"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-128555 → checkpoint-129176}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-128555 → checkpoint-129176}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-128555 → checkpoint-129176}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-128555 → checkpoint-129176}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-128555 → checkpoint-129176}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-128555 → checkpoint-129176}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-128555 → checkpoint-129176}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-128555 → checkpoint-129176}/trainer_state.json +792 -3
- model-bin/finetune/base/{checkpoint-128555 → checkpoint-129176}/training_args.bin +0 -0
- model-bin/finetune/base/log/1630156731.366123/events.out.tfevents.1630156731.86bb0ddabf9b.4092.171 +3 -0
- model-bin/finetune/base/log/1630157126.4497585/events.out.tfevents.1630157126.86bb0ddabf9b.4092.173 +3 -0
- model-bin/finetune/base/log/1630157634.5855007/events.out.tfevents.1630157634.86bb0ddabf9b.4092.175 +3 -0
- model-bin/finetune/base/log/1630158028.9414186/events.out.tfevents.1630158028.86bb0ddabf9b.4092.177 +3 -0
- model-bin/finetune/base/log/1630158417.3059762/events.out.tfevents.1630158417.86bb0ddabf9b.4092.179 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630156731.86bb0ddabf9b.4092.170 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630157126.86bb0ddabf9b.4092.172 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630157634.86bb0ddabf9b.4092.174 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630158028.86bb0ddabf9b.4092.176 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630158417.86bb0ddabf9b.4092.178 +3 -0
model-bin/finetune/base/{checkpoint-128555 → checkpoint-129176}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-128555 → checkpoint-129176}/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:eef39fa6cedd0a5dad2389a755c5767119494ff9ec018785ba316a051c1a86ae
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-128555 → checkpoint-129176}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-128555 → checkpoint-129176}/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:ddcdb21d1013c71e1d3b8def1a1cb5abfaa9e9109f124556dd93ae4623b89ef1
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-128555 → checkpoint-129176}/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 sha256:7ed24719502ecd1398b4779f77a1e1e9a27b5683bb3fb054c18a56f93676f447
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-128555 → checkpoint-129176}/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 sha256:6c4e080eeb88caf6658eef6af97f13224fe0c49f64a0c60f4f4bdea046def223
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-128555 → checkpoint-129176}/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:d0101e701c636c0d4340257ed6ec38f2fd70334f2118159c88c1eb5f2070956e
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-128555 → checkpoint-129176}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.17162025681719809,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-124947",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -262929,11 +262929,800 @@
|
|
| 262929 |
"eval_steps_per_second": 0.748,
|
| 262930 |
"eval_wer": 0.17821493624772314,
|
| 262931 |
"step": 128555
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 262932 |
}
|
| 262933 |
],
|
| 262934 |
"max_steps": 620000,
|
| 262935 |
"num_train_epochs": 5000,
|
| 262936 |
-
"total_flos": 3.
|
| 262937 |
"trial_name": null,
|
| 262938 |
"trial_params": null
|
| 262939 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.17162025681719809,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-124947",
|
| 4 |
+
"epoch": 1040.995983935743,
|
| 5 |
+
"global_step": 129176,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 262929 |
"eval_steps_per_second": 0.748,
