Training in progress, step 4000
Browse files- last-checkpoint/optimizer.pt +1 -1
- last-checkpoint/pytorch_model.bin +1 -1
- last-checkpoint/rng_state.pth +2 -2
- last-checkpoint/scheduler.pt +1 -1
- last-checkpoint/spiece.model +3 -0
- last-checkpoint/trainer_state.json +4383 -4
- last-checkpoint/training_args.bin +1 -1
- pytorch_model.bin +1 -1
- spiece.model +3 -0
- training_args.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 3545432
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2fad6b5d095a9b088a00a85229b1f30618b7c0da52a8a7c8e838e5bd8255acd3
|
| 3 |
size 3545432
|
last-checkpoint/pytorch_model.bin
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 1140034381
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c57d5e652de14e5f01a9d33e0167070f639ab8a97f101a3d3639e153530869ce
|
| 3 |
size 1140034381
|
last-checkpoint/rng_state.pth
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9339b45a3919543bcaf98f631cf143d045fb484ed307b40b221a329ab3d5cba8
|
| 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:af8f9447546e655346b41f85fcb5317c211b847c1a09e9a09f7ba308dccd60ed
|
| 3 |
size 627
|
last-checkpoint/spiece.model
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:fa8e856a5564ec886599b5ed6728e3bd0cf9e040a2a56dfb6bf87c63a32a5caf
|
| 3 |
+
size 2243614
|
last-checkpoint/trainer_state.json
CHANGED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": null,
|
| 3 |
"best_model_checkpoint": null,
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -537,11 +537,4390 @@
|
|
| 537 |
"eval_samples_per_second": 0.061,
|
| 538 |
"eval_steps_per_second": 0.032,
|
| 539 |
"step": 360
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 540 |
}
|
| 541 |
],
|
| 542 |
-
"max_steps":
|
| 543 |
"num_train_epochs": 3,
|
| 544 |
-
"total_flos": 1.
|
| 545 |
"trial_name": null,
|
| 546 |
"trial_params": null
|
| 547 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": null,
|
| 3 |
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 0.3233760459193985,
|
| 5 |
+
"global_step": 4000,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 537 |
"eval_samples_per_second": 0.061,
|
| 538 |
"eval_steps_per_second": 0.032,
|
| 539 |
"step": 360
|
| 540 |
+
},
|
| 541 |
+
{
|
| 542 |
+
"epoch": 0.03,
|
| 543 |
+
"learning_rate": 4.951491632306573e-05,
|
| 544 |
+
"loss": 3.6245,
|
| 545 |
+
"step": 365
|
| 546 |
+
},
|
| 547 |
+
{
|
| 548 |
+
"epoch": 0.03,
|
| 549 |
+
"learning_rate": 4.950817904977498e-05,
|
| 550 |
+
"loss": 3.3963,
|
| 551 |
+
"step": 370
|
| 552 |
+
},
|
| 553 |
+
{
|
| 554 |
+
"epoch": 0.03,
|
| 555 |
+
"learning_rate": 4.950144177648422e-05,
|
| 556 |
+
"loss": 3.1854,
|
| 557 |
+
"step": 375
|
| 558 |
+
},
|
| 559 |
+
{
|
| 560 |
+
"epoch": 0.03,
|
| 561 |
+
"learning_rate": 4.9494704503193465e-05,
|
| 562 |
+
"loss": 3.5305,
|
| 563 |
+
"step": 380
|
| 564 |
+
},
|
| 565 |
+
{
|
| 566 |
+
"epoch": 0.03,
|
| 567 |
+
"learning_rate": 4.9487967229902713e-05,
|
| 568 |
+
"loss": 3.3036,
|
| 569 |
+
"step": 385
|
| 570 |
+
},
|
| 571 |
+
{
|
| 572 |
+
"epoch": 0.03,
|
| 573 |
+
"learning_rate": 4.948122995661196e-05,
|
| 574 |
+
"loss": 3.8064,
|
| 575 |
+
"step": 390
|
| 576 |
+
},
|
| 577 |
+
{
|
| 578 |
+
"epoch": 0.03,
|
| 579 |
+
"learning_rate": 4.947449268332121e-05,
|
| 580 |
+
"loss": 3.3977,
|
| 581 |
+
"step": 395
|
| 582 |
+
},
|
| 583 |
+
{
|
| 584 |
+
"epoch": 0.03,
|
| 585 |
+
"learning_rate": 4.946775541003045e-05,
|
| 586 |
+
"loss": 3.453,
|
| 587 |
+
"step": 400
|
| 588 |
+
},
|
| 589 |
+
{
|
| 590 |
+
"epoch": 0.03,
|
| 591 |
+
"learning_rate": 4.94610181367397e-05,
|
| 592 |
+
"loss": 3.1789,
|
| 593 |
+
"step": 405
|
| 594 |
+
},
|
| 595 |
+
{
|
| 596 |
+
"epoch": 0.03,
|
| 597 |
+
"learning_rate": 4.945428086344895e-05,
|
| 598 |
+
"loss": 3.3437,
|
| 599 |
+
"step": 410
|
| 600 |
+
},
|
| 601 |
+
{
|
| 602 |
+
"epoch": 0.03,
|
| 603 |
+
"learning_rate": 4.944754359015819e-05,
|
| 604 |
+
"loss": 3.1314,
|
| 605 |
+
"step": 415
|
| 606 |
+
},
|
| 607 |
+
{
|
| 608 |
+
"epoch": 0.03,
|
| 609 |
+
"learning_rate": 4.9440806316867435e-05,
|
| 610 |
+
"loss": 3.5475,
|
| 611 |
+
"step": 420
|
| 612 |
+
},
|
| 613 |
+
{
|
| 614 |
+
"epoch": 0.03,
|
| 615 |
+
"learning_rate": 4.9434069043576684e-05,
|
| 616 |
+
"loss": 3.145,
|
| 617 |
+
"step": 425
|
| 618 |
+
},
|
| 619 |
+
{
|
| 620 |
+
"epoch": 0.03,
|
| 621 |
+
"learning_rate": 4.942733177028593e-05,
|
| 622 |
+
"loss": 3.3634,
|
| 623 |
+
"step": 430
|
| 624 |
+
},
|
| 625 |
+
{
|
| 626 |
+
"epoch": 0.04,
|
| 627 |
+
"learning_rate": 4.9420594496995175e-05,
|
| 628 |
+
"loss": 3.2653,
|
| 629 |
+
"step": 435
|
| 630 |
+
},
|
| 631 |
+
{
|
| 632 |
+
"epoch": 0.04,
|
| 633 |
+
"learning_rate": 4.9413857223704424e-05,
|
| 634 |
+
"loss": 3.4231,
|
| 635 |
+
"step": 440
|
| 636 |
+
},
|
| 637 |
+
{
|
| 638 |
+
"epoch": 0.04,
|
| 639 |
+
"learning_rate": 4.940711995041367e-05,
|
| 640 |
+
"loss": 3.4497,
|
| 641 |
+
"step": 445
|
| 642 |
+
},
|
| 643 |
+
{
|
| 644 |
+
"epoch": 0.04,
|
| 645 |
+
"learning_rate": 4.940173013178107e-05,
|
| 646 |
+
"loss": 3.6448,
|
| 647 |
+
"step": 450
|
| 648 |
+
},
|
| 649 |
+
{
|
| 650 |
+
"epoch": 0.04,
|
| 651 |
+
"learning_rate": 4.939499285849031e-05,
|
| 652 |
+
"loss": 3.1871,
|
| 653 |
+
"step": 455
|
| 654 |
+
},
|
| 655 |
+
{
|
| 656 |
+
"epoch": 0.04,
|
| 657 |
+
"learning_rate": 4.938825558519956e-05,
|
| 658 |
+
"loss": 3.0073,
|
| 659 |
+
"step": 460
|
| 660 |
+
},
|
| 661 |
+
{
|
| 662 |
+
"epoch": 0.04,
|
| 663 |
+
"learning_rate": 4.938151831190881e-05,
|
| 664 |
+
"loss": 3.2752,
|
| 665 |
+
"step": 465
|
| 666 |
+
},
|
| 667 |
+
{
|
| 668 |
+
"epoch": 0.04,
|
| 669 |
+
"learning_rate": 4.937478103861805e-05,
|
| 670 |
+
"loss": 3.3738,
|
| 671 |
+
"step": 470
|
| 672 |
+
},
|
| 673 |
+
{
|
| 674 |
+
"epoch": 0.04,
|
| 675 |
+
"learning_rate": 4.93680437653273e-05,
|
| 676 |
+
"loss": 3.1156,
|
| 677 |
+
"step": 475
|
| 678 |
+
},
|
| 679 |
+
{
|
| 680 |
+
"epoch": 0.04,
|
| 681 |
+
"learning_rate": 4.936130649203655e-05,
|
| 682 |
+
"loss": 3.089,
|
| 683 |
+
"step": 480
|
| 684 |
+
},
|
| 685 |
+
{
|
| 686 |
+
"epoch": 0.04,
|
| 687 |
+
"learning_rate": 4.9354569218745796e-05,
|
| 688 |
+
"loss": 3.2361,
|
| 689 |
+
"step": 485
|
| 690 |
+
},
|
| 691 |
+
{
|
| 692 |
+
"epoch": 0.04,
|
| 693 |
+
"learning_rate": 4.934783194545504e-05,
|
| 694 |
+
"loss": 3.0767,
|
| 695 |
+
"step": 490
|
| 696 |
+
},
|
| 697 |
+
{
|
| 698 |
+
"epoch": 0.04,
|
| 699 |
+
"learning_rate": 4.934109467216429e-05,
|
| 700 |
+
"loss": 3.4057,
|
| 701 |
+
"step": 495
|
| 702 |
+
},
|
| 703 |
+
{
|
| 704 |
+
"epoch": 0.04,
|
| 705 |
+
"learning_rate": 4.933435739887353e-05,
|
| 706 |
+
"loss": 2.9778,
|
| 707 |
+
"step": 500
|
| 708 |
+
},
|
| 709 |
+
{
|
| 710 |
+
"epoch": 0.04,
|
| 711 |
+
"learning_rate": 4.932762012558277e-05,
|
| 712 |
+
"loss": 3.1364,
|
| 713 |
+
"step": 505
|
| 714 |
+
},
|
| 715 |
+
{
|
| 716 |
+
"epoch": 0.04,
|
| 717 |
+
"learning_rate": 4.932088285229202e-05,
|
| 718 |
+
"loss": 3.5782,
|
| 719 |
+
"step": 510
|
| 720 |
+
},
|
| 721 |
+
{
|
| 722 |
+
"epoch": 0.04,
|
| 723 |
+
"learning_rate": 4.931414557900127e-05,
|
| 724 |
+
"loss": 3.056,
|
| 725 |
+
"step": 515
|
| 726 |
+
},
|
| 727 |
+
{
|
| 728 |
+
"epoch": 0.04,
|
| 729 |
+
"learning_rate": 4.930740830571052e-05,
|
| 730 |
+
"loss": 3.4149,
|
| 731 |
+
"step": 520
|
| 732 |
+
},
|
| 733 |
+
{
|
| 734 |
+
"epoch": 0.04,
|
| 735 |
+
"learning_rate": 4.930067103241976e-05,
|
| 736 |
+
"loss": 3.2642,
|
| 737 |
+
"step": 525
|
| 738 |
+
},
|
| 739 |
+
{
|
| 740 |
+
"epoch": 0.04,
|
| 741 |
+
"learning_rate": 4.929393375912901e-05,
|
| 742 |
+
"loss": 3.0618,
|
| 743 |
+
"step": 530
|
| 744 |
+
},
|
| 745 |
+
{
|
| 746 |
+
"epoch": 0.04,
|
| 747 |
+
"learning_rate": 4.928719648583826e-05,
|
| 748 |
+
"loss": 3.2814,
|
| 749 |
+
"step": 535
|
| 750 |
+
},
|
| 751 |
+
{
|
| 752 |
+
"epoch": 0.04,
|
| 753 |
+
"learning_rate": 4.92804592125475e-05,
|
| 754 |
+
"loss": 3.0362,
|
| 755 |
+
"step": 540
|
| 756 |
+
},
|
| 757 |
+
{
|
| 758 |
+
"epoch": 0.04,
|
| 759 |
+
"learning_rate": 4.927372193925674e-05,
|
| 760 |
+
"loss": 3.4842,
|
| 761 |
+
"step": 545
|
| 762 |
+
},
|
| 763 |
+
{
|
| 764 |
+
"epoch": 0.04,
|
| 765 |
+
"learning_rate": 4.926698466596599e-05,
|
| 766 |
+
"loss": 3.2595,
|
| 767 |
+
"step": 550
|
| 768 |
+
},
|
| 769 |
+
{
|
| 770 |
+
"epoch": 0.04,
|
| 771 |
+
"learning_rate": 4.926024739267524e-05,
|
| 772 |
+
"loss": 3.2634,
|
| 773 |
+
"step": 555
|
| 774 |
+
},
|
| 775 |
+
{
|
| 776 |
+
"epoch": 0.05,
|
| 777 |
+
"learning_rate": 4.925351011938448e-05,
|
| 778 |
+
"loss": 3.492,
|
| 779 |
+
"step": 560
|
| 780 |
+
},
|
| 781 |
+
{
|
| 782 |
+
"epoch": 0.05,
|
| 783 |
+
"learning_rate": 4.924677284609373e-05,
|
| 784 |
+
"loss": 3.5355,
|
| 785 |
+
"step": 565
|
| 786 |
+
},
|
| 787 |
+
{
|
| 788 |
+
"epoch": 0.05,
|
| 789 |
+
"learning_rate": 4.924003557280298e-05,
|
| 790 |
+
"loss": 3.3369,
|
| 791 |
+
"step": 570
|
| 792 |
+
},
|
| 793 |
+
{
|
| 794 |
+
"epoch": 0.05,
|
| 795 |
+
"learning_rate": 4.923329829951223e-05,
|
| 796 |
+
"loss": 3.3085,
|
| 797 |
+
"step": 575
|
| 798 |
+
},
|
| 799 |
+
{
|
| 800 |
+
"epoch": 0.05,
|
| 801 |
+
"learning_rate": 4.922656102622147e-05,
|
| 802 |
+
"loss": 3.0434,
|
| 803 |
+
"step": 580
|
| 804 |
+
},
|
| 805 |
+
{
|
| 806 |
+
"epoch": 0.05,
|
| 807 |
+
"learning_rate": 4.921982375293071e-05,
|
| 808 |
+
"loss": 3.1722,
|
| 809 |
+
"step": 585
|
| 810 |
+
},
|
| 811 |
+
{
|
| 812 |
+
"epoch": 0.05,
|
| 813 |
+
"learning_rate": 4.921308647963996e-05,
|
| 814 |
+
"loss": 3.2637,
|
| 815 |
+
"step": 590
|
| 816 |
+
},
|
| 817 |
+
{
|
| 818 |
+
"epoch": 0.05,
|
| 819 |
+
"learning_rate": 4.9206349206349204e-05,
|
| 820 |
+
"loss": 3.5246,
|
| 821 |
+
"step": 595
|
| 822 |
+
},
|
| 823 |
+
{
|
| 824 |
+
"epoch": 0.05,
|
| 825 |
+
"learning_rate": 4.919961193305845e-05,
|
| 826 |
+
"loss": 3.5051,
|
| 827 |
+
"step": 600
|
| 828 |
+
},
|
| 829 |
+
{
|
| 830 |
+
"epoch": 0.05,
|
| 831 |
+
"learning_rate": 4.91928746597677e-05,
|
| 832 |
+
"loss": 3.2346,
|
| 833 |
+
"step": 605
|
| 834 |
+
},
|
| 835 |
+
{
|
| 836 |
+
"epoch": 0.05,
|
| 837 |
+
"learning_rate": 4.918613738647695e-05,
|
| 838 |
+
"loss": 3.2653,
|
| 839 |
+
"step": 610
|
| 840 |
+
},
|
| 841 |
+
{
|
| 842 |
+
"epoch": 0.05,
|
| 843 |
+
"learning_rate": 4.917940011318619e-05,
|
| 844 |
+
"loss": 3.1633,
|
| 845 |
+
"step": 615
|
| 846 |
+
},
|
| 847 |
+
{
|
| 848 |
+
"epoch": 0.05,
|
| 849 |
+
"learning_rate": 4.917266283989544e-05,
|
| 850 |
+
"loss": 3.0785,
|
| 851 |
+
"step": 620
|
| 852 |
+
},
|
| 853 |
+
{
|
| 854 |
+
"epoch": 0.05,
|
| 855 |
+
"learning_rate": 4.916592556660469e-05,
|
| 856 |
+
"loss": 3.1973,
|
| 857 |
+
"step": 625
|
| 858 |
+
},
|
| 859 |
+
{
|
| 860 |
+
"epoch": 0.05,
|
| 861 |
+
"learning_rate": 4.915918829331393e-05,
|
| 862 |
+
"loss": 3.2443,
|
| 863 |
+
"step": 630
|
| 864 |
+
},
|
| 865 |
+
{
|
| 866 |
+
"epoch": 0.05,
|
| 867 |
+
"learning_rate": 4.9152451020023175e-05,
|
| 868 |
+
"loss": 3.3169,
|
| 869 |
+
"step": 635
|
| 870 |
+
},
|
| 871 |
+
{
|
| 872 |
+
"epoch": 0.05,
|
| 873 |
+
"learning_rate": 4.9145713746732423e-05,
|
| 874 |
+
"loss": 3.4131,
|
| 875 |
+
"step": 640
|
| 876 |
+
},
|
| 877 |
+
{
|
| 878 |
+
"epoch": 0.05,
|
| 879 |
+
"learning_rate": 4.913897647344167e-05,
|
| 880 |
+
"loss": 3.2112,
|
| 881 |
+
"step": 645
|
| 882 |
+
},
|
| 883 |
+
{
|
| 884 |
+
"epoch": 0.05,
|
| 885 |
+
"learning_rate": 4.9132239200150914e-05,
|
| 886 |
+
"loss": 2.9755,
|
| 887 |
+
"step": 650
|
| 888 |
+
},
|
| 889 |
+
{
|
| 890 |
+
"epoch": 0.05,
|
| 891 |
+
"learning_rate": 4.912550192686016e-05,
|
| 892 |
+
"loss": 3.5015,
|
| 893 |
+
"step": 655
|
| 894 |
+
},
|
| 895 |
+
{
|
| 896 |
+
"epoch": 0.05,
|
| 897 |
+
"learning_rate": 4.911876465356941e-05,
|
| 898 |
+
"loss": 3.2172,
|
| 899 |
+
"step": 660
|
| 900 |
+
},
|
| 901 |
+
{
|
| 902 |
+
"epoch": 0.05,
|
| 903 |
+
"learning_rate": 4.911202738027866e-05,
|
| 904 |
+
"loss": 3.0727,
