151422 |
+
"learning_rate": 9.36051282051282e-06,
|
| 151423 |
+
"loss": 1.2225,
|
| 151424 |
+
"step": 40920
|
| 151425 |
+
},
|
| 151426 |
+
{
|
| 151427 |
+
"epoch": 329.85,
|
| 151428 |
+
"learning_rate": 9.360432692307692e-06,
|
| 151429 |
+
"loss": 0.3677,
|
| 151430 |
+
"step": 40925
|
| 151431 |
+
},
|
| 151432 |
+
{
|
| 151433 |
+
"epoch": 329.89,
|
| 151434 |
+
"learning_rate": 9.360352564102565e-06,
|
| 151435 |
+
"loss": 0.4049,
|
| 151436 |
+
"step": 40930
|
| 151437 |
+
},
|
| 151438 |
+
{
|
| 151439 |
+
"epoch": 329.93,
|
| 151440 |
+
"learning_rate": 9.360272435897437e-06,
|
| 151441 |
+
"loss": 0.4002,
|
| 151442 |
+
"step": 40935
|
| 151443 |
+
},
|
| 151444 |
+
{
|
| 151445 |
+
"epoch": 329.97,
|
| 151446 |
+
"learning_rate": 9.360192307692308e-06,
|
| 151447 |
+
"loss": 0.75,
|
| 151448 |
+
"step": 40940
|
| 151449 |
+
},
|
| 151450 |
+
{
|
| 151451 |
+
"epoch": 330.0,
|
| 151452 |
+
"eval_loss": 0.43646708130836487,
|
| 151453 |
+
"eval_runtime": 38.8578,
|
| 151454 |
+
"eval_samples_per_second": 21.617,
|
| 151455 |
+
"eval_steps_per_second": 0.695,
|
| 151456 |
+
"eval_wer": 0.20797337487607986,
|
| 151457 |
+
"step": 40943
|
| 151458 |
+
},
|
| 151459 |
+
{
|
| 151460 |
+
"epoch": 330.02,
|
| 151461 |
+
"learning_rate": 9.360112179487181e-06,
|
| 151462 |
+
"loss": 0.4076,
|
| 151463 |
+
"step": 40945
|
| 151464 |
+
},
|
| 151465 |
+
{
|
| 151466 |
+
"epoch": 330.06,
|
| 151467 |
+
"learning_rate": 9.360032051282052e-06,
|
| 151468 |
+
"loss": 0.3221,
|
| 151469 |
+
"step": 40950
|
| 151470 |
+
},
|
| 151471 |
+
{
|
| 151472 |
+
"epoch": 330.1,
|
| 151473 |
+
"learning_rate": 9.359951923076924e-06,
|
| 151474 |
+
"loss": 0.4117,
|
| 151475 |
+
"step": 40955
|
| 151476 |
+
},
|
| 151477 |
+
{
|
| 151478 |
+
"epoch": 330.14,
|
| 151479 |
+
"learning_rate": 9.359871794871795e-06,
|
| 151480 |
+
"loss": 0.4201,
|
| 151481 |
+
"step": 40960
|
| 151482 |
+
},
|
| 151483 |
+
{
|
| 151484 |
+
"epoch": 330.18,
|
| 151485 |
+
"learning_rate": 9.359791666666668e-06,
|
| 151486 |
+
"loss": 0.8076,
|
| 151487 |
+
"step": 40965
|
| 151488 |
+
},
|
| 151489 |
+
{
|
| 151490 |
+
"epoch": 330.22,
|
| 151491 |
+
"learning_rate": 9.35971153846154e-06,
|
| 151492 |
+
"loss": 1.1731,
|
| 151493 |
+
"step": 40970
|
| 151494 |
+
},
|
| 151495 |
+
{
|
| 151496 |
+
"epoch": 330.26,
|
| 151497 |
+
"learning_rate": 9.35963141025641e-06,
|
| 151498 |
+
"loss": 0.3538,
|
| 151499 |
+
"step": 40975
|
| 151500 |
+
},
|
| 151501 |
+
{
|
| 151502 |
+
"epoch": 330.3,
|
| 151503 |
+
"learning_rate": 9.359551282051282e-06,
|
| 151504 |
+
"loss": 0.3134,
|
| 151505 |
+
"step": 40980
|
| 151506 |
+
},
|
| 151507 |
+
{
|
| 151508 |
+
"epoch": 330.34,
|
| 151509 |
+
"learning_rate": 9.359471153846155e-06,
|
| 151510 |
+
"loss": 0.4412,
|
| 151511 |
+
"step": 40985
|
| 151512 |
+
},
|
| 151513 |
+
{
|
| 151514 |
+
"epoch": 330.38,
|
| 151515 |
+