|
| 262930 |
"eval_wer": 0.17821493624772314,
|
| 262931 |
"step": 128555
|
| 262932 |
+
},
|
| 262933 |
+
{
|
| 262934 |
+
"epoch": 1036.04,
|
| 262935 |
+
"learning_rate": 7.956570512820513e-06,
|
| 262936 |
+
"loss": 0.3447,
|
| 262937 |
+
"step": 128560
|
| 262938 |
+
},
|
| 262939 |
+
{
|
| 262940 |
+
"epoch": 1036.08,
|
| 262941 |
+
"learning_rate": 7.956490384615384e-06,
|
| 262942 |
+
"loss": 0.32,
|
| 262943 |
+
"step": 128565
|
| 262944 |
+
},
|
| 262945 |
+
{
|
| 262946 |
+
"epoch": 1036.12,
|
| 262947 |
+
"learning_rate": 7.956410256410258e-06,
|
| 262948 |
+
"loss": 0.3023,
|
| 262949 |
+
"step": 128570
|
| 262950 |
+
},
|
| 262951 |
+
{
|
| 262952 |
+
"epoch": 1036.16,
|
| 262953 |
+
"learning_rate": 7.956330128205129e-06,
|
| 262954 |
+
"loss": 0.5107,
|
| 262955 |
+
"step": 128575
|
| 262956 |
+
},
|
| 262957 |
+
{
|
| 262958 |
+
"epoch": 1036.2,
|
| 262959 |
+
"learning_rate": 7.95625e-06,
|
| 262960 |
+
"loss": 1.1463,
|
| 262961 |
+
"step": 128580
|
| 262962 |
+
},
|
| 262963 |
+
{
|
| 262964 |
+
"epoch": 1036.24,
|
| 262965 |
+
"learning_rate": 7.956169871794872e-06,
|
| 262966 |
+
"loss": 0.2702,
|
| 262967 |
+
"step": 128585
|
| 262968 |
+
},
|
| 262969 |
+
{
|
| 262970 |
+
"epoch": 1036.28,
|
| 262971 |
+
"learning_rate": 7.956089743589745e-06,
|
| 262972 |
+
"loss": 0.2714,
|
| 262973 |
+
"step": 128590
|
| 262974 |
+
},
|
| 262975 |
+
{
|
| 262976 |
+
"epoch": 1036.32,
|
| 262977 |
+
"learning_rate": 7.956009615384616e-06,
|
| 262978 |
+
"loss": 0.3527,
|
| 262979 |
+
"step": 128595
|
| 262980 |
+
},
|
| 262981 |
+
{
|
| 262982 |
+
"epoch": 1036.36,
|
| 262983 |
+
"learning_rate": 7.955929487179487e-06,
|
| 262984 |
+
"loss": 0.4857,
|
| 262985 |
+
"step": 128600
|
| 262986 |
+
},
|
| 262987 |
+
{
|
| 262988 |
+
"epoch": 1036.4,
|
| 262989 |
+
"learning_rate": 7.95584935897436e-06,
|
| 262990 |
+
"loss": 1.304,
|
| 262991 |
+
"step": 128605
|
| 262992 |
+
},
|
| 262993 |
+
{
|
| 262994 |
+
"epoch": 1036.44,
|
| 262995 |
+
"learning_rate": 7.955769230769232e-06,
|
| 262996 |
+
"loss": 0.3013,
|
| 262997 |
+
"step": 128610
|
| 262998 |
+
},
|
| 262999 |
+
{
|
| 263000 |
+
"epoch": 1036.48,
|
| 263001 |
+
"learning_rate": 7.955689102564103e-06,
|
| 263002 |
+
"loss": 0.3293,
|
| 263003 |
+
"step": 128615
|
| 263004 |
+
},
|
| 263005 |
+
{
|
| 263006 |
+
"epoch": 1036.52,
|
| 263007 |
+
"learning_rate": 7.955608974358975e-06,
|
| 263008 |
+
"loss": 0.2689,
|
| 263009 |
+
"step": 128620
|
| 263010 |
+
},
|
| 263011 |
+
{
|
| 263012 |
+
"epoch": 1036.56,
|
| 263013 |
+
"learning_rate": 7.955528846153848e-06,
|
| 263014 |
+
"loss": 0.4438,
|
| 263015 |
+
"step": 128625
|
| 263016 |
+
},
|
| 263017 |
+
{
|
| 263018 |
+
"epoch": 1036.6,
|
| 263019 |
+
"learning_rate": 7.955448717948719e-06,
|
| 263020 |
+
"loss": 1.0367,
|
| 263021 |
+
"step": 128630
|
| 263022 |
+
},
|
| 263023 |
+
{
|
| 263024 |
+
"epoch": 1036.64,
|
| 263025 |
+
"learning_rate": 7.95536858974359e-06,
|
| 263026 |
+
"loss": 0.3354,
|
| 263027 |
+
"step": 128635
|
| 263028 |
+
},
|
| 263029 |
+
{
|
| 263030 |
+
"epoch": 1036.68,
|
| 263031 |
+
"learning_rate": 7.955288461538462e-06,
|
| 263032 |
+
"loss": 0.2889,
|
| 263033 |
+
"step": 128640
|
| 263034 |
+
},
|
| 263035 |
+
{
|
| 263036 |
+
"epoch": 1036.72,
|
| 263037 |
+
"learning_rate": 7.955208333333335e-06,
|
| 263038 |
+
"loss": 0.3395,
|
| 263039 |
+
"step": 128645
|
| 263040 |
+
},
|
| 263041 |
+
{
|
| 263042 |
+
"epoch": 1036.76,
|
| 263043 |
+
"learning_rate": 7.955128205128206e-06,
|
| 263044 |
+
"loss": 0.4001,
|
| 263045 |
+
"step": 128650
|
| 263046 |
+
},
|
| 263047 |
+
{
|
| 263048 |
+
"epoch": 1036.8,
|
| 263049 |
+
"learning_rate": 7.955048076923077e-06,
|
| 263050 |
+
"loss": 1.2136,
|
| 263051 |
+
"step": 128655
|
| 263052 |
+
},
|
| 263053 |
+
{
|
| 263054 |
+
"epoch": 1036.84,
|
| 263055 |
+
"learning_rate": 7.95496794871795e-06,