|
| 905 |
+
"step": 665
|
| 906 |
+
},
|
| 907 |
+
{
|
| 908 |
+
"epoch": 0.05,
|
| 909 |
+
"learning_rate": 4.91052901069879e-05,
|
| 910 |
+
"loss": 3.1883,
|
| 911 |
+
"step": 670
|
| 912 |
+
},
|
| 913 |
+
{
|
| 914 |
+
"epoch": 0.05,
|
| 915 |
+
"learning_rate": 4.9098552833697145e-05,
|
| 916 |
+
"loss": 3.6434,
|
| 917 |
+
"step": 675
|
| 918 |
+
},
|
| 919 |
+
{
|
| 920 |
+
"epoch": 0.05,
|
| 921 |
+
"learning_rate": 4.9091815560406394e-05,
|
| 922 |
+
"loss": 3.582,
|
| 923 |
+
"step": 680
|
| 924 |
+
},
|
| 925 |
+
{
|
| 926 |
+
"epoch": 0.06,
|
| 927 |
+
"learning_rate": 4.908507828711564e-05,
|
| 928 |
+
"loss": 3.6279,
|
| 929 |
+
"step": 685
|
| 930 |
+
},
|
| 931 |
+
{
|
| 932 |
+
"epoch": 0.06,
|
| 933 |
+
"learning_rate": 4.9078341013824885e-05,
|
| 934 |
+
"loss": 3.4427,
|
| 935 |
+
"step": 690
|
| 936 |
+
},
|
| 937 |
+
{
|
| 938 |
+
"epoch": 0.06,
|
| 939 |
+
"learning_rate": 4.9071603740534134e-05,
|
| 940 |
+
"loss": 3.2696,
|
| 941 |
+
"step": 695
|
| 942 |
+
},
|
| 943 |
+
{
|
| 944 |
+
"epoch": 0.06,
|
| 945 |
+
"learning_rate": 4.906486646724338e-05,
|
| 946 |
+
"loss": 3.2136,
|
| 947 |
+
"step": 700
|
| 948 |
+
},
|
| 949 |
+
{
|
| 950 |
+
"epoch": 0.06,
|
| 951 |
+
"learning_rate": 4.9058129193952625e-05,
|
| 952 |
+
"loss": 3.4226,
|
| 953 |
+
"step": 705
|
| 954 |
+
},
|
| 955 |
+
{
|
| 956 |
+
"epoch": 0.06,
|
| 957 |
+
"learning_rate": 4.9051391920661874e-05,
|
| 958 |
+
"loss": 3.0063,
|
| 959 |
+
"step": 710
|
| 960 |
+
},
|
| 961 |
+
{
|
| 962 |
+
"epoch": 0.06,
|
| 963 |
+
"learning_rate": 4.9044654647371116e-05,
|
| 964 |
+
"loss": 3.2971,
|
| 965 |
+
"step": 715
|
| 966 |
+
},
|
| 967 |
+
{
|
| 968 |
+
"epoch": 0.06,
|
| 969 |
+
"learning_rate": 4.9037917374080365e-05,
|
| 970 |
+
"loss": 3.4248,
|
| 971 |
+
"step": 720
|
| 972 |
+
},
|
| 973 |
+
{
|
| 974 |
+
"epoch": 0.06,
|
| 975 |
+
"learning_rate": 4.903118010078961e-05,
|
| 976 |
+
"loss": 3.3834,
|
| 977 |
+
"step": 725
|
| 978 |
+
},
|
| 979 |
+
{
|
| 980 |
+
"epoch": 0.06,
|
| 981 |
+
"learning_rate": 4.9024442827498856e-05,
|
| 982 |
+
"loss": 3.3403,
|
| 983 |
+
"step": 730
|
| 984 |
+
},
|
| 985 |
+
{
|
| 986 |
+
"epoch": 0.06,
|
| 987 |
+
"learning_rate": 4.9017705554208105e-05,
|
| 988 |
+
"loss": 3.1636,
|
| 989 |
+
"step": 735
|
| 990 |
+
},
|
| 991 |
+
{
|
| 992 |
+
"epoch": 0.06,
|
| 993 |
+
"learning_rate": 4.9010968280917353e-05,
|
| 994 |
+
"loss": 3.2906,
|
| 995 |
+
"step": 740
|
| 996 |
+
},
|
| 997 |
+
{
|
| 998 |
+
"epoch": 0.06,
|
| 999 |
+
"learning_rate": 4.9004231007626596e-05,
|
| 1000 |
+
"loss": 3.2874,
|
| 1001 |
+
"step": 745
|
| 1002 |
+
},
|
| 1003 |
+
{
|
| 1004 |
+
"epoch": 0.06,
|
| 1005 |
+
"learning_rate": 4.8997493734335844e-05,
|
| 1006 |
+
"loss": 3.1831,
|
| 1007 |
+
"step": 750
|
| 1008 |
+
},
|
| 1009 |
+
{
|
| 1010 |
+
"epoch": 0.06,
|
| 1011 |
+
"learning_rate": 4.8990756461045086e-05,
|
| 1012 |
+
"loss": 3.2486,
|
| 1013 |
+
"step": 755
|
| 1014 |
+
},
|
| 1015 |
+
{
|
| 1016 |
+
"epoch": 0.06,
|
| 1017 |
+
"learning_rate": 4.898401918775433e-05,
|
| 1018 |
+
"loss": 3.2148,
|
| 1019 |
+
"step": 760
|
| 1020 |
+
},
|
| 1021 |
+
{
|
| 1022 |
+
"epoch": 0.06,
|
| 1023 |
+
"learning_rate": 4.897728191446358e-05,
|
| 1024 |
+
"loss": 3.2995,
|
| 1025 |
+
"step": 765
|
| 1026 |
+
},
|
| 1027 |
+
{
|
| 1028 |
+
"epoch": 0.06,
|
| 1029 |
+
"learning_rate": 4.8970544641172826e-05,
|
| 1030 |
+
"loss": 3.3068,
|
| 1031 |
+
"step": 770
|
| 1032 |
+
},
|
| 1033 |
+
{
|
| 1034 |
+
"epoch": 0.06,
|
| 1035 |
+
"learning_rate": 4.8963807367882075e-05,
|
| 1036 |
+
"loss": 3.3789,
|
| 1037 |
+
"step": 775
|
| 1038 |
+
},
|
| 1039 |
+
{
|
| 1040 |
+
"epoch": 0.06,
|
| 1041 |
+
"learning_rate": 4.895707009459132e-05,
|
| 1042 |
+
"loss": 3.2501,
|
| 1043 |
+
"step": 780
|
| 1044 |
+
},
|
| 1045 |
+
{
|
| 1046 |
+
"epoch": 0.06,
|
| 1047 |
+
"learning_rate": 4.8950332821300566e-05,
|
| 1048 |
+
"loss": 3.3165,
|
| 1049 |
+
"step": 785
|
| 1050 |
+
},
|
| 1051 |
+
{
|
| 1052 |
+
"epoch": 0.06,
|
| 1053 |
+
"learning_rate": 4.8943595548009815e-05,
|
| 1054 |
+
"loss": 3.2548,
|
| 1055 |
+
"step": 790
|
| 1056 |
+
},
|
| 1057 |
+
{
|
| 1058 |
+
"epoch": 0.06,
|
| 1059 |
+
"learning_rate": 4.8936858274719064e-05,
|
| 1060 |
+
"loss": 3.0099,
|
| 1061 |
+
"step": 795
|
| 1062 |
+
},
|
| 1063 |
+
{
|
| 1064 |
+
"epoch": 0.06,
|
| 1065 |
+
"learning_rate": 4.8930121001428306e-05,
|
| 1066 |
+
"loss": 3.466,
|
| 1067 |
+
"step": 800
|
| 1068 |
+
},
|
| 1069 |
+
{
|
| 1070 |
+
"epoch": 0.07,
|
| 1071 |
+
"learning_rate": 4.892338372813755e-05,
|
| 1072 |
+
"loss": 3.106,
|
| 1073 |
+
"step": 805
|
| 1074 |
+
},
|
| 1075 |
+
{
|
| 1076 |
+
"epoch": 0.07,
|
| 1077 |
+
"learning_rate": 4.89166464548468e-05,
|
| 1078 |
+
"loss": 3.4483,
|
| 1079 |
+
"step": 810
|
| 1080 |
+
},
|
| 1081 |
+
{
|
| 1082 |
+
"epoch": 0.07,
|
| 1083 |
+
"learning_rate": 4.890990918155604e-05,
|
| 1084 |
+
"loss": 3.0073,
|
| 1085 |
+
"step": 815
|
| 1086 |
+
},
|
| 1087 |
+
{
|
| 1088 |
+
"epoch": 0.07,
|
| 1089 |
+
"learning_rate": 4.890317190826529e-05,
|
| 1090 |
+
"loss": 3.1426,
|
| 1091 |
+
"step": 820
|
| 1092 |
+
},
|
| 1093 |
+
{
|
| 1094 |
+
"epoch": 0.07,
|
| 1095 |
+
"learning_rate": 4.889643463497454e-05,
|
| 1096 |
+
"loss": 3.2155,
|
| 1097 |
+
"step": 825
|
| 1098 |
+
},
|
| 1099 |
+
{
|
| 1100 |
+
"epoch": 0.07,
|
| 1101 |
+
"learning_rate": 4.8889697361683786e-05,
|
| 1102 |
+
"loss": 3.2883,
|
| 1103 |
+
"step": 830
|
| 1104 |
+
},
|
| 1105 |
+
{
|
| 1106 |
+
"epoch": 0.07,
|
| 1107 |
+
"learning_rate": 4.888296008839303e-05,
|
| 1108 |
+
"loss": 3.3822,
|
| 1109 |
+
"step": 835
|
| 1110 |
+
},
|
| 1111 |
+
{
|
| 1112 |
+
"epoch": 0.07,
|
| 1113 |
+
"learning_rate": 4.887622281510228e-05,
|
| 1114 |
+
"loss": 3.1398,
|
| 1115 |
+
"step": 840
|
| 1116 |
+
},
|
| 1117 |
+
{
|
| 1118 |
+
"epoch": 0.07,
|
| 1119 |
+
"learning_rate": 4.886948554181152e-05,
|
| 1120 |
+
"loss": 3.252,
|
| 1121 |
+
"step": 845
|
| 1122 |
+
},
|
| 1123 |
+
{
|
| 1124 |
+
"epoch": 0.07,
|
| 1125 |
+
"learning_rate": 4.886274826852077e-05,
|
| 1126 |
+
"loss": 3.0845,
|
| 1127 |
+
"step": 850
|
| 1128 |
+
},
|
| 1129 |
+
{
|
| 1130 |
+
"epoch": 0.07,
|
| 1131 |
+
"learning_rate": 4.885601099523001e-05,
|
| 1132 |
+
"loss": 3.4627,
|
| 1133 |
+
"step": 855
|
| 1134 |
+
},
|
| 1135 |
+
{
|
| 1136 |
+
"epoch": 0.07,
|
| 1137 |
+
"learning_rate": 4.884927372193926e-05,
|
| 1138 |
+
"loss": 3.1775,
|
| 1139 |
+
"step": 860
|
| 1140 |
+
},
|
| 1141 |
+
{
|
| 1142 |
+
"epoch": 0.07,
|
| 1143 |
+
"learning_rate": 4.884253644864851e-05,
|
| 1144 |
+
"loss": 3.5839,
|
| 1145 |
+
"step": 865
|
| 1146 |
+
},
|
| 1147 |
+
{
|
| 1148 |
+
"epoch": 0.07,
|
| 1149 |
+
"learning_rate": 4.883579917535775e-05,
|
| 1150 |
+
"loss": 3.1501,
|
| 1151 |
+
"step": 870
|
| 1152 |
+
},
|
| 1153 |
+
{
|
| 1154 |
+
"epoch": 0.07,
|
| 1155 |
+
"learning_rate": 4.8829061902067e-05,
|
| 1156 |
+
"loss": 3.3487,
|
| 1157 |
+
"step": 875
|
| 1158 |
+
},
|
| 1159 |
+
{
|
| 1160 |
+
"epoch": 0.07,
|
| 1161 |
+
"learning_rate": 4.882232462877625e-05,
|
| 1162 |
+
"loss": 3.1447,
|
| 1163 |
+
"step": 880
|
| 1164 |
+
},
|
| 1165 |
+
{
|
| 1166 |
+
"epoch": 0.07,
|
| 1167 |
+
"learning_rate": 4.881558735548549e-05,
|
| 1168 |
+
"loss": 3.0964,
|
| 1169 |
+
"step": 885
|
| 1170 |
+
},
|
| 1171 |
+
{
|
| 1172 |
+
"epoch": 0.07,
|
| 1173 |
+
"learning_rate": 4.880885008219473e-05,
|
| 1174 |
+
"loss": 3.3593,
|
| 1175 |
+
"step": 890
|
| 1176 |
+
},
|
| 1177 |
+
{
|
| 1178 |
+
"epoch": 0.07,
|
| 1179 |
+
"learning_rate": 4.880211280890398e-05,
|
| 1180 |
+
"loss": 3.0965,
|
| 1181 |
+
"step": 895
|
| 1182 |
+
},
|
| 1183 |
+
{
|
| 1184 |
+
"epoch": 0.07,
|
| 1185 |
+
"learning_rate": 4.879537553561323e-05,
|
| 1186 |
+
"loss": 3.0485,
|
| 1187 |
+
"step": 900
|
| 1188 |
+
},
|
| 1189 |
+
{
|
| 1190 |
+
"epoch": 0.07,
|
| 1191 |
+
"learning_rate": 4.878863826232247e-05,
|
| 1192 |
+
"loss": 3.497,
|
| 1193 |
+
"step": 905
|
| 1194 |
+
},
|
| 1195 |
+
{
|
| 1196 |
+
"epoch": 0.07,
|
| 1197 |
+
"learning_rate": 4.878190098903172e-05,
|
| 1198 |
+
"loss": 3.2529,
|
| 1199 |
+
"step": 910
|
| 1200 |
+
},
|
| 1201 |
+
{
|
| 1202 |
+
"epoch": 0.07,
|
| 1203 |
+
"learning_rate": 4.877516371574097e-05,
|
| 1204 |
+
"loss": 3.2689,
|
| 1205 |
+
"step": 915
|
| 1206 |
+
},
|
| 1207 |
+
{
|
| 1208 |
+
"epoch": 0.07,
|
| 1209 |
+
"learning_rate": 4.876842644245022e-05,
|
| 1210 |
+
"loss": 3.2244,
|
| 1211 |
+
"step": 920
|
| 1212 |
+
},
|
| 1213 |
+
{
|
| 1214 |
+
"epoch": 0.07,
|
| 1215 |
+
"learning_rate": 4.876168916915946e-05,
|
| 1216 |
+
"loss": 3.3559,
|
| 1217 |
+
"step": 925
|
| 1218 |
+
},
|
| 1219 |
+
{
|
| 1220 |
+
"epoch": 0.08,
|
| 1221 |
+
"learning_rate": 4.875495189586871e-05,
|
| 1222 |
+
"loss": 3.1012,
|
| 1223 |
+
"step": 930
|
| 1224 |
+
},
|
| 1225 |
+
{
|
| 1226 |
+
"epoch": 0.08,
|
| 1227 |
+
"learning_rate": 4.8749562077236104e-05,
|
| 1228 |
+
"loss": 2.9742,
|
| 1229 |
+
"step": 935
|
| 1230 |
+
},
|
| 1231 |
+
{
|
| 1232 |
+
"epoch": 0.08,
|
| 1233 |
+
"learning_rate": 4.8742824803945346e-05,
|
| 1234 |
+
"loss": 3.3356,
|
| 1235 |
+
"step": 940
|
| 1236 |
+
},
|
| 1237 |
+
{
|
| 1238 |
+
"epoch": 0.08,
|
| 1239 |
+
"learning_rate": 4.8736087530654595e-05,
|
| 1240 |
+
"loss": 3.0625,
|
| 1241 |
+
"step": 945
|
| 1242 |
+
},
|
| 1243 |
+
{
|
| 1244 |
+
"epoch": 0.08,
|
| 1245 |
+
"learning_rate": 4.8729350257363844e-05,
|
| 1246 |
+
"loss": 3.8974,
|
| 1247 |
+
"step": 950
|
| 1248 |
+
},
|
| 1249 |
+
{
|
| 1250 |
+
"epoch": 0.08,
|
| 1251 |
+
"learning_rate": 4.872261298407309e-05,
|
| 1252 |
+
"loss": 3.2096,
|
| 1253 |
+
"step": 955
|
| 1254 |
+
},
|
| 1255 |
+
{
|
| 1256 |
+
"epoch": 0.08,
|
| 1257 |
+
"learning_rate": 4.8715875710782335e-05,
|
| 1258 |
+
"loss": 3.32,
|
| 1259 |
+
"step": 960
|
| 1260 |
+
},
|
| 1261 |
+
{
|
| 1262 |
+
"epoch": 0.08,
|
| 1263 |
+
"learning_rate": 4.8709138437491584e-05,
|
| 1264 |
+
"loss": 3.3169,
|
| 1265 |
+
"step": 965
|
| 1266 |
+
},
|
| 1267 |
+
{
|
| 1268 |
+
"epoch": 0.08,
|
| 1269 |
+
"learning_rate": 4.8702401164200826e-05,
|
| 1270 |
+
"loss": 3.2782,
|
| 1271 |
+
"step": 970
|
| 1272 |
+
},
|
| 1273 |
+
{
|
| 1274 |
+
"epoch": 0.08,
|
| 1275 |
+
"learning_rate": 4.8695663890910075e-05,
|
| 1276 |
+
"loss": 3.7628,
|
| 1277 |
+
"step": 975
|
| 1278 |
+
},
|
| 1279 |
+
{
|
| 1280 |
+
"epoch": 0.08,
|
| 1281 |
+
"learning_rate": 4.868892661761932e-05,
|
| 1282 |
+
"loss": 3.3816,
|
| 1283 |
+
"step": 980
|
| 1284 |
+
},
|
| 1285 |
+
{
|
| 1286 |
+
"epoch": 0.08,
|
| 1287 |
+
"learning_rate": 4.8682189344328566e-05,
|
| 1288 |
+
"loss": 3.3809,
|
| 1289 |
+
"step": 985
|
| 1290 |
+
},
|
| 1291 |
+
{
|
| 1292 |
+
"epoch": 0.08,
|
| 1293 |
+
"learning_rate": 4.8675452071037814e-05,
|
| 1294 |
+
"loss": 3.5458,
|
| 1295 |
+
"step": 990
|
| 1296 |
+
},
|
| 1297 |
+
{
|
| 1298 |
+
"epoch": 0.08,
|
| 1299 |
+
"learning_rate": 4.8668714797747057e-05,
|
| 1300 |
+
"loss": 3.3134,
|
| 1301 |
+
"step": 995
|
| 1302 |
+
},
|
| 1303 |
+
{
|
| 1304 |
+
"epoch": 0.08,
|
| 1305 |
+
"learning_rate": 4.8661977524456305e-05,
|
| 1306 |
+
"loss": 3.062,
|
| 1307 |
+
"step": 1000
|
| 1308 |
+
},
|
| 1309 |
+
{
|
| 1310 |
+
"epoch": 0.08,
|
| 1311 |
+
"learning_rate": 4.8655240251165554e-05,
|
| 1312 |
+
"loss": 3.0732,
|
| 1313 |
+
"step": 1005
|
| 1314 |
+
},
|
| 1315 |
+
{
|
| 1316 |
+
"epoch": 0.08,
|
| 1317 |
+
"learning_rate": 4.8648502977874796e-05,
|
| 1318 |
+
"loss": 3.325,
|
| 1319 |
+
"step": 1010
|
| 1320 |
+
},
|
| 1321 |
+
{
|
| 1322 |
+
"epoch": 0.08,
|
| 1323 |
+
"learning_rate": 4.864176570458404e-05,
|
| 1324 |
+
"loss": 3.1344,
|
| 1325 |
+
"step": 1015
|
| 1326 |
+
},
|
| 1327 |
+
{
|
| 1328 |
+
"epoch": 0.08,
|
| 1329 |
+
"learning_rate": 4.863502843129329e-05,
|
| 1330 |
+
"loss": 3.4551,
|
| 1331 |
+
"step": 1020
|
| 1332 |
+
},
|
| 1333 |
+
{
|
| 1334 |
+
"epoch": 0.08,
|
| 1335 |
+
"learning_rate": 4.8628291158002536e-05,
|
| 1336 |
+
"loss": 3.0637,
|
| 1337 |
+
"step": 1025
|
