Training in progress, step 3000
Browse files- last-checkpoint/optimizer.pt +1 -1
- last-checkpoint/pytorch_model.bin +1 -1
- last-checkpoint/rng_state.pth +1 -1
- last-checkpoint/scheduler.pt +1 -1
- last-checkpoint/trainer_state.json +3018 -5
- pytorch_model.bin +1 -1
last-checkpoint/optimizer.pt
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 4736616809
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f6de8b06ed44d4963b25bb376812b173405edc863a23e8867e0f736341d74a8d
|
| 3 |
size 4736616809
|
last-checkpoint/pytorch_model.bin
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 2368281769
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:02c85713dad1dcfeb5c32c03d2b6f448d20d550a87402f9d21be276100bce607
|
| 3 |
size 2368281769
|
last-checkpoint/rng_state.pth
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 14575
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:0a2eb101bb6b101af129672a2bea24581932e36cd77fc392a7cda646daab49ff
|
| 3 |
size 14575
|
last-checkpoint/scheduler.pt
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 627
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:cabb275feefad232b9c910702270f8d2a4ae8df759c0f8aeaf1aba436940e944
|
| 3 |
size 627
|
last-checkpoint/trainer_state.json
CHANGED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
-
"best_metric": 2.
|
| 3 |
-
"best_model_checkpoint": "output/checkpoint-
|
| 4 |
-
"epoch": 0.
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -15071,11 +15071,3024 @@
|
|
| 15071 |
"eval_samples_per_second": 1.188,
|
| 15072 |
"eval_steps_per_second": 0.166,
|
| 15073 |
"step": 2500
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 15074 |
}
|
| 15075 |
],
|
| 15076 |
"max_steps": 6000,
|
| 15077 |
"num_train_epochs": 9223372036854775807,
|
| 15078 |
-
"total_flos": 2.
|
| 15079 |
"trial_name": null,
|
| 15080 |
"trial_params": null
|
| 15081 |
}
|
|
|
|
| 1 |
{
|
| 2 |
+
"best_metric": 2.0400795936584473,
|
| 3 |
+
"best_model_checkpoint": "output/checkpoint-3000",
|
| 4 |
+
"epoch": 0.5,
|
| 5 |
+
"global_step": 3000,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 15071 |
"eval_samples_per_second": 1.188,
|
| 15072 |
"eval_steps_per_second": 0.166,
|
| 15073 |
"step": 2500
|
| 15074 |
+
},
|
| 15075 |
+
{
|
| 15076 |
+
"epoch": 0.42,
|
| 15077 |
+
"learning_rate": 0.00047444067796610173,
|
| 15078 |
+
"loss": 2.207,
|
| 15079 |
+
"step": 2501
|
| 15080 |
+
},
|
| 15081 |
+
{
|
| 15082 |
+
"epoch": 0.42,
|
| 15083 |
+
"learning_rate": 0.0004743050847457627,
|
| 15084 |
+
"loss": 2.1369,
|
| 15085 |
+
"step": 2502
|
| 15086 |
+
},
|
| 15087 |
+
{
|
| 15088 |
+
"epoch": 0.42,
|
| 15089 |
+
"learning_rate": 0.0004741694915254238,
|
| 15090 |
+
"loss": 2.1234,
|
| 15091 |
+
"step": 2503
|
| 15092 |
+
},
|
| 15093 |
+
{
|
| 15094 |
+
"epoch": 0.42,
|
| 15095 |
+
"learning_rate": 0.00047403389830508473,
|
| 15096 |
+
"loss": 2.1835,
|
| 15097 |
+
"step": 2504
|
| 15098 |
+
},
|
| 15099 |
+
{
|
| 15100 |
+
"epoch": 0.42,
|
| 15101 |
+
"learning_rate": 0.0004738983050847458,
|
| 15102 |
+
"loss": 2.1962,
|
| 15103 |
+
"step": 2505
|
| 15104 |
+
},
|
| 15105 |
+
{
|
| 15106 |
+
"epoch": 0.42,
|
| 15107 |
+
"learning_rate": 0.0004737627118644068,
|
| 15108 |
+
"loss": 2.1814,
|
| 15109 |
+
"step": 2506
|
| 15110 |
+
},
|
| 15111 |
+
{
|
| 15112 |
+
"epoch": 0.42,
|
| 15113 |
+
"learning_rate": 0.00047362711864406783,
|
| 15114 |
+
"loss": 2.1553,
|
| 15115 |
+
"step": 2507
|
| 15116 |
+
},
|
| 15117 |
+
{
|
| 15118 |
+
"epoch": 0.42,
|
| 15119 |
+
"learning_rate": 0.0004734915254237288,
|
| 15120 |
+
"loss": 2.1694,
|
| 15121 |
+
"step": 2508
|
| 15122 |
+
},
|
| 15123 |
+
{
|
| 15124 |
+
"epoch": 0.42,
|
| 15125 |
+
"learning_rate": 0.0004733559322033899,
|
| 15126 |
+
"loss": 2.2564,
|
| 15127 |
+
"step": 2509
|
| 15128 |
+
},
|
| 15129 |
+
{
|
| 15130 |
+
"epoch": 0.42,
|
| 15131 |
+
"learning_rate": 0.0004732203389830509,
|
| 15132 |
+
"loss": 2.1611,
|
| 15133 |
+
"step": 2510
|
| 15134 |
+
},
|
| 15135 |
+
{
|
| 15136 |
+
"epoch": 0.42,
|
| 15137 |
+
"learning_rate": 0.00047308474576271186,
|
| 15138 |
+
"loss": 2.3156,
|
| 15139 |
+
"step": 2511
|
| 15140 |
+
},
|
| 15141 |
+
{
|
| 15142 |
+
"epoch": 0.42,
|
| 15143 |
+
"learning_rate": 0.0004729491525423729,
|
| 15144 |
+
"loss": 2.1023,
|
| 15145 |
+
"step": 2512
|
| 15146 |
+
},
|
| 15147 |
+
{
|
| 15148 |
+
"epoch": 0.42,
|
| 15149 |
+
"learning_rate": 0.0004728135593220339,
|
| 15150 |
+
"loss": 2.2861,
|
| 15151 |
+
"step": 2513
|
| 15152 |
+
},
|
| 15153 |
+
{
|
| 15154 |
+
"epoch": 0.42,
|
| 15155 |
+
"learning_rate": 0.00047267796610169497,
|
| 15156 |
+
"loss": 2.1733,
|
| 15157 |
+
"step": 2514
|
| 15158 |
+
},
|
| 15159 |
+
{
|
| 15160 |
+
"epoch": 0.42,
|
| 15161 |
+
"learning_rate": 0.0004725423728813559,
|
| 15162 |
+
"loss": 2.1,
|
| 15163 |
+
"step": 2515
|
| 15164 |
+
},
|
| 15165 |
+
{
|
| 15166 |
+
"epoch": 0.42,
|
| 15167 |
+
"learning_rate": 0.000472406779661017,
|
| 15168 |
+
"loss": 2.1805,
|
| 15169 |
+
"step": 2516
|
| 15170 |
+
},
|
| 15171 |
+
{
|
| 15172 |
+
"epoch": 0.42,
|
| 15173 |
+
"learning_rate": 0.00047227118644067796,
|
| 15174 |
+
"loss": 2.1485,
|
| 15175 |
+
"step": 2517
|
| 15176 |
+
},
|
| 15177 |
+
{
|
| 15178 |
+
"epoch": 0.42,
|
| 15179 |
+
"learning_rate": 0.000472135593220339,
|
| 15180 |
+
"loss": 2.2094,
|
| 15181 |
+
"step": 2518
|
| 15182 |
+
},
|
| 15183 |
+
{
|
| 15184 |
+
"epoch": 0.42,
|
| 15185 |
+
"learning_rate": 0.000472,
|
| 15186 |
+
"loss": 2.1898,
|
| 15187 |
+
"step": 2519
|
| 15188 |
+
},
|
| 15189 |
+
{
|
| 15190 |
+
"epoch": 0.42,
|
| 15191 |
+
"learning_rate": 0.00047186440677966107,
|
| 15192 |
+
"loss": 2.2167,
|
| 15193 |
+
"step": 2520
|
| 15194 |
+
},
|
| 15195 |
+
{
|
| 15196 |
+
"epoch": 0.42,
|
| 15197 |
+
"learning_rate": 0.00047172881355932205,
|
| 15198 |
+
"loss": 2.1117,
|
| 15199 |
+
"step": 2521
|
| 15200 |
+
},
|
| 15201 |
+
{
|
| 15202 |
+
"epoch": 0.42,
|
| 15203 |
+
"learning_rate": 0.0004715932203389831,
|
| 15204 |
+
"loss": 2.1604,
|
| 15205 |
+
"step": 2522
|
| 15206 |
+
},
|
| 15207 |
+
{
|
| 15208 |
+
"epoch": 0.42,
|
| 15209 |
+
"learning_rate": 0.00047145762711864407,
|
| 15210 |
+
"loss": 2.1861,
|
| 15211 |
+
"step": 2523
|
| 15212 |
+
},
|
| 15213 |
+
{
|
| 15214 |
+
"epoch": 0.42,
|
| 15215 |
+
"learning_rate": 0.00047132203389830515,
|
| 15216 |
+
"loss": 2.1883,
|
| 15217 |
+
"step": 2524
|
| 15218 |
+
},
|
| 15219 |
+
{
|
| 15220 |
+
"epoch": 0.42,
|
| 15221 |
+
"learning_rate": 0.00047118644067796613,
|
| 15222 |
+
"loss": 2.1772,
|
| 15223 |
+
"step": 2525
|
| 15224 |
+
},
|
| 15225 |
+
{
|
| 15226 |
+
"epoch": 0.42,
|
| 15227 |
+
"learning_rate": 0.00047105084745762717,
|
| 15228 |
+
"loss": 2.1937,
|
| 15229 |
+
"step": 2526
|
| 15230 |
+
},
|
| 15231 |
+
{
|
| 15232 |
+
"epoch": 0.42,
|
| 15233 |
+
"learning_rate": 0.00047091525423728815,
|
| 15234 |
+
"loss": 2.2126,
|
| 15235 |
+
"step": 2527
|
| 15236 |
+
},
|
| 15237 |
+
{
|
| 15238 |
+
"epoch": 0.42,
|
| 15239 |
+
"learning_rate": 0.00047077966101694924,
|
| 15240 |
+
"loss": 2.2554,
|
| 15241 |
+
"step": 2528
|
| 15242 |
+
},
|
| 15243 |
+
{
|
| 15244 |
+
"epoch": 0.42,
|
| 15245 |
+
"learning_rate": 0.00047064406779661017,
|
| 15246 |
+
"loss": 2.1951,
|
| 15247 |
+
"step": 2529
|
| 15248 |
+
},
|
| 15249 |
+
{
|
| 15250 |
+
"epoch": 0.42,
|
| 15251 |
+
"learning_rate": 0.00047050847457627125,
|
| 15252 |
+
"loss": 2.1638,
|
| 15253 |
+
"step": 2530
|
| 15254 |
+
},
|
| 15255 |
+
{
|
| 15256 |
+
"epoch": 0.42,
|
| 15257 |
+
"learning_rate": 0.00047037288135593224,
|
| 15258 |
+
"loss": 2.1608,
|
| 15259 |
+
"step": 2531
|
| 15260 |
+
},
|
| 15261 |
+
{
|
| 15262 |
+
"epoch": 0.42,
|
| 15263 |
+
"learning_rate": 0.0004702372881355932,
|
| 15264 |
+
"loss": 2.092,
|
| 15265 |
+
"step": 2532
|
| 15266 |
+
},
|
| 15267 |
+
{
|
| 15268 |
+
"epoch": 0.42,
|
| 15269 |
+
"learning_rate": 0.00047010169491525425,
|
| 15270 |
+
"loss": 2.1712,
|
| 15271 |
+
"step": 2533
|
| 15272 |
+
},
|
| 15273 |
+
{
|
| 15274 |
+
"epoch": 0.42,
|
| 15275 |
+
"learning_rate": 0.00046996610169491523,
|
| 15276 |
+
"loss": 2.2195,
|
| 15277 |
+
"step": 2534
|
| 15278 |
+
},
|
| 15279 |
+
{
|
| 15280 |
+
"epoch": 0.42,
|
| 15281 |
+
"learning_rate": 0.0004698305084745763,
|
| 15282 |
+
"loss": 2.1822,
|
| 15283 |
+
"step": 2535
|
| 15284 |
+
},
|
| 15285 |
+
{
|
| 15286 |
+
"epoch": 0.42,
|
| 15287 |
+
"learning_rate": 0.0004696949152542373,
|
| 15288 |
+
"loss": 2.1229,
|
| 15289 |
+
"step": 2536
|
| 15290 |
+
},
|
| 15291 |
+
{
|
| 15292 |
+
"epoch": 0.42,
|
| 15293 |
+
"learning_rate": 0.00046955932203389834,
|
| 15294 |
+
"loss": 2.191,
|
| 15295 |
+
"step": 2537
|
| 15296 |
+
},
|
| 15297 |
+
{
|
| 15298 |
+
"epoch": 0.42,
|
| 15299 |
+
"learning_rate": 0.0004694237288135593,
|
| 15300 |
+
"loss": 2.196,
|
| 15301 |
+
"step": 2538
|
| 15302 |
+
},
|
| 15303 |
+
{
|
| 15304 |
+
"epoch": 0.42,
|
| 15305 |
+
"learning_rate": 0.0004692881355932204,
|
| 15306 |
+
"loss": 2.2301,
|
| 15307 |
+
"step": 2539
|
| 15308 |
+
},
|
| 15309 |
+
{
|
| 15310 |
+
"epoch": 0.42,
|
| 15311 |
+
"learning_rate": 0.00046915254237288133,
|
| 15312 |
+
"loss": 2.1972,
|
| 15313 |
+
"step": 2540
|
| 15314 |
+
},
|
| 15315 |
+
{
|
| 15316 |
+
"epoch": 0.42,
|
| 15317 |
+
"learning_rate": 0.0004690169491525424,
|
| 15318 |
+
"loss": 2.2244,
|
| 15319 |
+
"step": 2541
|
| 15320 |
+
},
|
| 15321 |
+
{
|
| 15322 |
+
"epoch": 0.42,
|
| 15323 |
+
"learning_rate": 0.0004688813559322034,
|
| 15324 |
+
"loss": 2.1232,
|
| 15325 |
+
"step": 2542
|
| 15326 |
+
},
|
| 15327 |
+
{
|
| 15328 |
+
"epoch": 0.42,
|
| 15329 |
+
"learning_rate": 0.00046874576271186444,
|
| 15330 |
+
"loss": 2.2249,
|
| 15331 |
+
"step": 2543
|
| 15332 |
+
},
|
| 15333 |
+
{
|
| 15334 |
+
"epoch": 0.42,
|
| 15335 |
+
"learning_rate": 0.0004686101694915254,
|
| 15336 |
+
"loss": 2.1743,
|
| 15337 |
+
"step": 2544
|
| 15338 |
+
},
|
| 15339 |
+
{
|
| 15340 |
+
"epoch": 0.42,
|
| 15341 |
+
"learning_rate": 0.0004684745762711865,
|
| 15342 |
+
"loss": 2.186,
|
| 15343 |
+
"step": 2545
|
| 15344 |
+
},
|
| 15345 |
+
{
|
| 15346 |
+
"epoch": 0.42,
|
| 15347 |
+
"learning_rate": 0.0004683389830508475,
|
| 15348 |
+
"loss": 2.1699,
|
| 15349 |
+
"step": 2546
|
| 15350 |
+
},
|
| 15351 |
+
{
|
| 15352 |
+
"epoch": 0.42,
|
| 15353 |
+
"learning_rate": 0.0004682033898305085,
|
| 15354 |
+
"loss": 2.188,
|
| 15355 |
+
"step": 2547
|
| 15356 |
+
},
|
| 15357 |
+
{
|
| 15358 |
+
"epoch": 0.42,
|
| 15359 |
+
"learning_rate": 0.0004680677966101695,
|
| 15360 |
+
"loss": 2.1757,
|
| 15361 |
+
"step": 2548
|
| 15362 |
+
},
|
| 15363 |
+
{
|
| 15364 |
+
"epoch": 0.42,
|
| 15365 |
+
"learning_rate": 0.0004679322033898306,
|
| 15366 |
+
"loss": 2.1531,
|
| 15367 |
+
"step": 2549
|
| 15368 |
+
},
|
| 15369 |
+
{
|
| 15370 |
+
"epoch": 0.42,
|
| 15371 |
+
"learning_rate": 0.00046779661016949157,
|
| 15372 |
+
"loss": 2.2214,
|
| 15373 |
+
"step": 2550
|
| 15374 |
+
},
|
| 15375 |
+
{
|
| 15376 |
+
"epoch": 0.43,
|
| 15377 |
+
"learning_rate": 0.0004676610169491526,
|
| 15378 |
+
"loss": 2.1826,
|
| 15379 |
+
"step": 2551
|
| 15380 |
+
},
|
| 15381 |
+
{
|
| 15382 |
+
"epoch": 0.43,
|
| 15383 |
+
"learning_rate": 0.0004675254237288136,
|
| 15384 |
+
"loss": 2.2332,
|
| 15385 |
+
"step": 2552
|
| 15386 |