|
| 263056 |
+
"loss": 0.3309,
|
| 263057 |
+
"step": 128660
|
| 263058 |
+
},
|
| 263059 |
+
{
|
| 263060 |
+
"epoch": 1036.88,
|
| 263061 |
+
"learning_rate": 7.954887820512822e-06,
|
| 263062 |
+
"loss": 0.297,
|
| 263063 |
+
"step": 128665
|
| 263064 |
+
},
|
| 263065 |
+
{
|
| 263066 |
+
"epoch": 1036.92,
|
| 263067 |
+
"learning_rate": 7.954807692307693e-06,
|
| 263068 |
+
"loss": 0.3402,
|
| 263069 |
+
"step": 128670
|
| 263070 |
+
},
|
| 263071 |
+
{
|
| 263072 |
+
"epoch": 1036.96,
|
| 263073 |
+
"learning_rate": 7.954727564102565e-06,
|
| 263074 |
+
"loss": 0.4656,
|
| 263075 |
+
"step": 128675
|
| 263076 |
+
},
|
| 263077 |
+
{
|
| 263078 |
+
"epoch": 1037.0,
|
| 263079 |
+
"eval_loss": 0.48478376865386963,
|
| 263080 |
+
"eval_runtime": 36.371,
|
| 263081 |
+
"eval_samples_per_second": 23.123,
|
| 263082 |
+
"eval_steps_per_second": 0.742,
|
| 263083 |
+
"eval_wer": 0.1743723319586137,
|
| 263084 |
+
"step": 128679
|
| 263085 |
+
},
|
| 263086 |
+
{
|
| 263087 |
+
"epoch": 1037.01,
|
| 263088 |
+
"learning_rate": 7.954647435897438e-06,
|
| 263089 |
+
"loss": 0.3782,
|
| 263090 |
+
"step": 128680
|
| 263091 |
+
},
|
| 263092 |
+
{
|
| 263093 |
+
"epoch": 1037.05,
|
| 263094 |
+
"learning_rate": 7.954567307692307e-06,
|
| 263095 |
+
"loss": 0.2696,
|
| 263096 |
+
"step": 128685
|
| 263097 |
+
},
|
| 263098 |
+
{
|
| 263099 |
+
"epoch": 1037.09,
|
| 263100 |
+
"learning_rate": 7.95448717948718e-06,
|
| 263101 |
+
"loss": 0.288,
|
| 263102 |
+
"step": 128690
|
| 263103 |
+
},
|
| 263104 |
+
{
|
| 263105 |
+
"epoch": 1037.13,
|
| 263106 |
+
"learning_rate": 7.954407051282052e-06,
|
| 263107 |
+
"loss": 0.3534,
|
| 263108 |
+
"step": 128695
|
| 263109 |
+
},
|
| 263110 |
+
{
|
| 263111 |
+
"epoch": 1037.17,
|
| 263112 |
+
"learning_rate": 7.954326923076923e-06,
|
| 263113 |
+
"loss": 0.5256,
|
| 263114 |
+
"step": 128700
|
| 263115 |
+
},
|
| 263116 |
+
{
|
| 263117 |
+
"epoch": 1037.21,
|
| 263118 |
+
"learning_rate": 7.954246794871796e-06,
|
| 263119 |
+
"loss": 1.0363,
|
| 263120 |
+
"step": 128705
|
| 263121 |
+
},
|
| 263122 |
+
{
|
| 263123 |
+
"epoch": 1037.25,
|
| 263124 |
+
"learning_rate": 7.954166666666667e-06,
|
| 263125 |
+
"loss": 0.3092,
|
| 263126 |
+
"step": 128710
|
| 263127 |
+
},
|
| 263128 |
+
{
|
| 263129 |
+
"epoch": 1037.29,
|
| 263130 |
+
"learning_rate": 7.954086538461539e-06,
|
| 263131 |
+
"loss": 0.2433,
|
| 263132 |
+
"step": 128715
|
| 263133 |
+
},
|
| 263134 |
+
{
|
| 263135 |
+
"epoch": 1037.33,
|
| 263136 |
+
"learning_rate": 7.95400641025641e-06,
|
| 263137 |
+
"loss": 0.3564,
|
| 263138 |
+
"step": 128720
|
| 263139 |
+
},
|
| 263140 |
+
{
|
| 263141 |
+
"epoch": 1037.37,
|
| 263142 |
+
"learning_rate": 7.953926282051283e-06,
|
| 263143 |
+
"loss": 0.5547,
|
| 263144 |
+
"step": 128725
|
| 263145 |
+
},
|
| 263146 |
+
{
|
| 263147 |
+
"epoch": 1037.41,
|
| 263148 |
+
"learning_rate": 7.953846153846155e-06,
|
| 263149 |
+
"loss": 1.236,
|
| 263150 |
+
"step": 128730
|
| 263151 |
+
},
|
| 263152 |
+
{
|
| 263153 |
+
"epoch": 1037.45,
|
| 263154 |
+
"learning_rate": 7.953766025641026e-06,
|
| 263155 |
+
"loss": 0.2611,
|
| 263156 |
+
"step": 128735
|
| 263157 |
+
},
|
| 263158 |
+
{
|
| 263159 |
+
"epoch": 1037.49,
|
| 263160 |
+
"learning_rate": 7.953685897435897e-06,
|
| 263161 |
+
"loss": 0.2709,
|
| 263162 |
+
"step": 128740
|
| 263163 |
+
},
|
| 263164 |
+
{
|
| 263165 |
+
"epoch": 1037.53,
|
| 263166 |
+
"learning_rate": 7.95360576923077e-06,
|
| 263167 |
+
"loss": 0.2772,
|
| 263168 |
+
"step": 128745
|
| 263169 |
+
},
|
| 263170 |
+
{
|
| 263171 |
+
"epoch": 1037.57,
|
| 263172 |
+
"learning_rate": 7.953525641025642e-06,
|
| 263173 |
+
"loss": 0.6421,
|
| 263174 |
+
"step": 128750
|
| 263175 |
+
},
|
| 263176 |
+
{
|
| 263177 |
+
"epoch": 1037.61,
|
| 263178 |
+
"learning_rate": 7.953445512820513e-06,
|
| 263179 |
+
"loss": 1.0382,
|
| 263180 |