| 1338 |
+
},
|
| 1339 |
+
{
|
| 1340 |
+
"epoch": 0.08,
|
| 1341 |
+
"learning_rate": 4.862155388471178e-05,
|
| 1342 |
+
"loss": 3.3949,
|
| 1343 |
+
"step": 1030
|
| 1344 |
+
},
|
| 1345 |
+
{
|
| 1346 |
+
"epoch": 0.08,
|
| 1347 |
+
"learning_rate": 4.861481661142103e-05,
|
| 1348 |
+
"loss": 3.172,
|
| 1349 |
+
"step": 1035
|
| 1350 |
+
},
|
| 1351 |
+
{
|
| 1352 |
+
"epoch": 0.08,
|
| 1353 |
+
"learning_rate": 4.8608079338130276e-05,
|
| 1354 |
+
"loss": 3.1154,
|
| 1355 |
+
"step": 1040
|
| 1356 |
+
},
|
| 1357 |
+
{
|
| 1358 |
+
"epoch": 0.08,
|
| 1359 |
+
"learning_rate": 4.8601342064839525e-05,
|
| 1360 |
+
"loss": 3.0203,
|
| 1361 |
+
"step": 1045
|
| 1362 |
+
},
|
| 1363 |
+
{
|
| 1364 |
+
"epoch": 0.08,
|
| 1365 |
+
"learning_rate": 4.859460479154877e-05,
|
| 1366 |
+
"loss": 3.3186,
|
| 1367 |
+
"step": 1050
|
| 1368 |
+
},
|
| 1369 |
+
{
|
| 1370 |
+
"epoch": 0.09,
|
| 1371 |
+
"learning_rate": 4.858786751825801e-05,
|
| 1372 |
+
"loss": 3.2057,
|
| 1373 |
+
"step": 1055
|
| 1374 |
+
},
|
| 1375 |
+
{
|
| 1376 |
+
"epoch": 0.09,
|
| 1377 |
+
"learning_rate": 4.858113024496726e-05,
|
| 1378 |
+
"loss": 3.2119,
|
| 1379 |
+
"step": 1060
|
| 1380 |
+
},
|
| 1381 |
+
{
|
| 1382 |
+
"epoch": 0.09,
|
| 1383 |
+
"learning_rate": 4.857439297167651e-05,
|
| 1384 |
+
"loss": 3.0146,
|
| 1385 |
+
"step": 1065
|
| 1386 |
+
},
|
| 1387 |
+
{
|
| 1388 |
+
"epoch": 0.09,
|
| 1389 |
+
"learning_rate": 4.856765569838575e-05,
|
| 1390 |
+
"loss": 3.0737,
|
| 1391 |
+
"step": 1070
|
| 1392 |
+
},
|
| 1393 |
+
{
|
| 1394 |
+
"epoch": 0.09,
|
| 1395 |
+
"learning_rate": 4.8560918425095e-05,
|
| 1396 |
+
"loss": 3.118,
|
| 1397 |
+
"step": 1075
|
| 1398 |
+
},
|
| 1399 |
+
{
|
| 1400 |
+
"epoch": 0.09,
|
| 1401 |
+
"learning_rate": 4.855418115180425e-05,
|
| 1402 |
+
"loss": 3.2081,
|
| 1403 |
+
"step": 1080
|
| 1404 |
+
},
|
| 1405 |
+
{
|
| 1406 |
+
"epoch": 0.09,
|
| 1407 |
+
"learning_rate": 4.854744387851349e-05,
|
| 1408 |
+
"loss": 3.2234,
|
| 1409 |
+
"step": 1085
|
| 1410 |
+
},
|
| 1411 |
+
{
|
| 1412 |
+
"epoch": 0.09,
|
| 1413 |
+
"learning_rate": 4.854070660522274e-05,
|
| 1414 |
+
"loss": 3.217,
|
| 1415 |
+
"step": 1090
|
| 1416 |
+
},
|
| 1417 |
+
{
|
| 1418 |
+
"epoch": 0.09,
|
| 1419 |
+
"learning_rate": 4.8533969331931987e-05,
|
| 1420 |
+
"loss": 3.4233,
|
| 1421 |
+
"step": 1095
|
| 1422 |
+
},
|
| 1423 |
+
{
|
| 1424 |
+
"epoch": 0.09,
|
| 1425 |
+
"learning_rate": 4.852723205864123e-05,
|
| 1426 |
+
"loss": 3.3069,
|
| 1427 |
+
"step": 1100
|
| 1428 |
+
},
|
| 1429 |
+
{
|
| 1430 |
+
"epoch": 0.09,
|
| 1431 |
+
"learning_rate": 4.852049478535047e-05,
|
| 1432 |
+
"loss": 3.2344,
|
| 1433 |
+
"step": 1105
|
| 1434 |
+
},
|
| 1435 |
+
{
|
| 1436 |
+
"epoch": 0.09,
|
| 1437 |
+
"learning_rate": 4.851375751205972e-05,
|
| 1438 |
+
"loss": 3.0722,
|
| 1439 |
+
"step": 1110
|
| 1440 |
+
},
|
| 1441 |
+
{
|
| 1442 |
+
"epoch": 0.09,
|
| 1443 |
+
"learning_rate": 4.850702023876897e-05,
|
| 1444 |
+
"loss": 2.9557,
|
| 1445 |
+
"step": 1115
|
| 1446 |
+
},
|
| 1447 |
+
{
|
| 1448 |
+
"epoch": 0.09,
|
| 1449 |
+
"learning_rate": 4.850028296547822e-05,
|
| 1450 |
+
"loss": 3.132,
|
| 1451 |
+
"step": 1120
|
| 1452 |
+
},
|
| 1453 |
+
{
|
| 1454 |
+
"epoch": 0.09,
|
| 1455 |
+
"learning_rate": 4.849354569218746e-05,
|
| 1456 |
+
"loss": 3.2406,
|
| 1457 |
+
"step": 1125
|
| 1458 |
+
},
|
| 1459 |
+
{
|
| 1460 |
+
"epoch": 0.09,
|
| 1461 |
+
"learning_rate": 4.848680841889671e-05,
|
| 1462 |
+
"loss": 3.1061,
|
| 1463 |
+
"step": 1130
|
| 1464 |
+
},
|
| 1465 |
+
{
|
| 1466 |
+
"epoch": 0.09,
|
| 1467 |
+
"learning_rate": 4.848007114560596e-05,
|
| 1468 |
+
"loss": 3.2933,
|
| 1469 |
+
"step": 1135
|
| 1470 |
+
},
|
| 1471 |
+
{
|
| 1472 |
+
"epoch": 0.09,
|
| 1473 |
+
"learning_rate": 4.84733338723152e-05,
|
| 1474 |
+
"loss": 2.9373,
|
| 1475 |
+
"step": 1140
|
| 1476 |
+
},
|
| 1477 |
+
{
|
| 1478 |
+
"epoch": 0.09,
|
| 1479 |
+
"learning_rate": 4.846659659902444e-05,
|
| 1480 |
+
"loss": 3.5017,
|
| 1481 |
+
"step": 1145
|
| 1482 |
+
},
|
| 1483 |
+
{
|
| 1484 |
+
"epoch": 0.09,
|
| 1485 |
+
"learning_rate": 4.845985932573369e-05,
|
| 1486 |
+
"loss": 3.1143,
|
| 1487 |
+
"step": 1150
|
| 1488 |
+
},
|
| 1489 |
+
{
|
| 1490 |
+
"epoch": 0.09,
|
| 1491 |
+
"learning_rate": 4.845312205244294e-05,
|
| 1492 |
+
"loss": 3.3182,
|
| 1493 |
+
"step": 1155
|
| 1494 |
+
},
|
| 1495 |
+
{
|
| 1496 |
+
"epoch": 0.09,
|
| 1497 |
+
"learning_rate": 4.844638477915218e-05,
|
| 1498 |
+
"loss": 3.3364,
|
| 1499 |
+
"step": 1160
|
| 1500 |
+
},
|
| 1501 |
+
{
|
| 1502 |
+
"epoch": 0.09,
|
| 1503 |
+
"learning_rate": 4.843964750586143e-05,
|
| 1504 |
+
"loss": 2.8652,
|
| 1505 |
+
"step": 1165
|
| 1506 |
+
},
|
| 1507 |
+
{
|
| 1508 |
+
"epoch": 0.09,
|
| 1509 |
+
"learning_rate": 4.843291023257068e-05,
|
| 1510 |
+
"loss": 3.3397,
|
| 1511 |
+
"step": 1170
|
| 1512 |
+
},
|
| 1513 |
+
{
|
| 1514 |
+
"epoch": 0.09,
|
| 1515 |
+
"learning_rate": 4.842617295927993e-05,
|
| 1516 |
+
"loss": 3.2569,
|
| 1517 |
+
"step": 1175
|
| 1518 |
+
},
|
| 1519 |
+
{
|
| 1520 |
+
"epoch": 0.1,
|
| 1521 |
+
"learning_rate": 4.841943568598917e-05,
|
| 1522 |
+
"loss": 3.1305,
|
| 1523 |
+
"step": 1180
|
| 1524 |
+
},
|
| 1525 |
+
{
|
| 1526 |
+
"epoch": 0.1,
|
| 1527 |
+
"learning_rate": 4.841269841269841e-05,
|
| 1528 |
+
"loss": 3.3159,
|
| 1529 |
+
"step": 1185
|
| 1530 |
+
},
|
| 1531 |
+
{
|
| 1532 |
+
"epoch": 0.1,
|
| 1533 |
+
"learning_rate": 4.840596113940766e-05,
|
| 1534 |
+
"loss": 3.2887,
|
| 1535 |
+
"step": 1190
|
| 1536 |
+
},
|
| 1537 |
+
{
|
| 1538 |
+
"epoch": 0.1,
|
| 1539 |
+
"learning_rate": 4.83992238661169e-05,
|
| 1540 |
+
"loss": 3.4384,
|
| 1541 |
+
"step": 1195
|
| 1542 |
+
},
|
| 1543 |
+
{
|
| 1544 |
+
"epoch": 0.1,
|
| 1545 |
+
"learning_rate": 4.839248659282615e-05,
|
| 1546 |
+
"loss": 2.9754,
|
| 1547 |
+
"step": 1200
|
| 1548 |
+
},
|
| 1549 |
+
{
|
| 1550 |
+
"epoch": 0.1,
|
| 1551 |
+
"learning_rate": 4.83857493195354e-05,
|
| 1552 |
+
"loss": 3.5198,
|
| 1553 |
+
"step": 1205
|
| 1554 |
+
},
|
| 1555 |
+
{
|
| 1556 |
+
"epoch": 0.1,
|
| 1557 |
+
"learning_rate": 4.837901204624465e-05,
|
| 1558 |
+
"loss": 3.1371,
|
| 1559 |
+
"step": 1210
|
| 1560 |
+
},
|
| 1561 |
+
{
|
| 1562 |
+
"epoch": 0.1,
|
| 1563 |
+
"learning_rate": 4.837227477295389e-05,
|
| 1564 |
+
"loss": 3.2225,
|
| 1565 |
+
"step": 1215
|
| 1566 |
+
},
|
| 1567 |
+
{
|
| 1568 |
+
"epoch": 0.1,
|
| 1569 |
+
"learning_rate": 4.836553749966314e-05,
|
| 1570 |
+
"loss": 3.0097,
|
| 1571 |
+
"step": 1220
|
| 1572 |
+
},
|
| 1573 |
+
{
|
| 1574 |
+
"epoch": 0.1,
|
| 1575 |
+
"learning_rate": 4.835880022637238e-05,
|
| 1576 |
+
"loss": 3.2081,
|
| 1577 |
+
"step": 1225
|
| 1578 |
+
},
|
| 1579 |
+
{
|
| 1580 |
+
"epoch": 0.1,
|
| 1581 |
+
"learning_rate": 4.835206295308163e-05,
|
| 1582 |
+
"loss": 3.3162,
|
| 1583 |
+
"step": 1230
|
| 1584 |
+
},
|
| 1585 |
+
{
|
| 1586 |
+
"epoch": 0.1,
|
| 1587 |
+
"learning_rate": 4.8345325679790874e-05,
|
| 1588 |
+
"loss": 3.1359,
|
| 1589 |
+
"step": 1235
|
| 1590 |
+
},
|
| 1591 |
+
{
|
| 1592 |
+
"epoch": 0.1,
|
| 1593 |
+
"learning_rate": 4.833858840650012e-05,
|
| 1594 |
+
"loss": 3.0363,
|
| 1595 |
+
"step": 1240
|
| 1596 |
+
},
|
| 1597 |
+
{
|
| 1598 |
+
"epoch": 0.1,
|
| 1599 |
+
"learning_rate": 4.833185113320937e-05,
|
| 1600 |
+
"loss": 3.2238,
|
| 1601 |
+
"step": 1245
|
| 1602 |
+
},
|
| 1603 |
+
{
|
| 1604 |
+
"epoch": 0.1,
|
| 1605 |
+
"learning_rate": 4.8325113859918614e-05,
|
| 1606 |
+
"loss": 3.3966,
|
| 1607 |
+
"step": 1250
|
| 1608 |
+
},
|
| 1609 |
+
{
|
| 1610 |
+
"epoch": 0.1,
|
| 1611 |
+
"learning_rate": 4.831837658662786e-05,
|
| 1612 |
+
"loss": 3.4219,
|
| 1613 |
+
"step": 1255
|
| 1614 |
+
},
|
| 1615 |
+
{
|
| 1616 |
+
"epoch": 0.1,
|
| 1617 |
+
"learning_rate": 4.831163931333711e-05,
|
| 1618 |
+
"loss": 3.2468,
|
| 1619 |
+
"step": 1260
|
| 1620 |
+
},
|
| 1621 |
+
{
|
| 1622 |
+
"epoch": 0.1,
|
| 1623 |
+
"learning_rate": 4.830490204004636e-05,
|
| 1624 |
+
"loss": 3.2864,
|
| 1625 |
+
"step": 1265
|
| 1626 |
+
},
|
| 1627 |
+
{
|
| 1628 |
+
"epoch": 0.1,
|
| 1629 |
+
"learning_rate": 4.82981647667556e-05,
|
| 1630 |
+
"loss": 3.2543,
|
| 1631 |
+
"step": 1270
|
| 1632 |
+
},
|
| 1633 |
+
{
|
| 1634 |
+
"epoch": 0.1,
|
| 1635 |
+
"learning_rate": 4.8291427493464844e-05,
|
| 1636 |
+
"loss": 3.3271,
|
| 1637 |
+
"step": 1275
|
| 1638 |
+
},
|
| 1639 |
+
{
|
| 1640 |
+
"epoch": 0.1,
|
| 1641 |
+
"learning_rate": 4.828469022017409e-05,
|
| 1642 |
+
"loss": 3.3202,
|
| 1643 |
+
"step": 1280
|
| 1644 |
+
},
|
| 1645 |
+
{
|
| 1646 |
+
"epoch": 0.1,
|
| 1647 |
+
"learning_rate": 4.8277952946883335e-05,
|
| 1648 |
+
"loss": 3.3111,
|
| 1649 |
+
"step": 1285
|
| 1650 |
+
},
|
| 1651 |
+
{
|
| 1652 |
+
"epoch": 0.1,
|
| 1653 |
+
"learning_rate": 4.8271215673592584e-05,
|
| 1654 |
+
"loss": 3.2664,
|
| 1655 |
+
"step": 1290
|
| 1656 |
+
},
|
| 1657 |
+
{
|
| 1658 |
+
"epoch": 0.1,
|
| 1659 |
+
"learning_rate": 4.826447840030183e-05,
|
| 1660 |
+
"loss": 3.2784,
|
| 1661 |
+
"step": 1295
|
| 1662 |
+
},
|
| 1663 |
+
{
|
| 1664 |
+
"epoch": 0.11,
|
| 1665 |
+
"learning_rate": 4.825774112701108e-05,
|
| 1666 |
+
"loss": 3.2833,
|
| 1667 |
+
"step": 1300
|
| 1668 |
+
},
|
| 1669 |
+
{
|
| 1670 |
+
"epoch": 0.11,
|
| 1671 |
+
"learning_rate": 4.8251003853720324e-05,
|
| 1672 |
+
"loss": 3.1219,
|
| 1673 |
+
"step": 1305
|
| 1674 |
+
},
|
| 1675 |
+
{
|
| 1676 |
+
"epoch": 0.11,
|
| 1677 |
+
"learning_rate": 4.824426658042957e-05,
|
| 1678 |
+
"loss": 3.1704,
|
| 1679 |
+
"step": 1310
|
| 1680 |
+
},
|
| 1681 |
+
{
|
| 1682 |
+
"epoch": 0.11,
|
| 1683 |
+
"learning_rate": 4.8237529307138815e-05,
|
| 1684 |
+
"loss": 3.1486,
|
| 1685 |
+
"step": 1315
|
| 1686 |
+
},
|
| 1687 |
+
{
|
| 1688 |
+
"epoch": 0.11,
|
| 1689 |
+
"learning_rate": 4.8230792033848064e-05,
|
| 1690 |
+
"loss": 3.2648,
|
| 1691 |
+
"step": 1320
|
| 1692 |
+
},
|
| 1693 |
+
{
|
| 1694 |
+
"epoch": 0.11,
|
| 1695 |
+
"learning_rate": 4.8224054760557306e-05,
|
| 1696 |
+
"loss": 3.1059,
|
| 1697 |
+
"step": 1325
|
| 1698 |
+
},
|
| 1699 |
+
{
|
| 1700 |
+
"epoch": 0.11,
|
| 1701 |
+
"learning_rate": 4.8217317487266555e-05,
|
| 1702 |
+
"loss": 3.3623,
|
| 1703 |
+
"step": 1330
|
| 1704 |
+
},
|
| 1705 |
+
{
|
| 1706 |
+
"epoch": 0.11,
|
| 1707 |
+
"learning_rate": 4.8210580213975804e-05,
|
| 1708 |
+
"loss": 3.5866,
|
| 1709 |
+
"step": 1335
|
| 1710 |
+
},
|
| 1711 |
+
{
|
| 1712 |
+
"epoch": 0.11,
|
| 1713 |
+
"learning_rate": 4.8203842940685046e-05,
|
| 1714 |
+
"loss": 3.3165,
|
| 1715 |
+
"step": 1340
|
| 1716 |
+
},
|
| 1717 |
+
{
|
| 1718 |
+
"epoch": 0.11,
|
| 1719 |
+
"learning_rate": 4.8197105667394295e-05,
|
| 1720 |
+
"loss": 3.3361,
|
| 1721 |
+
"step": 1345
|
| 1722 |
+
},
|
| 1723 |
+
{
|
| 1724 |
+
"epoch": 0.11,
|
| 1725 |
+
"learning_rate": 4.8190368394103544e-05,
|
| 1726 |
+
"loss": 3.3784,
|
| 1727 |
+
"step": 1350
|
| 1728 |
+
},
|
| 1729 |
+
{
|
| 1730 |
+
"epoch": 0.11,
|
| 1731 |
+
"learning_rate": 4.8183631120812786e-05,
|
| 1732 |
+
"loss": 2.9877,
|
| 1733 |
+
"step": 1355
|
| 1734 |
+
},
|
| 1735 |
+
{
|
| 1736 |
+
"epoch": 0.11,
|
| 1737 |
+
"learning_rate": 4.817689384752203e-05,
|
| 1738 |
+
"loss": 3.5171,
|
| 1739 |
+
"step": 1360
|
| 1740 |
+
},
|
| 1741 |
+
{
|
| 1742 |
+
"epoch": 0.11,
|
| 1743 |
+
"learning_rate": 4.817015657423128e-05,
|
| 1744 |
+
"loss": 3.1608,
|
| 1745 |
+
"step": 1365
|
| 1746 |
+
},
|
| 1747 |
+
{
|
| 1748 |
+
"epoch": 0.11,
|
| 1749 |
+
"learning_rate": 4.8163419300940526e-05,
|
| 1750 |
+
"loss": 3.1863,
|
| 1751 |
+
"step": 1370
|
| 1752 |
+
},
|
| 1753 |
+
{
|
| 1754 |
+
"epoch": 0.11,
|
| 1755 |
+
"learning_rate": 4.8156682027649774e-05,
|
| 1756 |
+
"loss": 3.1035,
|
| 1757 |
+
"step": 1375
|
| 1758 |
+
},
|
| 1759 |
+
{
|
| 1760 |
+
"epoch": 0.11,
|
| 1761 |
+
"learning_rate": 4.8149944754359017e-05,
|
| 1762 |
+
"loss": 3.4208,
|
| 1763 |
+
"step": 1380
|
| 1764 |
+
},
|
| 1765 |
+
{
|
| 1766 |
+
"epoch": 0.11,
|
| 1767 |
+