+
},
|
| 15387 |
+
{
|
| 15388 |
+
"epoch": 0.43,
|
| 15389 |
+
"learning_rate": 0.00046738983050847457,
|
| 15390 |
+
"loss": 2.1946,
|
| 15391 |
+
"step": 2553
|
| 15392 |
+
},
|
| 15393 |
+
{
|
| 15394 |
+
"epoch": 0.43,
|
| 15395 |
+
"learning_rate": 0.0004672542372881356,
|
| 15396 |
+
"loss": 2.1449,
|
| 15397 |
+
"step": 2554
|
| 15398 |
+
},
|
| 15399 |
+
{
|
| 15400 |
+
"epoch": 0.43,
|
| 15401 |
+
"learning_rate": 0.0004671186440677966,
|
| 15402 |
+
"loss": 2.1659,
|
| 15403 |
+
"step": 2555
|
| 15404 |
+
},
|
| 15405 |
+
{
|
| 15406 |
+
"epoch": 0.43,
|
| 15407 |
+
"learning_rate": 0.00046698305084745767,
|
| 15408 |
+
"loss": 2.2544,
|
| 15409 |
+
"step": 2556
|
| 15410 |
+
},
|
| 15411 |
+
{
|
| 15412 |
+
"epoch": 0.43,
|
| 15413 |
+
"learning_rate": 0.00046684745762711865,
|
| 15414 |
+
"loss": 2.1633,
|
| 15415 |
+
"step": 2557
|
| 15416 |
+
},
|
| 15417 |
+
{
|
| 15418 |
+
"epoch": 0.43,
|
| 15419 |
+
"learning_rate": 0.0004667118644067797,
|
| 15420 |
+
"loss": 2.169,
|
| 15421 |
+
"step": 2558
|
| 15422 |
+
},
|
| 15423 |
+
{
|
| 15424 |
+
"epoch": 0.43,
|
| 15425 |
+
"learning_rate": 0.00046657627118644067,
|
| 15426 |
+
"loss": 2.2566,
|
| 15427 |
+
"step": 2559
|
| 15428 |
+
},
|
| 15429 |
+
{
|
| 15430 |
+
"epoch": 0.43,
|
| 15431 |
+
"learning_rate": 0.00046644067796610176,
|
| 15432 |
+
"loss": 2.1982,
|
| 15433 |
+
"step": 2560
|
| 15434 |
+
},
|
| 15435 |
+
{
|
| 15436 |
+
"epoch": 0.43,
|
| 15437 |
+
"learning_rate": 0.00046630508474576274,
|
| 15438 |
+
"loss": 2.1657,
|
| 15439 |
+
"step": 2561
|
| 15440 |
+
},
|
| 15441 |
+
{
|
| 15442 |
+
"epoch": 0.43,
|
| 15443 |
+
"learning_rate": 0.00046616949152542377,
|
| 15444 |
+
"loss": 2.2471,
|
| 15445 |
+
"step": 2562
|
| 15446 |
+
},
|
| 15447 |
+
{
|
| 15448 |
+
"epoch": 0.43,
|
| 15449 |
+
"learning_rate": 0.00046603389830508475,
|
| 15450 |
+
"loss": 2.207,
|
| 15451 |
+
"step": 2563
|
| 15452 |
+
},
|
| 15453 |
+
{
|
| 15454 |
+
"epoch": 0.43,
|
| 15455 |
+
"learning_rate": 0.0004658983050847458,
|
| 15456 |
+
"loss": 2.1663,
|
| 15457 |
+
"step": 2564
|
| 15458 |
+
},
|
| 15459 |
+
{
|
| 15460 |
+
"epoch": 0.43,
|
| 15461 |
+
"learning_rate": 0.00046576271186440677,
|
| 15462 |
+
"loss": 2.087,
|
| 15463 |
+
"step": 2565
|
| 15464 |
+
},
|
| 15465 |
+
{
|
| 15466 |
+
"epoch": 0.43,
|
| 15467 |
+
"learning_rate": 0.00046562711864406786,
|
| 15468 |
+
"loss": 2.2147,
|
| 15469 |
+
"step": 2566
|
| 15470 |
+
},
|
| 15471 |
+
{
|
| 15472 |
+
"epoch": 0.43,
|
| 15473 |
+
"learning_rate": 0.00046549152542372884,
|
| 15474 |
+
"loss": 2.22,
|
| 15475 |
+
"step": 2567
|
| 15476 |
+
},
|
| 15477 |
+
{
|
| 15478 |
+
"epoch": 0.43,
|
| 15479 |
+
"learning_rate": 0.00046535593220338987,
|
| 15480 |
+
"loss": 2.2194,
|
| 15481 |
+
"step": 2568
|
| 15482 |
+
},
|
| 15483 |
+
{
|
| 15484 |
+
"epoch": 0.43,
|
| 15485 |
+
"learning_rate": 0.00046522033898305085,
|
| 15486 |
+
"loss": 2.1823,
|
| 15487 |
+
"step": 2569
|
| 15488 |
+
},
|
| 15489 |
+
{
|
| 15490 |
+
"epoch": 0.43,
|
| 15491 |
+
"learning_rate": 0.00046508474576271194,
|
| 15492 |
+
"loss": 2.1742,
|
| 15493 |
+
"step": 2570
|
| 15494 |
+
},
|
| 15495 |
+
{
|
| 15496 |
+
"epoch": 0.43,
|
| 15497 |
+
"learning_rate": 0.0004649491525423729,
|
| 15498 |
+
"loss": 2.2006,
|
| 15499 |
+
"step": 2571
|
| 15500 |
+
},
|
| 15501 |
+
{
|
| 15502 |
+
"epoch": 0.43,
|
| 15503 |
+
"learning_rate": 0.00046481355932203396,
|
| 15504 |
+
"loss": 2.2428,
|
| 15505 |
+
"step": 2572
|
| 15506 |
+
},
|
| 15507 |
+
{
|
| 15508 |
+
"epoch": 0.43,
|
| 15509 |
+
"learning_rate": 0.00046467796610169494,
|
| 15510 |
+
"loss": 2.1535,
|
| 15511 |
+
"step": 2573
|
| 15512 |
+
},
|
| 15513 |
+
{
|
| 15514 |
+
"epoch": 0.43,
|
| 15515 |
+
"learning_rate": 0.0004645423728813559,
|
| 15516 |
+
"loss": 2.1628,
|
| 15517 |
+
"step": 2574
|
| 15518 |
+
},
|
| 15519 |
+
{
|
| 15520 |
+
"epoch": 0.43,
|
| 15521 |
+
"learning_rate": 0.00046440677966101695,
|
| 15522 |
+
"loss": 2.1787,
|
| 15523 |
+
"step": 2575
|
| 15524 |
+
},
|
| 15525 |
+
{
|
| 15526 |
+
"epoch": 0.43,
|
| 15527 |
+
"learning_rate": 0.00046427118644067793,
|
| 15528 |
+
"loss": 2.1241,
|
| 15529 |
+
"step": 2576
|
| 15530 |
+
},
|
| 15531 |
+
{
|
| 15532 |
+
"epoch": 0.43,
|
| 15533 |
+
"learning_rate": 0.000464135593220339,
|
| 15534 |
+
"loss": 2.2536,
|
| 15535 |
+
"step": 2577
|
| 15536 |
+
},
|
| 15537 |
+
{
|
| 15538 |
+
"epoch": 0.43,
|
| 15539 |
+
"learning_rate": 0.000464,
|
| 15540 |
+
"loss": 2.0936,
|
| 15541 |
+
"step": 2578
|
| 15542 |
+
},
|
| 15543 |
+
{
|
| 15544 |
+
"epoch": 0.43,
|
| 15545 |
+
"learning_rate": 0.00046386440677966104,
|
| 15546 |
+
"loss": 2.2309,
|
| 15547 |
+
"step": 2579
|
| 15548 |
+
},
|
| 15549 |
+
{
|
| 15550 |
+
"epoch": 0.43,
|
| 15551 |
+
"learning_rate": 0.000463728813559322,
|
| 15552 |
+
"loss": 2.1871,
|
| 15553 |
+
"step": 2580
|
| 15554 |
+
},
|
| 15555 |
+
{
|
| 15556 |
+
"epoch": 0.43,
|
| 15557 |
+
"learning_rate": 0.0004635932203389831,
|
| 15558 |
+
"loss": 2.2268,
|
| 15559 |
+
"step": 2581
|
| 15560 |
+
},
|
| 15561 |
+
{
|
| 15562 |
+
"epoch": 0.43,
|
| 15563 |
+
"learning_rate": 0.0004634576271186441,
|
| 15564 |
+
"loss": 2.1735,
|
| 15565 |
+
"step": 2582
|
| 15566 |
+
},
|
| 15567 |
+
{
|
| 15568 |
+
"epoch": 0.43,
|
| 15569 |
+
"learning_rate": 0.0004633220338983051,
|
| 15570 |
+
"loss": 2.1171,
|
| 15571 |
+
"step": 2583
|
| 15572 |
+
},
|
| 15573 |
+
{
|
| 15574 |
+
"epoch": 0.43,
|
| 15575 |
+
"learning_rate": 0.0004631864406779661,
|
| 15576 |
+
"loss": 2.2784,
|
| 15577 |
+
"step": 2584
|
| 15578 |
+
},
|
| 15579 |
+
{
|
| 15580 |
+
"epoch": 0.43,
|
| 15581 |
+
"learning_rate": 0.0004630508474576272,
|
| 15582 |
+
"loss": 2.212,
|
| 15583 |
+
"step": 2585
|
| 15584 |
+
},
|
| 15585 |
+
{
|
| 15586 |
+
"epoch": 0.43,
|
| 15587 |
+
"learning_rate": 0.0004629152542372881,
|
| 15588 |
+
"loss": 2.3069,
|
| 15589 |
+
"step": 2586
|
| 15590 |
+
},
|
| 15591 |
+
{
|
| 15592 |
+
"epoch": 0.43,
|
| 15593 |
+
"learning_rate": 0.0004627796610169492,
|
| 15594 |
+
"loss": 2.2547,
|
| 15595 |
+
"step": 2587
|
| 15596 |
+
},
|
| 15597 |
+
{
|
| 15598 |
+
"epoch": 0.43,
|
| 15599 |
+
"learning_rate": 0.0004626440677966102,
|
| 15600 |
+
"loss": 2.1827,
|
| 15601 |
+
"step": 2588
|
| 15602 |
+
},
|
| 15603 |
+
{
|
| 15604 |
+
"epoch": 0.43,
|
| 15605 |
+
"learning_rate": 0.0004625084745762712,
|
| 15606 |
+
"loss": 2.1967,
|
| 15607 |
+
"step": 2589
|
| 15608 |
+
},
|
| 15609 |
+
{
|
| 15610 |
+
"epoch": 0.43,
|
| 15611 |
+
"learning_rate": 0.0004623728813559322,
|
| 15612 |
+
"loss": 2.1982,
|
| 15613 |
+
"step": 2590
|
| 15614 |
+
},
|
| 15615 |
+
{
|
| 15616 |
+
"epoch": 0.43,
|
| 15617 |
+
"learning_rate": 0.0004622372881355933,
|
| 15618 |
+
"loss": 2.2301,
|
| 15619 |
+
"step": 2591
|
| 15620 |
+
},
|
| 15621 |
+
{
|
| 15622 |
+
"epoch": 0.43,
|
| 15623 |
+
"learning_rate": 0.00046210169491525427,
|
| 15624 |
+
"loss": 2.1121,
|
| 15625 |
+
"step": 2592
|
| 15626 |
+
},
|
| 15627 |
+
{
|
| 15628 |
+
"epoch": 0.43,
|
| 15629 |
+
"learning_rate": 0.00046196610169491525,
|
| 15630 |
+
"loss": 2.1594,
|
| 15631 |
+
"step": 2593
|
| 15632 |
+
},
|
| 15633 |
+
{
|
| 15634 |
+
"epoch": 0.43,
|
| 15635 |
+
"learning_rate": 0.0004618305084745763,
|
| 15636 |
+
"loss": 2.2726,
|
| 15637 |
+
"step": 2594
|
| 15638 |
+
},
|
| 15639 |
+
{
|
| 15640 |
+
"epoch": 0.43,
|
| 15641 |
+
"learning_rate": 0.00046169491525423727,
|
| 15642 |
+
"loss": 2.1845,
|
| 15643 |
+
"step": 2595
|
| 15644 |
+
},
|
| 15645 |
+
{
|
| 15646 |
+
"epoch": 0.43,
|
| 15647 |
+
"learning_rate": 0.00046155932203389836,
|
| 15648 |
+
"loss": 2.131,
|
| 15649 |
+
"step": 2596
|
| 15650 |
+
},
|
| 15651 |
+
{
|
| 15652 |
+
"epoch": 0.43,
|
| 15653 |
+
"learning_rate": 0.0004614237288135593,
|
| 15654 |
+
"loss": 2.1921,
|
| 15655 |
+
"step": 2597
|
| 15656 |
+
},
|
| 15657 |
+
{
|
| 15658 |
+
"epoch": 0.43,
|
| 15659 |
+
"learning_rate": 0.0004612881355932204,
|
| 15660 |
+
"loss": 2.1942,
|
| 15661 |
+
"step": 2598
|
| 15662 |
+
},
|
| 15663 |
+
{
|
| 15664 |
+
"epoch": 0.43,
|
| 15665 |
+
"learning_rate": 0.00046115254237288135,
|
| 15666 |
+
"loss": 2.2077,
|
| 15667 |
+
"step": 2599
|
| 15668 |
+
},
|
| 15669 |
+
{
|
| 15670 |
+
"epoch": 0.43,
|
| 15671 |
+
"learning_rate": 0.0004610169491525424,
|
| 15672 |
+
"loss": 2.237,
|
| 15673 |
+
"step": 2600
|
| 15674 |
+
},
|
| 15675 |
+
{
|
| 15676 |
+
"epoch": 0.43,
|
| 15677 |
+
"learning_rate": 0.00046088135593220337,
|
| 15678 |
+
"loss": 2.1849,
|
| 15679 |
+
"step": 2601
|
| 15680 |
+
},
|
| 15681 |
+
{
|
| 15682 |
+
"epoch": 0.43,
|
| 15683 |
+
"learning_rate": 0.00046074576271186446,
|
| 15684 |
+
"loss": 2.2331,
|
| 15685 |
+
"step": 2602
|
| 15686 |
+
},
|
| 15687 |
+
{
|
| 15688 |
+
"epoch": 0.43,
|
| 15689 |
+
"learning_rate": 0.00046061016949152544,
|
| 15690 |
+
"loss": 2.1861,
|
| 15691 |
+
"step": 2603
|
| 15692 |
+
},
|
| 15693 |
+
{
|
| 15694 |
+
"epoch": 0.43,
|
| 15695 |
+
"learning_rate": 0.0004604745762711865,
|
| 15696 |
+
"loss": 2.1723,
|
| 15697 |
+
"step": 2604
|
| 15698 |
+
},
|
| 15699 |
+
{
|
| 15700 |
+
"epoch": 0.43,
|
| 15701 |
+
"learning_rate": 0.00046033898305084745,
|
| 15702 |
+
"loss": 2.2068,
|
| 15703 |
+
"step": 2605
|
| 15704 |
+
},
|
| 15705 |
+
{
|
| 15706 |
+
"epoch": 0.43,
|
| 15707 |
+
"learning_rate": 0.00046020338983050854,
|
| 15708 |
+
"loss": 2.1222,
|
| 15709 |
+
"step": 2606
|
| 15710 |
+
},
|
| 15711 |
+
{
|
| 15712 |
+
"epoch": 0.43,
|
| 15713 |
+
"learning_rate": 0.0004600677966101695,
|
| 15714 |
+
"loss": 2.1435,
|
| 15715 |
+
"step": 2607
|
| 15716 |
+
},
|
| 15717 |
+
{
|
| 15718 |
+
"epoch": 0.43,
|
| 15719 |
+
"learning_rate": 0.00045993220338983056,
|
| 15720 |
+
"loss": 2.2329,
|
| 15721 |
+
"step": 2608
|
| 15722 |
+
},
|
| 15723 |
+
{
|
| 15724 |
+
"epoch": 0.43,
|
| 15725 |
+
"learning_rate": 0.00045979661016949154,
|
| 15726 |
+
"loss": 2.1265,
|
| 15727 |
+
"step": 2609
|
| 15728 |
+
},
|
| 15729 |
+
{
|
| 15730 |
+
"epoch": 0.43,
|
| 15731 |
+
"learning_rate": 0.0004596610169491526,
|
| 15732 |
+
"loss": 2.2636,
|
| 15733 |
+
"step": 2610
|
| 15734 |
+
},
|
| 15735 |
+
{
|
| 15736 |
+
"epoch": 0.44,
|
| 15737 |
+
"learning_rate": 0.00045952542372881355,
|
| 15738 |
+
"loss": 2.2236,
|
| 15739 |
+
"step": 2611
|
| 15740 |
+
},
|
| 15741 |
+
{
|
| 15742 |
+
"epoch": 0.44,
|
| 15743 |
+
"learning_rate": 0.00045938983050847464,
|
| 15744 |
+
"loss": 2.2194,
|
| 15745 |
+
"step": 2612
|
| 15746 |
+
},
|
| 15747 |
+
{
|
| 15748 |
+
"epoch": 0.44,
|
| 15749 |
+
"learning_rate": 0.0004592542372881356,
|
| 15750 |
+
"loss": 2.1611,
|
| 15751 |
+
"step": 2613
|
| 15752 |
+
},
|
| 15753 |
+
{
|
| 15754 |
+
"epoch": 0.44,
|
| 15755 |
+
"learning_rate": 0.0004591186440677966,
|
| 15756 |
+
"loss": 2.1603,
|
| 15757 |
+
"step": 2614
|
| 15758 |
+
},
|
| 15759 |
+
{
|
| 15760 |
+
"epoch": 0.44,
|
| 15761 |
+
"learning_rate": 0.00045898305084745764,
|
| 15762 |
+
"loss": 2.1169,
|
| 15763 |
+
"step": 2615
|
| 15764 |
+
},
|
| 15765 |
+
{
|
| 15766 |
+
"epoch": 0.44,
|
| 15767 |
+
"learning_rate": 0.0004588474576271186,
|
| 15768 |
+
"loss": 2.2098,
|
| 15769 |
+
"step": 2616
|
| 15770 |
+
},
|
| 15771 |
+
{
|
| 15772 |
+
"epoch": 0.44,
|
| 15773 |
+
"learning_rate": 0.0004587118644067797,
|
| 15774 |
+
"loss": 2.237,