+
"step": 128755
|
| 263181 |
+
},
|
| 263182 |
+
{
|
| 263183 |
+
"epoch": 1037.65,
|
| 263184 |
+
"learning_rate": 7.953365384615386e-06,
|
| 263185 |
+
"loss": 0.3014,
|
| 263186 |
+
"step": 128760
|
| 263187 |
+
},
|
| 263188 |
+
{
|
| 263189 |
+
"epoch": 1037.69,
|
| 263190 |
+
"learning_rate": 7.953285256410257e-06,
|
| 263191 |
+
"loss": 0.2696,
|
| 263192 |
+
"step": 128765
|
| 263193 |
+
},
|
| 263194 |
+
{
|
| 263195 |
+
"epoch": 1037.73,
|
| 263196 |
+
"learning_rate": 7.953205128205129e-06,
|
| 263197 |
+
"loss": 0.3137,
|
| 263198 |
+
"step": 128770
|
| 263199 |
+
},
|
| 263200 |
+
{
|
| 263201 |
+
"epoch": 1037.77,
|
| 263202 |
+
"learning_rate": 7.953125e-06,
|
| 263203 |
+
"loss": 0.4803,
|
| 263204 |
+
"step": 128775
|
| 263205 |
+
},
|
| 263206 |
+
{
|
| 263207 |
+
"epoch": 1037.81,
|
| 263208 |
+
"learning_rate": 7.953044871794873e-06,
|
| 263209 |
+
"loss": 0.9273,
|
| 263210 |
+
"step": 128780
|
| 263211 |
+
},
|
| 263212 |
+
{
|
| 263213 |
+
"epoch": 1037.85,
|
| 263214 |
+
"learning_rate": 7.952964743589745e-06,
|
| 263215 |
+
"loss": 0.3498,
|
| 263216 |
+
"step": 128785
|
| 263217 |
+
},
|
| 263218 |
+
{
|
| 263219 |
+
"epoch": 1037.89,
|
| 263220 |
+
"learning_rate": 7.952884615384616e-06,
|
| 263221 |
+
"loss": 0.2619,
|
| 263222 |
+
"step": 128790
|
| 263223 |
+
},
|
| 263224 |
+
{
|
| 263225 |
+
"epoch": 1037.93,
|
| 263226 |
+
"learning_rate": 7.952804487179487e-06,
|
| 263227 |
+
"loss": 0.3481,
|
| 263228 |
+
"step": 128795
|
| 263229 |
+
},
|
| 263230 |
+
{
|
| 263231 |
+
"epoch": 1037.97,
|
| 263232 |
+
"learning_rate": 7.95272435897436e-06,
|
| 263233 |
+
"loss": 0.6731,
|
| 263234 |
+
"step": 128800
|
| 263235 |
+
},
|
| 263236 |
+
{
|
| 263237 |
+
"epoch": 1038.0,
|
| 263238 |
+
"eval_loss": 0.400423139333725,
|
| 263239 |
+
"eval_runtime": 52.5135,
|
| 263240 |
+
"eval_samples_per_second": 16.015,
|
| 263241 |
+
"eval_steps_per_second": 0.514,
|
| 263242 |
+
"eval_wer": 0.17735765779020157,
|
| 263243 |
+
"step": 128803
|
| 263244 |
+
},
|
| 263245 |
+
{
|
| 263246 |
+
"epoch": 1038.02,
|
| 263247 |
+
"learning_rate": 7.95264423076923e-06,
|
| 263248 |
+
"loss": 0.4173,
|
| 263249 |
+
"step": 128805
|
| 263250 |
+
},
|
| 263251 |
+
{
|
| 263252 |
+
"epoch": 1038.06,
|
| 263253 |
+
"learning_rate": 7.952564102564103e-06,
|
| 263254 |
+
"loss": 0.3741,
|
| 263255 |
+
"step": 128810
|
| 263256 |
+
},
|
| 263257 |
+
{
|
| 263258 |
+
"epoch": 1038.1,
|
| 263259 |
+
"learning_rate": 7.952483974358976e-06,
|
| 263260 |
+
"loss": 0.2887,
|
| 263261 |
+
"step": 128815
|
| 263262 |
+
},
|
| 263263 |
+
{
|
| 263264 |
+
"epoch": 1038.14,
|
| 263265 |
+
"learning_rate": 7.952403846153846e-06,
|
| 263266 |
+
"loss": 0.2953,
|
| 263267 |
+
"step": 128820
|
| 263268 |
+
},
|
| 263269 |
+
{
|
| 263270 |
+
"epoch": 1038.18,
|
| 263271 |
+
"learning_rate": 7.952323717948719e-06,
|
| 263272 |
+
"loss": 0.5716,
|
| 263273 |
+
"step": 128825
|
| 263274 |
+
},
|
| 263275 |
+
{
|
| 263276 |
+
"epoch": 1038.22,
|
| 263277 |
+
"learning_rate": 7.95224358974359e-06,
|
| 263278 |
+
"loss": 0.9941,
|
| 263279 |
+
"step": 128830
|
| 263280 |
+
},
|
| 263281 |
+
{
|
| 263282 |
+
"epoch": 1038.26,
|
| 263283 |
+
"learning_rate": 7.952163461538462e-06,
|
| 263284 |
+
"loss": 0.2912,
|
| 263285 |
+
"step": 128835
|
| 263286 |
+
},
|
| 263287 |
+
{
|
| 263288 |
+
"epoch": 1038.3,
|
| 263289 |
+
"learning_rate": 7.952083333333333e-06,
|
| 263290 |
+
"loss": 0.2764,
|
| 263291 |
+
"step": 128840
|
| 263292 |
+
},
|
| 263293 |
+
{
|
| 263294 |
+
"epoch": 1038.34,
|
| 263295 |
+
"learning_rate": 7.952003205128206e-06,
|
| 263296 |
+
"loss": 0.303,
|
| 263297 |
+
"step": 128845
|
| 263298 |
+
},
|
| 263299 |
+
{
|
| 263300 |
+
"epoch": 1038.38,
|
| 263301 |
+
"learning_rate": 7.951923076923077e-06,
|
| 263302 |
+
"loss": 0.5834,
|
| 263303 |
+
"step": 128850
|
| 263304 |
+
},
|
| 263305 |
+
{