"learning_rate": 9.359391025641027e-06,
|
| 151516 |
+
"loss": 0.7839,
|
| 151517 |
+
"step": 40990
|
| 151518 |
+
},
|
| 151519 |
+
{
|
| 151520 |
+
"epoch": 330.42,
|
| 151521 |
+
"learning_rate": 9.359310897435898e-06,
|
| 151522 |
+
"loss": 1.1787,
|
| 151523 |
+
"step": 40995
|
| 151524 |
+
},
|
| 151525 |
+
{
|
| 151526 |
+
"epoch": 330.46,
|
| 151527 |
+
"learning_rate": 9.359230769230771e-06,
|
| 151528 |
+
"loss": 0.3157,
|
| 151529 |
+
"step": 41000
|
| 151530 |
+
},
|
| 151531 |
+
{
|
| 151532 |
+
"epoch": 330.5,
|
| 151533 |
+
"learning_rate": 9.359150641025642e-06,
|
| 151534 |
+
"loss": 0.3251,
|
| 151535 |
+
"step": 41005
|
| 151536 |
+
},
|
| 151537 |
+
{
|
| 151538 |
+
"epoch": 330.54,
|
| 151539 |
+
"learning_rate": 9.359070512820514e-06,
|
| 151540 |
+
"loss": 0.4251,
|
| 151541 |
+
"step": 41010
|
| 151542 |
+
},
|
| 151543 |
+
{
|
| 151544 |
+
"epoch": 330.58,
|
| 151545 |
+
"learning_rate": 9.358990384615385e-06,
|
| 151546 |
+
"loss": 0.6777,
|
| 151547 |
+
"step": 41015
|
| 151548 |
+
},
|
| 151549 |
+
{
|
| 151550 |
+
"epoch": 330.62,
|
| 151551 |
+
"learning_rate": 9.358910256410258e-06,
|
| 151552 |
+
"loss": 1.1292,
|
| 151553 |
+
"step": 41020
|
| 151554 |
+
},
|
| 151555 |
+
{
|
| 151556 |
+
"epoch": 330.66,
|
| 151557 |
+
"learning_rate": 9.358830128205128e-06,
|
| 151558 |
+
"loss": 0.3494,
|
| 151559 |
+
"step": 41025
|
| 151560 |
+
},
|
| 151561 |
+
{
|
| 151562 |
+
"epoch": 330.7,
|
| 151563 |
+
"learning_rate": 9.35875e-06,
|
| 151564 |
+
"loss": 0.3142,
|
| 151565 |
+
"step": 41030
|
| 151566 |
+
},
|
| 151567 |
+
{
|
| 151568 |
+
"epoch": 330.74,
|
| 151569 |
+
"learning_rate": 9.358669871794872e-06,
|
| 151570 |
+
"loss": 0.4359,
|
| 151571 |
+
"step": 41035
|
| 151572 |
+
},
|
| 151573 |
+
{
|
| 151574 |
+
"epoch": 330.78,
|
| 151575 |
+
"learning_rate": 9.358589743589744e-06,
|
| 151576 |
+
"loss": 0.6908,
|
| 151577 |
+
"step": 41040
|
| 151578 |
+
},
|
| 151579 |
+
{
|
| 151580 |
+
"epoch": 330.82,
|
| 151581 |
+
"learning_rate": 9.358509615384617e-06,
|
| 151582 |
+
"loss": 1.1847,
|
| 151583 |
+
"step": 41045
|
| 151584 |
+
},
|
| 151585 |
+
{
|
| 151586 |
+
"epoch": 330.86,
|
| 151587 |
+
"learning_rate": 9.358429487179488e-06,
|
| 151588 |
+
"loss": 0.3698,
|
| 151589 |
+
"step": 41050
|
| 151590 |
+
},
|
| 151591 |
+
{
|
| 151592 |
+
"epoch": 330.9,
|
| 151593 |
+
"learning_rate": 9.35834935897436e-06,
|
| 151594 |
+
"loss": 0.3982,
|
| 151595 |
+
"step": 41055
|
| 151596 |
+
},
|
| 151597 |
+
{
|
| 151598 |
+
"epoch": 330.94,
|
| 151599 |
+
"learning_rate": 9.35826923076923e-06,
|
| 151600 |
+
"loss": 1.118,
|
| 151601 |
+
"step": 41060
|
| 151602 |
+
},
|
| 151603 |
+
{
|
| 151604 |
+
"epoch": 330.98,
|
| 151605 |
+
"learning_rate": 9.358189102564104e-06,
|
| 151606 |
+
"loss": 0.8147,
|
| 151607 |
+
"step": 41065
|
| 151608 |
+
},
|
| 151609 |
+
{
|
| 151610 |
+
"epoch": 331.0,
|
| 151611 |
+