|
| 15775 |
+
"step": 2617
|
| 15776 |
+
},
|
| 15777 |
+
{
|
| 15778 |
+
"epoch": 0.44,
|
| 15779 |
+
"learning_rate": 0.0004585762711864407,
|
| 15780 |
+
"loss": 2.1146,
|
| 15781 |
+
"step": 2618
|
| 15782 |
+
},
|
| 15783 |
+
{
|
| 15784 |
+
"epoch": 0.44,
|
| 15785 |
+
"learning_rate": 0.0004584406779661017,
|
| 15786 |
+
"loss": 2.2139,
|
| 15787 |
+
"step": 2619
|
| 15788 |
+
},
|
| 15789 |
+
{
|
| 15790 |
+
"epoch": 0.44,
|
| 15791 |
+
"learning_rate": 0.0004583050847457627,
|
| 15792 |
+
"loss": 2.1968,
|
| 15793 |
+
"step": 2620
|
| 15794 |
+
},
|
| 15795 |
+
{
|
| 15796 |
+
"epoch": 0.44,
|
| 15797 |
+
"learning_rate": 0.00045816949152542374,
|
| 15798 |
+
"loss": 2.1989,
|
| 15799 |
+
"step": 2621
|
| 15800 |
+
},
|
| 15801 |
+
{
|
| 15802 |
+
"epoch": 0.44,
|
| 15803 |
+
"learning_rate": 0.0004580338983050847,
|
| 15804 |
+
"loss": 2.2338,
|
| 15805 |
+
"step": 2622
|
| 15806 |
+
},
|
| 15807 |
+
{
|
| 15808 |
+
"epoch": 0.44,
|
| 15809 |
+
"learning_rate": 0.0004578983050847458,
|
| 15810 |
+
"loss": 2.2344,
|
| 15811 |
+
"step": 2623
|
| 15812 |
+
},
|
| 15813 |
+
{
|
| 15814 |
+
"epoch": 0.44,
|
| 15815 |
+
"learning_rate": 0.0004577627118644068,
|
| 15816 |
+
"loss": 2.1605,
|
| 15817 |
+
"step": 2624
|
| 15818 |
+
},
|
| 15819 |
+
{
|
| 15820 |
+
"epoch": 0.44,
|
| 15821 |
+
"learning_rate": 0.0004576271186440678,
|
| 15822 |
+
"loss": 2.1599,
|
| 15823 |
+
"step": 2625
|
| 15824 |
+
},
|
| 15825 |
+
{
|
| 15826 |
+
"epoch": 0.44,
|
| 15827 |
+
"learning_rate": 0.0004574915254237288,
|
| 15828 |
+
"loss": 2.2193,
|
| 15829 |
+
"step": 2626
|
| 15830 |
+
},
|
| 15831 |
+
{
|
| 15832 |
+
"epoch": 0.44,
|
| 15833 |
+
"learning_rate": 0.0004573559322033899,
|
| 15834 |
+
"loss": 2.199,
|
| 15835 |
+
"step": 2627
|
| 15836 |
+
},
|
| 15837 |
+
{
|
| 15838 |
+
"epoch": 0.44,
|
| 15839 |
+
"learning_rate": 0.0004572203389830509,
|
| 15840 |
+
"loss": 2.1502,
|
| 15841 |
+
"step": 2628
|
| 15842 |
+
},
|
| 15843 |
+
{
|
| 15844 |
+
"epoch": 0.44,
|
| 15845 |
+
"learning_rate": 0.0004570847457627119,
|
| 15846 |
+
"loss": 2.2478,
|
| 15847 |
+
"step": 2629
|
| 15848 |
+
},
|
| 15849 |
+
{
|
| 15850 |
+
"epoch": 0.44,
|
| 15851 |
+
"learning_rate": 0.0004569491525423729,
|
| 15852 |
+
"loss": 2.1512,
|
| 15853 |
+
"step": 2630
|
| 15854 |
+
},
|
| 15855 |
+
{
|
| 15856 |
+
"epoch": 0.44,
|
| 15857 |
+
"learning_rate": 0.000456813559322034,
|
| 15858 |
+
"loss": 2.2605,
|
| 15859 |
+
"step": 2631
|
| 15860 |
+
},
|
| 15861 |
+
{
|
| 15862 |
+
"epoch": 0.44,
|
| 15863 |
+
"learning_rate": 0.0004566779661016949,
|
| 15864 |
+
"loss": 2.258,
|
| 15865 |
+
"step": 2632
|
| 15866 |
+
},
|
| 15867 |
+
{
|
| 15868 |
+
"epoch": 0.44,
|
| 15869 |
+
"learning_rate": 0.000456542372881356,
|
| 15870 |
+
"loss": 2.1609,
|
| 15871 |
+
"step": 2633
|
| 15872 |
+
},
|
| 15873 |
+
{
|
| 15874 |
+
"epoch": 0.44,
|
| 15875 |
+
"learning_rate": 0.000456406779661017,
|
| 15876 |
+
"loss": 2.2546,
|
| 15877 |
+
"step": 2634
|
| 15878 |
+
},
|
| 15879 |
+
{
|
| 15880 |
+
"epoch": 0.44,
|
| 15881 |
+
"learning_rate": 0.00045627118644067796,
|
| 15882 |
+
"loss": 2.1307,
|
| 15883 |
+
"step": 2635
|
| 15884 |
+
},
|
| 15885 |
+
{
|
| 15886 |
+
"epoch": 0.44,
|
| 15887 |
+
"learning_rate": 0.000456135593220339,
|
| 15888 |
+
"loss": 2.1853,
|
| 15889 |
+
"step": 2636
|
| 15890 |
+
},
|
| 15891 |
+
{
|
| 15892 |
+
"epoch": 0.44,
|
| 15893 |
+
"learning_rate": 0.00045599999999999997,
|
| 15894 |
+
"loss": 2.2144,
|
| 15895 |
+
"step": 2637
|
| 15896 |
+
},
|
| 15897 |
+
{
|
| 15898 |
+
"epoch": 0.44,
|
| 15899 |
+
"learning_rate": 0.00045586440677966106,
|
| 15900 |
+
"loss": 2.2212,
|
| 15901 |
+
"step": 2638
|
| 15902 |
+
},
|
| 15903 |
+
{
|
| 15904 |
+
"epoch": 0.44,
|
| 15905 |
+
"learning_rate": 0.00045572881355932204,
|
| 15906 |
+
"loss": 2.1757,
|
| 15907 |
+
"step": 2639
|
| 15908 |
+
},
|
| 15909 |
+
{
|
| 15910 |
+
"epoch": 0.44,
|
| 15911 |
+
"learning_rate": 0.0004555932203389831,
|
| 15912 |
+
"loss": 2.1837,
|
| 15913 |
+
"step": 2640
|
| 15914 |
+
},
|
| 15915 |
+
{
|
| 15916 |
+
"epoch": 0.44,
|
| 15917 |
+
"learning_rate": 0.00045545762711864406,
|
| 15918 |
+
"loss": 2.1359,
|
| 15919 |
+
"step": 2641
|
| 15920 |
+
},
|
| 15921 |
+
{
|
| 15922 |
+
"epoch": 0.44,
|
| 15923 |
+
"learning_rate": 0.00045532203389830514,
|
| 15924 |
+
"loss": 2.2074,
|
| 15925 |
+
"step": 2642
|
| 15926 |
+
},
|
| 15927 |
+
{
|
| 15928 |
+
"epoch": 0.44,
|
| 15929 |
+
"learning_rate": 0.00045518644067796607,
|
| 15930 |
+
"loss": 2.1438,
|
| 15931 |
+
"step": 2643
|
| 15932 |
+
},
|
| 15933 |
+
{
|
| 15934 |
+
"epoch": 0.44,
|
| 15935 |
+
"learning_rate": 0.00045505084745762716,
|
| 15936 |
+
"loss": 2.1324,
|
| 15937 |
+
"step": 2644
|
| 15938 |
+
},
|
| 15939 |
+
{
|
| 15940 |
+
"epoch": 0.44,
|
| 15941 |
+
"learning_rate": 0.00045491525423728814,
|
| 15942 |
+
"loss": 2.225,
|
| 15943 |
+
"step": 2645
|
| 15944 |
+
},
|
| 15945 |
+
{
|
| 15946 |
+
"epoch": 0.44,
|
| 15947 |
+
"learning_rate": 0.0004547796610169492,
|
| 15948 |
+
"loss": 2.1935,
|
| 15949 |
+
"step": 2646
|
| 15950 |
+
},
|
| 15951 |
+
{
|
| 15952 |
+
"epoch": 0.44,
|
| 15953 |
+
"learning_rate": 0.00045464406779661016,
|
| 15954 |
+
"loss": 2.1445,
|
| 15955 |
+
"step": 2647
|
| 15956 |
+
},
|
| 15957 |
+
{
|
| 15958 |
+
"epoch": 0.44,
|
| 15959 |
+
"learning_rate": 0.00045450847457627125,
|
| 15960 |
+
"loss": 2.2461,
|
| 15961 |
+
"step": 2648
|
| 15962 |
+
},
|
| 15963 |
+
{
|
| 15964 |
+
"epoch": 0.44,
|
| 15965 |
+
"learning_rate": 0.0004543728813559322,
|
| 15966 |
+
"loss": 2.256,
|
| 15967 |
+
"step": 2649
|
| 15968 |
+
},
|
| 15969 |
+
{
|
| 15970 |
+
"epoch": 0.44,
|
| 15971 |
+
"learning_rate": 0.00045423728813559326,
|
| 15972 |
+
"loss": 2.1491,
|
| 15973 |
+
"step": 2650
|
| 15974 |
+
},
|
| 15975 |
+
{
|
| 15976 |
+
"epoch": 0.44,
|
| 15977 |
+
"learning_rate": 0.00045410169491525424,
|
| 15978 |
+
"loss": 2.139,
|
| 15979 |
+
"step": 2651
|
| 15980 |
+
},
|
| 15981 |
+
{
|
| 15982 |
+
"epoch": 0.44,
|
| 15983 |
+
"learning_rate": 0.00045396610169491533,
|
| 15984 |
+
"loss": 2.1354,
|
| 15985 |
+
"step": 2652
|
| 15986 |
+
},
|
| 15987 |
+
{
|
| 15988 |
+
"epoch": 0.44,
|
| 15989 |
+
"learning_rate": 0.0004538305084745763,
|
| 15990 |
+
"loss": 2.2622,
|
| 15991 |
+
"step": 2653
|
| 15992 |
+
},
|
| 15993 |
+
{
|
| 15994 |
+
"epoch": 0.44,
|
| 15995 |
+
"learning_rate": 0.00045369491525423735,
|
| 15996 |
+
"loss": 2.2424,
|
| 15997 |
+
"step": 2654
|
| 15998 |
+
},
|
| 15999 |
+
{
|
| 16000 |
+
"epoch": 0.44,
|
| 16001 |
+
"learning_rate": 0.0004535593220338983,
|
| 16002 |
+
"loss": 2.3128,
|
| 16003 |
+
"step": 2655
|
| 16004 |
+
},
|
| 16005 |
+
{
|
| 16006 |
+
"epoch": 0.44,
|
| 16007 |
+
"learning_rate": 0.0004534237288135593,
|
| 16008 |
+
"loss": 2.2169,
|
| 16009 |
+
"step": 2656
|
| 16010 |
+
},
|
| 16011 |
+
{
|
| 16012 |
+
"epoch": 0.44,
|
| 16013 |
+
"learning_rate": 0.00045328813559322034,
|
| 16014 |
+
"loss": 2.2816,
|
| 16015 |
+
"step": 2657
|
| 16016 |
+
},
|
| 16017 |
+
{
|
| 16018 |
+
"epoch": 0.44,
|
| 16019 |
+
"learning_rate": 0.0004531525423728813,
|
| 16020 |
+
"loss": 2.2674,
|
| 16021 |
+
"step": 2658
|
| 16022 |
+
},
|
| 16023 |
+
{
|
| 16024 |
+
"epoch": 0.44,
|
| 16025 |
+
"learning_rate": 0.0004530169491525424,
|
| 16026 |
+
"loss": 2.1776,
|
| 16027 |
+
"step": 2659
|
| 16028 |
+
},
|
| 16029 |
+
{
|
| 16030 |
+
"epoch": 0.44,
|
| 16031 |
+
"learning_rate": 0.0004528813559322034,
|
| 16032 |
+
"loss": 2.2076,
|
| 16033 |
+
"step": 2660
|
| 16034 |
+
},
|
| 16035 |
+
{
|
| 16036 |
+
"epoch": 0.44,
|
| 16037 |
+
"learning_rate": 0.0004527457627118644,
|
| 16038 |
+
"loss": 2.1904,
|
| 16039 |
+
"step": 2661
|
| 16040 |
+
},
|
| 16041 |
+
{
|
| 16042 |
+
"epoch": 0.44,
|
| 16043 |
+
"learning_rate": 0.0004526101694915254,
|
| 16044 |
+
"loss": 2.175,
|
| 16045 |
+
"step": 2662
|
| 16046 |
+
},
|
| 16047 |
+
{
|
| 16048 |
+
"epoch": 0.44,
|
| 16049 |
+
"learning_rate": 0.0004524745762711865,
|
| 16050 |
+
"loss": 2.2094,
|
| 16051 |
+
"step": 2663
|
| 16052 |
+
},
|
| 16053 |
+
{
|
| 16054 |
+
"epoch": 0.44,
|
| 16055 |
+
"learning_rate": 0.0004523389830508475,
|
| 16056 |
+
"loss": 2.0975,
|
| 16057 |
+
"step": 2664
|
| 16058 |
+
},
|
| 16059 |
+
{
|
| 16060 |
+
"epoch": 0.44,
|
| 16061 |
+
"learning_rate": 0.0004522033898305085,
|
| 16062 |
+
"loss": 2.1961,
|
| 16063 |
+
"step": 2665
|
| 16064 |
+
},
|
| 16065 |
+
{
|
| 16066 |
+
"epoch": 0.44,
|
| 16067 |
+
"learning_rate": 0.0004520677966101695,
|
| 16068 |
+
"loss": 2.1559,
|
| 16069 |
+
"step": 2666
|
| 16070 |
+
},
|
| 16071 |
+
{
|
| 16072 |
+
"epoch": 0.44,
|
| 16073 |
+
"learning_rate": 0.0004519322033898306,
|
| 16074 |
+
"loss": 2.1439,
|
| 16075 |
+
"step": 2667
|
| 16076 |
+
},
|
| 16077 |
+
{
|
| 16078 |
+
"epoch": 0.44,
|
| 16079 |
+
"learning_rate": 0.0004517966101694915,
|
| 16080 |
+
"loss": 2.1496,
|
| 16081 |
+
"step": 2668
|
| 16082 |
+
},
|
| 16083 |
+
{
|
| 16084 |
+
"epoch": 0.44,
|
| 16085 |
+
"learning_rate": 0.0004516610169491526,
|
| 16086 |
+
"loss": 2.1898,
|
| 16087 |
+
"step": 2669
|
| 16088 |
+
},
|
| 16089 |
+
{
|
| 16090 |
+
"epoch": 0.45,
|
| 16091 |
+
"learning_rate": 0.0004515254237288136,
|
| 16092 |
+
"loss": 2.1843,
|
| 16093 |
+
"step": 2670
|
| 16094 |
+
},
|
| 16095 |
+
{
|
| 16096 |
+
"epoch": 0.45,
|
| 16097 |
+
"learning_rate": 0.0004513898305084746,
|
| 16098 |
+
"loss": 2.1139,
|
| 16099 |
+
"step": 2671
|
| 16100 |
+
},
|
| 16101 |
+
{
|
| 16102 |
+
"epoch": 0.45,
|
| 16103 |
+
"learning_rate": 0.0004512542372881356,
|
| 16104 |
+
"loss": 2.1933,
|
| 16105 |
+
"step": 2672
|
| 16106 |
+
},
|
| 16107 |
+
{
|
| 16108 |
+
"epoch": 0.45,
|
| 16109 |
+
"learning_rate": 0.0004511186440677967,
|
| 16110 |
+
"loss": 2.1813,
|
| 16111 |
+
"step": 2673
|
| 16112 |
+
},
|
| 16113 |
+
{
|
| 16114 |
+
"epoch": 0.45,
|
| 16115 |
+
"learning_rate": 0.00045098305084745766,
|
| 16116 |
+
"loss": 2.1483,
|
| 16117 |
+
"step": 2674
|
| 16118 |
+
},
|
| 16119 |
+
{
|
| 16120 |
+
"epoch": 0.45,
|
| 16121 |
+
"learning_rate": 0.0004508474576271187,
|
| 16122 |
+
"loss": 2.1978,
|
| 16123 |
+
"step": 2675
|
| 16124 |
+
},
|
| 16125 |
+
{
|
| 16126 |
+
"epoch": 0.45,
|
| 16127 |
+
"learning_rate": 0.0004507118644067797,
|
| 16128 |
+
"loss": 2.1171,
|
| 16129 |
+
"step": 2676
|
| 16130 |
+
},
|
| 16131 |
+
{
|
| 16132 |
+
"epoch": 0.45,
|
| 16133 |
+
"learning_rate": 0.00045057627118644066,
|
| 16134 |
+
"loss": 2.1211,
|
| 16135 |
+
"step": 2677
|
| 16136 |
+
},
|
| 16137 |
+
{
|
| 16138 |
+
"epoch": 0.45,
|
| 16139 |
+
"learning_rate": 0.00045044067796610175,
|
| 16140 |
+
"loss": 2.1748,
|
| 16141 |
+
"step": 2678
|
| 16142 |
+
},
|
| 16143 |
+
{
|
| 16144 |
+
"epoch": 0.45,
|
| 16145 |
+
"learning_rate": 0.0004503050847457627,
|
| 16146 |
+
"loss": 2.1855,
|
| 16147 |
+
"step": 2679
|
| 16148 |
+
},
|
| 16149 |
+
{
|
| 16150 |
+
"epoch": 0.45,
|
| 16151 |
+
"learning_rate": 0.00045016949152542376,
|
| 16152 |
+
"loss": 2.1501,
|
| 16153 |
+
"step": 2680
|
| 16154 |
+
},
|
| 16155 |
+
{
|
| 16156 |
+
"epoch": 0.45,
|
| 16157 |
+
"learning_rate": 0.00045003389830508474,
|
| 16158 |
+
"loss": 2.2428,
|
| 16159 |
+
"step": 2681
|
| 16160 |
+
},
|
| 16161 |
+
{
|
| 16162 |
+
"epoch": 0.45,
|
| 16163 |
+
"learning_rate": 0.0004498983050847458,