|
| 263306 |
+
"epoch": 1038.42,
|
| 263307 |
+
"learning_rate": 7.951842948717949e-06,
|
| 263308 |
+
"loss": 0.9354,
|
| 263309 |
+
"step": 128855
|
| 263310 |
+
},
|
| 263311 |
+
{
|
| 263312 |
+
"epoch": 1038.46,
|
| 263313 |
+
"learning_rate": 7.951762820512822e-06,
|
| 263314 |
+
"loss": 0.2896,
|
| 263315 |
+
"step": 128860
|
| 263316 |
+
},
|
| 263317 |
+
{
|
| 263318 |
+
"epoch": 1038.5,
|
| 263319 |
+
"learning_rate": 7.951682692307693e-06,
|
| 263320 |
+
"loss": 0.277,
|
| 263321 |
+
"step": 128865
|
| 263322 |
+
},
|
| 263323 |
+
{
|
| 263324 |
+
"epoch": 1038.54,
|
| 263325 |
+
"learning_rate": 7.951602564102564e-06,
|
| 263326 |
+
"loss": 0.3579,
|
| 263327 |
+
"step": 128870
|
| 263328 |
+
},
|
| 263329 |
+
{
|
| 263330 |
+
"epoch": 1038.58,
|
| 263331 |
+
"learning_rate": 7.951522435897436e-06,
|
| 263332 |
+
"loss": 0.6319,
|
| 263333 |
+
"step": 128875
|
| 263334 |
+
},
|
| 263335 |
+
{
|
| 263336 |
+
"epoch": 1038.62,
|
| 263337 |
+
"learning_rate": 7.951442307692309e-06,
|
| 263338 |
+
"loss": 1.0096,
|
| 263339 |
+
"step": 128880
|
| 263340 |
+
},
|
| 263341 |
+
{
|
| 263342 |
+
"epoch": 1038.66,
|
| 263343 |
+
"learning_rate": 7.95136217948718e-06,
|
| 263344 |
+
"loss": 0.2903,
|
| 263345 |
+
"step": 128885
|
| 263346 |
+
},
|
| 263347 |
+
{
|
| 263348 |
+
"epoch": 1038.7,
|
| 263349 |
+
"learning_rate": 7.951282051282052e-06,
|
| 263350 |
+
"loss": 0.285,
|
| 263351 |
+
"step": 128890
|
| 263352 |
+
},
|
| 263353 |
+
{
|
| 263354 |
+
"epoch": 1038.74,
|
| 263355 |
+
"learning_rate": 7.951201923076923e-06,
|
| 263356 |
+
"loss": 0.3683,
|
| 263357 |
+
"step": 128895
|
| 263358 |
+
},
|
| 263359 |
+
{
|
| 263360 |
+
"epoch": 1038.78,
|
| 263361 |
+
"learning_rate": 7.951121794871796e-06,
|
| 263362 |
+
"loss": 0.599,
|
| 263363 |
+
"step": 128900
|
| 263364 |
+
},
|
| 263365 |
+
{
|
| 263366 |
+
"epoch": 1038.82,
|
| 263367 |
+
"learning_rate": 7.951041666666667e-06,
|
| 263368 |
+
"loss": 0.9766,
|
| 263369 |
+
"step": 128905
|
| 263370 |
+
},
|
| 263371 |
+
{
|
| 263372 |
+
"epoch": 1038.86,
|
| 263373 |
+
"learning_rate": 7.950961538461539e-06,
|
| 263374 |
+
"loss": 0.2988,
|
| 263375 |
+
"step": 128910
|
| 263376 |
+
},
|
| 263377 |
+
{
|
| 263378 |
+
"epoch": 1038.9,
|
| 263379 |
+
"learning_rate": 7.950881410256412e-06,
|
| 263380 |
+
"loss": 0.2757,
|
| 263381 |
+
"step": 128915
|
| 263382 |
+
},
|
| 263383 |
+
{
|
| 263384 |
+
"epoch": 1038.94,
|
| 263385 |
+
"learning_rate": 7.950801282051283e-06,
|
| 263386 |
+
"loss": 0.3093,
|
| 263387 |
+
"step": 128920
|
| 263388 |
+
},
|
| 263389 |
+
{
|
| 263390 |
+
"epoch": 1038.98,
|
| 263391 |
+
"learning_rate": 7.950721153846155e-06,
|
| 263392 |
+
"loss": 0.8108,
|
| 263393 |
+
"step": 128925
|
| 263394 |
+
},
|
| 263395 |
+
{
|
| 263396 |
+
"epoch": 1039.0,
|
| 263397 |
+
"eval_loss": 0.4010806381702423,
|
| 263398 |
+
"eval_runtime": 35.4704,
|
| 263399 |
+
"eval_samples_per_second": 23.541,
|
| 263400 |
+
"eval_steps_per_second": 0.761,
|
| 263401 |
+
"eval_wer": 0.17762962962962964,
|
| 263402 |
+
"step": 128927
|
| 263403 |
+
},
|
| 263404 |
+
{
|
| 263405 |
+
"epoch": 1031.02,
|
| 263406 |
+
"learning_rate": 7.950641025641026e-06,
|
| 263407 |
+
"loss": 0.3348,
|
| 263408 |
+
"step": 128930
|
| 263409 |
+
},
|
| 263410 |
+
{
|
| 263411 |
+
"epoch": 1031.06,
|
| 263412 |
+
"learning_rate": 7.950560897435899e-06,
|
| 263413 |
+
"loss": 0.2639,
|
| 263414 |
+
"step": 128935
|
| 263415 |
+
},
|
| 263416 |
+
{
|
| 263417 |
+
"epoch": 1031.1,
|
| 263418 |
+
"learning_rate": 7.95048076923077e-06,
|
| 263419 |
+
"loss": 0.2835,
|
| 263420 |
+
"step": 128940
|
| 263421 |
+
},
|
| 263422 |
+
{
|
| 263423 |
+
"epoch": 1031.14,
|
| 263424 |
+
"learning_rate": 7.950400641025642e-06,
|
| 263425 |
+
"loss": 0.3394,
|
| 263426 |
+
"step": 128945
|
| 263427 |
+
},
|
| 263428 |
+
{
|
| 263429 |
+
"epoch": 1031.18,
|
| 263430 |
+