"eval_loss": 0.4650619626045227,
|
| 151612 |
+
"eval_runtime": 40.2043,
|
| 151613 |
+
"eval_samples_per_second": 20.893,
|
| 151614 |
+
"eval_steps_per_second": 0.672,
|
| 151615 |
+
"eval_wer": 0.2111755635509215,
|
| 151616 |
+
"step": 41067
|
| 151617 |
+
},
|
| 151618 |
+
{
|
| 151619 |
+
"epoch": 331.02,
|
| 151620 |
+
"learning_rate": 9.358108974358975e-06,
|
| 151621 |
+
"loss": 0.5312,
|
| 151622 |
+
"step": 41070
|
| 151623 |
+
},
|
| 151624 |
+
{
|
| 151625 |
+
"epoch": 331.06,
|
| 151626 |
+
"learning_rate": 9.358028846153846e-06,
|
| 151627 |
+
"loss": 0.4109,
|
| 151628 |
+
"step": 41075
|
| 151629 |
+
},
|
| 151630 |
+
{
|
| 151631 |
+
"epoch": 331.1,
|
| 151632 |
+
"learning_rate": 9.357948717948718e-06,
|
| 151633 |
+
"loss": 0.4092,
|
| 151634 |
+
"step": 41080
|
| 151635 |
+
},
|
| 151636 |
+
{
|
| 151637 |
+
"epoch": 331.14,
|
| 151638 |
+
"learning_rate": 9.35786858974359e-06,
|
| 151639 |
+
"loss": 0.427,
|
| 151640 |
+
"step": 41085
|
| 151641 |
+
},
|
| 151642 |
+
{
|
| 151643 |
+
"epoch": 331.18,
|
| 151644 |
+
"learning_rate": 9.357788461538462e-06,
|
| 151645 |
+
"loss": 0.941,
|
| 151646 |
+
"step": 41090
|
| 151647 |
+
},
|
| 151648 |
+
{
|
| 151649 |
+
"epoch": 331.22,
|
| 151650 |
+
"learning_rate": 9.357708333333334e-06,
|
| 151651 |
+
"loss": 0.8317,
|
| 151652 |
+
"step": 41095
|
| 151653 |
+
},
|
| 151654 |
+
{
|
| 151655 |
+
"epoch": 331.27,
|
| 151656 |
+
"learning_rate": 9.357628205128207e-06,
|
| 151657 |
+
"loss": 0.3703,
|
| 151658 |
+
"step": 41100
|
| 151659 |
+
},
|
| 151660 |
+
{
|
| 151661 |
+
"epoch": 331.31,
|
| 151662 |
+
"learning_rate": 9.357548076923078e-06,
|
| 151663 |
+
"loss": 0.3447,
|
| 151664 |
+
"step": 41105
|
| 151665 |
+
},
|
| 151666 |
+
{
|
| 151667 |
+
"epoch": 331.35,
|
| 151668 |
+
"learning_rate": 9.35746794871795e-06,
|
| 151669 |
+
"loss": 0.3982,
|
| 151670 |
+
"step": 41110
|
| 151671 |
+
},
|
| 151672 |
+
{
|
| 151673 |
+
"epoch": 331.39,
|
| 151674 |
+
"learning_rate": 9.35738782051282e-06,
|
| 151675 |
+
"loss": 0.681,
|
| 151676 |
+
"step": 41115
|
| 151677 |
+
},
|
| 151678 |
+
{
|
| 151679 |
+
"epoch": 331.43,
|
| 151680 |
+
"learning_rate": 9.357307692307694e-06,
|
| 151681 |
+
"loss": 0.9477,
|
| 151682 |
+
"step": 41120
|
| 151683 |
+
},
|
| 151684 |
+
{
|
| 151685 |
+
"epoch": 331.47,
|
| 151686 |
+
"learning_rate": 9.357227564102565e-06,
|
| 151687 |
+
"loss": 0.3787,
|
| 151688 |
+
"step": 41125
|
| 151689 |
+
},
|
| 151690 |
+
{
|
| 151691 |
+
"epoch": 331.51,
|
| 151692 |
+
"learning_rate": 9.357147435897436e-06,
|
| 151693 |
+
"loss": 0.3256,
|
| 151694 |
+
"step": 41130
|
| 151695 |
+
},
|
| 151696 |
+
{
|
| 151697 |
+
"epoch": 331.55,
|
| 151698 |
+
"learning_rate": 9.357067307692308e-06,
|
| 151699 |
+
"loss": 0.4219,
|
| 151700 |
+
"step": 41135
|
| 151701 |
+
},
|
| 151702 |
+
{
|
| 151703 |
+
"epoch": 331.59,
|
| 151704 |
+
"learning_rate": 9.356987179487181e-06,