|
| 16164 |
+
"loss": 2.1357,
|
| 16165 |
+
"step": 2682
|
| 16166 |
+
},
|
| 16167 |
+
{
|
| 16168 |
+
"epoch": 0.45,
|
| 16169 |
+
"learning_rate": 0.00044976271186440676,
|
| 16170 |
+
"loss": 2.1712,
|
| 16171 |
+
"step": 2683
|
| 16172 |
+
},
|
| 16173 |
+
{
|
| 16174 |
+
"epoch": 0.45,
|
| 16175 |
+
"learning_rate": 0.00044962711864406785,
|
| 16176 |
+
"loss": 2.1691,
|
| 16177 |
+
"step": 2684
|
| 16178 |
+
},
|
| 16179 |
+
{
|
| 16180 |
+
"epoch": 0.45,
|
| 16181 |
+
"learning_rate": 0.00044949152542372883,
|
| 16182 |
+
"loss": 2.2532,
|
| 16183 |
+
"step": 2685
|
| 16184 |
+
},
|
| 16185 |
+
{
|
| 16186 |
+
"epoch": 0.45,
|
| 16187 |
+
"learning_rate": 0.00044935593220338986,
|
| 16188 |
+
"loss": 2.1859,
|
| 16189 |
+
"step": 2686
|
| 16190 |
+
},
|
| 16191 |
+
{
|
| 16192 |
+
"epoch": 0.45,
|
| 16193 |
+
"learning_rate": 0.00044922033898305084,
|
| 16194 |
+
"loss": 2.2379,
|
| 16195 |
+
"step": 2687
|
| 16196 |
+
},
|
| 16197 |
+
{
|
| 16198 |
+
"epoch": 0.45,
|
| 16199 |
+
"learning_rate": 0.00044908474576271193,
|
| 16200 |
+
"loss": 2.2257,
|
| 16201 |
+
"step": 2688
|
| 16202 |
+
},
|
| 16203 |
+
{
|
| 16204 |
+
"epoch": 0.45,
|
| 16205 |
+
"learning_rate": 0.0004489491525423729,
|
| 16206 |
+
"loss": 2.1396,
|
| 16207 |
+
"step": 2689
|
| 16208 |
+
},
|
| 16209 |
+
{
|
| 16210 |
+
"epoch": 0.45,
|
| 16211 |
+
"learning_rate": 0.00044881355932203395,
|
| 16212 |
+
"loss": 2.2046,
|
| 16213 |
+
"step": 2690
|
| 16214 |
+
},
|
| 16215 |
+
{
|
| 16216 |
+
"epoch": 0.45,
|
| 16217 |
+
"learning_rate": 0.00044867796610169493,
|
| 16218 |
+
"loss": 2.1902,
|
| 16219 |
+
"step": 2691
|
| 16220 |
+
},
|
| 16221 |
+
{
|
| 16222 |
+
"epoch": 0.45,
|
| 16223 |
+
"learning_rate": 0.00044854237288135596,
|
| 16224 |
+
"loss": 2.1787,
|
| 16225 |
+
"step": 2692
|
| 16226 |
+
},
|
| 16227 |
+
{
|
| 16228 |
+
"epoch": 0.45,
|
| 16229 |
+
"learning_rate": 0.00044840677966101694,
|
| 16230 |
+
"loss": 2.2116,
|
| 16231 |
+
"step": 2693
|
| 16232 |
+
},
|
| 16233 |
+
{
|
| 16234 |
+
"epoch": 0.45,
|
| 16235 |
+
"learning_rate": 0.00044827118644067803,
|
| 16236 |
+
"loss": 2.196,
|
| 16237 |
+
"step": 2694
|
| 16238 |
+
},
|
| 16239 |
+
{
|
| 16240 |
+
"epoch": 0.45,
|
| 16241 |
+
"learning_rate": 0.000448135593220339,
|
| 16242 |
+
"loss": 2.1512,
|
| 16243 |
+
"step": 2695
|
| 16244 |
+
},
|
| 16245 |
+
{
|
| 16246 |
+
"epoch": 0.45,
|
| 16247 |
+
"learning_rate": 0.00044800000000000005,
|
| 16248 |
+
"loss": 2.2,
|
| 16249 |
+
"step": 2696
|
| 16250 |
+
},
|
| 16251 |
+
{
|
| 16252 |
+
"epoch": 0.45,
|
| 16253 |
+
"learning_rate": 0.00044786440677966103,
|
| 16254 |
+
"loss": 2.2264,
|
| 16255 |
+
"step": 2697
|
| 16256 |
+
},
|
| 16257 |
+
{
|
| 16258 |
+
"epoch": 0.45,
|
| 16259 |
+
"learning_rate": 0.000447728813559322,
|
| 16260 |
+
"loss": 2.2349,
|
| 16261 |
+
"step": 2698
|
| 16262 |
+
},
|
| 16263 |
+
{
|
| 16264 |
+
"epoch": 0.45,
|
| 16265 |
+
"learning_rate": 0.0004475932203389831,
|
| 16266 |
+
"loss": 2.191,
|
| 16267 |
+
"step": 2699
|
| 16268 |
+
},
|
| 16269 |
+
{
|
| 16270 |
+
"epoch": 0.45,
|
| 16271 |
+
"learning_rate": 0.0004474576271186441,
|
| 16272 |
+
"loss": 2.1978,
|
| 16273 |
+
"step": 2700
|
| 16274 |
+
},
|
| 16275 |
+
{
|
| 16276 |
+
"epoch": 0.45,
|
| 16277 |
+
"learning_rate": 0.0004473220338983051,
|
| 16278 |
+
"loss": 2.2675,
|
| 16279 |
+
"step": 2701
|
| 16280 |
+
},
|
| 16281 |
+
{
|
| 16282 |
+
"epoch": 0.45,
|
| 16283 |
+
"learning_rate": 0.0004471864406779661,
|
| 16284 |
+
"loss": 2.178,
|
| 16285 |
+
"step": 2702
|
| 16286 |
+
},
|
| 16287 |
+
{
|
| 16288 |
+
"epoch": 0.45,
|
| 16289 |
+
"learning_rate": 0.00044705084745762713,
|
| 16290 |
+
"loss": 2.2941,
|
| 16291 |
+
"step": 2703
|
| 16292 |
+
},
|
| 16293 |
+
{
|
| 16294 |
+
"epoch": 0.45,
|
| 16295 |
+
"learning_rate": 0.0004469152542372881,
|
| 16296 |
+
"loss": 2.2429,
|
| 16297 |
+
"step": 2704
|
| 16298 |
+
},
|
| 16299 |
+
{
|
| 16300 |
+
"epoch": 0.45,
|
| 16301 |
+
"learning_rate": 0.0004467796610169492,
|
| 16302 |
+
"loss": 2.2242,
|
| 16303 |
+
"step": 2705
|
| 16304 |
+
},
|
| 16305 |
+
{
|
| 16306 |
+
"epoch": 0.45,
|
| 16307 |
+
"learning_rate": 0.0004466440677966102,
|
| 16308 |
+
"loss": 2.2783,
|
| 16309 |
+
"step": 2706
|
| 16310 |
+
},
|
| 16311 |
+
{
|
| 16312 |
+
"epoch": 0.45,
|
| 16313 |
+
"learning_rate": 0.0004465084745762712,
|
| 16314 |
+
"loss": 2.241,
|
| 16315 |
+
"step": 2707
|
| 16316 |
+
},
|
| 16317 |
+
{
|
| 16318 |
+
"epoch": 0.45,
|
| 16319 |
+
"learning_rate": 0.0004463728813559322,
|
| 16320 |
+
"loss": 2.2416,
|
| 16321 |
+
"step": 2708
|
| 16322 |
+
},
|
| 16323 |
+
{
|
| 16324 |
+
"epoch": 0.45,
|
| 16325 |
+
"learning_rate": 0.0004462372881355933,
|
| 16326 |
+
"loss": 2.1677,
|
| 16327 |
+
"step": 2709
|
| 16328 |
+
},
|
| 16329 |
+
{
|
| 16330 |
+
"epoch": 0.45,
|
| 16331 |
+
"learning_rate": 0.00044610169491525426,
|
| 16332 |
+
"loss": 2.2013,
|
| 16333 |
+
"step": 2710
|
| 16334 |
+
},
|
| 16335 |
+
{
|
| 16336 |
+
"epoch": 0.45,
|
| 16337 |
+
"learning_rate": 0.0004459661016949153,
|
| 16338 |
+
"loss": 2.1218,
|
| 16339 |
+
"step": 2711
|
| 16340 |
+
},
|
| 16341 |
+
{
|
| 16342 |
+
"epoch": 0.45,
|
| 16343 |
+
"learning_rate": 0.0004458305084745763,
|
| 16344 |
+
"loss": 2.1403,
|
| 16345 |
+
"step": 2712
|
| 16346 |
+
},
|
| 16347 |
+
{
|
| 16348 |
+
"epoch": 0.45,
|
| 16349 |
+
"learning_rate": 0.00044569491525423737,
|
| 16350 |
+
"loss": 2.2065,
|
| 16351 |
+
"step": 2713
|
| 16352 |
+
},
|
| 16353 |
+
{
|
| 16354 |
+
"epoch": 0.45,
|
| 16355 |
+
"learning_rate": 0.0004455593220338983,
|
| 16356 |
+
"loss": 2.1355,
|
| 16357 |
+
"step": 2714
|
| 16358 |
+
},
|
| 16359 |
+
{
|
| 16360 |
+
"epoch": 0.45,
|
| 16361 |
+
"learning_rate": 0.0004454237288135594,
|
| 16362 |
+
"loss": 2.1722,
|
| 16363 |
+
"step": 2715
|
| 16364 |
+
},
|
| 16365 |
+
{
|
| 16366 |
+
"epoch": 0.45,
|
| 16367 |
+
"learning_rate": 0.00044528813559322036,
|
| 16368 |
+
"loss": 2.2116,
|
| 16369 |
+
"step": 2716
|
| 16370 |
+
},
|
| 16371 |
+
{
|
| 16372 |
+
"epoch": 0.45,
|
| 16373 |
+
"learning_rate": 0.0004451525423728814,
|
| 16374 |
+
"loss": 2.1821,
|
| 16375 |
+
"step": 2717
|
| 16376 |
+
},
|
| 16377 |
+
{
|
| 16378 |
+
"epoch": 0.45,
|
| 16379 |
+
"learning_rate": 0.0004450169491525424,
|
| 16380 |
+
"loss": 2.2321,
|
| 16381 |
+
"step": 2718
|
| 16382 |
+
},
|
| 16383 |
+
{
|
| 16384 |
+
"epoch": 0.45,
|
| 16385 |
+
"learning_rate": 0.00044488135593220336,
|
| 16386 |
+
"loss": 2.1749,
|
| 16387 |
+
"step": 2719
|
| 16388 |
+
},
|
| 16389 |
+
{
|
| 16390 |
+
"epoch": 0.45,
|
| 16391 |
+
"learning_rate": 0.00044474576271186445,
|
| 16392 |
+
"loss": 2.219,
|
| 16393 |
+
"step": 2720
|
| 16394 |
+
},
|
| 16395 |
+
{
|
| 16396 |
+
"epoch": 0.45,
|
| 16397 |
+
"learning_rate": 0.00044461016949152543,
|
| 16398 |
+
"loss": 2.1305,
|
| 16399 |
+
"step": 2721
|
| 16400 |
+
},
|
| 16401 |
+
{
|
| 16402 |
+
"epoch": 0.45,
|
| 16403 |
+
"learning_rate": 0.00044447457627118646,
|
| 16404 |
+
"loss": 2.1545,
|
| 16405 |
+
"step": 2722
|
| 16406 |
+
},
|
| 16407 |
+
{
|
| 16408 |
+
"epoch": 0.45,
|
| 16409 |
+
"learning_rate": 0.00044433898305084744,
|
| 16410 |
+
"loss": 2.2897,
|
| 16411 |
+
"step": 2723
|
| 16412 |
+
},
|
| 16413 |
+
{
|
| 16414 |
+
"epoch": 0.45,
|
| 16415 |
+
"learning_rate": 0.00044420338983050853,
|
| 16416 |
+
"loss": 2.1997,
|
| 16417 |
+
"step": 2724
|
| 16418 |
+
},
|
| 16419 |
+
{
|
| 16420 |
+
"epoch": 0.45,
|
| 16421 |
+
"learning_rate": 0.00044406779661016946,
|
| 16422 |
+
"loss": 2.2596,
|
| 16423 |
+
"step": 2725
|
| 16424 |
+
},
|
| 16425 |
+
{
|
| 16426 |
+
"epoch": 0.45,
|
| 16427 |
+
"learning_rate": 0.00044393220338983055,
|
| 16428 |
+
"loss": 2.2346,
|
| 16429 |
+
"step": 2726
|
| 16430 |
+
},
|
| 16431 |
+
{
|
| 16432 |
+
"epoch": 0.45,
|
| 16433 |
+
"learning_rate": 0.00044379661016949153,
|
| 16434 |
+
"loss": 2.1678,
|
| 16435 |
+
"step": 2727
|
| 16436 |
+
},
|
| 16437 |
+
{
|
| 16438 |
+
"epoch": 0.45,
|
| 16439 |
+
"learning_rate": 0.00044366101694915256,
|
| 16440 |
+
"loss": 2.1903,
|
| 16441 |
+
"step": 2728
|
| 16442 |
+
},
|
| 16443 |
+
{
|
| 16444 |
+
"epoch": 0.45,
|
| 16445 |
+
"learning_rate": 0.00044352542372881354,
|
| 16446 |
+
"loss": 2.1976,
|
| 16447 |
+
"step": 2729
|
| 16448 |
+
},
|
| 16449 |
+
{
|
| 16450 |
+
"epoch": 0.46,
|
| 16451 |
+
"learning_rate": 0.00044338983050847463,
|
| 16452 |
+
"loss": 2.1003,
|
| 16453 |
+
"step": 2730
|
| 16454 |
+
},
|
| 16455 |
+
{
|
| 16456 |
+
"epoch": 0.46,
|
| 16457 |
+
"learning_rate": 0.0004432542372881356,
|
| 16458 |
+
"loss": 2.0941,
|
| 16459 |
+
"step": 2731
|
| 16460 |
+
},
|
| 16461 |
+
{
|
| 16462 |
+
"epoch": 0.46,
|
| 16463 |
+
"learning_rate": 0.00044311864406779665,
|
| 16464 |
+
"loss": 2.2205,
|
| 16465 |
+
"step": 2732
|
| 16466 |
+
},
|
| 16467 |
+
{
|
| 16468 |
+
"epoch": 0.46,
|
| 16469 |
+
"learning_rate": 0.00044298305084745763,
|
| 16470 |
+
"loss": 2.108,
|
| 16471 |
+
"step": 2733
|
| 16472 |
+
},
|
| 16473 |
+
{
|
| 16474 |
+
"epoch": 0.46,
|
| 16475 |
+
"learning_rate": 0.0004428474576271187,
|
| 16476 |
+
"loss": 2.1486,
|
| 16477 |
+
"step": 2734
|
| 16478 |
+
},
|
| 16479 |
+
{
|
| 16480 |
+
"epoch": 0.46,
|
| 16481 |
+
"learning_rate": 0.0004427118644067797,
|
| 16482 |
+
"loss": 2.1967,
|
| 16483 |
+
"step": 2735
|
| 16484 |
+
},
|
| 16485 |
+
{
|
| 16486 |
+
"epoch": 0.46,
|
| 16487 |
+
"learning_rate": 0.00044257627118644073,
|
| 16488 |
+
"loss": 2.2078,
|
| 16489 |
+
"step": 2736
|
| 16490 |
+
},
|
| 16491 |
+
{
|
| 16492 |
+
"epoch": 0.46,
|
| 16493 |
+
"learning_rate": 0.0004424406779661017,
|
| 16494 |
+
"loss": 2.2254,
|
| 16495 |
+
"step": 2737
|
| 16496 |
+
},
|
| 16497 |
+
{
|
| 16498 |
+
"epoch": 0.46,
|
| 16499 |
+
"learning_rate": 0.00044230508474576275,
|
| 16500 |
+
"loss": 2.2598,
|
| 16501 |
+
"step": 2738
|
| 16502 |
+
},
|
| 16503 |
+
{
|
| 16504 |
+
"epoch": 0.46,
|
| 16505 |
+
"learning_rate": 0.00044216949152542373,
|
| 16506 |
+
"loss": 2.2207,
|
| 16507 |
+
"step": 2739
|
| 16508 |
+
},
|
| 16509 |
+
{
|
| 16510 |
+
"epoch": 0.46,
|
| 16511 |
+
"learning_rate": 0.0004420338983050847,
|
| 16512 |
+
"loss": 2.1272,
|
| 16513 |
+
"step": 2740
|
| 16514 |
+
},
|
| 16515 |
+
{
|
| 16516 |
+
"epoch": 0.46,
|
| 16517 |
+
"learning_rate": 0.0004418983050847458,
|
| 16518 |
+
"loss": 2.2438,
|
| 16519 |
+
"step": 2741
|
| 16520 |
+
},
|
| 16521 |
+
{
|
| 16522 |
+
"epoch": 0.46,
|
| 16523 |
+
"learning_rate": 0.0004417627118644068,
|
| 16524 |
+
"loss": 2.1494,
|
| 16525 |
+
"step": 2742
|
| 16526 |
+
},
|
| 16527 |
+
{
|
| 16528 |
+
"epoch": 0.46,
|
| 16529 |
+
"learning_rate": 0.0004416271186440678,
|
| 16530 |
+
"loss": 2.2687,
|
| 16531 |
+
"step": 2743
|
| 16532 |
+
},
|
| 16533 |
+
{
|
| 16534 |
+
"epoch": 0.46,
|
| 16535 |
+
"learning_rate": 0.0004414915254237288,
|
| 16536 |
+
"loss": 2.1845,
|
| 16537 |
+
"step": 2744
|
| 16538 |
+
},
|
| 16539 |
+
{
|
| 16540 |
+
"epoch": 0.46,
|
| 16541 |
+
"learning_rate": 0.0004413559322033899,
|
| 16542 |
+
"loss": 2.1781,
|
| 16543 |
+
"step": 2745
|
| 16544 |
+
},
|
| 16545 |
+
{
|
| 16546 |
+
"epoch": 0.46,
|
| 16547 |
+
"learning_rate": 0.00044122033898305087,
|
| 16548 |
+
"loss": 2.1787,
|
| 16549 |
+
"step": 2746
|
| 16550 |
+
},
|
| 16551 |
+
{
|
| 16552 |
+