"learning_rate": 7.950320512820513e-06,
|
| 263431 |
+
"loss": 0.828,
|
| 263432 |
+
"step": 128950
|
| 263433 |
+
},
|
| 263434 |
+
{
|
| 263435 |
+
"epoch": 1031.22,
|
| 263436 |
+
"learning_rate": 7.950240384615386e-06,
|
| 263437 |
+
"loss": 0.7796,
|
| 263438 |
+
"step": 128955
|
| 263439 |
+
},
|
| 263440 |
+
{
|
| 263441 |
+
"epoch": 1031.26,
|
| 263442 |
+
"learning_rate": 7.950160256410257e-06,
|
| 263443 |
+
"loss": 0.2815,
|
| 263444 |
+
"step": 128960
|
| 263445 |
+
},
|
| 263446 |
+
{
|
| 263447 |
+
"epoch": 1031.3,
|
| 263448 |
+
"learning_rate": 7.950080128205129e-06,
|
| 263449 |
+
"loss": 0.2814,
|
| 263450 |
+
"step": 128965
|
| 263451 |
+
},
|
| 263452 |
+
{
|
| 263453 |
+
"epoch": 1031.34,
|
| 263454 |
+
"learning_rate": 7.950000000000002e-06,
|
| 263455 |
+
"loss": 0.3541,
|
| 263456 |
+
"step": 128970
|
| 263457 |
+
},
|
| 263458 |
+
{
|
| 263459 |
+
"epoch": 1031.38,
|
| 263460 |
+
"learning_rate": 7.949919871794871e-06,
|
| 263461 |
+
"loss": 0.655,
|
| 263462 |
+
"step": 128975
|
| 263463 |
+
},
|
| 263464 |
+
{
|
| 263465 |
+
"epoch": 1031.42,
|
| 263466 |
+
"learning_rate": 7.949839743589745e-06,
|
| 263467 |
+
"loss": 0.8656,
|
| 263468 |
+
"step": 128980
|
| 263469 |
+
},
|
| 263470 |
+
{
|
| 263471 |
+
"epoch": 1031.46,
|
| 263472 |
+
"learning_rate": 7.949759615384616e-06,
|
| 263473 |
+
"loss": 0.2695,
|
| 263474 |
+
"step": 128985
|
| 263475 |
+
},
|
| 263476 |
+
{
|
| 263477 |
+
"epoch": 1031.5,
|
| 263478 |
+
"learning_rate": 7.949679487179487e-06,
|
| 263479 |
+
"loss": 0.3559,
|
| 263480 |
+
"step": 128990
|
| 263481 |
+
},
|
| 263482 |
+
{
|
| 263483 |
+
"epoch": 1031.54,
|
| 263484 |
+
"learning_rate": 7.949599358974359e-06,
|
| 263485 |
+
"loss": 0.3306,
|
| 263486 |
+
"step": 128995
|
| 263487 |
+
},
|
| 263488 |
+
{
|
| 263489 |
+
"epoch": 1031.58,
|
| 263490 |
+
"learning_rate": 7.949519230769232e-06,
|
| 263491 |
+
"loss": 0.7098,
|
| 263492 |
+
"step": 129000
|
| 263493 |
+
},
|
| 263494 |
+
{
|
| 263495 |
+
"epoch": 1031.62,
|
| 263496 |
+
"learning_rate": 7.949439102564103e-06,
|
| 263497 |
+
"loss": 0.8704,
|
| 263498 |
+
"step": 129005
|
| 263499 |
+
},
|
| 263500 |
+
{
|
| 263501 |
+
"epoch": 1031.66,
|
| 263502 |
+
"learning_rate": 7.949358974358974e-06,
|
| 263503 |
+
"loss": 0.2563,
|
| 263504 |
+
"step": 129010
|
| 263505 |
+
},
|
| 263506 |
+
{
|
| 263507 |
+
"epoch": 1031.7,
|
| 263508 |
+
"learning_rate": 7.949278846153847e-06,
|
| 263509 |
+
"loss": 0.2483,
|
| 263510 |
+
"step": 129015
|
| 263511 |
+
},
|
| 263512 |
+
{
|
| 263513 |
+
"epoch": 1031.74,
|
| 263514 |
+
"learning_rate": 7.949198717948719e-06,
|
| 263515 |
+
"loss": 0.4822,
|
| 263516 |
+
"step": 129020
|
| 263517 |
+
},
|
| 263518 |
+
{
|
| 263519 |
+
"epoch": 1031.78,
|
| 263520 |
+
"learning_rate": 7.94911858974359e-06,
|
| 263521 |
+
"loss": 0.7116,
|
| 263522 |
+
"step": 129025
|
| 263523 |
+
},
|
| 263524 |
+
{
|
| 263525 |
+
"epoch": 1031.82,
|
| 263526 |
+
"learning_rate": 7.949038461538462e-06,
|
| 263527 |
+
"loss": 0.7471,
|
| 263528 |
+
"step": 129030
|
| 263529 |
+
},
|
| 263530 |
+
{
|
| 263531 |
+
"epoch": 1031.86,
|
| 263532 |
+
"learning_rate": 7.948958333333335e-06,
|
| 263533 |
+
"loss": 0.2788,
|
| 263534 |
+
"step": 129035
|
| 263535 |
+
},
|
| 263536 |
+
{
|
| 263537 |
+
"epoch": 1031.9,
|
| 263538 |
+
"learning_rate": 7.948878205128206e-06,
|
| 263539 |
+
"loss": 0.2955,
|
| 263540 |
+
"step": 129040
|
| 263541 |
+
},
|
| 263542 |
+
{
|
| 263543 |
+
"epoch": 1031.94,
|
| 263544 |
+
"learning_rate": 7.948798076923077e-06,
|
| 263545 |
+
"loss": 0.4194,
|
| 263546 |
+
"step": 129045
|
| 263547 |
+
},
|
| 263548 |
+
{
|
| 263549 |
+
"epoch": 1031.98,
|
| 263550 |
+
"learning_rate": 7.948717948717949e-06,
|
| 263551 |
+
"loss": 0.7488,
|
| 263552 |
+
"step": 129050
|
| 263553 |
+
},
|
| 263554 |
+
{
|
| 263555 |
+
"epoch": 1032.0,
|
| 263556 |
+