|
| 151705 |
+
"loss": 0.8923,
|
| 151706 |
+
"step": 41140
|
| 151707 |
+
},
|
| 151708 |
+
{
|
| 151709 |
+
"epoch": 331.63,
|
| 151710 |
+
"learning_rate": 9.356907051282052e-06,
|
| 151711 |
+
"loss": 1.0381,
|
| 151712 |
+
"step": 41145
|
| 151713 |
+
},
|
| 151714 |
+
{
|
| 151715 |
+
"epoch": 331.67,
|
| 151716 |
+
"learning_rate": 9.356826923076924e-06,
|
| 151717 |
+
"loss": 0.3512,
|
| 151718 |
+
"step": 41150
|
| 151719 |
+
},
|
| 151720 |
+
{
|
| 151721 |
+
"epoch": 331.71,
|
| 151722 |
+
"learning_rate": 9.356746794871797e-06,
|
| 151723 |
+
"loss": 0.4129,
|
| 151724 |
+
"step": 41155
|
| 151725 |
+
},
|
| 151726 |
+
{
|
| 151727 |
+
"epoch": 331.75,
|
| 151728 |
+
"learning_rate": 9.356666666666668e-06,
|
| 151729 |
+
"loss": 0.4651,
|
| 151730 |
+
"step": 41160
|
| 151731 |
+
},
|
| 151732 |
+
{
|
| 151733 |
+
"epoch": 331.79,
|
| 151734 |
+
"learning_rate": 9.35658653846154e-06,
|
| 151735 |
+
"loss": 0.9417,
|
| 151736 |
+
"step": 41165
|
| 151737 |
+
},
|
| 151738 |
+
{
|
| 151739 |
+
"epoch": 331.83,
|
| 151740 |
+
"learning_rate": 9.35650641025641e-06,
|
| 151741 |
+
"loss": 1.3069,
|
| 151742 |
+
"step": 41170
|
| 151743 |
+
},
|
| 151744 |
+
{
|
| 151745 |
+
"epoch": 331.87,
|
| 151746 |
+
"learning_rate": 9.356426282051284e-06,
|
| 151747 |
+
"loss": 0.6687,
|
| 151748 |
+
"step": 41175
|
| 151749 |
+
},
|
| 151750 |
+
{
|
| 151751 |
+
"epoch": 331.91,
|
| 151752 |
+
"learning_rate": 9.356346153846153e-06,
|
| 151753 |
+
"loss": 0.3321,
|
| 151754 |
+
"step": 41180
|
| 151755 |
+
},
|
| 151756 |
+
{
|
| 151757 |
+
"epoch": 331.95,
|
| 151758 |
+
"learning_rate": 9.356266025641026e-06,
|
| 151759 |
+
"loss": 0.4239,
|
| 151760 |
+
"step": 41185
|
| 151761 |
+
},
|
| 151762 |
+
{
|
| 151763 |
+
"epoch": 331.99,
|
| 151764 |
+
"learning_rate": 9.356185897435898e-06,
|
| 151765 |
+
"loss": 0.8613,
|
| 151766 |
+
"step": 41190
|
| 151767 |
+
},
|
| 151768 |
+
{
|
| 151769 |
+
"epoch": 332.0,
|
| 151770 |
+
"eval_loss": 0.42474114894866943,
|
| 151771 |
+
"eval_runtime": 39.6558,
|
| 151772 |
+
"eval_samples_per_second": 21.207,
|
| 151773 |
+
"eval_steps_per_second": 0.681,
|
| 151774 |
+
"eval_wer": 0.21611721611721613,
|
| 151775 |
+
"step": 41191
|
| 151776 |
+
},
|
| 151777 |
+
{
|
| 151778 |
+
"epoch": 329.03,
|
| 151779 |
+
"learning_rate": 9.35610576923077e-06,
|
| 151780 |
+
"loss": 0.4057,
|
| 151781 |
+
"step": 41195
|
| 151782 |
+
},
|
| 151783 |
+
{
|
| 151784 |
+
"epoch": 329.07,
|
| 151785 |
+
"learning_rate": 9.356025641025642e-06,
|
| 151786 |
+
"loss": 0.4501,
|
| 151787 |
+
"step": 41200
|
| 151788 |
+
},
|
| 151789 |
+
{
|
| 151790 |
+
"epoch": 329.11,
|
| 151791 |
+
"learning_rate": 9.355945512820514e-06,
|
| 151792 |
+
"loss": 0.3767,
|
| 151793 |
+
"step": 41205
|
| 151794 |
+
},
|
| 151795 |
+
{
|
| 151796 |
+
"epoch": 329.15,
|
| 151797 |
+
"learning_rate": 9.355865384615385e-06,
|
| 151798 |