"epoch": 0.46,
|
| 16553 |
+
"learning_rate": 0.0004410847457627119,
|
| 16554 |
+
"loss": 2.1107,
|
| 16555 |
+
"step": 2747
|
| 16556 |
+
},
|
| 16557 |
+
{
|
| 16558 |
+
"epoch": 0.46,
|
| 16559 |
+
"learning_rate": 0.0004409491525423729,
|
| 16560 |
+
"loss": 2.1027,
|
| 16561 |
+
"step": 2748
|
| 16562 |
+
},
|
| 16563 |
+
{
|
| 16564 |
+
"epoch": 0.46,
|
| 16565 |
+
"learning_rate": 0.0004408135593220339,
|
| 16566 |
+
"loss": 2.2099,
|
| 16567 |
+
"step": 2749
|
| 16568 |
+
},
|
| 16569 |
+
{
|
| 16570 |
+
"epoch": 0.46,
|
| 16571 |
+
"learning_rate": 0.0004406779661016949,
|
| 16572 |
+
"loss": 2.1507,
|
| 16573 |
+
"step": 2750
|
| 16574 |
+
},
|
| 16575 |
+
{
|
| 16576 |
+
"epoch": 0.46,
|
| 16577 |
+
"learning_rate": 0.000440542372881356,
|
| 16578 |
+
"loss": 2.168,
|
| 16579 |
+
"step": 2751
|
| 16580 |
+
},
|
| 16581 |
+
{
|
| 16582 |
+
"epoch": 0.46,
|
| 16583 |
+
"learning_rate": 0.00044040677966101697,
|
| 16584 |
+
"loss": 2.1331,
|
| 16585 |
+
"step": 2752
|
| 16586 |
+
},
|
| 16587 |
+
{
|
| 16588 |
+
"epoch": 0.46,
|
| 16589 |
+
"learning_rate": 0.000440271186440678,
|
| 16590 |
+
"loss": 2.286,
|
| 16591 |
+
"step": 2753
|
| 16592 |
+
},
|
| 16593 |
+
{
|
| 16594 |
+
"epoch": 0.46,
|
| 16595 |
+
"learning_rate": 0.000440135593220339,
|
| 16596 |
+
"loss": 2.1822,
|
| 16597 |
+
"step": 2754
|
| 16598 |
+
},
|
| 16599 |
+
{
|
| 16600 |
+
"epoch": 0.46,
|
| 16601 |
+
"learning_rate": 0.00044000000000000007,
|
| 16602 |
+
"loss": 2.1692,
|
| 16603 |
+
"step": 2755
|
| 16604 |
+
},
|
| 16605 |
+
{
|
| 16606 |
+
"epoch": 0.46,
|
| 16607 |
+
"learning_rate": 0.00043986440677966105,
|
| 16608 |
+
"loss": 2.1917,
|
| 16609 |
+
"step": 2756
|
| 16610 |
+
},
|
| 16611 |
+
{
|
| 16612 |
+
"epoch": 0.46,
|
| 16613 |
+
"learning_rate": 0.0004397288135593221,
|
| 16614 |
+
"loss": 2.1979,
|
| 16615 |
+
"step": 2757
|
| 16616 |
+
},
|
| 16617 |
+
{
|
| 16618 |
+
"epoch": 0.46,
|
| 16619 |
+
"learning_rate": 0.00043959322033898307,
|
| 16620 |
+
"loss": 2.1971,
|
| 16621 |
+
"step": 2758
|
| 16622 |
+
},
|
| 16623 |
+
{
|
| 16624 |
+
"epoch": 0.46,
|
| 16625 |
+
"learning_rate": 0.00043945762711864415,
|
| 16626 |
+
"loss": 2.1208,
|
| 16627 |
+
"step": 2759
|
| 16628 |
+
},
|
| 16629 |
+
{
|
| 16630 |
+
"epoch": 0.46,
|
| 16631 |
+
"learning_rate": 0.0004393220338983051,
|
| 16632 |
+
"loss": 2.2626,
|
| 16633 |
+
"step": 2760
|
| 16634 |
+
},
|
| 16635 |
+
{
|
| 16636 |
+
"epoch": 0.46,
|
| 16637 |
+
"learning_rate": 0.00043918644067796606,
|
| 16638 |
+
"loss": 2.2041,
|
| 16639 |
+
"step": 2761
|
| 16640 |
+
},
|
| 16641 |
+
{
|
| 16642 |
+
"epoch": 0.46,
|
| 16643 |
+
"learning_rate": 0.00043905084745762715,
|
| 16644 |
+
"loss": 2.1519,
|
| 16645 |
+
"step": 2762
|
| 16646 |
+
},
|
| 16647 |
+
{
|
| 16648 |
+
"epoch": 0.46,
|
| 16649 |
+
"learning_rate": 0.00043891525423728813,
|
| 16650 |
+
"loss": 2.2102,
|
| 16651 |
+
"step": 2763
|
| 16652 |
+
},
|
| 16653 |
+
{
|
| 16654 |
+
"epoch": 0.46,
|
| 16655 |
+
"learning_rate": 0.00043877966101694917,
|
| 16656 |
+
"loss": 2.1933,
|
| 16657 |
+
"step": 2764
|
| 16658 |
+
},
|
| 16659 |
+
{
|
| 16660 |
+
"epoch": 0.46,
|
| 16661 |
+
"learning_rate": 0.00043864406779661015,
|
| 16662 |
+
"loss": 2.124,
|
| 16663 |
+
"step": 2765
|
| 16664 |
+
},
|
| 16665 |
+
{
|
| 16666 |
+
"epoch": 0.46,
|
| 16667 |
+
"learning_rate": 0.00043850847457627124,
|
| 16668 |
+
"loss": 2.1205,
|
| 16669 |
+
"step": 2766
|
| 16670 |
+
},
|
| 16671 |
+
{
|
| 16672 |
+
"epoch": 0.46,
|
| 16673 |
+
"learning_rate": 0.0004383728813559322,
|
| 16674 |
+
"loss": 2.306,
|
| 16675 |
+
"step": 2767
|
| 16676 |
+
},
|
| 16677 |
+
{
|
| 16678 |
+
"epoch": 0.46,
|
| 16679 |
+
"learning_rate": 0.00043823728813559325,
|
| 16680 |
+
"loss": 2.196,
|
| 16681 |
+
"step": 2768
|
| 16682 |
+
},
|
| 16683 |
+
{
|
| 16684 |
+
"epoch": 0.46,
|
| 16685 |
+
"learning_rate": 0.00043810169491525423,
|
| 16686 |
+
"loss": 2.2451,
|
| 16687 |
+
"step": 2769
|
| 16688 |
+
},
|
| 16689 |
+
{
|
| 16690 |
+
"epoch": 0.46,
|
| 16691 |
+
"learning_rate": 0.0004379661016949153,
|
| 16692 |
+
"loss": 2.1794,
|
| 16693 |
+
"step": 2770
|
| 16694 |
+
},
|
| 16695 |
+
{
|
| 16696 |
+
"epoch": 0.46,
|
| 16697 |
+
"learning_rate": 0.00043783050847457625,
|
| 16698 |
+
"loss": 2.1634,
|
| 16699 |
+
"step": 2771
|
| 16700 |
+
},
|
| 16701 |
+
{
|
| 16702 |
+
"epoch": 0.46,
|
| 16703 |
+
"learning_rate": 0.00043769491525423734,
|
| 16704 |
+
"loss": 2.1123,
|
| 16705 |
+
"step": 2772
|
| 16706 |
+
},
|
| 16707 |
+
{
|
| 16708 |
+
"epoch": 0.46,
|
| 16709 |
+
"learning_rate": 0.0004375593220338983,
|
| 16710 |
+
"loss": 2.2193,
|
| 16711 |
+
"step": 2773
|
| 16712 |
+
},
|
| 16713 |
+
{
|
| 16714 |
+
"epoch": 0.46,
|
| 16715 |
+
"learning_rate": 0.00043742372881355935,
|
| 16716 |
+
"loss": 2.1491,
|
| 16717 |
+
"step": 2774
|
| 16718 |
+
},
|
| 16719 |
+
{
|
| 16720 |
+
"epoch": 0.46,
|
| 16721 |
+
"learning_rate": 0.00043728813559322033,
|
| 16722 |
+
"loss": 2.2288,
|
| 16723 |
+
"step": 2775
|
| 16724 |
+
},
|
| 16725 |
+
{
|
| 16726 |
+
"epoch": 0.46,
|
| 16727 |
+
"learning_rate": 0.0004371525423728814,
|
| 16728 |
+
"loss": 2.1587,
|
| 16729 |
+
"step": 2776
|
| 16730 |
+
},
|
| 16731 |
+
{
|
| 16732 |
+
"epoch": 0.46,
|
| 16733 |
+
"learning_rate": 0.0004370169491525424,
|
| 16734 |
+
"loss": 2.2145,
|
| 16735 |
+
"step": 2777
|
| 16736 |
+
},
|
| 16737 |
+
{
|
| 16738 |
+
"epoch": 0.46,
|
| 16739 |
+
"learning_rate": 0.00043688135593220344,
|
| 16740 |
+
"loss": 2.1669,
|
| 16741 |
+
"step": 2778
|
| 16742 |
+
},
|
| 16743 |
+
{
|
| 16744 |
+
"epoch": 0.46,
|
| 16745 |
+
"learning_rate": 0.0004367457627118644,
|
| 16746 |
+
"loss": 2.1546,
|
| 16747 |
+
"step": 2779
|
| 16748 |
+
},
|
| 16749 |
+
{
|
| 16750 |
+
"epoch": 0.46,
|
| 16751 |
+
"learning_rate": 0.0004366101694915254,
|
| 16752 |
+
"loss": 2.1785,
|
| 16753 |
+
"step": 2780
|
| 16754 |
+
},
|
| 16755 |
+
{
|
| 16756 |
+
"epoch": 0.46,
|
| 16757 |
+
"learning_rate": 0.0004364745762711865,
|
| 16758 |
+
"loss": 2.1258,
|
| 16759 |
+
"step": 2781
|
| 16760 |
+
},
|
| 16761 |
+
{
|
| 16762 |
+
"epoch": 0.46,
|
| 16763 |
+
"learning_rate": 0.0004363389830508474,
|
| 16764 |
+
"loss": 2.2338,
|
| 16765 |
+
"step": 2782
|
| 16766 |
+
},
|
| 16767 |
+
{
|
| 16768 |
+
"epoch": 0.46,
|
| 16769 |
+
"learning_rate": 0.0004362033898305085,
|
| 16770 |
+
"loss": 2.1971,
|
| 16771 |
+
"step": 2783
|
| 16772 |
+
},
|
| 16773 |
+
{
|
| 16774 |
+
"epoch": 0.46,
|
| 16775 |
+
"learning_rate": 0.0004360677966101695,
|
| 16776 |
+
"loss": 2.1648,
|
| 16777 |
+
"step": 2784
|
| 16778 |
+
},
|
| 16779 |
+
{
|
| 16780 |
+
"epoch": 0.46,
|
| 16781 |
+
"learning_rate": 0.0004359322033898305,
|
| 16782 |
+
"loss": 2.1698,
|
| 16783 |
+
"step": 2785
|
| 16784 |
+
},
|
| 16785 |
+
{
|
| 16786 |
+
"epoch": 0.46,
|
| 16787 |
+
"learning_rate": 0.0004357966101694915,
|
| 16788 |
+
"loss": 2.1884,
|
| 16789 |
+
"step": 2786
|
| 16790 |
+
},
|
| 16791 |
+
{
|
| 16792 |
+
"epoch": 0.46,
|
| 16793 |
+
"learning_rate": 0.0004356610169491526,
|
| 16794 |
+
"loss": 2.1965,
|
| 16795 |
+
"step": 2787
|
| 16796 |
+
},
|
| 16797 |
+
{
|
| 16798 |
+
"epoch": 0.46,
|
| 16799 |
+
"learning_rate": 0.00043552542372881357,
|
| 16800 |
+
"loss": 2.263,
|
| 16801 |
+
"step": 2788
|
| 16802 |
+
},
|
| 16803 |
+
{
|
| 16804 |
+
"epoch": 0.46,
|
| 16805 |
+
"learning_rate": 0.0004353898305084746,
|
| 16806 |
+
"loss": 2.2093,
|
| 16807 |
+
"step": 2789
|
| 16808 |
+
},
|
| 16809 |
+
{
|
| 16810 |
+
"epoch": 0.47,
|
| 16811 |
+
"learning_rate": 0.0004352542372881356,
|
| 16812 |
+
"loss": 2.1458,
|
| 16813 |
+
"step": 2790
|
| 16814 |
+
},
|
| 16815 |
+
{
|
| 16816 |
+
"epoch": 0.47,
|
| 16817 |
+
"learning_rate": 0.00043511864406779667,
|
| 16818 |
+
"loss": 2.1612,
|
| 16819 |
+
"step": 2791
|
| 16820 |
+
},
|
| 16821 |
+
{
|
| 16822 |
+
"epoch": 0.47,
|
| 16823 |
+
"learning_rate": 0.00043498305084745765,
|
| 16824 |
+
"loss": 2.2038,
|
| 16825 |
+
"step": 2792
|
| 16826 |
+
},
|
| 16827 |
+
{
|
| 16828 |
+
"epoch": 0.47,
|
| 16829 |
+
"learning_rate": 0.0004348474576271187,
|
| 16830 |
+
"loss": 2.1496,
|
| 16831 |
+
"step": 2793
|
| 16832 |
+
},
|
| 16833 |
+
{
|
| 16834 |
+
"epoch": 0.47,
|
| 16835 |
+
"learning_rate": 0.00043471186440677967,
|
| 16836 |
+
"loss": 2.1516,
|
| 16837 |
+
"step": 2794
|
| 16838 |
+
},
|
| 16839 |
+
{
|
| 16840 |
+
"epoch": 0.47,
|
| 16841 |
+
"learning_rate": 0.00043457627118644076,
|
| 16842 |
+
"loss": 2.1516,
|
| 16843 |
+
"step": 2795
|
| 16844 |
+
},
|
| 16845 |
+
{
|
| 16846 |
+
"epoch": 0.47,
|
| 16847 |
+
"learning_rate": 0.0004344406779661017,
|
| 16848 |
+
"loss": 2.2392,
|
| 16849 |
+
"step": 2796
|
| 16850 |
+
},
|
| 16851 |
+
{
|
| 16852 |
+
"epoch": 0.47,
|
| 16853 |
+
"learning_rate": 0.00043430508474576277,
|
| 16854 |
+
"loss": 2.1482,
|
| 16855 |
+
"step": 2797
|
| 16856 |
+
},
|
| 16857 |
+
{
|
| 16858 |
+
"epoch": 0.47,
|
| 16859 |
+
"learning_rate": 0.00043416949152542375,
|
| 16860 |
+
"loss": 2.1994,
|
| 16861 |
+
"step": 2798
|
| 16862 |
+
},
|
| 16863 |
+
{
|
| 16864 |
+
"epoch": 0.47,
|
| 16865 |
+
"learning_rate": 0.0004340338983050848,
|
| 16866 |
+
"loss": 2.1625,
|
| 16867 |
+
"step": 2799
|
| 16868 |
+
},
|
| 16869 |
+
{
|
| 16870 |
+
"epoch": 0.47,
|
| 16871 |
+
"learning_rate": 0.00043389830508474577,
|
| 16872 |
+
"loss": 2.1373,
|
| 16873 |
+
"step": 2800
|
| 16874 |
+
},
|
| 16875 |
+
{
|
| 16876 |
+
"epoch": 0.47,
|
| 16877 |
+
"learning_rate": 0.00043376271186440675,
|
| 16878 |
+
"loss": 2.2308,
|
| 16879 |
+
"step": 2801
|
| 16880 |
+
},
|
| 16881 |
+
{
|
| 16882 |
+
"epoch": 0.47,
|
| 16883 |
+
"learning_rate": 0.00043362711864406784,
|
| 16884 |
+
"loss": 2.132,
|
| 16885 |
+
"step": 2802
|
| 16886 |
+
},
|
| 16887 |
+
{
|
| 16888 |
+
"epoch": 0.47,
|
| 16889 |
+
"learning_rate": 0.0004334915254237288,
|
| 16890 |
+
"loss": 2.1553,
|
| 16891 |
+
"step": 2803
|
| 16892 |
+
},
|
| 16893 |
+
{
|
| 16894 |
+
"epoch": 0.47,
|
| 16895 |
+
"learning_rate": 0.00043335593220338985,
|
| 16896 |
+
"loss": 2.1216,
|
| 16897 |
+
"step": 2804
|
| 16898 |
+
},
|
| 16899 |
+
{
|
| 16900 |
+
"epoch": 0.47,
|
| 16901 |
+
"learning_rate": 0.00043322033898305083,
|
| 16902 |
+
"loss": 2.1978,
|
| 16903 |
+
"step": 2805
|
| 16904 |
+
},
|
| 16905 |
+
{
|
| 16906 |
+
"epoch": 0.47,
|
| 16907 |
+
"learning_rate": 0.0004330847457627119,
|
| 16908 |
+
"loss": 2.2046,
|
| 16909 |
+
"step": 2806
|
| 16910 |
+
},
|
| 16911 |
+
{
|
| 16912 |
+
"epoch": 0.47,
|
| 16913 |
+
"learning_rate": 0.00043294915254237285,
|
| 16914 |
+
"loss": 2.1302,
|
| 16915 |
+
"step": 2807
|
| 16916 |
+
},
|
| 16917 |
+
{
|
| 16918 |
+
"epoch": 0.47,
|
| 16919 |
+
"learning_rate": 0.00043281355932203394,
|
| 16920 |
+
"loss": 2.1334,
|
| 16921 |
+
"step": 2808
|
| 16922 |
+
},
|
| 16923 |
+
{
|
| 16924 |
+
"epoch": 0.47,
|
| 16925 |
+
"learning_rate": 0.0004326779661016949,
|
| 16926 |
+
"loss": 2.1931,
|
| 16927 |
+
"step": 2809
|
| 16928 |
+
},
|
| 16929 |
+
{
|
| 16930 |
+
"epoch": 0.47,
|
| 16931 |
+
"learning_rate": 0.00043254237288135595,
|
| 16932 |
+
"loss": 2.0879,
|
| 16933 |
+
"step": 2810
|
| 16934 |
+
},
|
| 16935 |
+
{
|
| 16936 |
+
"epoch": 0.47,
|
| 16937 |
+
"learning_rate": 0.00043240677966101693,
|
| 16938 |
+
"loss": 2.1784,
|
| 16939 |
+
"step": 2811