"learning_rate": 4.8143207481068265e-05,
|
| 1768 |
+
"loss": 3.0785,
|
| 1769 |
+
"step": 1385
|
| 1770 |
+
},
|
| 1771 |
+
{
|
| 1772 |
+
"epoch": 0.11,
|
| 1773 |
+
"learning_rate": 4.8136470207777514e-05,
|
| 1774 |
+
"loss": 3.1869,
|
| 1775 |
+
"step": 1390
|
| 1776 |
+
},
|
| 1777 |
+
{
|
| 1778 |
+
"epoch": 0.11,
|
| 1779 |
+
"learning_rate": 4.8129732934486756e-05,
|
| 1780 |
+
"loss": 3.0122,
|
| 1781 |
+
"step": 1395
|
| 1782 |
+
},
|
| 1783 |
+
{
|
| 1784 |
+
"epoch": 0.11,
|
| 1785 |
+
"learning_rate": 4.8122995661196005e-05,
|
| 1786 |
+
"loss": 3.5176,
|
| 1787 |
+
"step": 1400
|
| 1788 |
+
},
|
| 1789 |
+
{
|
| 1790 |
+
"epoch": 0.11,
|
| 1791 |
+
"learning_rate": 4.811625838790525e-05,
|
| 1792 |
+
"loss": 3.3739,
|
| 1793 |
+
"step": 1405
|
| 1794 |
+
},
|
| 1795 |
+
{
|
| 1796 |
+
"epoch": 0.11,
|
| 1797 |
+
"learning_rate": 4.8109521114614496e-05,
|
| 1798 |
+
"loss": 3.1952,
|
| 1799 |
+
"step": 1410
|
| 1800 |
+
},
|
| 1801 |
+
{
|
| 1802 |
+
"epoch": 0.11,
|
| 1803 |
+
"learning_rate": 4.810278384132374e-05,
|
| 1804 |
+
"loss": 3.2811,
|
| 1805 |
+
"step": 1415
|
| 1806 |
+
},
|
| 1807 |
+
{
|
| 1808 |
+
"epoch": 0.11,
|
| 1809 |
+
"learning_rate": 4.809604656803299e-05,
|
| 1810 |
+
"loss": 3.4939,
|
| 1811 |
+
"step": 1420
|
| 1812 |
+
},
|
| 1813 |
+
{
|
| 1814 |
+
"epoch": 0.12,
|
| 1815 |
+
"learning_rate": 4.8089309294742236e-05,
|
| 1816 |
+
"loss": 3.3097,
|
| 1817 |
+
"step": 1425
|
| 1818 |
+
},
|
| 1819 |
+
{
|
| 1820 |
+
"epoch": 0.12,
|
| 1821 |
+
"learning_rate": 4.8082572021451485e-05,
|
| 1822 |
+
"loss": 3.1636,
|
| 1823 |
+
"step": 1430
|
| 1824 |
+
},
|
| 1825 |
+
{
|
| 1826 |
+
"epoch": 0.12,
|
| 1827 |
+
"learning_rate": 4.807583474816073e-05,
|
| 1828 |
+
"loss": 3.1452,
|
| 1829 |
+
"step": 1435
|
| 1830 |
+
},
|
| 1831 |
+
{
|
| 1832 |
+
"epoch": 0.12,
|
| 1833 |
+
"learning_rate": 4.8069097474869976e-05,
|
| 1834 |
+
"loss": 3.3203,
|
| 1835 |
+
"step": 1440
|
| 1836 |
+
},
|
| 1837 |
+
{
|
| 1838 |
+
"epoch": 0.12,
|
| 1839 |
+
"learning_rate": 4.806236020157922e-05,
|
| 1840 |
+
"loss": 3.1017,
|
| 1841 |
+
"step": 1445
|
| 1842 |
+
},
|
| 1843 |
+
{
|
| 1844 |
+
"epoch": 0.12,
|
| 1845 |
+
"learning_rate": 4.805562292828846e-05,
|
| 1846 |
+
"loss": 3.1715,
|
| 1847 |
+
"step": 1450
|
| 1848 |
+
},
|
| 1849 |
+
{
|
| 1850 |
+
"epoch": 0.12,
|
| 1851 |
+
"learning_rate": 4.804888565499771e-05,
|
| 1852 |
+
"loss": 2.8621,
|
| 1853 |
+
"step": 1455
|
| 1854 |
+
},
|
| 1855 |
+
{
|
| 1856 |
+
"epoch": 0.12,
|
| 1857 |
+
"learning_rate": 4.804214838170696e-05,
|
| 1858 |
+
"loss": 3.1176,
|
| 1859 |
+
"step": 1460
|
| 1860 |
+
},
|
| 1861 |
+
{
|
| 1862 |
+
"epoch": 0.12,
|
| 1863 |
+
"learning_rate": 4.803541110841621e-05,
|
| 1864 |
+
"loss": 3.5175,
|
| 1865 |
+
"step": 1465
|
| 1866 |
+
},
|
| 1867 |
+
{
|
| 1868 |
+
"epoch": 0.12,
|
| 1869 |
+
"learning_rate": 4.802867383512545e-05,
|
| 1870 |
+
"loss": 3.245,
|
| 1871 |
+
"step": 1470
|
| 1872 |
+
},
|
| 1873 |
+
{
|
| 1874 |
+
"epoch": 0.12,
|
| 1875 |
+
"learning_rate": 4.80219365618347e-05,
|
| 1876 |
+
"loss": 3.0326,
|
| 1877 |
+
"step": 1475
|
| 1878 |
+
},
|
| 1879 |
+
{
|
| 1880 |
+
"epoch": 0.12,
|
| 1881 |
+
"learning_rate": 4.8015199288543947e-05,
|
| 1882 |
+
"loss": 3.2505,
|
| 1883 |
+
"step": 1480
|
| 1884 |
+
},
|
| 1885 |
+
{
|
| 1886 |
+
"epoch": 0.12,
|
| 1887 |
+
"learning_rate": 4.800846201525319e-05,
|
| 1888 |
+
"loss": 3.36,
|
| 1889 |
+
"step": 1485
|
| 1890 |
+
},
|
| 1891 |
+
{
|
| 1892 |
+
"epoch": 0.12,
|
| 1893 |
+
"learning_rate": 4.800172474196243e-05,
|
| 1894 |
+
"loss": 3.3422,
|
| 1895 |
+
"step": 1490
|
| 1896 |
+
},
|
| 1897 |
+
{
|
| 1898 |
+
"epoch": 0.12,
|
| 1899 |
+
"learning_rate": 4.799498746867168e-05,
|
| 1900 |
+
"loss": 3.2268,
|
| 1901 |
+
"step": 1495
|
| 1902 |
+
},
|
| 1903 |
+
{
|
| 1904 |
+
"epoch": 0.12,
|
| 1905 |
+
"learning_rate": 4.798825019538093e-05,
|
| 1906 |
+
"loss": 3.1994,
|
| 1907 |
+
"step": 1500
|
| 1908 |
+
},
|
| 1909 |
+
{
|
| 1910 |
+
"epoch": 0.12,
|
| 1911 |
+
"learning_rate": 4.7982860376748324e-05,
|
| 1912 |
+
"loss": 3.1365,
|
| 1913 |
+
"step": 1505
|
| 1914 |
+
},
|
| 1915 |
+
{
|
| 1916 |
+
"epoch": 0.12,
|
| 1917 |
+
"learning_rate": 4.797612310345757e-05,
|
| 1918 |
+
"loss": 3.2696,
|
| 1919 |
+
"step": 1510
|
| 1920 |
+
},
|
| 1921 |
+
{
|
| 1922 |
+
"epoch": 0.12,
|
| 1923 |
+
"learning_rate": 4.796938583016682e-05,
|
| 1924 |
+
"loss": 2.9565,
|
| 1925 |
+
"step": 1515
|
| 1926 |
+
},
|
| 1927 |
+
{
|
| 1928 |
+
"epoch": 0.12,
|
| 1929 |
+
"learning_rate": 4.796264855687606e-05,
|
| 1930 |
+
"loss": 2.9312,
|
| 1931 |
+
"step": 1520
|
| 1932 |
+
},
|
| 1933 |
+
{
|
| 1934 |
+
"epoch": 0.12,
|
| 1935 |
+
"learning_rate": 4.7955911283585305e-05,
|
| 1936 |
+
"loss": 2.901,
|
| 1937 |
+
"step": 1525
|
| 1938 |
+
},
|
| 1939 |
+
{
|
| 1940 |
+
"epoch": 0.12,
|
| 1941 |
+
"learning_rate": 4.7949174010294554e-05,
|
| 1942 |
+
"loss": 3.3423,
|
| 1943 |
+
"step": 1530
|
| 1944 |
+
},
|
| 1945 |
+
{
|
| 1946 |
+
"epoch": 0.12,
|
| 1947 |
+
"learning_rate": 4.79424367370038e-05,
|
| 1948 |
+
"loss": 3.4224,
|
| 1949 |
+
"step": 1535
|
| 1950 |
+
},
|
| 1951 |
+
{
|
| 1952 |
+
"epoch": 0.12,
|
| 1953 |
+
"learning_rate": 4.7935699463713045e-05,
|
| 1954 |
+
"loss": 3.1498,
|
| 1955 |
+
"step": 1540
|
| 1956 |
+
},
|
| 1957 |
+
{
|
| 1958 |
+
"epoch": 0.12,
|
| 1959 |
+
"learning_rate": 4.7928962190422294e-05,
|
| 1960 |
+
"loss": 3.31,
|
| 1961 |
+
"step": 1545
|
| 1962 |
+
},
|
| 1963 |
+
{
|
| 1964 |
+
"epoch": 0.13,
|
| 1965 |
+
"learning_rate": 4.792222491713154e-05,
|
| 1966 |
+
"loss": 3.3928,
|
| 1967 |
+
"step": 1550
|
| 1968 |
+
},
|
| 1969 |
+
{
|
| 1970 |
+
"epoch": 0.13,
|
| 1971 |
+
"learning_rate": 4.791548764384079e-05,
|
| 1972 |
+
"loss": 3.3894,
|
| 1973 |
+
"step": 1555
|
| 1974 |
+
},
|
| 1975 |
+
{
|
| 1976 |
+
"epoch": 0.13,
|
| 1977 |
+
"learning_rate": 4.7908750370550034e-05,
|
| 1978 |
+
"loss": 3.056,
|
| 1979 |
+
"step": 1560
|
| 1980 |
+
},
|
| 1981 |
+
{
|
| 1982 |
+
"epoch": 0.13,
|
| 1983 |
+
"learning_rate": 4.790201309725928e-05,
|
| 1984 |
+
"loss": 3.1893,
|
| 1985 |
+
"step": 1565
|
| 1986 |
+
},
|
| 1987 |
+
{
|
| 1988 |
+
"epoch": 0.13,
|
| 1989 |
+
"learning_rate": 4.7895275823968525e-05,
|
| 1990 |
+
"loss": 3.1102,
|
| 1991 |
+
"step": 1570
|
| 1992 |
+
},
|
| 1993 |
+
{
|
| 1994 |
+
"epoch": 0.13,
|
| 1995 |
+
"learning_rate": 4.788853855067777e-05,
|
| 1996 |
+
"loss": 3.2477,
|
| 1997 |
+
"step": 1575
|
| 1998 |
+
},
|
| 1999 |
+
{
|
| 2000 |
+
"epoch": 0.13,
|
| 2001 |
+
"learning_rate": 4.7881801277387016e-05,
|
| 2002 |
+
"loss": 3.1685,
|
| 2003 |
+
"step": 1580
|
| 2004 |
+
},
|
| 2005 |
+
{
|
| 2006 |
+
"epoch": 0.13,
|
| 2007 |
+
"learning_rate": 4.7875064004096265e-05,
|
| 2008 |
+
"loss": 3.285,
|
| 2009 |
+
"step": 1585
|
| 2010 |
+
},
|
| 2011 |
+
{
|
| 2012 |
+
"epoch": 0.13,
|
| 2013 |
+
"learning_rate": 4.7868326730805514e-05,
|
| 2014 |
+
"loss": 3.3288,
|
| 2015 |
+
"step": 1590
|
| 2016 |
+
},
|
| 2017 |
+
{
|
| 2018 |
+
"epoch": 0.13,
|
| 2019 |
+
"learning_rate": 4.7861589457514756e-05,
|
| 2020 |
+
"loss": 2.9146,
|
| 2021 |
+
"step": 1595
|
| 2022 |
+
},
|
| 2023 |
+
{
|
| 2024 |
+
"epoch": 0.13,
|
| 2025 |
+
"learning_rate": 4.7854852184224005e-05,
|
| 2026 |
+
"loss": 3.3004,
|
| 2027 |
+
"step": 1600
|
| 2028 |
+
},
|
| 2029 |
+
{
|
| 2030 |
+
"epoch": 0.13,
|
| 2031 |
+
"learning_rate": 4.7848114910933254e-05,
|
| 2032 |
+
"loss": 3.3265,
|
| 2033 |
+
"step": 1605
|
| 2034 |
+
},
|
| 2035 |
+
{
|
| 2036 |
+
"epoch": 0.13,
|
| 2037 |
+
"learning_rate": 4.7841377637642496e-05,
|
| 2038 |
+
"loss": 3.2319,
|
| 2039 |
+
"step": 1610
|
| 2040 |
+
},
|
| 2041 |
+
{
|
| 2042 |
+
"epoch": 0.13,
|
| 2043 |
+
"learning_rate": 4.783464036435174e-05,
|
| 2044 |
+
"loss": 3.2694,
|
| 2045 |
+
"step": 1615
|
| 2046 |
+
},
|
| 2047 |
+
{
|
| 2048 |
+
"epoch": 0.13,
|
| 2049 |
+
"learning_rate": 4.782790309106099e-05,
|
| 2050 |
+
"loss": 3.3163,
|
| 2051 |
+
"step": 1620
|
| 2052 |
+
},
|
| 2053 |
+
{
|
| 2054 |
+
"epoch": 0.13,
|
| 2055 |
+
"learning_rate": 4.7821165817770235e-05,
|
| 2056 |
+
"loss": 2.9109,
|
| 2057 |
+
"step": 1625
|
| 2058 |
+
},
|
| 2059 |
+
{
|
| 2060 |
+
"epoch": 0.13,
|
| 2061 |
+
"learning_rate": 4.781442854447948e-05,
|
| 2062 |
+
"loss": 3.1788,
|
| 2063 |
+
"step": 1630
|
| 2064 |
+
},
|
| 2065 |
+
{
|
| 2066 |
+
"epoch": 0.13,
|
| 2067 |
+
"learning_rate": 4.7807691271188726e-05,
|
| 2068 |
+
"loss": 3.2638,
|
| 2069 |
+
"step": 1635
|
| 2070 |
+
},
|
| 2071 |
+
{
|
| 2072 |
+
"epoch": 0.13,
|
| 2073 |
+
"learning_rate": 4.7800953997897975e-05,
|
| 2074 |
+
"loss": 3.1715,
|
| 2075 |
+
"step": 1640
|
| 2076 |
+
},
|
| 2077 |
+
{
|
| 2078 |
+
"epoch": 0.13,
|
| 2079 |
+
"learning_rate": 4.7794216724607224e-05,
|
| 2080 |
+
"loss": 3.0204,
|
| 2081 |
+
"step": 1645
|
| 2082 |
+
},
|
| 2083 |
+
{
|
| 2084 |
+
"epoch": 0.13,
|
| 2085 |
+
"learning_rate": 4.7787479451316466e-05,
|
| 2086 |
+
"loss": 3.1442,
|
| 2087 |
+
"step": 1650
|
| 2088 |
+
},
|
| 2089 |
+
{
|
| 2090 |
+
"epoch": 0.13,
|
| 2091 |
+
"learning_rate": 4.778074217802571e-05,
|
| 2092 |
+
"loss": 3.1554,
|
| 2093 |
+
"step": 1655
|
| 2094 |
+
},
|
| 2095 |
+
{
|
| 2096 |
+
"epoch": 0.13,
|
| 2097 |
+
"learning_rate": 4.777400490473496e-05,
|
| 2098 |
+
"loss": 3.0041,
|
| 2099 |
+
"step": 1660
|
| 2100 |
+
},
|
| 2101 |
+
{
|
| 2102 |
+
"epoch": 0.13,
|
| 2103 |
+
"learning_rate": 4.7767267631444206e-05,
|
| 2104 |
+
"loss": 3.4228,
|
| 2105 |
+
"step": 1665
|
| 2106 |
+
},
|
| 2107 |
+
{
|
| 2108 |
+
"epoch": 0.14,
|
| 2109 |
+
"learning_rate": 4.776053035815345e-05,
|
| 2110 |
+
"loss": 3.1159,
|
| 2111 |
+
"step": 1670
|
| 2112 |
+
},
|
| 2113 |
+
{
|
| 2114 |
+
"epoch": 0.14,
|
| 2115 |
+
"learning_rate": 4.77537930848627e-05,
|
| 2116 |
+
"loss": 3.3559,
|
| 2117 |
+
"step": 1675
|
| 2118 |
+
},
|
| 2119 |
+
{
|
| 2120 |
+
"epoch": 0.14,
|
| 2121 |
+
"learning_rate": 4.7747055811571946e-05,
|
| 2122 |
+
"loss": 2.8674,
|
| 2123 |
+
"step": 1680
|
| 2124 |
+
},
|
| 2125 |
+
{
|
| 2126 |
+
"epoch": 0.14,
|
| 2127 |
+
"learning_rate": 4.774031853828119e-05,
|
| 2128 |
+
"loss": 3.1701,
|
| 2129 |
+
"step": 1685
|
| 2130 |
+
},
|
| 2131 |
+
{
|
| 2132 |
+
"epoch": 0.14,
|
| 2133 |
+
"learning_rate": 4.773358126499044e-05,
|
| 2134 |
+
"loss": 3.3984,
|
| 2135 |
+
"step": 1690
|
| 2136 |
+
},
|
| 2137 |
+
{
|
| 2138 |
+
"epoch": 0.14,
|
| 2139 |
+
"learning_rate": 4.772684399169968e-05,
|
| 2140 |
+
"loss": 3.2335,
|
| 2141 |
+
"step": 1695
|
| 2142 |
+
},
|
| 2143 |
+
{
|
| 2144 |
+
"epoch": 0.14,
|
| 2145 |
+
"learning_rate": 4.772010671840893e-05,
|
| 2146 |
+
"loss": 3.0153,
|
| 2147 |
+
"step": 1700
|
| 2148 |
+
},
|
| 2149 |
+
{
|
| 2150 |
+
"epoch": 0.14,
|
| 2151 |
+
"learning_rate": 4.771336944511817e-05,
|
| 2152 |
+
"loss": 3.2157,
|
| 2153 |
+
"step": 1705
|
| 2154 |
+
},
|
| 2155 |
+
{
|
| 2156 |
+
"epoch": 0.14,
|
| 2157 |
+
"learning_rate": 4.770663217182742e-05,
|
| 2158 |
+
"loss": 3.2443,
|
| 2159 |
+
"step": 1710
|
| 2160 |
+
},
|
| 2161 |
+
{
|
| 2162 |
+
"epoch": 0.14,
|
| 2163 |
+
"learning_rate": 4.769989489853667e-05,
|
| 2164 |
+
"loss": 3.2683,
|
| 2165 |
+
"step": 1715
|
| 2166 |
+
},
|
| 2167 |
+
{
|
| 2168 |
+
"epoch": 0.14,
|
| 2169 |
+
"learning_rate": 4.769315762524591e-05,
|
| 2170 |
+
"loss": 3.3998,
|
| 2171 |
+
"step": 1720
|
| 2172 |
+
},
|
| 2173 |
+
{
|
| 2174 |
+
"epoch": 0.14,
|
| 2175 |
+
"learning_rate": 4.768642035195516e-05,
|
| 2176 |
+
"loss": 3.2701,
|
| 2177 |
+
"step": 1725
|
| 2178 |
+
},
|
| 2179 |
+
{
|
| 2180 |
+
"epoch": 0.14,
|
| 2181 |
+
"learning_rate": 4.767968307866441e-05,
|
| 2182 |
+
"loss": 3.2304,
|
| 2183 |
+
"step": 1730
|
| 2184 |
+
},
|
| 2185 |
+
{
|
| 2186 |
+
"epoch": 0.14,
|
| 2187 |
+
"learning_rate": 4.7672945805373656e-05,
|
| 2188 |
+
"loss": 3.4887,
|
| 2189 |
+
"step": 1735
|
| 2190 |
+
},
|
| 2191 |
+
{
|
| 2192 |
+
"epoch": 0.14,
|
| 2193 |
+
"learning_rate": 4.76662085320829e-05,
|
| 2194 |
+