"eval_loss": 0.3547717332839966,
|
| 263557 |
+
"eval_runtime": 34.5607,
|
| 263558 |
+
"eval_samples_per_second": 24.16,
|
| 263559 |
+
"eval_steps_per_second": 0.781,
|
| 263560 |
+
"eval_wer": 0.18015820211428993,
|
| 263561 |
+
"step": 129052
|
| 263562 |
+
},
|
| 263563 |
+
{
|
| 263564 |
+
"epoch": 1040.02,
|
| 263565 |
+
"learning_rate": 7.948637820512822e-06,
|
| 263566 |
+
"loss": 0.3579,
|
| 263567 |
+
"step": 129055
|
| 263568 |
+
},
|
| 263569 |
+
{
|
| 263570 |
+
"epoch": 1040.06,
|
| 263571 |
+
"learning_rate": 7.948557692307693e-06,
|
| 263572 |
+
"loss": 0.2747,
|
| 263573 |
+
"step": 129060
|
| 263574 |
+
},
|
| 263575 |
+
{
|
| 263576 |
+
"epoch": 1040.1,
|
| 263577 |
+
"learning_rate": 7.948477564102564e-06,
|
| 263578 |
+
"loss": 0.3033,
|
| 263579 |
+
"step": 129065
|
| 263580 |
+
},
|
| 263581 |
+
{
|
| 263582 |
+
"epoch": 1040.14,
|
| 263583 |
+
"learning_rate": 7.948397435897437e-06,
|
| 263584 |
+
"loss": 0.3235,
|
| 263585 |
+
"step": 129070
|
| 263586 |
+
},
|
| 263587 |
+
{
|
| 263588 |
+
"epoch": 1040.18,
|
| 263589 |
+
"learning_rate": 7.948317307692309e-06,
|
| 263590 |
+
"loss": 0.7648,
|
| 263591 |
+
"step": 129075
|
| 263592 |
+
},
|
| 263593 |
+
{
|
| 263594 |
+
"epoch": 1040.22,
|
| 263595 |
+
"learning_rate": 7.94823717948718e-06,
|
| 263596 |
+
"loss": 0.7604,
|
| 263597 |
+
"step": 129080
|
| 263598 |
+
},
|
| 263599 |
+
{
|
| 263600 |
+
"epoch": 1040.27,
|
| 263601 |
+
"learning_rate": 7.948157051282052e-06,
|
| 263602 |
+
"loss": 0.2639,
|
| 263603 |
+
"step": 129085
|
| 263604 |
+
},
|
| 263605 |
+
{
|
| 263606 |
+
"epoch": 1040.31,
|
| 263607 |
+
"learning_rate": 7.948076923076925e-06,
|
| 263608 |
+
"loss": 0.2748,
|
| 263609 |
+
"step": 129090
|
| 263610 |
+
},
|
| 263611 |
+
{
|
| 263612 |
+
"epoch": 1040.35,
|
| 263613 |
+
"learning_rate": 7.947996794871796e-06,
|
| 263614 |
+
"loss": 0.3545,
|
| 263615 |
+
"step": 129095
|
| 263616 |
+
},
|
| 263617 |
+
{
|
| 263618 |
+
"epoch": 1040.39,
|
| 263619 |
+
"learning_rate": 7.947916666666667e-06,
|
| 263620 |
+
"loss": 0.6764,
|
| 263621 |
+
"step": 129100
|
| 263622 |
+
},
|
| 263623 |
+
{
|
| 263624 |
+
"epoch": 1040.43,
|
| 263625 |
+
"learning_rate": 7.947836538461539e-06,
|
| 263626 |
+
"loss": 0.8452,
|
| 263627 |
+
"step": 129105
|
| 263628 |
+
},
|
| 263629 |
+
{
|
| 263630 |
+
"epoch": 1040.47,
|
| 263631 |
+
"learning_rate": 7.947756410256412e-06,
|
| 263632 |
+
"loss": 0.2781,
|
| 263633 |
+
"step": 129110
|
| 263634 |
+
},
|
| 263635 |
+
{
|
| 263636 |
+
"epoch": 1040.51,
|
| 263637 |
+
"learning_rate": 7.947676282051283e-06,
|
| 263638 |
+
"loss": 0.3401,
|
| 263639 |
+
"step": 129115
|
| 263640 |
+
},
|
| 263641 |
+
{
|
| 263642 |
+
"epoch": 1040.55,
|
| 263643 |
+
"learning_rate": 7.947596153846154e-06,
|
| 263644 |
+
"loss": 0.349,
|
| 263645 |
+
"step": 129120
|
| 263646 |
+
},
|
| 263647 |
+
{
|
| 263648 |
+
"epoch": 1040.59,
|
| 263649 |
+
"learning_rate": 7.947516025641027e-06,
|
| 263650 |
+
"loss": 0.7103,
|
| 263651 |
+
"step": 129125
|
| 263652 |
+
},
|
| 263653 |
+
{
|
| 263654 |
+
"epoch": 1040.63,
|
| 263655 |
+
"learning_rate": 7.947435897435897e-06,
|
| 263656 |
+
"loss": 0.878,
|
| 263657 |
+
"step": 129130
|
| 263658 |
+
},
|
| 263659 |
+
{
|
| 263660 |
+
"epoch": 1040.67,
|
| 263661 |
+
"learning_rate": 7.94735576923077e-06,
|
| 263662 |
+
"loss": 0.2839,
|
| 263663 |
+
"step": 129135
|
| 263664 |
+
},
|
| 263665 |
+
{
|
| 263666 |
+
"epoch": 1040.71,
|
| 263667 |
+
"learning_rate": 7.947275641025642e-06,
|
| 263668 |
+
"loss": 0.2753,
|
| 263669 |
+
"step": 129140
|
| 263670 |
+
},
|
| 263671 |
+
{
|
| 263672 |
+
"epoch": 1040.75,
|
| 263673 |
+
"learning_rate": 7.947195512820513e-06,
|
| 263674 |
+
"loss": 0.3219,
|
| 263675 |
+
"step": 129145
|
| 263676 |
+
},
|
| 263677 |
+
{
|
| 263678 |
+
"epoch": 1040.79,
|
| 263679 |
+
"learning_rate": 7.947115384615384e-06,
|
| 263680 |