|
| 16940 |
+
},
|
| 16941 |
+
{
|
| 16942 |
+
"epoch": 0.47,
|
| 16943 |
+
"learning_rate": 0.000432271186440678,
|
| 16944 |
+
"loss": 2.18,
|
| 16945 |
+
"step": 2812
|
| 16946 |
+
},
|
| 16947 |
+
{
|
| 16948 |
+
"epoch": 0.47,
|
| 16949 |
+
"learning_rate": 0.000432135593220339,
|
| 16950 |
+
"loss": 2.1735,
|
| 16951 |
+
"step": 2813
|
| 16952 |
+
},
|
| 16953 |
+
{
|
| 16954 |
+
"epoch": 0.47,
|
| 16955 |
+
"learning_rate": 0.00043200000000000004,
|
| 16956 |
+
"loss": 2.0985,
|
| 16957 |
+
"step": 2814
|
| 16958 |
+
},
|
| 16959 |
+
{
|
| 16960 |
+
"epoch": 0.47,
|
| 16961 |
+
"learning_rate": 0.000431864406779661,
|
| 16962 |
+
"loss": 2.1894,
|
| 16963 |
+
"step": 2815
|
| 16964 |
+
},
|
| 16965 |
+
{
|
| 16966 |
+
"epoch": 0.47,
|
| 16967 |
+
"learning_rate": 0.0004317288135593221,
|
| 16968 |
+
"loss": 2.1497,
|
| 16969 |
+
"step": 2816
|
| 16970 |
+
},
|
| 16971 |
+
{
|
| 16972 |
+
"epoch": 0.47,
|
| 16973 |
+
"learning_rate": 0.0004315932203389831,
|
| 16974 |
+
"loss": 2.1981,
|
| 16975 |
+
"step": 2817
|
| 16976 |
+
},
|
| 16977 |
+
{
|
| 16978 |
+
"epoch": 0.47,
|
| 16979 |
+
"learning_rate": 0.0004314576271186441,
|
| 16980 |
+
"loss": 2.2008,
|
| 16981 |
+
"step": 2818
|
| 16982 |
+
},
|
| 16983 |
+
{
|
| 16984 |
+
"epoch": 0.47,
|
| 16985 |
+
"learning_rate": 0.0004313220338983051,
|
| 16986 |
+
"loss": 2.1571,
|
| 16987 |
+
"step": 2819
|
| 16988 |
+
},
|
| 16989 |
+
{
|
| 16990 |
+
"epoch": 0.47,
|
| 16991 |
+
"learning_rate": 0.00043118644067796614,
|
| 16992 |
+
"loss": 2.2743,
|
| 16993 |
+
"step": 2820
|
| 16994 |
+
},
|
| 16995 |
+
{
|
| 16996 |
+
"epoch": 0.47,
|
| 16997 |
+
"learning_rate": 0.0004310508474576271,
|
| 16998 |
+
"loss": 2.1833,
|
| 16999 |
+
"step": 2821
|
| 17000 |
+
},
|
| 17001 |
+
{
|
| 17002 |
+
"epoch": 0.47,
|
| 17003 |
+
"learning_rate": 0.0004309152542372881,
|
| 17004 |
+
"loss": 2.1796,
|
| 17005 |
+
"step": 2822
|
| 17006 |
+
},
|
| 17007 |
+
{
|
| 17008 |
+
"epoch": 0.47,
|
| 17009 |
+
"learning_rate": 0.0004307796610169492,
|
| 17010 |
+
"loss": 2.2698,
|
| 17011 |
+
"step": 2823
|
| 17012 |
+
},
|
| 17013 |
+
{
|
| 17014 |
+
"epoch": 0.47,
|
| 17015 |
+
"learning_rate": 0.00043064406779661017,
|
| 17016 |
+
"loss": 2.2112,
|
| 17017 |
+
"step": 2824
|
| 17018 |
+
},
|
| 17019 |
+
{
|
| 17020 |
+
"epoch": 0.47,
|
| 17021 |
+
"learning_rate": 0.0004305084745762712,
|
| 17022 |
+
"loss": 2.1944,
|
| 17023 |
+
"step": 2825
|
| 17024 |
+
},
|
| 17025 |
+
{
|
| 17026 |
+
"epoch": 0.47,
|
| 17027 |
+
"learning_rate": 0.0004303728813559322,
|
| 17028 |
+
"loss": 2.1812,
|
| 17029 |
+
"step": 2826
|
| 17030 |
+
},
|
| 17031 |
+
{
|
| 17032 |
+
"epoch": 0.47,
|
| 17033 |
+
"learning_rate": 0.0004302372881355933,
|
| 17034 |
+
"loss": 2.2353,
|
| 17035 |
+
"step": 2827
|
| 17036 |
+
},
|
| 17037 |
+
{
|
| 17038 |
+
"epoch": 0.47,
|
| 17039 |
+
"learning_rate": 0.00043010169491525425,
|
| 17040 |
+
"loss": 2.2165,
|
| 17041 |
+
"step": 2828
|
| 17042 |
+
},
|
| 17043 |
+
{
|
| 17044 |
+
"epoch": 0.47,
|
| 17045 |
+
"learning_rate": 0.0004299661016949153,
|
| 17046 |
+
"loss": 2.1878,
|
| 17047 |
+
"step": 2829
|
| 17048 |
+
},
|
| 17049 |
+
{
|
| 17050 |
+
"epoch": 0.47,
|
| 17051 |
+
"learning_rate": 0.00042983050847457627,
|
| 17052 |
+
"loss": 2.1921,
|
| 17053 |
+
"step": 2830
|
| 17054 |
+
},
|
| 17055 |
+
{
|
| 17056 |
+
"epoch": 0.47,
|
| 17057 |
+
"learning_rate": 0.0004296949152542373,
|
| 17058 |
+
"loss": 2.1842,
|
| 17059 |
+
"step": 2831
|
| 17060 |
+
},
|
| 17061 |
+
{
|
| 17062 |
+
"epoch": 0.47,
|
| 17063 |
+
"learning_rate": 0.0004295593220338983,
|
| 17064 |
+
"loss": 2.2485,
|
| 17065 |
+
"step": 2832
|
| 17066 |
+
},
|
| 17067 |
+
{
|
| 17068 |
+
"epoch": 0.47,
|
| 17069 |
+
"learning_rate": 0.0004294237288135594,
|
| 17070 |
+
"loss": 2.1801,
|
| 17071 |
+
"step": 2833
|
| 17072 |
+
},
|
| 17073 |
+
{
|
| 17074 |
+
"epoch": 0.47,
|
| 17075 |
+
"learning_rate": 0.00042928813559322035,
|
| 17076 |
+
"loss": 2.2013,
|
| 17077 |
+
"step": 2834
|
| 17078 |
+
},
|
| 17079 |
+
{
|
| 17080 |
+
"epoch": 0.47,
|
| 17081 |
+
"learning_rate": 0.0004291525423728814,
|
| 17082 |
+
"loss": 2.2313,
|
| 17083 |
+
"step": 2835
|
| 17084 |
+
},
|
| 17085 |
+
{
|
| 17086 |
+
"epoch": 0.47,
|
| 17087 |
+
"learning_rate": 0.00042901694915254237,
|
| 17088 |
+
"loss": 2.1955,
|
| 17089 |
+
"step": 2836
|
| 17090 |
+
},
|
| 17091 |
+
{
|
| 17092 |
+
"epoch": 0.47,
|
| 17093 |
+
"learning_rate": 0.00042888135593220346,
|
| 17094 |
+
"loss": 2.1234,
|
| 17095 |
+
"step": 2837
|
| 17096 |
+
},
|
| 17097 |
+
{
|
| 17098 |
+
"epoch": 0.47,
|
| 17099 |
+
"learning_rate": 0.00042874576271186444,
|
| 17100 |
+
"loss": 2.1559,
|
| 17101 |
+
"step": 2838
|
| 17102 |
+
},
|
| 17103 |
+
{
|
| 17104 |
+
"epoch": 0.47,
|
| 17105 |
+
"learning_rate": 0.0004286101694915255,
|
| 17106 |
+
"loss": 2.1444,
|
| 17107 |
+
"step": 2839
|
| 17108 |
+
},
|
| 17109 |
+
{
|
| 17110 |
+
"epoch": 0.47,
|
| 17111 |
+
"learning_rate": 0.00042847457627118645,
|
| 17112 |
+
"loss": 2.1511,
|
| 17113 |
+
"step": 2840
|
| 17114 |
+
},
|
| 17115 |
+
{
|
| 17116 |
+
"epoch": 0.47,
|
| 17117 |
+
"learning_rate": 0.00042833898305084754,
|
| 17118 |
+
"loss": 2.1536,
|
| 17119 |
+
"step": 2841
|
| 17120 |
+
},
|
| 17121 |
+
{
|
| 17122 |
+
"epoch": 0.47,
|
| 17123 |
+
"learning_rate": 0.00042820338983050847,
|
| 17124 |
+
"loss": 2.2474,
|
| 17125 |
+
"step": 2842
|
| 17126 |
+
},
|
| 17127 |
+
{
|
| 17128 |
+
"epoch": 0.47,
|
| 17129 |
+
"learning_rate": 0.00042806779661016945,
|
| 17130 |
+
"loss": 2.1643,
|
| 17131 |
+
"step": 2843
|
| 17132 |
+
},
|
| 17133 |
+
{
|
| 17134 |
+
"epoch": 0.47,
|
| 17135 |
+
"learning_rate": 0.00042793220338983054,
|
| 17136 |
+
"loss": 2.1671,
|
| 17137 |
+
"step": 2844
|
| 17138 |
+
},
|
| 17139 |
+
{
|
| 17140 |
+
"epoch": 0.47,
|
| 17141 |
+
"learning_rate": 0.0004277966101694915,
|
| 17142 |
+
"loss": 2.1588,
|
| 17143 |
+
"step": 2845
|
| 17144 |
+
},
|
| 17145 |
+
{
|
| 17146 |
+
"epoch": 0.47,
|
| 17147 |
+
"learning_rate": 0.00042766101694915255,
|
| 17148 |
+
"loss": 2.1799,
|
| 17149 |
+
"step": 2846
|
| 17150 |
+
},
|
| 17151 |
+
{
|
| 17152 |
+
"epoch": 0.47,
|
| 17153 |
+
"learning_rate": 0.00042752542372881354,
|
| 17154 |
+
"loss": 2.1565,
|
| 17155 |
+
"step": 2847
|
| 17156 |
+
},
|
| 17157 |
+
{
|
| 17158 |
+
"epoch": 0.47,
|
| 17159 |
+
"learning_rate": 0.0004273898305084746,
|
| 17160 |
+
"loss": 2.1751,
|
| 17161 |
+
"step": 2848
|
| 17162 |
+
},
|
| 17163 |
+
{
|
| 17164 |
+
"epoch": 0.47,
|
| 17165 |
+
"learning_rate": 0.0004272542372881356,
|
| 17166 |
+
"loss": 2.1091,
|
| 17167 |
+
"step": 2849
|
| 17168 |
+
},
|
| 17169 |
+
{
|
| 17170 |
+
"epoch": 0.47,
|
| 17171 |
+
"learning_rate": 0.00042711864406779664,
|
| 17172 |
+
"loss": 2.2063,
|
| 17173 |
+
"step": 2850
|
| 17174 |
+
},
|
| 17175 |
+
{
|
| 17176 |
+
"epoch": 0.48,
|
| 17177 |
+
"learning_rate": 0.0004269830508474576,
|
| 17178 |
+
"loss": 2.0999,
|
| 17179 |
+
"step": 2851
|
| 17180 |
+
},
|
| 17181 |
+
{
|
| 17182 |
+
"epoch": 0.48,
|
| 17183 |
+
"learning_rate": 0.0004268474576271187,
|
| 17184 |
+
"loss": 2.1041,
|
| 17185 |
+
"step": 2852
|
| 17186 |
+
},
|
| 17187 |
+
{
|
| 17188 |
+
"epoch": 0.48,
|
| 17189 |
+
"learning_rate": 0.00042671186440677964,
|
| 17190 |
+
"loss": 2.1839,
|
| 17191 |
+
"step": 2853
|
| 17192 |
+
},
|
| 17193 |
+
{
|
| 17194 |
+
"epoch": 0.48,
|
| 17195 |
+
"learning_rate": 0.0004265762711864407,
|
| 17196 |
+
"loss": 2.0895,
|
| 17197 |
+
"step": 2854
|
| 17198 |
+
},
|
| 17199 |
+
{
|
| 17200 |
+
"epoch": 0.48,
|
| 17201 |
+
"learning_rate": 0.0004264406779661017,
|
| 17202 |
+
"loss": 2.1978,
|
| 17203 |
+
"step": 2855
|
| 17204 |
+
},
|
| 17205 |
+
{
|
| 17206 |
+
"epoch": 0.48,
|
| 17207 |
+
"learning_rate": 0.00042630508474576274,
|
| 17208 |
+
"loss": 2.1899,
|
| 17209 |
+
"step": 2856
|
| 17210 |
+
},
|
| 17211 |
+
{
|
| 17212 |
+
"epoch": 0.48,
|
| 17213 |
+
"learning_rate": 0.0004261694915254237,
|
| 17214 |
+
"loss": 2.1461,
|
| 17215 |
+
"step": 2857
|
| 17216 |
+
},
|
| 17217 |
+
{
|
| 17218 |
+
"epoch": 0.48,
|
| 17219 |
+
"learning_rate": 0.0004260338983050848,
|
| 17220 |
+
"loss": 2.2126,
|
| 17221 |
+
"step": 2858
|
| 17222 |
+
},
|
| 17223 |
+
{
|
| 17224 |
+
"epoch": 0.48,
|
| 17225 |
+
"learning_rate": 0.0004258983050847458,
|
| 17226 |
+
"loss": 2.1616,
|
| 17227 |
+
"step": 2859
|
| 17228 |
+
},
|
| 17229 |
+
{
|
| 17230 |
+
"epoch": 0.48,
|
| 17231 |
+
"learning_rate": 0.0004257627118644068,
|
| 17232 |
+
"loss": 2.1526,
|
| 17233 |
+
"step": 2860
|
| 17234 |
+
},
|
| 17235 |
+
{
|
| 17236 |
+
"epoch": 0.48,
|
| 17237 |
+
"learning_rate": 0.0004256271186440678,
|
| 17238 |
+
"loss": 2.143,
|
| 17239 |
+
"step": 2861
|
| 17240 |
+
},
|
| 17241 |
+
{
|
| 17242 |
+
"epoch": 0.48,
|
| 17243 |
+
"learning_rate": 0.0004254915254237289,
|
| 17244 |
+
"loss": 2.1868,
|
| 17245 |
+
"step": 2862
|
| 17246 |
+
},
|
| 17247 |
+
{
|
| 17248 |
+
"epoch": 0.48,
|
| 17249 |
+
"learning_rate": 0.0004253559322033899,
|
| 17250 |
+
"loss": 2.0951,
|
| 17251 |
+
"step": 2863
|
| 17252 |
+
},
|
| 17253 |
+
{
|
| 17254 |
+
"epoch": 0.48,
|
| 17255 |
+
"learning_rate": 0.0004252203389830508,
|
| 17256 |
+
"loss": 2.206,
|
| 17257 |
+
"step": 2864
|
| 17258 |
+
},
|
| 17259 |
+
{
|
| 17260 |
+
"epoch": 0.48,
|
| 17261 |
+
"learning_rate": 0.0004250847457627119,
|
| 17262 |
+
"loss": 2.1904,
|
| 17263 |
+
"step": 2865
|
| 17264 |
+
},
|
| 17265 |
+
{
|
| 17266 |
+
"epoch": 0.48,
|
| 17267 |
+
"learning_rate": 0.00042494915254237287,
|
| 17268 |
+
"loss": 2.2698,
|
| 17269 |
+
"step": 2866
|
| 17270 |
+
},
|
| 17271 |
+
{
|
| 17272 |
+
"epoch": 0.48,
|
| 17273 |
+
"learning_rate": 0.0004248135593220339,
|
| 17274 |
+
"loss": 2.1582,
|
| 17275 |
+
"step": 2867
|
| 17276 |
+
},
|
| 17277 |
+
{
|
| 17278 |
+
"epoch": 0.48,
|
| 17279 |
+
"learning_rate": 0.0004246779661016949,
|
| 17280 |
+
"loss": 2.1616,
|
| 17281 |
+
"step": 2868
|
| 17282 |
+
},
|
| 17283 |
+
{
|
| 17284 |
+
"epoch": 0.48,
|
| 17285 |
+
"learning_rate": 0.000424542372881356,
|
| 17286 |
+
"loss": 2.2028,
|
| 17287 |
+
"step": 2869
|
| 17288 |
+
},
|
| 17289 |
+
{
|
| 17290 |
+
"epoch": 0.48,
|
| 17291 |
+
"learning_rate": 0.00042440677966101696,
|
| 17292 |
+
"loss": 2.2078,
|
| 17293 |
+
"step": 2870
|
| 17294 |
+
},
|
| 17295 |
+
{
|
| 17296 |
+
"epoch": 0.48,
|
| 17297 |
+
"learning_rate": 0.000424271186440678,
|
| 17298 |
+
"loss": 2.1194,
|
| 17299 |
+
"step": 2871
|
| 17300 |
+
},
|
| 17301 |
+
{
|
| 17302 |
+
"epoch": 0.48,
|
| 17303 |
+
"learning_rate": 0.00042413559322033897,
|
| 17304 |
+
"loss": 2.1585,
|
| 17305 |
+
"step": 2872
|
| 17306 |
+
},
|
| 17307 |
+
{
|
| 17308 |
+
"epoch": 0.48,
|
| 17309 |
+
"learning_rate": 0.00042400000000000006,
|
| 17310 |
+
"loss": 2.1722,
|
| 17311 |
+
"step": 2873
|
| 17312 |
+
},
|
| 17313 |
+
{
|
| 17314 |
+
"epoch": 0.48,
|
| 17315 |
+
"learning_rate": 0.00042386440677966104,
|
| 17316 |
+
"loss": 2.1138,
|
| 17317 |
+
"step": 2874
|
| 17318 |
+
},
|
| 17319 |
+
{
|
| 17320 |
+
"epoch": 0.48,
|
| 17321 |
+
"learning_rate": 0.0004237288135593221,
|
| 17322 |
+
"loss": 2.118,
|
| 17323 |
+
"step": 2875
|
| 17324 |
+
},
|
| 17325 |
+
{
|
| 17326 |
+
"epoch": 0.48,
|
| 17327 |
+
"learning_rate": 0.00042359322033898306,
|
| 17328 |
+