"loss": 3.1407,
|
| 2195 |
+
"step": 1740
|
| 2196 |
+
},
|
| 2197 |
+
{
|
| 2198 |
+
"epoch": 0.14,
|
| 2199 |
+
"learning_rate": 4.765947125879214e-05,
|
| 2200 |
+
"loss": 3.3806,
|
| 2201 |
+
"step": 1745
|
| 2202 |
+
},
|
| 2203 |
+
{
|
| 2204 |
+
"epoch": 0.14,
|
| 2205 |
+
"learning_rate": 4.765273398550139e-05,
|
| 2206 |
+
"loss": 2.9394,
|
| 2207 |
+
"step": 1750
|
| 2208 |
+
},
|
| 2209 |
+
{
|
| 2210 |
+
"epoch": 0.14,
|
| 2211 |
+
"learning_rate": 4.764599671221064e-05,
|
| 2212 |
+
"loss": 3.0697,
|
| 2213 |
+
"step": 1755
|
| 2214 |
+
},
|
| 2215 |
+
{
|
| 2216 |
+
"epoch": 0.14,
|
| 2217 |
+
"learning_rate": 4.763925943891988e-05,
|
| 2218 |
+
"loss": 3.1154,
|
| 2219 |
+
"step": 1760
|
| 2220 |
+
},
|
| 2221 |
+
{
|
| 2222 |
+
"epoch": 0.14,
|
| 2223 |
+
"learning_rate": 4.763252216562913e-05,
|
| 2224 |
+
"loss": 3.308,
|
| 2225 |
+
"step": 1765
|
| 2226 |
+
},
|
| 2227 |
+
{
|
| 2228 |
+
"epoch": 0.14,
|
| 2229 |
+
"learning_rate": 4.762578489233838e-05,
|
| 2230 |
+
"loss": 3.1236,
|
| 2231 |
+
"step": 1770
|
| 2232 |
+
},
|
| 2233 |
+
{
|
| 2234 |
+
"epoch": 0.14,
|
| 2235 |
+
"learning_rate": 4.761904761904762e-05,
|
| 2236 |
+
"loss": 3.2146,
|
| 2237 |
+
"step": 1775
|
| 2238 |
+
},
|
| 2239 |
+
{
|
| 2240 |
+
"epoch": 0.14,
|
| 2241 |
+
"learning_rate": 4.761231034575687e-05,
|
| 2242 |
+
"loss": 3.0922,
|
| 2243 |
+
"step": 1780
|
| 2244 |
+
},
|
| 2245 |
+
{
|
| 2246 |
+
"epoch": 0.14,
|
| 2247 |
+
"learning_rate": 4.760557307246611e-05,
|
| 2248 |
+
"loss": 3.2447,
|
| 2249 |
+
"step": 1785
|
| 2250 |
+
},
|
| 2251 |
+
{
|
| 2252 |
+
"epoch": 0.14,
|
| 2253 |
+
"learning_rate": 4.759883579917536e-05,
|
| 2254 |
+
"loss": 3.2431,
|
| 2255 |
+
"step": 1790
|
| 2256 |
+
},
|
| 2257 |
+
{
|
| 2258 |
+
"epoch": 0.15,
|
| 2259 |
+
"learning_rate": 4.75920985258846e-05,
|
| 2260 |
+
"loss": 3.2935,
|
| 2261 |
+
"step": 1795
|
| 2262 |
+
},
|
| 2263 |
+
{
|
| 2264 |
+
"epoch": 0.15,
|
| 2265 |
+
"learning_rate": 4.758536125259385e-05,
|
| 2266 |
+
"loss": 3.0225,
|
| 2267 |
+
"step": 1800
|
| 2268 |
+
},
|
| 2269 |
+
{
|
| 2270 |
+
"epoch": 0.15,
|
| 2271 |
+
"learning_rate": 4.75786239793031e-05,
|
| 2272 |
+
"loss": 3.1884,
|
| 2273 |
+
"step": 1805
|
| 2274 |
+
},
|
| 2275 |
+
{
|
| 2276 |
+
"epoch": 0.15,
|
| 2277 |
+
"learning_rate": 4.757188670601235e-05,
|
| 2278 |
+
"loss": 3.1911,
|
| 2279 |
+
"step": 1810
|
| 2280 |
+
},
|
| 2281 |
+
{
|
| 2282 |
+
"epoch": 0.15,
|
| 2283 |
+
"learning_rate": 4.756514943272159e-05,
|
| 2284 |
+
"loss": 3.1588,
|
| 2285 |
+
"step": 1815
|
| 2286 |
+
},
|
| 2287 |
+
{
|
| 2288 |
+
"epoch": 0.15,
|
| 2289 |
+
"learning_rate": 4.755841215943084e-05,
|
| 2290 |
+
"loss": 3.0782,
|
| 2291 |
+
"step": 1820
|
| 2292 |
+
},
|
| 2293 |
+
{
|
| 2294 |
+
"epoch": 0.15,
|
| 2295 |
+
"learning_rate": 4.755167488614008e-05,
|
| 2296 |
+
"loss": 3.0288,
|
| 2297 |
+
"step": 1825
|
| 2298 |
+
},
|
| 2299 |
+
{
|
| 2300 |
+
"epoch": 0.15,
|
| 2301 |
+
"learning_rate": 4.7544937612849324e-05,
|
| 2302 |
+
"loss": 3.0151,
|
| 2303 |
+
"step": 1830
|
| 2304 |
+
},
|
| 2305 |
+
{
|
| 2306 |
+
"epoch": 0.15,
|
| 2307 |
+
"learning_rate": 4.753820033955857e-05,
|
| 2308 |
+
"loss": 3.058,
|
| 2309 |
+
"step": 1835
|
| 2310 |
+
},
|
| 2311 |
+
{
|
| 2312 |
+
"epoch": 0.15,
|
| 2313 |
+
"learning_rate": 4.753146306626782e-05,
|
| 2314 |
+
"loss": 3.1809,
|
| 2315 |
+
"step": 1840
|
| 2316 |
+
},
|
| 2317 |
+
{
|
| 2318 |
+
"epoch": 0.15,
|
| 2319 |
+
"learning_rate": 4.752472579297707e-05,
|
| 2320 |
+
"loss": 3.1528,
|
| 2321 |
+
"step": 1845
|
| 2322 |
+
},
|
| 2323 |
+
{
|
| 2324 |
+
"epoch": 0.15,
|
| 2325 |
+
"learning_rate": 4.751798851968631e-05,
|
| 2326 |
+
"loss": 3.1677,
|
| 2327 |
+
"step": 1850
|
| 2328 |
+
},
|
| 2329 |
+
{
|
| 2330 |
+
"epoch": 0.15,
|
| 2331 |
+
"learning_rate": 4.751125124639556e-05,
|
| 2332 |
+
"loss": 2.9781,
|
| 2333 |
+
"step": 1855
|
| 2334 |
+
},
|
| 2335 |
+
{
|
| 2336 |
+
"epoch": 0.15,
|
| 2337 |
+
"learning_rate": 4.750451397310481e-05,
|
| 2338 |
+
"loss": 3.4727,
|
| 2339 |
+
"step": 1860
|
| 2340 |
+
},
|
| 2341 |
+
{
|
| 2342 |
+
"epoch": 0.15,
|
| 2343 |
+
"learning_rate": 4.749777669981406e-05,
|
| 2344 |
+
"loss": 3.0776,
|
| 2345 |
+
"step": 1865
|
| 2346 |
+
},
|
| 2347 |
+
{
|
| 2348 |
+
"epoch": 0.15,
|
| 2349 |
+
"learning_rate": 4.74910394265233e-05,
|
| 2350 |
+
"loss": 3.4074,
|
| 2351 |
+
"step": 1870
|
| 2352 |
+
},
|
| 2353 |
+
{
|
| 2354 |
+
"epoch": 0.15,
|
| 2355 |
+
"learning_rate": 4.7484302153232544e-05,
|
| 2356 |
+
"loss": 3.0213,
|
| 2357 |
+
"step": 1875
|
| 2358 |
+
},
|
| 2359 |
+
{
|
| 2360 |
+
"epoch": 0.15,
|
| 2361 |
+
"learning_rate": 4.747756487994179e-05,
|
| 2362 |
+
"loss": 3.2027,
|
| 2363 |
+
"step": 1880
|
| 2364 |
+
},
|
| 2365 |
+
{
|
| 2366 |
+
"epoch": 0.15,
|
| 2367 |
+
"learning_rate": 4.7470827606651035e-05,
|
| 2368 |
+
"loss": 3.1548,
|
| 2369 |
+
"step": 1885
|
| 2370 |
+
},
|
| 2371 |
+
{
|
| 2372 |
+
"epoch": 0.15,
|
| 2373 |
+
"learning_rate": 4.7464090333360283e-05,
|
| 2374 |
+
"loss": 3.3882,
|
| 2375 |
+
"step": 1890
|
| 2376 |
+
},
|
| 2377 |
+
{
|
| 2378 |
+
"epoch": 0.15,
|
| 2379 |
+
"learning_rate": 4.745735306006953e-05,
|
| 2380 |
+
"loss": 3.0727,
|
| 2381 |
+
"step": 1895
|
| 2382 |
+
},
|
| 2383 |
+
{
|
| 2384 |
+
"epoch": 0.15,
|
| 2385 |
+
"learning_rate": 4.745061578677878e-05,
|
| 2386 |
+
"loss": 2.8628,
|
| 2387 |
+
"step": 1900
|
| 2388 |
+
},
|
| 2389 |
+
{
|
| 2390 |
+
"epoch": 0.15,
|
| 2391 |
+
"learning_rate": 4.744387851348802e-05,
|
| 2392 |
+
"loss": 3.4395,
|
| 2393 |
+
"step": 1905
|
| 2394 |
+
},
|
| 2395 |
+
{
|
| 2396 |
+
"epoch": 0.15,
|
| 2397 |
+
"learning_rate": 4.743714124019727e-05,
|
| 2398 |
+
"loss": 3.1199,
|
| 2399 |
+
"step": 1910
|
| 2400 |
+
},
|
| 2401 |
+
{
|
| 2402 |
+
"epoch": 0.15,
|
| 2403 |
+
"learning_rate": 4.7430403966906514e-05,
|
| 2404 |
+
"loss": 3.2914,
|
| 2405 |
+
"step": 1915
|
| 2406 |
+
},
|
| 2407 |
+
{
|
| 2408 |
+
"epoch": 0.16,
|
| 2409 |
+
"learning_rate": 4.742366669361576e-05,
|
| 2410 |
+
"loss": 3.3878,
|
| 2411 |
+
"step": 1920
|
| 2412 |
+
},
|
| 2413 |
+
{
|
| 2414 |
+
"epoch": 0.16,
|
| 2415 |
+
"learning_rate": 4.7416929420325005e-05,
|
| 2416 |
+
"loss": 3.0218,
|
| 2417 |
+
"step": 1925
|
| 2418 |
+
},
|
| 2419 |
+
{
|
| 2420 |
+
"epoch": 0.16,
|
| 2421 |
+
"learning_rate": 4.7410192147034254e-05,
|
| 2422 |
+
"loss": 3.2209,
|
| 2423 |
+
"step": 1930
|
| 2424 |
+
},
|
| 2425 |
+
{
|
| 2426 |
+
"epoch": 0.16,
|
| 2427 |
+
"learning_rate": 4.74034548737435e-05,
|
| 2428 |
+
"loss": 3.049,
|
| 2429 |
+
"step": 1935
|
| 2430 |
+
},
|
| 2431 |
+
{
|
| 2432 |
+
"epoch": 0.16,
|
| 2433 |
+
"learning_rate": 4.7396717600452745e-05,
|
| 2434 |
+
"loss": 3.1507,
|
| 2435 |
+
"step": 1940
|
| 2436 |
+
},
|
| 2437 |
+
{
|
| 2438 |
+
"epoch": 0.16,
|
| 2439 |
+
"learning_rate": 4.7389980327161994e-05,
|
| 2440 |
+
"loss": 3.2019,
|
| 2441 |
+
"step": 1945
|
| 2442 |
+
},
|
| 2443 |
+
{
|
| 2444 |
+
"epoch": 0.16,
|
| 2445 |
+
"learning_rate": 4.738324305387124e-05,
|
| 2446 |
+
"loss": 3.1197,
|
| 2447 |
+
"step": 1950
|
| 2448 |
+
},
|
| 2449 |
+
{
|
| 2450 |
+
"epoch": 0.16,
|
| 2451 |
+
"learning_rate": 4.7376505780580485e-05,
|
| 2452 |
+
"loss": 3.4023,
|
| 2453 |
+
"step": 1955
|
| 2454 |
+
},
|
| 2455 |
+
{
|
| 2456 |
+
"epoch": 0.16,
|
| 2457 |
+
"learning_rate": 4.736976850728973e-05,
|
| 2458 |
+
"loss": 3.0877,
|
| 2459 |
+
"step": 1960
|
| 2460 |
+
},
|
| 2461 |
+
{
|
| 2462 |
+
"epoch": 0.16,
|
| 2463 |
+
"learning_rate": 4.7363031233998976e-05,
|
| 2464 |
+
"loss": 3.2648,
|
| 2465 |
+
"step": 1965
|
| 2466 |
+
},
|
| 2467 |
+
{
|
| 2468 |
+
"epoch": 0.16,
|
| 2469 |
+
"learning_rate": 4.7356293960708225e-05,
|
| 2470 |
+
"loss": 2.9336,
|
| 2471 |
+
"step": 1970
|
| 2472 |
+
},
|
| 2473 |
+
{
|
| 2474 |
+
"epoch": 0.16,
|
| 2475 |
+
"learning_rate": 4.7349556687417474e-05,
|
| 2476 |
+
"loss": 3.0737,
|
| 2477 |
+
"step": 1975
|
| 2478 |
+
},
|
| 2479 |
+
{
|
| 2480 |
+
"epoch": 0.16,
|
| 2481 |
+
"learning_rate": 4.7342819414126716e-05,
|
| 2482 |
+
"loss": 2.8234,
|
| 2483 |
+
"step": 1980
|
| 2484 |
+
},
|
| 2485 |
+
{
|
| 2486 |
+
"epoch": 0.16,
|
| 2487 |
+
"learning_rate": 4.7336082140835965e-05,
|
| 2488 |
+
"loss": 3.2141,
|
| 2489 |
+
"step": 1985
|
| 2490 |
+
},
|
| 2491 |
+
{
|
| 2492 |
+
"epoch": 0.16,
|
| 2493 |
+
"learning_rate": 4.7329344867545213e-05,
|
| 2494 |
+
"loss": 3.1565,
|
| 2495 |
+
"step": 1990
|
| 2496 |
+
},
|
| 2497 |
+
{
|
| 2498 |
+
"epoch": 0.16,
|
| 2499 |
+
"learning_rate": 4.7322607594254456e-05,
|
| 2500 |
+
"loss": 2.9412,
|
| 2501 |
+
"step": 1995
|
| 2502 |
+
},
|
| 2503 |
+
{
|
| 2504 |
+
"epoch": 0.16,
|
| 2505 |
+
"learning_rate": 4.7315870320963704e-05,
|
| 2506 |
+
"loss": 3.2878,
|
| 2507 |
+
"step": 2000
|
| 2508 |
+
},
|
| 2509 |
+
{
|
| 2510 |
+
"epoch": 0.16,
|
| 2511 |
+
"learning_rate": 4.7309133047672947e-05,
|
| 2512 |
+
"loss": 3.0794,
|
| 2513 |
+
"step": 2005
|
| 2514 |
+
},
|
| 2515 |
+
{
|
| 2516 |
+
"epoch": 0.16,
|
| 2517 |
+
"learning_rate": 4.7302395774382195e-05,
|
| 2518 |
+
"loss": 3.397,
|
| 2519 |
+
"step": 2010
|
| 2520 |
+
},
|
| 2521 |
+
{
|
| 2522 |
+
"epoch": 0.16,
|
| 2523 |
+
"learning_rate": 4.729565850109144e-05,
|
| 2524 |
+
"loss": 3.0795,
|
| 2525 |
+
"step": 2015
|
| 2526 |
+
},
|
| 2527 |
+
{
|
| 2528 |
+
"epoch": 0.16,
|
| 2529 |
+
"learning_rate": 4.7288921227800686e-05,
|
| 2530 |
+
"loss": 3.5331,
|
| 2531 |
+
"step": 2020
|
| 2532 |
+
},
|
| 2533 |
+
{
|
| 2534 |
+
"epoch": 0.16,
|
| 2535 |
+
"learning_rate": 4.7282183954509935e-05,
|
| 2536 |
+
"loss": 3.1282,
|
| 2537 |
+
"step": 2025
|
| 2538 |
+
},
|
| 2539 |
+
{
|
| 2540 |
+
"epoch": 0.16,
|
| 2541 |
+
"learning_rate": 4.727544668121918e-05,
|
| 2542 |
+
"loss": 3.1138,
|
| 2543 |
+
"step": 2030
|
| 2544 |
+
},
|
| 2545 |
+
{
|
| 2546 |
+
"epoch": 0.16,
|
| 2547 |
+
"learning_rate": 4.7268709407928426e-05,
|
| 2548 |
+
"loss": 3.1175,
|
| 2549 |
+
"step": 2035
|
| 2550 |
+
},
|
| 2551 |
+
{
|
| 2552 |
+
"epoch": 0.16,
|
| 2553 |
+
"learning_rate": 4.7261972134637675e-05,
|
| 2554 |
+
"loss": 3.0259,
|
| 2555 |
+
"step": 2040
|
| 2556 |
+
},
|
| 2557 |
+
{
|
| 2558 |
+
"epoch": 0.17,
|
| 2559 |
+
"learning_rate": 4.725523486134692e-05,
|
| 2560 |
+
"loss": 3.4783,
|
| 2561 |
+
"step": 2045
|
| 2562 |
+
},
|
| 2563 |
+
{
|
| 2564 |
+
"epoch": 0.17,
|
| 2565 |
+
"learning_rate": 4.724849758805616e-05,
|
| 2566 |
+
"loss": 3.1384,
|
| 2567 |
+
"step": 2050
|
| 2568 |
+
},
|
| 2569 |
+
{
|
| 2570 |
+
"epoch": 0.17,
|
| 2571 |
+
"learning_rate": 4.724176031476541e-05,
|
| 2572 |
+
"loss": 3.2739,
|
| 2573 |
+
"step": 2055
|
| 2574 |
+
},
|
| 2575 |
+
{
|
| 2576 |
+
"epoch": 0.17,
|
| 2577 |
+
"learning_rate": 4.723502304147466e-05,
|
| 2578 |
+
"loss": 3.0155,
|
| 2579 |
+
"step": 2060
|
| 2580 |
+
},
|
| 2581 |
+
{
|
| 2582 |
+
"epoch": 0.17,
|
| 2583 |
+
"learning_rate": 4.7228285768183906e-05,
|
| 2584 |
+
"loss": 3.1998,
|
| 2585 |
+
"step": 2065
|
| 2586 |
+
},
|
| 2587 |
+
{
|
| 2588 |
+
"epoch": 0.17,
|
| 2589 |
+
"learning_rate": 4.722154849489315e-05,
|
| 2590 |
+
"loss": 3.0975,
|
| 2591 |
+
"step": 2070
|
| 2592 |
+
},
|
| 2593 |
+
{
|
| 2594 |
+
"epoch": 0.17,
|
| 2595 |
+
"learning_rate": 4.72148112216024e-05,
|
| 2596 |
+
"loss": 3.2449,
|
| 2597 |
+
"step": 2075
|
| 2598 |
+
},
|
| 2599 |
+
{
|
| 2600 |
+
"epoch": 0.17,
|
| 2601 |
+
"learning_rate": 4.7208073948311646e-05,
|
| 2602 |
+
"loss": 2.9318,
|
| 2603 |
+
"step": 2080
|
| 2604 |
+
},
|
| 2605 |
+
{
|
| 2606 |
+
"epoch": 0.17,
|
| 2607 |
+
"learning_rate": 4.720133667502089e-05,
|
| 2608 |
+
"loss": 3.0997,
|
| 2609 |
+
"step": 2085
|
| 2610 |
+
},
|
| 2611 |
+
{
|
| 2612 |
+
"epoch": 0.17,
|
| 2613 |
+
"learning_rate": 4.719459940173013e-05,
|
| 2614 |
+
"loss": 2.9982,
|
| 2615 |
+
"step": 2090
|
| 2616 |
+
},
|
| 2617 |
+
{
|
| 2618 |
+
"epoch": 0.17,
|
| 2619 |
+
"learning_rate": 4.718786212843938e-05,
|
| 2620 |
+
"loss": 2.9443,
|
| 2621 |
+
"step": 2095
|