+
"loss": 0.6957,
|
| 263681 |
+
"step": 129150
|
| 263682 |
+
},
|
| 263683 |
+
{
|
| 263684 |
+
"epoch": 1040.83,
|
| 263685 |
+
"learning_rate": 7.947035256410257e-06,
|
| 263686 |
+
"loss": 0.8732,
|
| 263687 |
+
"step": 129155
|
| 263688 |
+
},
|
| 263689 |
+
{
|
| 263690 |
+
"epoch": 1040.87,
|
| 263691 |
+
"learning_rate": 7.946955128205129e-06,
|
| 263692 |
+
"loss": 0.3013,
|
| 263693 |
+
"step": 129160
|
| 263694 |
+
},
|
| 263695 |
+
{
|
| 263696 |
+
"epoch": 1040.91,
|
| 263697 |
+
"learning_rate": 7.946875e-06,
|
| 263698 |
+
"loss": 0.3378,
|
| 263699 |
+
"step": 129165
|
| 263700 |
+
},
|
| 263701 |
+
{
|
| 263702 |
+
"epoch": 1040.95,
|
| 263703 |
+
"learning_rate": 7.946794871794873e-06,
|
| 263704 |
+
"loss": 0.2846,
|
| 263705 |
+
"step": 129170
|
| 263706 |
+
},
|
| 263707 |
+
{
|
| 263708 |
+
"epoch": 1040.99,
|
| 263709 |
+
"learning_rate": 7.946714743589744e-06,
|
| 263710 |
+
"loss": 0.8823,
|
| 263711 |
+
"step": 129175
|
| 263712 |
+
},
|
| 263713 |
+
{
|
| 263714 |
+
"epoch": 1041.0,
|
| 263715 |
+
"eval_loss": 0.422131210565567,
|
| 263716 |
+
"eval_runtime": 37.4615,
|
| 263717 |
+
"eval_samples_per_second": 22.316,
|
| 263718 |
+
"eval_steps_per_second": 0.721,
|
| 263719 |
+
"eval_wer": 0.1812391430225825,
|
| 263720 |
+
"step": 129176
|
| 263721 |
}
|
| 263722 |
],
|
| 263723 |
"max_steps": 620000,
|
| 263724 |
"num_train_epochs": 5000,
|
| 263725 |
+
"total_flos": 3.635111378931759e+20,
|
| 263726 |
"trial_name": null,
|
| 263727 |
"trial_params": null
|
| 263728 |
}
|
model-bin/finetune/base/{checkpoint-128555 → checkpoint-129176}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1630156731.366123/events.out.tfevents.1630156731.86bb0ddabf9b.4092.171
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d4af5900df9ed7f98f73e0ec00a461bd0d31ffe0b5577ecc6226e09f3bd6e658
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630157126.4497585/events.out.tfevents.1630157126.86bb0ddabf9b.4092.173
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9f0394a98fbe8f5a95dbaecdd6d8261f3108dc827a039649f6f23a8e428949f8
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630157634.5855007/events.out.tfevents.1630157634.86bb0ddabf9b.4092.175
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:76ea3e996dd72e449d4ca1b3f41d18851b2919b780613db232220f7d28c16a7c
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630158028.9414186/events.out.tfevents.1630158028.86bb0ddabf9b.4092.177
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6accbe4c2c163d73b2fb91cd2ed615b924c7dde91fcb6f2d0572b744c0f60a71
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630158417.3059762/events.out.tfevents.1630158417.86bb0ddabf9b.4092.179
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:496451c4008ba073cb1edd55e4e10300cfcd64e05bcac53af544bab1e22b667d
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1630156731.86bb0ddabf9b.4092.170
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:0f56498a951638a94985c01993ff4381e1566d350e8f2fd63e00786758835eb3
|
| 3 |
+
size 8462
|
model-bin/finetune/base/log/events.out.tfevents.1630157126.86bb0ddabf9b.4092.172
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:222b4eb05396fd71a23bb97b2207da10b7f89e51a20f0042016909391d798059
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630157634.86bb0ddabf9b.4092.174
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:daffb5f7a01d23c66a017b02bb98ce072d55ef1ebaef9f013d754252809ab5b4
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630158028.86bb0ddabf9b.4092.176
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:12a06b3cd1ff2f4bcd262af367b4609c42804a80d68f968448a252e8a2f089c4
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630158417.86bb0ddabf9b.4092.178
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:7a851e79ff7457da6c2dab27e9a89727e148bf719512dfb727990413641b7792
|
| 3 |
+
size 8622
|