+
"loss": 0.637,
|
| 151799 |
+
"step": 41210
|
| 151800 |
+
},
|
| 151801 |
+
{
|
| 151802 |
+
"epoch": 329.19,
|
| 151803 |
+
"learning_rate": 9.355785256410256e-06,
|
| 151804 |
+
"loss": 1.1709,
|
| 151805 |
+
"step": 41215
|
| 151806 |
+
},
|
| 151807 |
+
{
|
| 151808 |
+
"epoch": 329.23,
|
| 151809 |
+
"learning_rate": 9.35570512820513e-06,
|
| 151810 |
+
"loss": 0.774,
|
| 151811 |
+
"step": 41220
|
| 151812 |
+
},
|
| 151813 |
+
{
|
| 151814 |
+
"epoch": 329.27,
|
| 151815 |
+
"learning_rate": 9.355625e-06,
|
| 151816 |
+
"loss": 0.331,
|
| 151817 |
+
"step": 41225
|
| 151818 |
+
},
|
| 151819 |
+
{
|
| 151820 |
+
"epoch": 329.31,
|
| 151821 |
+
"learning_rate": 9.355544871794872e-06,
|
| 151822 |
+
"loss": 0.3707,
|
| 151823 |
+
"step": 41230
|
| 151824 |
+
},
|
| 151825 |
+
{
|
| 151826 |
+
"epoch": 329.35,
|
| 151827 |
+
"learning_rate": 9.355464743589743e-06,
|
| 151828 |
+
"loss": 0.5142,
|
| 151829 |
+
"step": 41235
|
| 151830 |
+
},
|
| 151831 |
+
{
|
| 151832 |
+
"epoch": 329.39,
|
| 151833 |
+
"learning_rate": 9.355384615384617e-06,
|
| 151834 |
+
"loss": 1.0853,
|
| 151835 |
+
"step": 41240
|
| 151836 |
+
},
|
| 151837 |
+
{
|
| 151838 |
+
"epoch": 329.43,
|
| 151839 |
+
"learning_rate": 9.355304487179488e-06,
|
| 151840 |
+
"loss": 0.7162,
|
| 151841 |
+
"step": 41245
|
| 151842 |
+
},
|
| 151843 |
+
{
|
| 151844 |
+
"epoch": 329.47,
|
| 151845 |
+
"learning_rate": 9.35522435897436e-06,
|
| 151846 |
+
"loss": 0.3704,
|
| 151847 |
+
"step": 41250
|
| 151848 |
+
},
|
| 151849 |
+
{
|
| 151850 |
+
"epoch": 329.51,
|
| 151851 |
+
"learning_rate": 9.355144230769232e-06,
|
| 151852 |
+
"loss": 0.4155,
|
| 151853 |
+
"step": 41255
|
| 151854 |
+
},
|
| 151855 |
+
{
|
| 151856 |
+
"epoch": 329.55,
|
| 151857 |
+
"learning_rate": 9.355064102564104e-06,
|
| 151858 |
+
"loss": 0.3764,
|
| 151859 |
+
"step": 41260
|
| 151860 |
+
},
|
| 151861 |
+
{
|
| 151862 |
+
"epoch": 329.59,
|
| 151863 |
+
"learning_rate": 9.354983974358975e-06,
|
| 151864 |
+
"loss": 1.0982,
|
| 151865 |
+
"step": 41265
|
| 151866 |
+
},
|
| 151867 |
+
{
|
| 151868 |
+
"epoch": 329.63,
|
| 151869 |
+
"learning_rate": 9.354903846153846e-06,
|
| 151870 |
+
"loss": 0.8679,
|
| 151871 |
+
"step": 41270
|
| 151872 |
+
},
|
| 151873 |
+
{
|
| 151874 |
+
"epoch": 329.67,
|
| 151875 |
+
"learning_rate": 9.35482371794872e-06,
|
| 151876 |
+
"loss": 0.3203,
|
| 151877 |
+
"step": 41275
|
| 151878 |
+
},
|
| 151879 |
+
{
|
| 151880 |
+
"epoch": 329.71,
|
| 151881 |
+
"learning_rate": 9.35474358974359e-06,
|
| 151882 |
+
"loss": 0.3592,
|
| 151883 |
+
"step": 41280
|
| 151884 |
+
},
|
| 151885 |
+
{
|
| 151886 |
+
"epoch": 329.75,
|
| 151887 |
+
"learning_rate": 9.354663461538462e-06,
|
| 151888 |
+
"loss": 0.437,
|
| 151889 |
+
"step": 41285
|
| 151890 |
+
},
|
| 151891 |
+
{
|
| 151892 |
+
"epoch": 329.79,
|
| 151893 |
+
"learning_rate": 9.354583333333333e-06,
|
| 151894 |