| 2622 |
+
},
|
| 2623 |
+
{
|
| 2624 |
+
"epoch": 0.17,
|
| 2625 |
+
"learning_rate": 4.718112485514863e-05,
|
| 2626 |
+
"loss": 2.889,
|
| 2627 |
+
"step": 2100
|
| 2628 |
+
},
|
| 2629 |
+
{
|
| 2630 |
+
"epoch": 0.17,
|
| 2631 |
+
"learning_rate": 4.717438758185787e-05,
|
| 2632 |
+
"loss": 3.4088,
|
| 2633 |
+
"step": 2105
|
| 2634 |
+
},
|
| 2635 |
+
{
|
| 2636 |
+
"epoch": 0.17,
|
| 2637 |
+
"learning_rate": 4.716765030856712e-05,
|
| 2638 |
+
"loss": 3.184,
|
| 2639 |
+
"step": 2110
|
| 2640 |
+
},
|
| 2641 |
+
{
|
| 2642 |
+
"epoch": 0.17,
|
| 2643 |
+
"learning_rate": 4.716091303527637e-05,
|
| 2644 |
+
"loss": 3.0915,
|
| 2645 |
+
"step": 2115
|
| 2646 |
+
},
|
| 2647 |
+
{
|
| 2648 |
+
"epoch": 0.17,
|
| 2649 |
+
"learning_rate": 4.7154175761985616e-05,
|
| 2650 |
+
"loss": 3.2379,
|
| 2651 |
+
"step": 2120
|
| 2652 |
+
},
|
| 2653 |
+
{
|
| 2654 |
+
"epoch": 0.17,
|
| 2655 |
+
"learning_rate": 4.714743848869486e-05,
|
| 2656 |
+
"loss": 3.1446,
|
| 2657 |
+
"step": 2125
|
| 2658 |
+
},
|
| 2659 |
+
{
|
| 2660 |
+
"epoch": 0.17,
|
| 2661 |
+
"learning_rate": 4.71407012154041e-05,
|
| 2662 |
+
"loss": 3.1694,
|
| 2663 |
+
"step": 2130
|
| 2664 |
+
},
|
| 2665 |
+
{
|
| 2666 |
+
"epoch": 0.17,
|
| 2667 |
+
"learning_rate": 4.713396394211335e-05,
|
| 2668 |
+
"loss": 3.8897,
|
| 2669 |
+
"step": 2135
|
| 2670 |
+
},
|
| 2671 |
+
{
|
| 2672 |
+
"epoch": 0.17,
|
| 2673 |
+
"learning_rate": 4.712722666882259e-05,
|
| 2674 |
+
"loss": 3.0934,
|
| 2675 |
+
"step": 2140
|
| 2676 |
+
},
|
| 2677 |
+
{
|
| 2678 |
+
"epoch": 0.17,
|
| 2679 |
+
"learning_rate": 4.712048939553184e-05,
|
| 2680 |
+
"loss": 3.1613,
|
| 2681 |
+
"step": 2145
|
| 2682 |
+
},
|
| 2683 |
+
{
|
| 2684 |
+
"epoch": 0.17,
|
| 2685 |
+
"learning_rate": 4.711375212224109e-05,
|
| 2686 |
+
"loss": 3.3189,
|
| 2687 |
+
"step": 2150
|
| 2688 |
+
},
|
| 2689 |
+
{
|
| 2690 |
+
"epoch": 0.17,
|
| 2691 |
+
"learning_rate": 4.710701484895034e-05,
|
| 2692 |
+
"loss": 3.218,
|
| 2693 |
+
"step": 2155
|
| 2694 |
+
},
|
| 2695 |
+
{
|
| 2696 |
+
"epoch": 0.17,
|
| 2697 |
+
"learning_rate": 4.710027757565958e-05,
|
| 2698 |
+
"loss": 3.1543,
|
| 2699 |
+
"step": 2160
|
| 2700 |
+
},
|
| 2701 |
+
{
|
| 2702 |
+
"epoch": 0.18,
|
| 2703 |
+
"learning_rate": 4.709354030236883e-05,
|
| 2704 |
+
"loss": 3.1454,
|
| 2705 |
+
"step": 2165
|
| 2706 |
+
},
|
| 2707 |
+
{
|
| 2708 |
+
"epoch": 0.18,
|
| 2709 |
+
"learning_rate": 4.708680302907808e-05,
|
| 2710 |
+
"loss": 2.9892,
|
| 2711 |
+
"step": 2170
|
| 2712 |
+
},
|
| 2713 |
+
{
|
| 2714 |
+
"epoch": 0.18,
|
| 2715 |
+
"learning_rate": 4.708006575578732e-05,
|
| 2716 |
+
"loss": 3.2591,
|
| 2717 |
+
"step": 2175
|
| 2718 |
+
},
|
| 2719 |
+
{
|
| 2720 |
+
"epoch": 0.18,
|
| 2721 |
+
"learning_rate": 4.707332848249656e-05,
|
| 2722 |
+
"loss": 2.9655,
|
| 2723 |
+
"step": 2180
|
| 2724 |
+
},
|
| 2725 |
+
{
|
| 2726 |
+
"epoch": 0.18,
|
| 2727 |
+
"learning_rate": 4.706659120920581e-05,
|
| 2728 |
+
"loss": 3.1065,
|
| 2729 |
+
"step": 2185
|
| 2730 |
+
},
|
| 2731 |
+
{
|
| 2732 |
+
"epoch": 0.18,
|
| 2733 |
+
"learning_rate": 4.705985393591506e-05,
|
| 2734 |
+
"loss": 2.8932,
|
| 2735 |
+
"step": 2190
|
| 2736 |
+
},
|
| 2737 |
+
{
|
| 2738 |
+
"epoch": 0.18,
|
| 2739 |
+
"learning_rate": 4.70531166626243e-05,
|
| 2740 |
+
"loss": 3.2734,
|
| 2741 |
+
"step": 2195
|
| 2742 |
+
},
|
| 2743 |
+
{
|
| 2744 |
+
"epoch": 0.18,
|
| 2745 |
+
"learning_rate": 4.704637938933355e-05,
|
| 2746 |
+
"loss": 3.4112,
|
| 2747 |
+
"step": 2200
|
| 2748 |
+
},
|
| 2749 |
+
{
|
| 2750 |
+
"epoch": 0.18,
|
| 2751 |
+
"learning_rate": 4.70396421160428e-05,
|
| 2752 |
+
"loss": 3.039,
|
| 2753 |
+
"step": 2205
|
| 2754 |
+
},
|
| 2755 |
+
{
|
| 2756 |
+
"epoch": 0.18,
|
| 2757 |
+
"learning_rate": 4.703290484275205e-05,
|
| 2758 |
+
"loss": 3.3834,
|
| 2759 |
+
"step": 2210
|
| 2760 |
+
},
|
| 2761 |
+
{
|
| 2762 |
+
"epoch": 0.18,
|
| 2763 |
+
"learning_rate": 4.702616756946129e-05,
|
| 2764 |
+
"loss": 3.2542,
|
| 2765 |
+
"step": 2215
|
| 2766 |
+
},
|
| 2767 |
+
{
|
| 2768 |
+
"epoch": 0.18,
|
| 2769 |
+
"learning_rate": 4.701943029617053e-05,
|
| 2770 |
+
"loss": 3.2943,
|
| 2771 |
+
"step": 2220
|
| 2772 |
+
},
|
| 2773 |
+
{
|
| 2774 |
+
"epoch": 0.18,
|
| 2775 |
+
"learning_rate": 4.701269302287978e-05,
|
| 2776 |
+
"loss": 3.1251,
|
| 2777 |
+
"step": 2225
|
| 2778 |
+
},
|
| 2779 |
+
{
|
| 2780 |
+
"epoch": 0.18,
|
| 2781 |
+
"learning_rate": 4.700595574958903e-05,
|
| 2782 |
+
"loss": 3.4983,
|
| 2783 |
+
"step": 2230
|
| 2784 |
+
},
|
| 2785 |
+
{
|
| 2786 |
+
"epoch": 0.18,
|
| 2787 |
+
"learning_rate": 4.699921847629827e-05,
|
| 2788 |
+
"loss": 3.0634,
|
| 2789 |
+
"step": 2235
|
| 2790 |
+
},
|
| 2791 |
+
{
|
| 2792 |
+
"epoch": 0.18,
|
| 2793 |
+
"learning_rate": 4.699248120300752e-05,
|
| 2794 |
+
"loss": 3.2068,
|
| 2795 |
+
"step": 2240
|
| 2796 |
+
},
|
| 2797 |
+
{
|
| 2798 |
+
"epoch": 0.18,
|
| 2799 |
+
"learning_rate": 4.698574392971677e-05,
|
| 2800 |
+
"loss": 3.0052,
|
| 2801 |
+
"step": 2245
|
| 2802 |
+
},
|
| 2803 |
+
{
|
| 2804 |
+
"epoch": 0.18,
|
| 2805 |
+
"learning_rate": 4.697900665642601e-05,
|
| 2806 |
+
"loss": 2.9226,
|
| 2807 |
+
"step": 2250
|
| 2808 |
+
},
|
| 2809 |
+
{
|
| 2810 |
+
"epoch": 0.18,
|
| 2811 |
+
"learning_rate": 4.697226938313526e-05,
|
| 2812 |
+
"loss": 3.0226,
|
| 2813 |
+
"step": 2255
|
| 2814 |
+
},
|
| 2815 |
+
{
|
| 2816 |
+
"epoch": 0.18,
|
| 2817 |
+
"learning_rate": 4.6965532109844504e-05,
|
| 2818 |
+
"loss": 3.5132,
|
| 2819 |
+
"step": 2260
|
| 2820 |
+
},
|
| 2821 |
+
{
|
| 2822 |
+
"epoch": 0.18,
|
| 2823 |
+
"learning_rate": 4.695879483655375e-05,
|
| 2824 |
+
"loss": 3.0079,
|
| 2825 |
+
"step": 2265
|
| 2826 |
+
},
|
| 2827 |
+
{
|
| 2828 |
+
"epoch": 0.18,
|
| 2829 |
+
"learning_rate": 4.6952057563262995e-05,
|
| 2830 |
+
"loss": 3.1758,
|
| 2831 |
+
"step": 2270
|
| 2832 |
+
},
|
| 2833 |
+
{
|
| 2834 |
+
"epoch": 0.18,
|
| 2835 |
+
"learning_rate": 4.6945320289972243e-05,
|
| 2836 |
+
"loss": 3.2862,
|
| 2837 |
+
"step": 2275
|
| 2838 |
+
},
|
| 2839 |
+
{
|
| 2840 |
+
"epoch": 0.18,
|
| 2841 |
+
"learning_rate": 4.693858301668149e-05,
|
| 2842 |
+
"loss": 3.2161,
|
| 2843 |
+
"step": 2280
|
| 2844 |
+
},
|
| 2845 |
+
{
|
| 2846 |
+
"epoch": 0.18,
|
| 2847 |
+
"learning_rate": 4.693184574339074e-05,
|
| 2848 |
+
"loss": 2.9678,
|
| 2849 |
+
"step": 2285
|
| 2850 |
+
},
|
| 2851 |
+
{
|
| 2852 |
+
"epoch": 0.19,
|
| 2853 |
+
"learning_rate": 4.692510847009998e-05,
|
| 2854 |
+
"loss": 3.1297,
|
| 2855 |
+
"step": 2290
|
| 2856 |
+
},
|
| 2857 |
+
{
|
| 2858 |
+
"epoch": 0.19,
|
| 2859 |
+
"learning_rate": 4.691837119680923e-05,
|
| 2860 |
+
"loss": 3.2068,
|
| 2861 |
+
"step": 2295
|
| 2862 |
+
},
|
| 2863 |
+
{
|
| 2864 |
+
"epoch": 0.19,
|
| 2865 |
+
"learning_rate": 4.691163392351848e-05,
|
| 2866 |
+
"loss": 2.9587,
|
| 2867 |
+
"step": 2300
|
| 2868 |
+
},
|
| 2869 |
+
{
|
| 2870 |
+
"epoch": 0.19,
|
| 2871 |
+
"learning_rate": 4.690489665022772e-05,
|
| 2872 |
+
"loss": 3.2795,
|
| 2873 |
+
"step": 2305
|
| 2874 |
+
},
|
| 2875 |
+
{
|
| 2876 |
+
"epoch": 0.19,
|
| 2877 |
+
"learning_rate": 4.6898159376936965e-05,
|
| 2878 |
+
"loss": 3.5198,
|
| 2879 |
+
"step": 2310
|
| 2880 |
+
},
|
| 2881 |
+
{
|
| 2882 |
+
"epoch": 0.19,
|
| 2883 |
+
"learning_rate": 4.6891422103646214e-05,
|
| 2884 |
+
"loss": 3.1278,
|
| 2885 |
+
"step": 2315
|
| 2886 |
+
},
|
| 2887 |
+
{
|
| 2888 |
+
"epoch": 0.19,
|
| 2889 |
+
"learning_rate": 4.688468483035546e-05,
|
| 2890 |
+
"loss": 3.1983,
|
| 2891 |
+
"step": 2320
|
| 2892 |
+
},
|
| 2893 |
+
{
|
| 2894 |
+
"epoch": 0.19,
|
| 2895 |
+
"learning_rate": 4.6877947557064705e-05,
|
| 2896 |
+
"loss": 2.9866,
|
| 2897 |
+
"step": 2325
|
| 2898 |
+
},
|
| 2899 |
+
{
|
| 2900 |
+
"epoch": 0.19,
|
| 2901 |
+
"learning_rate": 4.6871210283773954e-05,
|
| 2902 |
+
"loss": 3.1094,
|
| 2903 |
+
"step": 2330
|
| 2904 |
+
},
|
| 2905 |
+
{
|
| 2906 |
+
"epoch": 0.19,
|
| 2907 |
+
"learning_rate": 4.68644730104832e-05,
|
| 2908 |
+
"loss": 2.8821,
|
| 2909 |
+
"step": 2335
|
| 2910 |
+
},
|
| 2911 |
+
{
|
| 2912 |
+
"epoch": 0.19,
|
| 2913 |
+
"learning_rate": 4.6857735737192445e-05,
|
| 2914 |
+
"loss": 3.0871,
|
| 2915 |
+
"step": 2340
|
| 2916 |
+
},
|
| 2917 |
+
{
|
| 2918 |
+
"epoch": 0.19,
|
| 2919 |
+
"learning_rate": 4.6850998463901694e-05,
|
| 2920 |
+
"loss": 3.3086,
|
| 2921 |
+
"step": 2345
|
| 2922 |
+
},
|
| 2923 |
+
{
|
| 2924 |
+
"epoch": 0.19,
|
| 2925 |
+
"learning_rate": 4.6844261190610936e-05,
|
| 2926 |
+
"loss": 3.1617,
|
| 2927 |
+
"step": 2350
|
| 2928 |
+
},
|
| 2929 |
+
{
|
| 2930 |
+
"epoch": 0.19,
|
| 2931 |
+
"learning_rate": 4.6837523917320185e-05,
|
| 2932 |
+
"loss": 3.0705,
|
| 2933 |
+
"step": 2355
|
| 2934 |
+
},
|
| 2935 |
+
{
|
| 2936 |
+
"epoch": 0.19,
|
| 2937 |
+
"learning_rate": 4.683078664402943e-05,
|
| 2938 |
+
"loss": 3.3558,
|
| 2939 |
+
"step": 2360
|
| 2940 |
+
},
|
| 2941 |
+
{
|
| 2942 |
+
"epoch": 0.19,
|
| 2943 |
+
"learning_rate": 4.6824049370738676e-05,
|
| 2944 |
+
"loss": 3.2012,
|
| 2945 |
+
"step": 2365
|
| 2946 |
+
},
|
| 2947 |
+
{
|
| 2948 |
+
"epoch": 0.19,
|
| 2949 |
+
"learning_rate": 4.6817312097447925e-05,
|
| 2950 |
+
"loss": 3.1229,
|
| 2951 |
+
"step": 2370
|
| 2952 |
+
},
|
| 2953 |
+
{
|
| 2954 |
+
"epoch": 0.19,
|
| 2955 |
+
"learning_rate": 4.6810574824157173e-05,
|
| 2956 |
+
"loss": 3.1233,
|
| 2957 |
+
"step": 2375
|
| 2958 |
+
},
|
| 2959 |
+
{
|
| 2960 |
+
"epoch": 0.19,
|
| 2961 |
+
"learning_rate": 4.6803837550866416e-05,
|
| 2962 |
+
"loss": 3.4662,
|
| 2963 |
+
"step": 2380
|
| 2964 |
+
},
|
| 2965 |
+
{
|
| 2966 |
+
"epoch": 0.19,
|
| 2967 |
+
"learning_rate": 4.6797100277575664e-05,
|
| 2968 |
+
"loss": 3.5225,
|
| 2969 |
+
"step": 2385
|
| 2970 |
+
},
|
| 2971 |
+
{
|
| 2972 |
+
"epoch": 0.19,
|
| 2973 |
+
"learning_rate": 4.6790363004284906e-05,
|
| 2974 |
+
"loss": 3.1466,
|
| 2975 |
+
"step": 2390
|
| 2976 |
+
},
|
| 2977 |
+
{
|
| 2978 |
+
"epoch": 0.19,
|
| 2979 |
+
"learning_rate": 4.678362573099415e-05,
|
| 2980 |
+
"loss": 3.3767,
|
| 2981 |
+
"step": 2395
|
| 2982 |
+
},
|
| 2983 |
+
{
|
| 2984 |
+
"epoch": 0.19,
|
| 2985 |
+
"learning_rate": 4.67768884577034e-05,
|
| 2986 |
+
"loss": 3.1327,
|
| 2987 |
+
"step": 2400
|
| 2988 |
+
},
|
| 2989 |
+
{
|
| 2990 |
+
"epoch": 0.19,
|
| 2991 |
+
"learning_rate": 4.6770151184412646e-05,
|
| 2992 |
+
"loss": 3.0657,
|
| 2993 |
+
"step": 2405
|
| 2994 |
+
},
|
| 2995 |
+
{
|
| 2996 |
+
"epoch": 0.19,
|
| 2997 |
+
"learning_rate": 4.6763413911121895e-05,
|
| 2998 |
+
"loss": 3.238,
|
| 2999 |
+
"step": 2410
|
| 3000 |
+
},
|
| 3001 |
+
{
|
| 3002 |
+
"epoch": 0.2,
|
| 3003 |
+
"learning_rate": 4.675667663783114e-05,
|
| 3004 |
+
"loss": 3.0663,
|
| 3005 |
+
"step": 2415
|
| 3006 |
+
},
|
| 3007 |
+
{
|
| 3008 |
+
"epoch": 0.2,
|
| 3009 |
+
"learning_rate": 4.6749939364540386e-05,
|
| 3010 |
+
"loss": 3.2997,
|
| 3011 |
+
"step": 2420
|
| 3012 |
+
},
|
| 3013 |
+
{
|
| 3014 |
+
"epoch": 0.2,
|
| 3015 |
+
"learning_rate": 4.6743202091249635e-05,
|
| 3016 |
+
"loss": 3.1138,
|
| 3017 |
+
"step": 2425
|
| 3018 |
+
},
|
| 3019 |
+
{
|
| 3020 |
+
"epoch": 0.2,
|
| 3021 |
+
"learning_rate": 4.673646481795888e-05,
|
| 3022 |
+
"loss": 3.1481,
|
| 3023 |
+
"step": 2430
|
| 3024 |
+
},
|
| 3025 |
+
{
|
| 3026 |
+
"epoch": 0.2,
|
| 3027 |
+
"learning_rate": 4.6729727544668126e-05,
|
| 3028 |
+
"loss": 3.1374,
|
| 3029 |
+
"step": 2435
|
| 3030 |
+
},
|
| 3031 |
+
{
|
| 3032 |
+
"epoch": 0.2,
|
| 3033 |
+
"learning_rate": 4.672299027137737e-05,
|
| 3034 |
+
"loss": 3.137,
|
| 3035 |
+
"step": 2440
|
| 3036 |
+
},
|
| 3037 |
+
{
|
| 3038 |
+
"epoch": 0.2,
|
| 3039 |
+
"learning_rate": 4.671625299808662e-05,
|
| 3040 |
+
"loss": 3.0878,
|
| 3041 |
+
"step": 2445
|
| 3042 |
+
},
|
| 3043 |
+
{
|
| 3044 |
+
"epoch": 0.2,
|
| 3045 |
+
"learning_rate": 4.670951572479586e-05,
|
| 3046 |
+
"loss": 3.0358,
|
| 3047 |
+
"step": 2450
|
| 3048 |
+
},
|
| 3049 |
+
{
|
| 3050 |
+
"epoch": 0.2,
|