"loss": 2.1648,
|
| 17329 |
+
"step": 2876
|
| 17330 |
+
},
|
| 17331 |
+
{
|
| 17332 |
+
"epoch": 0.48,
|
| 17333 |
+
"learning_rate": 0.0004234576271186441,
|
| 17334 |
+
"loss": 2.1642,
|
| 17335 |
+
"step": 2877
|
| 17336 |
+
},
|
| 17337 |
+
{
|
| 17338 |
+
"epoch": 0.48,
|
| 17339 |
+
"learning_rate": 0.00042332203389830507,
|
| 17340 |
+
"loss": 2.1489,
|
| 17341 |
+
"step": 2878
|
| 17342 |
+
},
|
| 17343 |
+
{
|
| 17344 |
+
"epoch": 0.48,
|
| 17345 |
+
"learning_rate": 0.00042318644067796616,
|
| 17346 |
+
"loss": 2.1387,
|
| 17347 |
+
"step": 2879
|
| 17348 |
+
},
|
| 17349 |
+
{
|
| 17350 |
+
"epoch": 0.48,
|
| 17351 |
+
"learning_rate": 0.00042305084745762714,
|
| 17352 |
+
"loss": 2.1737,
|
| 17353 |
+
"step": 2880
|
| 17354 |
+
},
|
| 17355 |
+
{
|
| 17356 |
+
"epoch": 0.48,
|
| 17357 |
+
"learning_rate": 0.0004229152542372882,
|
| 17358 |
+
"loss": 2.1702,
|
| 17359 |
+
"step": 2881
|
| 17360 |
+
},
|
| 17361 |
+
{
|
| 17362 |
+
"epoch": 0.48,
|
| 17363 |
+
"learning_rate": 0.00042277966101694916,
|
| 17364 |
+
"loss": 2.0842,
|
| 17365 |
+
"step": 2882
|
| 17366 |
+
},
|
| 17367 |
+
{
|
| 17368 |
+
"epoch": 0.48,
|
| 17369 |
+
"learning_rate": 0.00042264406779661025,
|
| 17370 |
+
"loss": 2.1522,
|
| 17371 |
+
"step": 2883
|
| 17372 |
+
},
|
| 17373 |
+
{
|
| 17374 |
+
"epoch": 0.48,
|
| 17375 |
+
"learning_rate": 0.0004225084745762712,
|
| 17376 |
+
"loss": 2.2546,
|
| 17377 |
+
"step": 2884
|
| 17378 |
+
},
|
| 17379 |
+
{
|
| 17380 |
+
"epoch": 0.48,
|
| 17381 |
+
"learning_rate": 0.0004223728813559322,
|
| 17382 |
+
"loss": 2.1756,
|
| 17383 |
+
"step": 2885
|
| 17384 |
+
},
|
| 17385 |
+
{
|
| 17386 |
+
"epoch": 0.48,
|
| 17387 |
+
"learning_rate": 0.00042223728813559324,
|
| 17388 |
+
"loss": 2.2049,
|
| 17389 |
+
"step": 2886
|
| 17390 |
+
},
|
| 17391 |
+
{
|
| 17392 |
+
"epoch": 0.48,
|
| 17393 |
+
"learning_rate": 0.0004221016949152542,
|
| 17394 |
+
"loss": 2.1915,
|
| 17395 |
+
"step": 2887
|
| 17396 |
+
},
|
| 17397 |
+
{
|
| 17398 |
+
"epoch": 0.48,
|
| 17399 |
+
"learning_rate": 0.00042196610169491526,
|
| 17400 |
+
"loss": 2.2158,
|
| 17401 |
+
"step": 2888
|
| 17402 |
+
},
|
| 17403 |
+
{
|
| 17404 |
+
"epoch": 0.48,
|
| 17405 |
+
"learning_rate": 0.00042183050847457624,
|
| 17406 |
+
"loss": 2.1971,
|
| 17407 |
+
"step": 2889
|
| 17408 |
+
},
|
| 17409 |
+
{
|
| 17410 |
+
"epoch": 0.48,
|
| 17411 |
+
"learning_rate": 0.0004216949152542373,
|
| 17412 |
+
"loss": 2.1891,
|
| 17413 |
+
"step": 2890
|
| 17414 |
+
},
|
| 17415 |
+
{
|
| 17416 |
+
"epoch": 0.48,
|
| 17417 |
+
"learning_rate": 0.0004215593220338983,
|
| 17418 |
+
"loss": 2.1599,
|
| 17419 |
+
"step": 2891
|
| 17420 |
+
},
|
| 17421 |
+
{
|
| 17422 |
+
"epoch": 0.48,
|
| 17423 |
+
"learning_rate": 0.00042142372881355934,
|
| 17424 |
+
"loss": 2.2192,
|
| 17425 |
+
"step": 2892
|
| 17426 |
+
},
|
| 17427 |
+
{
|
| 17428 |
+
"epoch": 0.48,
|
| 17429 |
+
"learning_rate": 0.0004212881355932203,
|
| 17430 |
+
"loss": 2.1888,
|
| 17431 |
+
"step": 2893
|
| 17432 |
+
},
|
| 17433 |
+
{
|
| 17434 |
+
"epoch": 0.48,
|
| 17435 |
+
"learning_rate": 0.0004211525423728814,
|
| 17436 |
+
"loss": 2.1549,
|
| 17437 |
+
"step": 2894
|
| 17438 |
+
},
|
| 17439 |
+
{
|
| 17440 |
+
"epoch": 0.48,
|
| 17441 |
+
"learning_rate": 0.0004210169491525424,
|
| 17442 |
+
"loss": 2.0905,
|
| 17443 |
+
"step": 2895
|
| 17444 |
+
},
|
| 17445 |
+
{
|
| 17446 |
+
"epoch": 0.48,
|
| 17447 |
+
"learning_rate": 0.0004208813559322034,
|
| 17448 |
+
"loss": 2.1428,
|
| 17449 |
+
"step": 2896
|
| 17450 |
+
},
|
| 17451 |
+
{
|
| 17452 |
+
"epoch": 0.48,
|
| 17453 |
+
"learning_rate": 0.0004207457627118644,
|
| 17454 |
+
"loss": 2.0789,
|
| 17455 |
+
"step": 2897
|
| 17456 |
+
},
|
| 17457 |
+
{
|
| 17458 |
+
"epoch": 0.48,
|
| 17459 |
+
"learning_rate": 0.0004206101694915255,
|
| 17460 |
+
"loss": 2.2274,
|
| 17461 |
+
"step": 2898
|
| 17462 |
+
},
|
| 17463 |
+
{
|
| 17464 |
+
"epoch": 0.48,
|
| 17465 |
+
"learning_rate": 0.0004204745762711864,
|
| 17466 |
+
"loss": 2.2216,
|
| 17467 |
+
"step": 2899
|
| 17468 |
+
},
|
| 17469 |
+
{
|
| 17470 |
+
"epoch": 0.48,
|
| 17471 |
+
"learning_rate": 0.0004203389830508475,
|
| 17472 |
+
"loss": 2.1846,
|
| 17473 |
+
"step": 2900
|
| 17474 |
+
},
|
| 17475 |
+
{
|
| 17476 |
+
"epoch": 0.48,
|
| 17477 |
+
"learning_rate": 0.0004202033898305085,
|
| 17478 |
+
"loss": 2.2441,
|
| 17479 |
+
"step": 2901
|
| 17480 |
+
},
|
| 17481 |
+
{
|
| 17482 |
+
"epoch": 0.48,
|
| 17483 |
+
"learning_rate": 0.00042006779661016953,
|
| 17484 |
+
"loss": 2.1295,
|
| 17485 |
+
"step": 2902
|
| 17486 |
+
},
|
| 17487 |
+
{
|
| 17488 |
+
"epoch": 0.48,
|
| 17489 |
+
"learning_rate": 0.0004199322033898305,
|
| 17490 |
+
"loss": 2.2114,
|
| 17491 |
+
"step": 2903
|
| 17492 |
+
},
|
| 17493 |
+
{
|
| 17494 |
+
"epoch": 0.48,
|
| 17495 |
+
"learning_rate": 0.0004197966101694916,
|
| 17496 |
+
"loss": 2.1641,
|
| 17497 |
+
"step": 2904
|
| 17498 |
+
},
|
| 17499 |
+
{
|
| 17500 |
+
"epoch": 0.48,
|
| 17501 |
+
"learning_rate": 0.0004196610169491526,
|
| 17502 |
+
"loss": 2.1816,
|
| 17503 |
+
"step": 2905
|
| 17504 |
+
},
|
| 17505 |
+
{
|
| 17506 |
+
"epoch": 0.48,
|
| 17507 |
+
"learning_rate": 0.00041952542372881356,
|
| 17508 |
+
"loss": 2.2033,
|
| 17509 |
+
"step": 2906
|
| 17510 |
+
},
|
| 17511 |
+
{
|
| 17512 |
+
"epoch": 0.48,
|
| 17513 |
+
"learning_rate": 0.0004193898305084746,
|
| 17514 |
+
"loss": 2.1715,
|
| 17515 |
+
"step": 2907
|
| 17516 |
+
},
|
| 17517 |
+
{
|
| 17518 |
+
"epoch": 0.48,
|
| 17519 |
+
"learning_rate": 0.0004192542372881356,
|
| 17520 |
+
"loss": 2.1575,
|
| 17521 |
+
"step": 2908
|
| 17522 |
+
},
|
| 17523 |
+
{
|
| 17524 |
+
"epoch": 0.48,
|
| 17525 |
+
"learning_rate": 0.00041911864406779666,
|
| 17526 |
+
"loss": 2.2248,
|
| 17527 |
+
"step": 2909
|
| 17528 |
+
},
|
| 17529 |
+
{
|
| 17530 |
+
"epoch": 0.48,
|
| 17531 |
+
"learning_rate": 0.0004189830508474576,
|
| 17532 |
+
"loss": 2.281,
|
| 17533 |
+
"step": 2910
|
| 17534 |
+
},
|
| 17535 |
+
{
|
| 17536 |
+
"epoch": 0.49,
|
| 17537 |
+
"learning_rate": 0.0004188474576271187,
|
| 17538 |
+
"loss": 2.2173,
|
| 17539 |
+
"step": 2911
|
| 17540 |
+
},
|
| 17541 |
+
{
|
| 17542 |
+
"epoch": 0.49,
|
| 17543 |
+
"learning_rate": 0.00041871186440677966,
|
| 17544 |
+
"loss": 2.1685,
|
| 17545 |
+
"step": 2912
|
| 17546 |
+
},
|
| 17547 |
+
{
|
| 17548 |
+
"epoch": 0.49,
|
| 17549 |
+
"learning_rate": 0.0004185762711864407,
|
| 17550 |
+
"loss": 2.1741,
|
| 17551 |
+
"step": 2913
|
| 17552 |
+
},
|
| 17553 |
+
{
|
| 17554 |
+
"epoch": 0.49,
|
| 17555 |
+
"learning_rate": 0.0004184406779661017,
|
| 17556 |
+
"loss": 2.1727,
|
| 17557 |
+
"step": 2914
|
| 17558 |
+
},
|
| 17559 |
+
{
|
| 17560 |
+
"epoch": 0.49,
|
| 17561 |
+
"learning_rate": 0.00041830508474576276,
|
| 17562 |
+
"loss": 2.1974,
|
| 17563 |
+
"step": 2915
|
| 17564 |
+
},
|
| 17565 |
+
{
|
| 17566 |
+
"epoch": 0.49,
|
| 17567 |
+
"learning_rate": 0.00041816949152542374,
|
| 17568 |
+
"loss": 2.1526,
|
| 17569 |
+
"step": 2916
|
| 17570 |
+
},
|
| 17571 |
+
{
|
| 17572 |
+
"epoch": 0.49,
|
| 17573 |
+
"learning_rate": 0.0004180338983050848,
|
| 17574 |
+
"loss": 2.2312,
|
| 17575 |
+
"step": 2917
|
| 17576 |
+
},
|
| 17577 |
+
{
|
| 17578 |
+
"epoch": 0.49,
|
| 17579 |
+
"learning_rate": 0.00041789830508474576,
|
| 17580 |
+
"loss": 2.1809,
|
| 17581 |
+
"step": 2918
|
| 17582 |
+
},
|
| 17583 |
+
{
|
| 17584 |
+
"epoch": 0.49,
|
| 17585 |
+
"learning_rate": 0.00041776271186440685,
|
| 17586 |
+
"loss": 2.2041,
|
| 17587 |
+
"step": 2919
|
| 17588 |
+
},
|
| 17589 |
+
{
|
| 17590 |
+
"epoch": 0.49,
|
| 17591 |
+
"learning_rate": 0.00041762711864406783,
|
| 17592 |
+
"loss": 2.1479,
|
| 17593 |
+
"step": 2920
|
| 17594 |
+
},
|
| 17595 |
+
{
|
| 17596 |
+
"epoch": 0.49,
|
| 17597 |
+
"learning_rate": 0.00041749152542372886,
|
| 17598 |
+
"loss": 2.1227,
|
| 17599 |
+
"step": 2921
|
| 17600 |
+
},
|
| 17601 |
+
{
|
| 17602 |
+
"epoch": 0.49,
|
| 17603 |
+
"learning_rate": 0.00041735593220338984,
|
| 17604 |
+
"loss": 2.2373,
|
| 17605 |
+
"step": 2922
|
| 17606 |
+
},
|
| 17607 |
+
{
|
| 17608 |
+
"epoch": 0.49,
|
| 17609 |
+
"learning_rate": 0.00041722033898305093,
|
| 17610 |
+
"loss": 2.1932,
|
| 17611 |
+
"step": 2923
|
| 17612 |
+
},
|
| 17613 |
+
{
|
| 17614 |
+
"epoch": 0.49,
|
| 17615 |
+
"learning_rate": 0.00041708474576271186,
|
| 17616 |
+
"loss": 2.1475,
|
| 17617 |
+
"step": 2924
|
| 17618 |
+
},
|
| 17619 |
+
{
|
| 17620 |
+
"epoch": 0.49,
|
| 17621 |
+
"learning_rate": 0.00041694915254237295,
|
| 17622 |
+
"loss": 2.229,
|
| 17623 |
+
"step": 2925
|
| 17624 |
+
},
|
| 17625 |
+
{
|
| 17626 |
+
"epoch": 0.49,
|
| 17627 |
+
"learning_rate": 0.00041681355932203393,
|
| 17628 |
+
"loss": 2.2022,
|
| 17629 |
+
"step": 2926
|
| 17630 |
+
},
|
| 17631 |
+
{
|
| 17632 |
+
"epoch": 0.49,
|
| 17633 |
+
"learning_rate": 0.0004166779661016949,
|
| 17634 |
+
"loss": 2.1027,
|
| 17635 |
+
"step": 2927
|
| 17636 |
+
},
|
| 17637 |
+
{
|
| 17638 |
+
"epoch": 0.49,
|
| 17639 |
+
"learning_rate": 0.00041654237288135594,
|
| 17640 |
+
"loss": 2.1726,
|
| 17641 |
+
"step": 2928
|
| 17642 |
+
},
|
| 17643 |
+
{
|
| 17644 |
+
"epoch": 0.49,
|
| 17645 |
+
"learning_rate": 0.0004164067796610169,
|
| 17646 |
+
"loss": 2.1762,
|
| 17647 |
+
"step": 2929
|
| 17648 |
+
},
|
| 17649 |
+
{
|
| 17650 |
+
"epoch": 0.49,
|
| 17651 |
+
"learning_rate": 0.000416271186440678,
|
| 17652 |
+
"loss": 2.1976,
|
| 17653 |
+
"step": 2930
|
| 17654 |
+
},
|
| 17655 |
+
{
|
| 17656 |
+
"epoch": 0.49,
|
| 17657 |
+
"learning_rate": 0.000416135593220339,
|
| 17658 |
+
"loss": 2.1517,
|
| 17659 |
+
"step": 2931
|
| 17660 |
+
},
|
| 17661 |
+
{
|
| 17662 |
+
"epoch": 0.49,
|
| 17663 |
+
"learning_rate": 0.00041600000000000003,
|
| 17664 |
+
"loss": 2.1961,
|
| 17665 |
+
"step": 2932
|
| 17666 |
+
},
|
| 17667 |
+
{
|
| 17668 |
+
"epoch": 0.49,
|
| 17669 |
+
"learning_rate": 0.000415864406779661,
|
| 17670 |
+
"loss": 2.1523,
|
| 17671 |
+
"step": 2933
|
| 17672 |
+
},
|
| 17673 |
+
{
|
| 17674 |
+
"epoch": 0.49,
|
| 17675 |
+
"learning_rate": 0.0004157288135593221,
|
| 17676 |
+
"loss": 2.1589,
|
| 17677 |
+
"step": 2934
|
| 17678 |
+
},
|
| 17679 |
+
{
|
| 17680 |
+
"epoch": 0.49,
|
| 17681 |
+
"learning_rate": 0.000415593220338983,
|
| 17682 |
+
"loss": 2.2447,
|
| 17683 |
+
"step": 2935
|
| 17684 |
+
},
|
| 17685 |
+
{
|
| 17686 |
+
"epoch": 0.49,
|
| 17687 |
+
"learning_rate": 0.0004154576271186441,
|
| 17688 |
+
"loss": 2.1304,
|
| 17689 |
+
"step": 2936
|
| 17690 |
+
},
|
| 17691 |
+
{
|
| 17692 |
+
"epoch": 0.49,
|
| 17693 |
+
"learning_rate": 0.0004153220338983051,
|
| 17694 |
+
"loss": 2.137,
|
| 17695 |
+
"step": 2937
|
| 17696 |
+
},
|
| 17697 |
+
{
|
| 17698 |
+
"epoch": 0.49,
|
| 17699 |
+
"learning_rate": 0.00041518644067796613,
|
| 17700 |
+
"loss": 2.1495,
|
| 17701 |
+
"step": 2938
|
| 17702 |
+
},
|
| 17703 |
+
{
|
| 17704 |
+
"epoch": 0.49,
|
| 17705 |
+
"learning_rate": 0.0004150508474576271,
|
| 17706 |
+
"loss": 2.1051,
|
| 17707 |
+
"step": 2939
|
| 17708 |
+
},
|
| 17709 |
+
{
|
| 17710 |
+
"epoch": 0.49,
|
| 17711 |
+
"learning_rate": 0.0004149152542372882,
|
| 17712 |
+
"loss": 2.1827,
|
| 17713 |
+
"step": 2940
|
| 17714 |
+
},
|
| 17715 |
+
{
|
| 17716 |
+
"epoch": 0.49,
|
| 17717 |
+