+
"loss": 0.9893,
|
| 151895 |
+
"step": 41290
|
| 151896 |
+
},
|
| 151897 |
+
{
|
| 151898 |
+
"epoch": 329.83,
|
| 151899 |
+
"learning_rate": 9.354503205128207e-06,
|
| 151900 |
+
"loss": 0.7921,
|
| 151901 |
+
"step": 41295
|
| 151902 |
+
},
|
| 151903 |
+
{
|
| 151904 |
+
"epoch": 329.87,
|
| 151905 |
+
"learning_rate": 9.354423076923078e-06,
|
| 151906 |
+
"loss": 0.3596,
|
| 151907 |
+
"step": 41300
|
| 151908 |
+
},
|
| 151909 |
+
{
|
| 151910 |
+
"epoch": 329.91,
|
| 151911 |
+
"learning_rate": 9.35434294871795e-06,
|
| 151912 |
+
"loss": 0.4186,
|
| 151913 |
+
"step": 41305
|
| 151914 |
+
},
|
| 151915 |
+
{
|
| 151916 |
+
"epoch": 329.95,
|
| 151917 |
+
"learning_rate": 9.354262820512822e-06,
|
| 151918 |
+
"loss": 0.4717,
|
| 151919 |
+
"step": 41310
|
| 151920 |
+
},
|
| 151921 |
+
{
|
| 151922 |
+
"epoch": 329.99,
|
| 151923 |
+
"learning_rate": 9.354182692307694e-06,
|
| 151924 |
+
"loss": 1.6913,
|
| 151925 |
+
"step": 41315
|
| 151926 |
+
},
|
| 151927 |
+
{
|
| 151928 |
+
"epoch": 330.0,
|
| 151929 |
+
"eval_loss": 0.4328790009021759,
|
| 151930 |
+
"eval_runtime": 39.648,
|
| 151931 |
+
"eval_samples_per_second": 21.212,
|
| 151932 |
+
"eval_steps_per_second": 0.681,
|
| 151933 |
+
"eval_wer": 0.20647272727272728,
|
| 151934 |
+
"step": 41316
|
| 151935 |
+
},
|
| 151936 |
+
{
|
| 151937 |
+
"epoch": 330.03,
|
| 151938 |
+
"learning_rate": 9.354102564102565e-06,
|
| 151939 |
+
"loss": 0.4291,
|
| 151940 |
+
"step": 41320
|
| 151941 |
+
},
|
| 151942 |
+
{
|
| 151943 |
+
"epoch": 330.07,
|
| 151944 |
+
"learning_rate": 9.354022435897436e-06,
|
| 151945 |
+
"loss": 0.3432,
|
| 151946 |
+
"step": 41325
|
| 151947 |
+
},
|
| 151948 |
+
{
|
| 151949 |
+
"epoch": 330.11,
|
| 151950 |
+
"learning_rate": 9.35394230769231e-06,
|
| 151951 |
+
"loss": 0.4129,
|
| 151952 |
+
"step": 41330
|
| 151953 |
+
},
|
| 151954 |
+
{
|
| 151955 |
+
"epoch": 330.15,
|
| 151956 |
+
"learning_rate": 9.353862179487179e-06,
|
| 151957 |
+
"loss": 0.4743,
|
| 151958 |
+
"step": 41335
|
| 151959 |
+
},
|
| 151960 |
+
{
|
| 151961 |
+
"epoch": 330.19,
|
| 151962 |
+
"learning_rate": 9.353782051282052e-06,
|
| 151963 |
+
"loss": 1.0624,
|
| 151964 |
+
"step": 41340
|
| 151965 |
+
},
|
| 151966 |
+
{
|
| 151967 |
+
"epoch": 330.23,
|
| 151968 |
+
"learning_rate": 9.353701923076925e-06,
|
| 151969 |
+
"loss": 0.9663,
|
| 151970 |
+
"step": 41345
|
| 151971 |
+
},
|
| 151972 |
+
{
|
| 151973 |
+
"epoch": 330.27,
|
| 151974 |
+
"learning_rate": 9.353621794871795e-06,
|
| 151975 |
+
"loss": 0.3923,
|
| 151976 |
+
"step": 41350
|
| 151977 |
+
},
|
| 151978 |
+
{
|
| 151979 |
+
"epoch": 330.31,
|
| 151980 |
+
"learning_rate": 9.353541666666668e-06,
|
| 151981 |
+
"loss": 0.3644,
|
| 151982 |
+
"step": 41355
|
| 151983 |
+
},
|
| 151984 |
+
{
|
| 151985 |
+
"epoch": 330.35,
|
| 151986 |
+
"learning_rate": 9.35346153846154e-06,
|
| 151987 |
+
"loss": 0.4644,