| 3051 |
+
"learning_rate": 4.670277845150511e-05,
|
| 3052 |
+
"loss": 3.0953,
|
| 3053 |
+
"step": 2455
|
| 3054 |
+
},
|
| 3055 |
+
{
|
| 3056 |
+
"epoch": 0.2,
|
| 3057 |
+
"learning_rate": 4.669604117821436e-05,
|
| 3058 |
+
"loss": 3.0776,
|
| 3059 |
+
"step": 2460
|
| 3060 |
+
},
|
| 3061 |
+
{
|
| 3062 |
+
"epoch": 0.2,
|
| 3063 |
+
"learning_rate": 4.6689303904923606e-05,
|
| 3064 |
+
"loss": 3.4543,
|
| 3065 |
+
"step": 2465
|
| 3066 |
+
},
|
| 3067 |
+
{
|
| 3068 |
+
"epoch": 0.2,
|
| 3069 |
+
"learning_rate": 4.668256663163285e-05,
|
| 3070 |
+
"loss": 3.1356,
|
| 3071 |
+
"step": 2470
|
| 3072 |
+
},
|
| 3073 |
+
{
|
| 3074 |
+
"epoch": 0.2,
|
| 3075 |
+
"learning_rate": 4.66758293583421e-05,
|
| 3076 |
+
"loss": 3.3752,
|
| 3077 |
+
"step": 2475
|
| 3078 |
+
},
|
| 3079 |
+
{
|
| 3080 |
+
"epoch": 0.2,
|
| 3081 |
+
"learning_rate": 4.666909208505134e-05,
|
| 3082 |
+
"loss": 3.0158,
|
| 3083 |
+
"step": 2480
|
| 3084 |
+
},
|
| 3085 |
+
{
|
| 3086 |
+
"epoch": 0.2,
|
| 3087 |
+
"learning_rate": 4.666235481176059e-05,
|
| 3088 |
+
"loss": 3.1832,
|
| 3089 |
+
"step": 2485
|
| 3090 |
+
},
|
| 3091 |
+
{
|
| 3092 |
+
"epoch": 0.2,
|
| 3093 |
+
"learning_rate": 4.665561753846983e-05,
|
| 3094 |
+
"loss": 3.2367,
|
| 3095 |
+
"step": 2490
|
| 3096 |
+
},
|
| 3097 |
+
{
|
| 3098 |
+
"epoch": 0.2,
|
| 3099 |
+
"learning_rate": 4.664888026517908e-05,
|
| 3100 |
+
"loss": 3.2553,
|
| 3101 |
+
"step": 2495
|
| 3102 |
+
},
|
| 3103 |
+
{
|
| 3104 |
+
"epoch": 0.2,
|
| 3105 |
+
"learning_rate": 4.664214299188833e-05,
|
| 3106 |
+
"loss": 3.3684,
|
| 3107 |
+
"step": 2500
|
| 3108 |
+
},
|
| 3109 |
+
{
|
| 3110 |
+
"epoch": 0.2,
|
| 3111 |
+
"learning_rate": 4.663540571859757e-05,
|
| 3112 |
+
"loss": 3.157,
|
| 3113 |
+
"step": 2505
|
| 3114 |
+
},
|
| 3115 |
+
{
|
| 3116 |
+
"epoch": 0.2,
|
| 3117 |
+
"learning_rate": 4.662866844530682e-05,
|
| 3118 |
+
"loss": 3.3339,
|
| 3119 |
+
"step": 2510
|
| 3120 |
+
},
|
| 3121 |
+
{
|
| 3122 |
+
"epoch": 0.2,
|
| 3123 |
+
"learning_rate": 4.662193117201607e-05,
|
| 3124 |
+
"loss": 3.142,
|
| 3125 |
+
"step": 2515
|
| 3126 |
+
},
|
| 3127 |
+
{
|
| 3128 |
+
"epoch": 0.2,
|
| 3129 |
+
"learning_rate": 4.661519389872531e-05,
|
| 3130 |
+
"loss": 3.0608,
|
| 3131 |
+
"step": 2520
|
| 3132 |
+
},
|
| 3133 |
+
{
|
| 3134 |
+
"epoch": 0.2,
|
| 3135 |
+
"learning_rate": 4.660845662543455e-05,
|
| 3136 |
+
"loss": 3.0232,
|
| 3137 |
+
"step": 2525
|
| 3138 |
+
},
|
| 3139 |
+
{
|
| 3140 |
+
"epoch": 0.2,
|
| 3141 |
+
"learning_rate": 4.66017193521438e-05,
|
| 3142 |
+
"loss": 3.0909,
|
| 3143 |
+
"step": 2530
|
| 3144 |
+
},
|
| 3145 |
+
{
|
| 3146 |
+
"epoch": 0.2,
|
| 3147 |
+
"learning_rate": 4.659498207885305e-05,
|
| 3148 |
+
"loss": 3.0394,
|
| 3149 |
+
"step": 2535
|
| 3150 |
+
},
|
| 3151 |
+
{
|
| 3152 |
+
"epoch": 0.21,
|
| 3153 |
+
"learning_rate": 4.65882448055623e-05,
|
| 3154 |
+
"loss": 3.3022,
|
| 3155 |
+
"step": 2540
|
| 3156 |
+
},
|
| 3157 |
+
{
|
| 3158 |
+
"epoch": 0.21,
|
| 3159 |
+
"learning_rate": 4.658150753227154e-05,
|
| 3160 |
+
"loss": 3.0824,
|
| 3161 |
+
"step": 2545
|
| 3162 |
+
},
|
| 3163 |
+
{
|
| 3164 |
+
"epoch": 0.21,
|
| 3165 |
+
"learning_rate": 4.657477025898079e-05,
|
| 3166 |
+
"loss": 3.0483,
|
| 3167 |
+
"step": 2550
|
| 3168 |
+
},
|
| 3169 |
+
{
|
| 3170 |
+
"epoch": 0.21,
|
| 3171 |
+
"learning_rate": 4.656803298569004e-05,
|
| 3172 |
+
"loss": 3.0755,
|
| 3173 |
+
"step": 2555
|
| 3174 |
+
},
|
| 3175 |
+
{
|
| 3176 |
+
"epoch": 0.21,
|
| 3177 |
+
"learning_rate": 4.656129571239928e-05,
|
| 3178 |
+
"loss": 3.4422,
|
| 3179 |
+
"step": 2560
|
| 3180 |
+
},
|
| 3181 |
+
{
|
| 3182 |
+
"epoch": 0.21,
|
| 3183 |
+
"learning_rate": 4.655455843910852e-05,
|
| 3184 |
+
"loss": 3.1593,
|
| 3185 |
+
"step": 2565
|
| 3186 |
+
},
|
| 3187 |
+
{
|
| 3188 |
+
"epoch": 0.21,
|
| 3189 |
+
"learning_rate": 4.654782116581777e-05,
|
| 3190 |
+
"loss": 3.2496,
|
| 3191 |
+
"step": 2570
|
| 3192 |
+
},
|
| 3193 |
+
{
|
| 3194 |
+
"epoch": 0.21,
|
| 3195 |
+
"learning_rate": 4.654108389252702e-05,
|
| 3196 |
+
"loss": 3.3281,
|
| 3197 |
+
"step": 2575
|
| 3198 |
+
},
|
| 3199 |
+
{
|
| 3200 |
+
"epoch": 0.21,
|
| 3201 |
+
"learning_rate": 4.653434661923626e-05,
|
| 3202 |
+
"loss": 2.9645,
|
| 3203 |
+
"step": 2580
|
| 3204 |
+
},
|
| 3205 |
+
{
|
| 3206 |
+
"epoch": 0.21,
|
| 3207 |
+
"learning_rate": 4.652760934594551e-05,
|
| 3208 |
+
"loss": 3.4181,
|
| 3209 |
+
"step": 2585
|
| 3210 |
+
},
|
| 3211 |
+
{
|
| 3212 |
+
"epoch": 0.21,
|
| 3213 |
+
"learning_rate": 4.652087207265476e-05,
|
| 3214 |
+
"loss": 3.1505,
|
| 3215 |
+
"step": 2590
|
| 3216 |
+
},
|
| 3217 |
+
{
|
| 3218 |
+
"epoch": 0.21,
|
| 3219 |
+
"learning_rate": 4.6514134799364e-05,
|
| 3220 |
+
"loss": 3.2713,
|
| 3221 |
+
"step": 2595
|
| 3222 |
+
},
|
| 3223 |
+
{
|
| 3224 |
+
"epoch": 0.21,
|
| 3225 |
+
"learning_rate": 4.650739752607325e-05,
|
| 3226 |
+
"loss": 3.0737,
|
| 3227 |
+
"step": 2600
|
| 3228 |
+
},
|
| 3229 |
+
{
|
| 3230 |
+
"epoch": 0.21,
|
| 3231 |
+
"learning_rate": 4.65006602527825e-05,
|
| 3232 |
+
"loss": 3.1046,
|
| 3233 |
+
"step": 2605
|
| 3234 |
+
},
|
| 3235 |
+
{
|
| 3236 |
+
"epoch": 0.21,
|
| 3237 |
+
"learning_rate": 4.649392297949174e-05,
|
| 3238 |
+
"loss": 3.0764,
|
| 3239 |
+
"step": 2610
|
| 3240 |
+
},
|
| 3241 |
+
{
|
| 3242 |
+
"epoch": 0.21,
|
| 3243 |
+
"learning_rate": 4.6487185706200984e-05,
|
| 3244 |
+
"loss": 3.1572,
|
| 3245 |
+
"step": 2615
|
| 3246 |
+
},
|
| 3247 |
+
{
|
| 3248 |
+
"epoch": 0.21,
|
| 3249 |
+
"learning_rate": 4.648044843291023e-05,
|
| 3250 |
+
"loss": 3.1946,
|
| 3251 |
+
"step": 2620
|
| 3252 |
+
},
|
| 3253 |
+
{
|
| 3254 |
+
"epoch": 0.21,
|
| 3255 |
+
"learning_rate": 4.647371115961948e-05,
|
| 3256 |
+
"loss": 3.0024,
|
| 3257 |
+
"step": 2625
|
| 3258 |
+
},
|
| 3259 |
+
{
|
| 3260 |
+
"epoch": 0.21,
|
| 3261 |
+
"learning_rate": 4.646697388632873e-05,
|
| 3262 |
+
"loss": 3.1511,
|
| 3263 |
+
"step": 2630
|
| 3264 |
+
},
|
| 3265 |
+
{
|
| 3266 |
+
"epoch": 0.21,
|
| 3267 |
+
"learning_rate": 4.646023661303797e-05,
|
| 3268 |
+
"loss": 3.333,
|
| 3269 |
+
"step": 2635
|
| 3270 |
+
},
|
| 3271 |
+
{
|
| 3272 |
+
"epoch": 0.21,
|
| 3273 |
+
"learning_rate": 4.645349933974722e-05,
|
| 3274 |
+
"loss": 2.9438,
|
| 3275 |
+
"step": 2640
|
| 3276 |
+
},
|
| 3277 |
+
{
|
| 3278 |
+
"epoch": 0.21,
|
| 3279 |
+
"learning_rate": 4.644676206645647e-05,
|
| 3280 |
+
"loss": 3.327,
|
| 3281 |
+
"step": 2645
|
| 3282 |
+
},
|
| 3283 |
+
{
|
| 3284 |
+
"epoch": 0.21,
|
| 3285 |
+
"learning_rate": 4.644002479316571e-05,
|
| 3286 |
+
"loss": 3.0695,
|
| 3287 |
+
"step": 2650
|
| 3288 |
+
},
|
| 3289 |
+
{
|
| 3290 |
+
"epoch": 0.21,
|
| 3291 |
+
"learning_rate": 4.6433287519874954e-05,
|
| 3292 |
+
"loss": 3.1144,
|
| 3293 |
+
"step": 2655
|
| 3294 |
+
},
|
| 3295 |
+
{
|
| 3296 |
+
"epoch": 0.22,
|
| 3297 |
+
"learning_rate": 4.64265502465842e-05,
|
| 3298 |
+
"loss": 3.4706,
|
| 3299 |
+
"step": 2660
|
| 3300 |
+
},
|
| 3301 |
+
{
|
| 3302 |
+
"epoch": 0.22,
|
| 3303 |
+
"learning_rate": 4.641981297329345e-05,
|
| 3304 |
+
"loss": 3.1941,
|
| 3305 |
+
"step": 2665
|
| 3306 |
+
},
|
| 3307 |
+
{
|
| 3308 |
+
"epoch": 0.22,
|
| 3309 |
+
"learning_rate": 4.6413075700002694e-05,
|
| 3310 |
+
"loss": 3.1067,
|
| 3311 |
+
"step": 2670
|
| 3312 |
+
},
|
| 3313 |
+
{
|
| 3314 |
+
"epoch": 0.22,
|
| 3315 |
+
"learning_rate": 4.640633842671194e-05,
|
| 3316 |
+
"loss": 3.1044,
|
| 3317 |
+
"step": 2675
|
| 3318 |
+
},
|
| 3319 |
+
{
|
| 3320 |
+
"epoch": 0.22,
|
| 3321 |
+
"learning_rate": 4.639960115342119e-05,
|
| 3322 |
+
"loss": 3.1814,
|
| 3323 |
+
"step": 2680
|
| 3324 |
+
},
|
| 3325 |
+
{
|
| 3326 |
+
"epoch": 0.22,
|
| 3327 |
+
"learning_rate": 4.639286388013044e-05,
|
| 3328 |
+
"loss": 3.2541,
|
| 3329 |
+
"step": 2685
|
| 3330 |
+
},
|
| 3331 |
+
{
|
| 3332 |
+
"epoch": 0.22,
|
| 3333 |
+
"learning_rate": 4.638612660683968e-05,
|
| 3334 |
+
"loss": 3.1815,
|
| 3335 |
+
"step": 2690
|
| 3336 |
+
},
|
| 3337 |
+
{
|
| 3338 |
+
"epoch": 0.22,
|
| 3339 |
+
"learning_rate": 4.6379389333548925e-05,
|
| 3340 |
+
"loss": 3.1444,
|
| 3341 |
+
"step": 2695
|
| 3342 |
+
},
|
| 3343 |
+
{
|
| 3344 |
+
"epoch": 0.22,
|
| 3345 |
+
"learning_rate": 4.6372652060258174e-05,
|
| 3346 |
+
"loss": 3.0728,
|
| 3347 |
+
"step": 2700
|
| 3348 |
+
},
|
| 3349 |
+
{
|
| 3350 |
+
"epoch": 0.22,
|
| 3351 |
+
"learning_rate": 4.6365914786967416e-05,
|
| 3352 |
+
"loss": 3.2032,
|
| 3353 |
+
"step": 2705
|
| 3354 |
+
},
|
| 3355 |
+
{
|
| 3356 |
+
"epoch": 0.22,
|
| 3357 |
+
"learning_rate": 4.6359177513676665e-05,
|
| 3358 |
+
"loss": 2.8255,
|
| 3359 |
+
"step": 2710
|
| 3360 |
+
},
|
| 3361 |
+
{
|
| 3362 |
+
"epoch": 0.22,
|
| 3363 |
+
"learning_rate": 4.6352440240385914e-05,
|
| 3364 |
+
"loss": 3.1325,
|
| 3365 |
+
"step": 2715
|
| 3366 |
+
},
|
| 3367 |
+
{
|
| 3368 |
+
"epoch": 0.22,
|
| 3369 |
+
"learning_rate": 4.634570296709516e-05,
|
| 3370 |
+
"loss": 3.2602,
|
| 3371 |
+
"step": 2720
|
| 3372 |
+
},
|
| 3373 |
+
{
|
| 3374 |
+
"epoch": 0.22,
|
| 3375 |
+
"learning_rate": 4.6338965693804405e-05,
|
| 3376 |
+
"loss": 2.9978,
|
| 3377 |
+
"step": 2725
|
| 3378 |
+
},
|
| 3379 |
+
{
|
| 3380 |
+
"epoch": 0.22,
|
| 3381 |
+
"learning_rate": 4.6332228420513654e-05,
|
| 3382 |
+
"loss": 3.3716,
|
| 3383 |
+
"step": 2730
|
| 3384 |
+
},
|
| 3385 |
+
{
|
| 3386 |
+
"epoch": 0.22,
|
| 3387 |
+
"learning_rate": 4.63254911472229e-05,
|
| 3388 |
+
"loss": 3.2488,
|
| 3389 |
+
"step": 2735
|
| 3390 |
+
},
|
| 3391 |
+
{
|
| 3392 |
+
"epoch": 0.22,
|
| 3393 |
+
"learning_rate": 4.6318753873932145e-05,
|
| 3394 |
+
"loss": 2.9851,
|
| 3395 |
+
"step": 2740
|
| 3396 |
+
},
|
| 3397 |
+
{
|
| 3398 |
+
"epoch": 0.22,
|
| 3399 |
+
"learning_rate": 4.631201660064139e-05,
|
| 3400 |
+
"loss": 2.9942,
|
| 3401 |
+
"step": 2745
|
| 3402 |
+
},
|
| 3403 |
+
{
|
| 3404 |
+
"epoch": 0.22,
|
| 3405 |
+
"learning_rate": 4.6305279327350636e-05,
|
| 3406 |
+
"loss": 3.2165,
|
| 3407 |
+
"step": 2750
|
| 3408 |
+
},
|
| 3409 |
+
{
|
| 3410 |
+
"epoch": 0.22,
|
| 3411 |
+
"learning_rate": 4.6298542054059884e-05,
|
| 3412 |
+
"loss": 3.0885,
|
| 3413 |
+
"step": 2755
|
| 3414 |
+
},
|
| 3415 |
+
{
|
| 3416 |
+
"epoch": 0.22,
|
| 3417 |
+
"learning_rate": 4.6291804780769127e-05,
|
| 3418 |
+
"loss": 3.0966,
|
| 3419 |
+
"step": 2760
|
| 3420 |
+
},
|
| 3421 |
+
{
|
| 3422 |
+
"epoch": 0.22,
|
| 3423 |
+
"learning_rate": 4.6285067507478375e-05,
|
| 3424 |
+
"loss": 3.2696,
|
| 3425 |
+
"step": 2765
|
| 3426 |
+
},
|
| 3427 |
+
{
|
| 3428 |
+
"epoch": 0.22,
|
| 3429 |
+
"learning_rate": 4.6278330234187624e-05,
|
| 3430 |
+
"loss": 3.0802,
|
| 3431 |
+
"step": 2770
|
| 3432 |
+
},
|
| 3433 |
+
{
|
| 3434 |
+
"epoch": 0.22,
|
| 3435 |
+
"learning_rate": 4.627159296089687e-05,
|
| 3436 |
+
"loss": 3.2361,
|
| 3437 |
+
"step": 2775
|
| 3438 |
+
},
|
| 3439 |
+
{
|
| 3440 |
+
"epoch": 0.22,
|
| 3441 |
+
"learning_rate": 4.6264855687606115e-05,
|
| 3442 |
+
"loss": 3.0949,
|
| 3443 |
+
"step": 2780
|
| 3444 |
+
},
|
| 3445 |
+
{
|
| 3446 |
+
"epoch": 0.23,
|
| 3447 |
+
"learning_rate": 4.625811841431536e-05,
|
| 3448 |
+
"loss": 3.3435,
|
| 3449 |
+
"step": 2785
|
| 3450 |
+
},
|
| 3451 |
+
{
|
| 3452 |
+
"epoch": 0.23,
|
| 3453 |
+
"learning_rate": 4.6251381141024606e-05,
|
| 3454 |
+
"loss": 3.3882,
|
| 3455 |
+
"step": 2790
|
| 3456 |
+
},
|
| 3457 |
+
{
|
| 3458 |
+
"epoch": 0.23,
|
| 3459 |
+
"learning_rate": 4.6244643867733855e-05,
|
| 3460 |
+
"loss": 2.8436,
|
| 3461 |
+
"step": 2795
|
| 3462 |
+
},
|
| 3463 |
+
{
|
| 3464 |
+
"epoch": 0.23,
|
| 3465 |
+
"learning_rate": 4.62379065944431e-05,
|
| 3466 |
+
"loss": 3.3431,
|
| 3467 |
+
"step": 2800
|
| 3468 |
+
},
|
| 3469 |
+
{
|
| 3470 |
+
"epoch": 0.23,
|
| 3471 |
+
"learning_rate": 4.6231169321152346e-05,
|
| 3472 |
+
"loss": 3.4118,
|
| 3473 |
+
"step": 2805
|
| 3474 |
+
},
|
| 3475 |
+
{
|
| 3476 |
+
"epoch": 0.23,
|
| 3477 |
+
"learning_rate": 4.6224432047861595e-05,
|
| 3478 |
+
"loss": 3.1914,