"learning_rate": 0.0004147796610169492,
|
| 17718 |
+
"loss": 2.2329,
|
| 17719 |
+
"step": 2941
|
| 17720 |
+
},
|
| 17721 |
+
{
|
| 17722 |
+
"epoch": 0.49,
|
| 17723 |
+
"learning_rate": 0.0004146440677966102,
|
| 17724 |
+
"loss": 2.1234,
|
| 17725 |
+
"step": 2942
|
| 17726 |
+
},
|
| 17727 |
+
{
|
| 17728 |
+
"epoch": 0.49,
|
| 17729 |
+
"learning_rate": 0.0004145084745762712,
|
| 17730 |
+
"loss": 2.1709,
|
| 17731 |
+
"step": 2943
|
| 17732 |
+
},
|
| 17733 |
+
{
|
| 17734 |
+
"epoch": 0.49,
|
| 17735 |
+
"learning_rate": 0.0004143728813559323,
|
| 17736 |
+
"loss": 2.1335,
|
| 17737 |
+
"step": 2944
|
| 17738 |
+
},
|
| 17739 |
+
{
|
| 17740 |
+
"epoch": 0.49,
|
| 17741 |
+
"learning_rate": 0.00041423728813559326,
|
| 17742 |
+
"loss": 2.1752,
|
| 17743 |
+
"step": 2945
|
| 17744 |
+
},
|
| 17745 |
+
{
|
| 17746 |
+
"epoch": 0.49,
|
| 17747 |
+
"learning_rate": 0.0004141016949152543,
|
| 17748 |
+
"loss": 2.1176,
|
| 17749 |
+
"step": 2946
|
| 17750 |
+
},
|
| 17751 |
+
{
|
| 17752 |
+
"epoch": 0.49,
|
| 17753 |
+
"learning_rate": 0.0004139661016949153,
|
| 17754 |
+
"loss": 2.1184,
|
| 17755 |
+
"step": 2947
|
| 17756 |
+
},
|
| 17757 |
+
{
|
| 17758 |
+
"epoch": 0.49,
|
| 17759 |
+
"learning_rate": 0.00041383050847457626,
|
| 17760 |
+
"loss": 2.0973,
|
| 17761 |
+
"step": 2948
|
| 17762 |
+
},
|
| 17763 |
+
{
|
| 17764 |
+
"epoch": 0.49,
|
| 17765 |
+
"learning_rate": 0.0004136949152542373,
|
| 17766 |
+
"loss": 2.1017,
|
| 17767 |
+
"step": 2949
|
| 17768 |
+
},
|
| 17769 |
+
{
|
| 17770 |
+
"epoch": 0.49,
|
| 17771 |
+
"learning_rate": 0.0004135593220338983,
|
| 17772 |
+
"loss": 2.1777,
|
| 17773 |
+
"step": 2950
|
| 17774 |
+
},
|
| 17775 |
+
{
|
| 17776 |
+
"epoch": 0.49,
|
| 17777 |
+
"learning_rate": 0.00041342372881355936,
|
| 17778 |
+
"loss": 2.1708,
|
| 17779 |
+
"step": 2951
|
| 17780 |
+
},
|
| 17781 |
+
{
|
| 17782 |
+
"epoch": 0.49,
|
| 17783 |
+
"learning_rate": 0.00041328813559322034,
|
| 17784 |
+
"loss": 2.2064,
|
| 17785 |
+
"step": 2952
|
| 17786 |
+
},
|
| 17787 |
+
{
|
| 17788 |
+
"epoch": 0.49,
|
| 17789 |
+
"learning_rate": 0.0004131525423728814,
|
| 17790 |
+
"loss": 2.1337,
|
| 17791 |
+
"step": 2953
|
| 17792 |
+
},
|
| 17793 |
+
{
|
| 17794 |
+
"epoch": 0.49,
|
| 17795 |
+
"learning_rate": 0.00041301694915254236,
|
| 17796 |
+
"loss": 2.1547,
|
| 17797 |
+
"step": 2954
|
| 17798 |
+
},
|
| 17799 |
+
{
|
| 17800 |
+
"epoch": 0.49,
|
| 17801 |
+
"learning_rate": 0.00041288135593220345,
|
| 17802 |
+
"loss": 2.1788,
|
| 17803 |
+
"step": 2955
|
| 17804 |
+
},
|
| 17805 |
+
{
|
| 17806 |
+
"epoch": 0.49,
|
| 17807 |
+
"learning_rate": 0.00041274576271186443,
|
| 17808 |
+
"loss": 2.2068,
|
| 17809 |
+
"step": 2956
|
| 17810 |
+
},
|
| 17811 |
+
{
|
| 17812 |
+
"epoch": 0.49,
|
| 17813 |
+
"learning_rate": 0.00041261016949152546,
|
| 17814 |
+
"loss": 2.1452,
|
| 17815 |
+
"step": 2957
|
| 17816 |
+
},
|
| 17817 |
+
{
|
| 17818 |
+
"epoch": 0.49,
|
| 17819 |
+
"learning_rate": 0.00041247457627118645,
|
| 17820 |
+
"loss": 2.1059,
|
| 17821 |
+
"step": 2958
|
| 17822 |
+
},
|
| 17823 |
+
{
|
| 17824 |
+
"epoch": 0.49,
|
| 17825 |
+
"learning_rate": 0.0004123389830508475,
|
| 17826 |
+
"loss": 2.1151,
|
| 17827 |
+
"step": 2959
|
| 17828 |
+
},
|
| 17829 |
+
{
|
| 17830 |
+
"epoch": 0.49,
|
| 17831 |
+
"learning_rate": 0.00041220338983050846,
|
| 17832 |
+
"loss": 2.2186,
|
| 17833 |
+
"step": 2960
|
| 17834 |
+
},
|
| 17835 |
+
{
|
| 17836 |
+
"epoch": 0.49,
|
| 17837 |
+
"learning_rate": 0.00041206779661016955,
|
| 17838 |
+
"loss": 2.2221,
|
| 17839 |
+
"step": 2961
|
| 17840 |
+
},
|
| 17841 |
+
{
|
| 17842 |
+
"epoch": 0.49,
|
| 17843 |
+
"learning_rate": 0.00041193220338983053,
|
| 17844 |
+
"loss": 2.1348,
|
| 17845 |
+
"step": 2962
|
| 17846 |
+
},
|
| 17847 |
+
{
|
| 17848 |
+
"epoch": 0.49,
|
| 17849 |
+
"learning_rate": 0.00041179661016949156,
|
| 17850 |
+
"loss": 2.1694,
|
| 17851 |
+
"step": 2963
|
| 17852 |
+
},
|
| 17853 |
+
{
|
| 17854 |
+
"epoch": 0.49,
|
| 17855 |
+
"learning_rate": 0.00041166101694915255,
|
| 17856 |
+
"loss": 2.1813,
|
| 17857 |
+
"step": 2964
|
| 17858 |
+
},
|
| 17859 |
+
{
|
| 17860 |
+
"epoch": 0.49,
|
| 17861 |
+
"learning_rate": 0.00041152542372881363,
|
| 17862 |
+
"loss": 2.2421,
|
| 17863 |
+
"step": 2965
|
| 17864 |
+
},
|
| 17865 |
+
{
|
| 17866 |
+
"epoch": 0.49,
|
| 17867 |
+
"learning_rate": 0.0004113898305084746,
|
| 17868 |
+
"loss": 2.2113,
|
| 17869 |
+
"step": 2966
|
| 17870 |
+
},
|
| 17871 |
+
{
|
| 17872 |
+
"epoch": 0.49,
|
| 17873 |
+
"learning_rate": 0.0004112542372881356,
|
| 17874 |
+
"loss": 2.146,
|
| 17875 |
+
"step": 2967
|
| 17876 |
+
},
|
| 17877 |
+
{
|
| 17878 |
+
"epoch": 0.49,
|
| 17879 |
+
"learning_rate": 0.00041111864406779663,
|
| 17880 |
+
"loss": 2.1209,
|
| 17881 |
+
"step": 2968
|
| 17882 |
+
},
|
| 17883 |
+
{
|
| 17884 |
+
"epoch": 0.49,
|
| 17885 |
+
"learning_rate": 0.0004109830508474576,
|
| 17886 |
+
"loss": 2.2126,
|
| 17887 |
+
"step": 2969
|
| 17888 |
+
},
|
| 17889 |
+
{
|
| 17890 |
+
"epoch": 0.49,
|
| 17891 |
+
"learning_rate": 0.00041084745762711865,
|
| 17892 |
+
"loss": 2.1748,
|
| 17893 |
+
"step": 2970
|
| 17894 |
+
},
|
| 17895 |
+
{
|
| 17896 |
+
"epoch": 0.5,
|
| 17897 |
+
"learning_rate": 0.0004107118644067796,
|
| 17898 |
+
"loss": 2.0626,
|
| 17899 |
+
"step": 2971
|
| 17900 |
+
},
|
| 17901 |
+
{
|
| 17902 |
+
"epoch": 0.5,
|
| 17903 |
+
"learning_rate": 0.0004105762711864407,
|
| 17904 |
+
"loss": 2.1513,
|
| 17905 |
+
"step": 2972
|
| 17906 |
+
},
|
| 17907 |
+
{
|
| 17908 |
+
"epoch": 0.5,
|
| 17909 |
+
"learning_rate": 0.0004104406779661017,
|
| 17910 |
+
"loss": 2.1045,
|
| 17911 |
+
"step": 2973
|
| 17912 |
+
},
|
| 17913 |
+
{
|
| 17914 |
+
"epoch": 0.5,
|
| 17915 |
+
"learning_rate": 0.00041030508474576273,
|
| 17916 |
+
"loss": 2.218,
|
| 17917 |
+
"step": 2974
|
| 17918 |
+
},
|
| 17919 |
+
{
|
| 17920 |
+
"epoch": 0.5,
|
| 17921 |
+
"learning_rate": 0.0004101694915254237,
|
| 17922 |
+
"loss": 2.1739,
|
| 17923 |
+
"step": 2975
|
| 17924 |
+
},
|
| 17925 |
+
{
|
| 17926 |
+
"epoch": 0.5,
|
| 17927 |
+
"learning_rate": 0.0004100338983050848,
|
| 17928 |
+
"loss": 2.1967,
|
| 17929 |
+
"step": 2976
|
| 17930 |
+
},
|
| 17931 |
+
{
|
| 17932 |
+
"epoch": 0.5,
|
| 17933 |
+
"learning_rate": 0.0004098983050847458,
|
| 17934 |
+
"loss": 2.1824,
|
| 17935 |
+
"step": 2977
|
| 17936 |
+
},
|
| 17937 |
+
{
|
| 17938 |
+
"epoch": 0.5,
|
| 17939 |
+
"learning_rate": 0.0004097627118644068,
|
| 17940 |
+
"loss": 2.1947,
|
| 17941 |
+
"step": 2978
|
| 17942 |
+
},
|
| 17943 |
+
{
|
| 17944 |
+
"epoch": 0.5,
|
| 17945 |
+
"learning_rate": 0.0004096271186440678,
|
| 17946 |
+
"loss": 2.1325,
|
| 17947 |
+
"step": 2979
|
| 17948 |
+
},
|
| 17949 |
+
{
|
| 17950 |
+
"epoch": 0.5,
|
| 17951 |
+
"learning_rate": 0.0004094915254237289,
|
| 17952 |
+
"loss": 2.2497,
|
| 17953 |
+
"step": 2980
|
| 17954 |
+
},
|
| 17955 |
+
{
|
| 17956 |
+
"epoch": 0.5,
|
| 17957 |
+
"learning_rate": 0.0004093559322033898,
|
| 17958 |
+
"loss": 2.1418,
|
| 17959 |
+
"step": 2981
|
| 17960 |
+
},
|
| 17961 |
+
{
|
| 17962 |
+
"epoch": 0.5,
|
| 17963 |
+
"learning_rate": 0.0004092203389830509,
|
| 17964 |
+
"loss": 2.1621,
|
| 17965 |
+
"step": 2982
|
| 17966 |
+
},
|
| 17967 |
+
{
|
| 17968 |
+
"epoch": 0.5,
|
| 17969 |
+
"learning_rate": 0.0004090847457627119,
|
| 17970 |
+
"loss": 2.1295,
|
| 17971 |
+
"step": 2983
|
| 17972 |
+
},
|
| 17973 |
+
{
|
| 17974 |
+
"epoch": 0.5,
|
| 17975 |
+
"learning_rate": 0.0004089491525423729,
|
| 17976 |
+
"loss": 2.1085,
|
| 17977 |
+
"step": 2984
|
| 17978 |
+
},
|
| 17979 |
+
{
|
| 17980 |
+
"epoch": 0.5,
|
| 17981 |
+
"learning_rate": 0.0004088135593220339,
|
| 17982 |
+
"loss": 2.1285,
|
| 17983 |
+
"step": 2985
|
| 17984 |
+
},
|
| 17985 |
+
{
|
| 17986 |
+
"epoch": 0.5,
|
| 17987 |
+
"learning_rate": 0.000408677966101695,
|
| 17988 |
+
"loss": 2.0908,
|
| 17989 |
+
"step": 2986
|
| 17990 |
+
},
|
| 17991 |
+
{
|
| 17992 |
+
"epoch": 0.5,
|
| 17993 |
+
"learning_rate": 0.00040854237288135597,
|
| 17994 |
+
"loss": 2.151,
|
| 17995 |
+
"step": 2987
|
| 17996 |
+
},
|
| 17997 |
+
{
|
| 17998 |
+
"epoch": 0.5,
|
| 17999 |
+
"learning_rate": 0.00040840677966101695,
|
| 18000 |
+
"loss": 2.2518,
|
| 18001 |
+
"step": 2988
|
| 18002 |
+
},
|
| 18003 |
+
{
|
| 18004 |
+
"epoch": 0.5,
|
| 18005 |
+
"learning_rate": 0.000408271186440678,
|
| 18006 |
+
"loss": 2.0897,
|
| 18007 |
+
"step": 2989
|
| 18008 |
+
},
|
| 18009 |
+
{
|
| 18010 |
+
"epoch": 0.5,
|
| 18011 |
+
"learning_rate": 0.00040813559322033896,
|
| 18012 |
+
"loss": 2.2226,
|
| 18013 |
+
"step": 2990
|
| 18014 |
+
},
|
| 18015 |
+
{
|
| 18016 |
+
"epoch": 0.5,
|
| 18017 |
+
"learning_rate": 0.00040800000000000005,
|
| 18018 |
+
"loss": 2.1714,
|
| 18019 |
+
"step": 2991
|
| 18020 |
+
},
|
| 18021 |
+
{
|
| 18022 |
+
"epoch": 0.5,
|
| 18023 |
+
"learning_rate": 0.000407864406779661,
|
| 18024 |
+
"loss": 2.1256,
|
| 18025 |
+
"step": 2992
|
| 18026 |
+
},
|
| 18027 |
+
{
|
| 18028 |
+
"epoch": 0.5,
|
| 18029 |
+
"learning_rate": 0.00040772881355932207,
|
| 18030 |
+
"loss": 2.2306,
|
| 18031 |
+
"step": 2993
|
| 18032 |
+
},
|
| 18033 |
+
{
|
| 18034 |
+
"epoch": 0.5,
|
| 18035 |
+
"learning_rate": 0.00040759322033898305,
|
| 18036 |
+
"loss": 2.18,
|
| 18037 |
+
"step": 2994
|
| 18038 |
+
},
|
| 18039 |
+
{
|
| 18040 |
+
"epoch": 0.5,
|
| 18041 |
+
"learning_rate": 0.0004074576271186441,
|
| 18042 |
+
"loss": 2.2124,
|
| 18043 |
+
"step": 2995
|
| 18044 |
+
},
|
| 18045 |
+
{
|
| 18046 |
+
"epoch": 0.5,
|
| 18047 |
+
"learning_rate": 0.00040732203389830506,
|
| 18048 |
+
"loss": 2.1893,
|
| 18049 |
+
"step": 2996
|
| 18050 |
+
},
|
| 18051 |
+
{
|
| 18052 |
+
"epoch": 0.5,
|
| 18053 |
+
"learning_rate": 0.00040718644067796615,
|
| 18054 |
+
"loss": 2.1891,
|
| 18055 |
+
"step": 2997
|
| 18056 |
+
},
|
| 18057 |
+
{
|
| 18058 |
+
"epoch": 0.5,
|
| 18059 |
+
"learning_rate": 0.00040705084745762713,
|
| 18060 |
+
"loss": 2.2257,
|
| 18061 |
+
"step": 2998
|
| 18062 |
+
},
|
| 18063 |
+
{
|
| 18064 |
+
"epoch": 0.5,
|
| 18065 |
+
"learning_rate": 0.00040691525423728817,
|
| 18066 |
+
"loss": 2.0973,
|
| 18067 |
+
"step": 2999
|
| 18068 |
+
},
|
| 18069 |
+
{
|
| 18070 |
+
"epoch": 0.5,
|
| 18071 |
+
"learning_rate": 0.00040677966101694915,
|
| 18072 |
+
"loss": 2.221,
|
| 18073 |
+
"step": 3000
|
| 18074 |
+
},
|
| 18075 |
+
{
|
| 18076 |
+
"epoch": 0.5,
|
| 18077 |
+
"eval_gen_len": 19.0,
|
| 18078 |
+
"eval_loss": 2.0400795936584473,
|
| 18079 |
+
"eval_rouge1": 0.2657,
|
| 18080 |
+
"eval_rouge2": 0.1001,
|
| 18081 |
+
"eval_rougeL": 0.2227,
|
| 18082 |
+
"eval_rougeLsum": 0.2226,
|
| 18083 |
+
"eval_runtime": 21.6863,
|
| 18084 |
+
"eval_samples_per_second": 2.306,
|
| 18085 |
+
"eval_steps_per_second": 0.323,
|
| 18086 |
+
"step": 3000
|
| 18087 |
}
|
| 18088 |
],
|
| 18089 |
"max_steps": 6000,
|
| 18090 |
"num_train_epochs": 9223372036854775807,
|
| 18091 |
+
"total_flos": 2.8570333609822003e+18,
|
| 18092 |
"trial_name": null,
|
| 18093 |
"trial_params": null
|
| 18094 |
}
|
pytorch_model.bin
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 2368281769
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:02c85713dad1dcfeb5c32c03d2b6f448d20d550a87402f9d21be276100bce607
|
| 3 |
size 2368281769
|