|
| 151988 |
+
"step": 41360
|
| 151989 |
+
},
|
| 151990 |
+
{
|
| 151991 |
+
"epoch": 330.39,
|
| 151992 |
+
"learning_rate": 9.35338141025641e-06,
|
| 151993 |
+
"loss": 1.1278,
|
| 151994 |
+
"step": 41365
|
| 151995 |
+
},
|
| 151996 |
+
{
|
| 151997 |
+
"epoch": 330.43,
|
| 151998 |
+
"learning_rate": 9.353301282051282e-06,
|
| 151999 |
+
"loss": 0.7478,
|
| 152000 |
+
"step": 41370
|
| 152001 |
+
},
|
| 152002 |
+
{
|
| 152003 |
+
"epoch": 330.47,
|
| 152004 |
+
"learning_rate": 9.353221153846155e-06,
|
| 152005 |
+
"loss": 0.3346,
|
| 152006 |
+
"step": 41375
|
| 152007 |
+
},
|
| 152008 |
+
{
|
| 152009 |
+
"epoch": 330.51,
|
| 152010 |
+
"learning_rate": 9.353141025641026e-06,
|
| 152011 |
+
"loss": 0.4023,
|
| 152012 |
+
"step": 41380
|
| 152013 |
+
},
|
| 152014 |
+
{
|
| 152015 |
+
"epoch": 330.55,
|
| 152016 |
+
"learning_rate": 9.353060897435898e-06,
|
| 152017 |
+
"loss": 0.4962,
|
| 152018 |
+
"step": 41385
|
| 152019 |
+
},
|
| 152020 |
+
{
|
| 152021 |
+
"epoch": 330.59,
|
| 152022 |
+
"learning_rate": 9.352980769230769e-06,
|
| 152023 |
+
"loss": 1.0702,
|
| 152024 |
+
"step": 41390
|
| 152025 |
+
},
|
| 152026 |
+
{
|
| 152027 |
+
"epoch": 330.63,
|
| 152028 |
+
"learning_rate": 9.352900641025642e-06,
|
| 152029 |
+
"loss": 0.6454,
|
| 152030 |
+
"step": 41395
|
| 152031 |
+
},
|
| 152032 |
+
{
|
| 152033 |
+
"epoch": 330.67,
|
| 152034 |
+
"learning_rate": 9.352820512820514e-06,
|
| 152035 |
+
"loss": 0.356,
|
| 152036 |
+
"step": 41400
|
| 152037 |
+
},
|
| 152038 |
+
{
|
| 152039 |
+
"epoch": 330.71,
|
| 152040 |
+
"learning_rate": 9.352740384615385e-06,
|
| 152041 |
+
"loss": 0.4445,
|
| 152042 |
+
"step": 41405
|
| 152043 |
+
},
|
| 152044 |
+
{
|
| 152045 |
+
"epoch": 330.75,
|
| 152046 |
+
"learning_rate": 9.352660256410258e-06,
|
| 152047 |
+
"loss": 0.5141,
|
| 152048 |
+
"step": 41410
|
| 152049 |
+
},
|
| 152050 |
+
{
|
| 152051 |
+
"epoch": 330.79,
|
| 152052 |
+
"learning_rate": 9.35258012820513e-06,
|
| 152053 |
+
"loss": 1.0236,
|
| 152054 |
+
"step": 41415
|
| 152055 |
+
},
|
| 152056 |
+
{
|
| 152057 |
+
"epoch": 330.83,
|
| 152058 |
+
"learning_rate": 9.3525e-06,
|
| 152059 |
+
"loss": 0.7786,
|
| 152060 |
+
"step": 41420
|
| 152061 |
+
},
|
| 152062 |
+
{
|
| 152063 |
+
"epoch": 330.87,
|
| 152064 |
+
"learning_rate": 9.352419871794872e-06,
|
| 152065 |
+
"loss": 0.3173,
|
| 152066 |
+
"step": 41425
|
| 152067 |
+
},
|
| 152068 |
+
{
|
| 152069 |
+
"epoch": 330.91,
|
| 152070 |
+
"learning_rate": 9.352339743589745e-06,
|
| 152071 |
+
"loss": 0.4374,
|
| 152072 |
+
"step": 41430
|
| 152073 |
+
},
|
| 152074 |
+
{
|
| 152075 |
+
"epoch": 330.95,
|
| 152076 |
+
"learning_rate": 9.352259615384616e-06,
|
| 152077 |
+
"loss": 0.487,
|
| 152078 |
+
"step": 41435
|
| 152079 |
+
},
|
| 152080 |
+
{
|
| 152081 |
+
"epoch": 330.99,
|
| 152082 |
+
"learning_rate": 9.352179487179488e-06,
|
| 152083 |
+
"loss": 1.052,
|
| 152084 |
+
"step": 41440
|
| 152085 |
+
},
|
| 152086 |
+
{
|
| 152087 |
+
"epoch": 331.0,
|
| 152088 |
+
"eval_loss": 0.42769986391067505,
|
| 152089 |
+
"eval_runtime": 36.4211,
|
| 152090 |
+
"eval_samples_per_second": 23.064,
|
| 152091 |
+
"eval_steps_per_second": 0.741,
|
| 152092 |
+
"eval_wer": 0.21098834569396094,
|
| 152093 |
+
"step": 41441
|
| 152094 |
}
|
| 152095 |
],
|
| 152096 |
"max_steps": 625000,
|
| 152097 |
"num_train_epochs": 5000,
|
| 152098 |
+
"total_flos": 1.1661123101288271e+20,
|
| 152099 |
"trial_name": null,
|
| 152100 |
"trial_params": null
|
| 152101 |
}
|
model-bin/finetune/base/{checkpoint-40819 → checkpoint-41441}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629693366.2436578/events.out.tfevents.1629693366.74272264b15c.932.43
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f492597dc62fc0dee306f57a9e867e567732259c7d9230ed14a82457749fbeaf
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629694012.331258/events.out.tfevents.1629694012.74272264b15c.932.45
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6e2d8e983308d47a0e62c264bc578d8f4bb3a1ccd76d8482add53b60d247660e
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629694653.2687066/events.out.tfevents.1629694653.74272264b15c.932.47
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:bcb347f3daf0f623265cbcd9bed860656f083c9a48d702432fcca891d0c5e0ed
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629695278.107289/events.out.tfevents.1629695278.74272264b15c.932.49
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:0d8e41dfbb7d3b598efe89fa107a489663c560fd7800025221ede7d420633742
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629695916.3520517/events.out.tfevents.1629695916.74272264b15c.932.51
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:302855ab0ec522e403a9eb2e40a7db85e89937d3cfbee26e1029af7b0e9f6f16
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629693366.74272264b15c.932.42
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c2410684b26cc9d18cceb84cabe032f289b0fb32af93f34c09dcbc29490cc1e8
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629694012.74272264b15c.932.44
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:3b91fc3d9b072d9e15ee496b8772ede845ba57c9155e2b3b546d5736fc2dcbf3
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629694653.74272264b15c.932.46
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:7ed283e2514d22e74c220529ce1a8a3e15d4e4bca178d6b4b8585420f07cb138
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629695278.74272264b15c.932.48
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:014eb791db4742b1f113e80372a65b25cd813ca85abe0bfa13981d7855c96ee6
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629695916.74272264b15c.932.50
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6c3b9c67660275cbd4efc3b650d5c6668a4adabdb9010018e2bcf3099c37bb68
|
| 3 |
+
size 8622
|