|
| 3479 |
+
"step": 2810
|
| 3480 |
+
},
|
| 3481 |
+
{
|
| 3482 |
+
"epoch": 0.23,
|
| 3483 |
+
"learning_rate": 4.621769477457084e-05,
|
| 3484 |
+
"loss": 3.061,
|
| 3485 |
+
"step": 2815
|
| 3486 |
+
},
|
| 3487 |
+
{
|
| 3488 |
+
"epoch": 0.23,
|
| 3489 |
+
"learning_rate": 4.6210957501280086e-05,
|
| 3490 |
+
"loss": 3.1028,
|
| 3491 |
+
"step": 2820
|
| 3492 |
+
},
|
| 3493 |
+
{
|
| 3494 |
+
"epoch": 0.23,
|
| 3495 |
+
"learning_rate": 4.620422022798933e-05,
|
| 3496 |
+
"loss": 3.0601,
|
| 3497 |
+
"step": 2825
|
| 3498 |
+
},
|
| 3499 |
+
{
|
| 3500 |
+
"epoch": 0.23,
|
| 3501 |
+
"learning_rate": 4.619748295469858e-05,
|
| 3502 |
+
"loss": 3.3313,
|
| 3503 |
+
"step": 2830
|
| 3504 |
+
},
|
| 3505 |
+
{
|
| 3506 |
+
"epoch": 0.23,
|
| 3507 |
+
"learning_rate": 4.619074568140782e-05,
|
| 3508 |
+
"loss": 3.1585,
|
| 3509 |
+
"step": 2835
|
| 3510 |
+
},
|
| 3511 |
+
{
|
| 3512 |
+
"epoch": 0.23,
|
| 3513 |
+
"learning_rate": 4.618400840811707e-05,
|
| 3514 |
+
"loss": 2.9804,
|
| 3515 |
+
"step": 2840
|
| 3516 |
+
},
|
| 3517 |
+
{
|
| 3518 |
+
"epoch": 0.23,
|
| 3519 |
+
"learning_rate": 4.617727113482632e-05,
|
| 3520 |
+
"loss": 3.197,
|
| 3521 |
+
"step": 2845
|
| 3522 |
+
},
|
| 3523 |
+
{
|
| 3524 |
+
"epoch": 0.23,
|
| 3525 |
+
"learning_rate": 4.6170533861535566e-05,
|
| 3526 |
+
"loss": 2.823,
|
| 3527 |
+
"step": 2850
|
| 3528 |
+
},
|
| 3529 |
+
{
|
| 3530 |
+
"epoch": 0.23,
|
| 3531 |
+
"learning_rate": 4.616379658824481e-05,
|
| 3532 |
+
"loss": 3.0995,
|
| 3533 |
+
"step": 2855
|
| 3534 |
+
},
|
| 3535 |
+
{
|
| 3536 |
+
"epoch": 0.23,
|
| 3537 |
+
"learning_rate": 4.6157059314954057e-05,
|
| 3538 |
+
"loss": 3.2473,
|
| 3539 |
+
"step": 2860
|
| 3540 |
+
},
|
| 3541 |
+
{
|
| 3542 |
+
"epoch": 0.23,
|
| 3543 |
+
"learning_rate": 4.61503220416633e-05,
|
| 3544 |
+
"loss": 3.3361,
|
| 3545 |
+
"step": 2865
|
| 3546 |
+
},
|
| 3547 |
+
{
|
| 3548 |
+
"epoch": 0.23,
|
| 3549 |
+
"learning_rate": 4.614358476837254e-05,
|
| 3550 |
+
"loss": 2.9229,
|
| 3551 |
+
"step": 2870
|
| 3552 |
+
},
|
| 3553 |
+
{
|
| 3554 |
+
"epoch": 0.23,
|
| 3555 |
+
"learning_rate": 4.613684749508179e-05,
|
| 3556 |
+
"loss": 3.2953,
|
| 3557 |
+
"step": 2875
|
| 3558 |
+
},
|
| 3559 |
+
{
|
| 3560 |
+
"epoch": 0.23,
|
| 3561 |
+
"learning_rate": 4.613011022179104e-05,
|
| 3562 |
+
"loss": 3.1187,
|
| 3563 |
+
"step": 2880
|
| 3564 |
+
},
|
| 3565 |
+
{
|
| 3566 |
+
"epoch": 0.23,
|
| 3567 |
+
"learning_rate": 4.612337294850029e-05,
|
| 3568 |
+
"loss": 3.009,
|
| 3569 |
+
"step": 2885
|
| 3570 |
+
},
|
| 3571 |
+
{
|
| 3572 |
+
"epoch": 0.23,
|
| 3573 |
+
"learning_rate": 4.611663567520953e-05,
|
| 3574 |
+
"loss": 3.5421,
|
| 3575 |
+
"step": 2890
|
| 3576 |
+
},
|
| 3577 |
+
{
|
| 3578 |
+
"epoch": 0.23,
|
| 3579 |
+
"learning_rate": 4.610989840191878e-05,
|
| 3580 |
+
"loss": 3.0521,
|
| 3581 |
+
"step": 2895
|
| 3582 |
+
},
|
| 3583 |
+
{
|
| 3584 |
+
"epoch": 0.23,
|
| 3585 |
+
"learning_rate": 4.610316112862803e-05,
|
| 3586 |
+
"loss": 2.9707,
|
| 3587 |
+
"step": 2900
|
| 3588 |
+
},
|
| 3589 |
+
{
|
| 3590 |
+
"epoch": 0.23,
|
| 3591 |
+
"learning_rate": 4.609642385533727e-05,
|
| 3592 |
+
"loss": 3.0522,
|
| 3593 |
+
"step": 2905
|
| 3594 |
+
},
|
| 3595 |
+
{
|
| 3596 |
+
"epoch": 0.24,
|
| 3597 |
+
"learning_rate": 4.608968658204652e-05,
|
| 3598 |
+
"loss": 3.1429,
|
| 3599 |
+
"step": 2910
|
| 3600 |
+
},
|
| 3601 |
+
{
|
| 3602 |
+
"epoch": 0.24,
|
| 3603 |
+
"learning_rate": 4.608294930875576e-05,
|
| 3604 |
+
"loss": 3.1433,
|
| 3605 |
+
"step": 2915
|
| 3606 |
+
},
|
| 3607 |
+
{
|
| 3608 |
+
"epoch": 0.24,
|
| 3609 |
+
"learning_rate": 4.607621203546501e-05,
|
| 3610 |
+
"loss": 3.0362,
|
| 3611 |
+
"step": 2920
|
| 3612 |
+
},
|
| 3613 |
+
{
|
| 3614 |
+
"epoch": 0.24,
|
| 3615 |
+
"learning_rate": 4.606947476217425e-05,
|
| 3616 |
+
"loss": 3.0698,
|
| 3617 |
+
"step": 2925
|
| 3618 |
+
},
|
| 3619 |
+
{
|
| 3620 |
+
"epoch": 0.24,
|
| 3621 |
+
"learning_rate": 4.60627374888835e-05,
|
| 3622 |
+
"loss": 2.9708,
|
| 3623 |
+
"step": 2930
|
| 3624 |
+
},
|
| 3625 |
+
{
|
| 3626 |
+
"epoch": 0.24,
|
| 3627 |
+
"learning_rate": 4.605600021559275e-05,
|
| 3628 |
+
"loss": 2.929,
|
| 3629 |
+
"step": 2935
|
| 3630 |
+
},
|
| 3631 |
+
{
|
| 3632 |
+
"epoch": 0.24,
|
| 3633 |
+
"learning_rate": 4.6049262942302e-05,
|
| 3634 |
+
"loss": 3.0652,
|
| 3635 |
+
"step": 2940
|
| 3636 |
+
},
|
| 3637 |
+
{
|
| 3638 |
+
"epoch": 0.24,
|
| 3639 |
+
"learning_rate": 4.604252566901124e-05,
|
| 3640 |
+
"loss": 3.0888,
|
| 3641 |
+
"step": 2945
|
| 3642 |
+
},
|
| 3643 |
+
{
|
| 3644 |
+
"epoch": 0.24,
|
| 3645 |
+
"learning_rate": 4.603578839572049e-05,
|
| 3646 |
+
"loss": 3.0333,
|
| 3647 |
+
"step": 2950
|
| 3648 |
+
},
|
| 3649 |
+
{
|
| 3650 |
+
"epoch": 0.24,
|
| 3651 |
+
"learning_rate": 4.602905112242973e-05,
|
| 3652 |
+
"loss": 3.1151,
|
| 3653 |
+
"step": 2955
|
| 3654 |
+
},
|
| 3655 |
+
{
|
| 3656 |
+
"epoch": 0.24,
|
| 3657 |
+
"learning_rate": 4.602231384913897e-05,
|
| 3658 |
+
"loss": 3.1866,
|
| 3659 |
+
"step": 2960
|
| 3660 |
+
},
|
| 3661 |
+
{
|
| 3662 |
+
"epoch": 0.24,
|
| 3663 |
+
"learning_rate": 4.601557657584822e-05,
|
| 3664 |
+
"loss": 3.0419,
|
| 3665 |
+
"step": 2965
|
| 3666 |
+
},
|
| 3667 |
+
{
|
| 3668 |
+
"epoch": 0.24,
|
| 3669 |
+
"learning_rate": 4.600883930255747e-05,
|
| 3670 |
+
"loss": 2.9358,
|
| 3671 |
+
"step": 2970
|
| 3672 |
+
},
|
| 3673 |
+
{
|
| 3674 |
+
"epoch": 0.24,
|
| 3675 |
+
"learning_rate": 4.600210202926672e-05,
|
| 3676 |
+
"loss": 3.3567,
|
| 3677 |
+
"step": 2975
|
| 3678 |
+
},
|
| 3679 |
+
{
|
| 3680 |
+
"epoch": 0.24,
|
| 3681 |
+
"learning_rate": 4.599536475597596e-05,
|
| 3682 |
+
"loss": 2.9169,
|
| 3683 |
+
"step": 2980
|
| 3684 |
+
},
|
| 3685 |
+
{
|
| 3686 |
+
"epoch": 0.24,
|
| 3687 |
+
"learning_rate": 4.598862748268521e-05,
|
| 3688 |
+
"loss": 3.0725,
|
| 3689 |
+
"step": 2985
|
| 3690 |
+
},
|
| 3691 |
+
{
|
| 3692 |
+
"epoch": 0.24,
|
| 3693 |
+
"learning_rate": 4.598189020939446e-05,
|
| 3694 |
+
"loss": 3.4152,
|
| 3695 |
+
"step": 2990
|
| 3696 |
+
},
|
| 3697 |
+
{
|
| 3698 |
+
"epoch": 0.24,
|
| 3699 |
+
"learning_rate": 4.59751529361037e-05,
|
| 3700 |
+
"loss": 3.0889,
|
| 3701 |
+
"step": 2995
|
| 3702 |
+
},
|
| 3703 |
+
{
|
| 3704 |
+
"epoch": 0.24,
|
| 3705 |
+
"learning_rate": 4.5968415662812944e-05,
|
| 3706 |
+
"loss": 3.3419,
|
| 3707 |
+
"step": 3000
|
| 3708 |
+
},
|
| 3709 |
+
{
|
| 3710 |
+
"epoch": 0.24,
|
| 3711 |
+
"learning_rate": 4.596167838952219e-05,
|
| 3712 |
+
"loss": 3.3641,
|
| 3713 |
+
"step": 3005
|
| 3714 |
+
},
|
| 3715 |
+
{
|
| 3716 |
+
"epoch": 0.24,
|
| 3717 |
+
"learning_rate": 4.595494111623144e-05,
|
| 3718 |
+
"loss": 3.0926,
|
| 3719 |
+
"step": 3010
|
| 3720 |
+
},
|
| 3721 |
+
{
|
| 3722 |
+
"epoch": 0.24,
|
| 3723 |
+
"learning_rate": 4.5948203842940684e-05,
|
| 3724 |
+
"loss": 3.0744,
|
| 3725 |
+
"step": 3015
|
| 3726 |
+
},
|
| 3727 |
+
{
|
| 3728 |
+
"epoch": 0.24,
|
| 3729 |
+
"learning_rate": 4.594146656964993e-05,
|
| 3730 |
+
"loss": 3.0896,
|
| 3731 |
+
"step": 3020
|
| 3732 |
+
},
|
| 3733 |
+
{
|
| 3734 |
+
"epoch": 0.24,
|
| 3735 |
+
"learning_rate": 4.593472929635918e-05,
|
| 3736 |
+
"loss": 3.2848,
|
| 3737 |
+
"step": 3025
|
| 3738 |
+
},
|
| 3739 |
+
{
|
| 3740 |
+
"epoch": 0.24,
|
| 3741 |
+
"learning_rate": 4.592799202306843e-05,
|
| 3742 |
+
"loss": 3.1748,
|
| 3743 |
+
"step": 3030
|
| 3744 |
+
},
|
| 3745 |
+
{
|
| 3746 |
+
"epoch": 0.25,
|
| 3747 |
+
"learning_rate": 4.592125474977767e-05,
|
| 3748 |
+
"loss": 3.3514,
|
| 3749 |
+
"step": 3035
|
| 3750 |
+
},
|
| 3751 |
+
{
|
| 3752 |
+
"epoch": 0.25,
|
| 3753 |
+
"learning_rate": 4.591451747648692e-05,
|
| 3754 |
+
"loss": 3.2087,
|
| 3755 |
+
"step": 3040
|
| 3756 |
+
},
|
| 3757 |
+
{
|
| 3758 |
+
"epoch": 0.25,
|
| 3759 |
+
"learning_rate": 4.590778020319616e-05,
|
| 3760 |
+
"loss": 3.1578,
|
| 3761 |
+
"step": 3045
|
| 3762 |
+
},
|
| 3763 |
+
{
|
| 3764 |
+
"epoch": 0.25,
|
| 3765 |
+
"learning_rate": 4.590104292990541e-05,
|
| 3766 |
+
"loss": 3.0495,
|
| 3767 |
+
"step": 3050
|
| 3768 |
+
},
|
| 3769 |
+
{
|
| 3770 |
+
"epoch": 0.25,
|
| 3771 |
+
"learning_rate": 4.5894305656614654e-05,
|
| 3772 |
+
"loss": 3.0536,
|
| 3773 |
+
"step": 3055
|
| 3774 |
+
},
|
| 3775 |
+
{
|
| 3776 |
+
"epoch": 0.25,
|
| 3777 |
+
"learning_rate": 4.58875683833239e-05,
|
| 3778 |
+
"loss": 3.1117,
|
| 3779 |
+
"step": 3060
|
| 3780 |
+
},
|
| 3781 |
+
{
|
| 3782 |
+
"epoch": 0.25,
|
| 3783 |
+
"learning_rate": 4.588083111003315e-05,
|
| 3784 |
+
"loss": 3.3505,
|
| 3785 |
+
"step": 3065
|
| 3786 |
+
},
|
| 3787 |
+
{
|
| 3788 |
+
"epoch": 0.25,
|
| 3789 |
+
"learning_rate": 4.5874093836742394e-05,
|
| 3790 |
+
"loss": 3.3008,
|
| 3791 |
+
"step": 3070
|
| 3792 |
+
},
|
| 3793 |
+
{
|
| 3794 |
+
"epoch": 0.25,
|
| 3795 |
+
"learning_rate": 4.586735656345164e-05,
|
| 3796 |
+
"loss": 2.9855,
|
| 3797 |
+
"step": 3075
|
| 3798 |
+
},
|
| 3799 |
+
{
|
| 3800 |
+
"epoch": 0.25,
|
| 3801 |
+
"learning_rate": 4.586061929016089e-05,
|
| 3802 |
+
"loss": 3.1741,
|
| 3803 |
+
"step": 3080
|
| 3804 |
+
},
|
| 3805 |
+
{
|
| 3806 |
+
"epoch": 0.25,
|
| 3807 |
+
"learning_rate": 4.5853882016870134e-05,
|
| 3808 |
+
"loss": 3.1377,
|
| 3809 |
+
"step": 3085
|
| 3810 |
+
},
|
| 3811 |
+
{
|
| 3812 |
+
"epoch": 0.25,
|
| 3813 |
+
"learning_rate": 4.5847144743579376e-05,
|
| 3814 |
+
"loss": 3.2383,
|
| 3815 |
+
"step": 3090
|
| 3816 |
+
},
|
| 3817 |
+
{
|
| 3818 |
+
"epoch": 0.25,
|
| 3819 |
+
"learning_rate": 4.5840407470288625e-05,
|
| 3820 |
+
"loss": 3.0129,
|
| 3821 |
+
"step": 3095
|
| 3822 |
+
},
|
| 3823 |
+
{
|
| 3824 |
+
"epoch": 0.25,
|
| 3825 |
+
"learning_rate": 4.5833670196997874e-05,
|
| 3826 |
+
"loss": 3.1036,
|
| 3827 |
+
"step": 3100
|
| 3828 |
+
},
|
| 3829 |
+
{
|
| 3830 |
+
"epoch": 0.25,
|
| 3831 |
+
"learning_rate": 4.582693292370712e-05,
|
| 3832 |
+
"loss": 3.1,
|
| 3833 |
+
"step": 3105
|
| 3834 |
+
},
|
| 3835 |
+
{
|
| 3836 |
+
"epoch": 0.25,
|
| 3837 |
+
"learning_rate": 4.5820195650416365e-05,
|
| 3838 |
+
"loss": 3.21,
|
| 3839 |
+
"step": 3110
|
| 3840 |
+
},
|
| 3841 |
+
{
|
| 3842 |
+
"epoch": 0.25,
|
| 3843 |
+
"learning_rate": 4.5813458377125614e-05,
|
| 3844 |
+
"loss": 3.1048,
|
| 3845 |
+
"step": 3115
|
| 3846 |
+
},
|
| 3847 |
+
{
|
| 3848 |
+
"epoch": 0.25,
|
| 3849 |
+
"learning_rate": 4.580672110383486e-05,
|
| 3850 |
+
"loss": 3.2457,
|
| 3851 |
+
"step": 3120
|
| 3852 |
+
},
|
| 3853 |
+
{
|
| 3854 |
+
"epoch": 0.25,
|
| 3855 |
+
"learning_rate": 4.5799983830544105e-05,
|
| 3856 |
+
"loss": 3.1279,
|
| 3857 |
+
"step": 3125
|
| 3858 |
+
},
|
| 3859 |
+
{
|
| 3860 |
+
"epoch": 0.25,
|
| 3861 |
+
"learning_rate": 4.579324655725335e-05,
|
| 3862 |
+
"loss": 3.247,
|
| 3863 |
+
"step": 3130
|
| 3864 |
+
},
|
| 3865 |
+
{
|
| 3866 |
+
"epoch": 0.25,
|
| 3867 |
+
"learning_rate": 4.5786509283962596e-05,
|
| 3868 |
+
"loss": 3.2605,
|
| 3869 |
+
"step": 3135
|
| 3870 |
+
},
|
| 3871 |
+
{
|
| 3872 |
+
"epoch": 0.25,
|
| 3873 |
+
"learning_rate": 4.5779772010671844e-05,
|
| 3874 |
+
"loss": 3.3872,
|
| 3875 |
+
"step": 3140
|
| 3876 |
+
},
|
| 3877 |
+
{
|
| 3878 |
+
"epoch": 0.25,
|
| 3879 |
+
"learning_rate": 4.5773034737381087e-05,
|
| 3880 |
+
"loss": 2.9347,
|
| 3881 |
+
"step": 3145
|
| 3882 |
+
},
|
| 3883 |
+
{
|
| 3884 |
+
"epoch": 0.25,
|
| 3885 |
+
"learning_rate": 4.5766297464090335e-05,
|
| 3886 |
+
"loss": 3.0918,
|
| 3887 |
+
"step": 3150
|
| 3888 |
+
},
|
| 3889 |
+
{
|
| 3890 |
+
"epoch": 0.26,
|
| 3891 |
+
"learning_rate": 4.5759560190799584e-05,
|
| 3892 |
+
"loss": 3.1318,
|
| 3893 |
+
"step": 3155
|
| 3894 |
+
},
|
| 3895 |
+
{
|
| 3896 |
+
"epoch": 0.26,
|
| 3897 |
+
"learning_rate": 4.575282291750883e-05,
|
| 3898 |
+
"loss": 2.9989,
|
| 3899 |
+
"step": 3160
|
| 3900 |
+
},
|
| 3901 |
+
{
|
| 3902 |
+
"epoch": 0.26,
|
| 3903 |
+
"learning_rate": 4.5746085644218075e-05,
|
| 3904 |
+
"loss": 3.2661,
|
| 3905 |
+
"step": 3165
|
| 3906 |
+
},
|
| 3907 |
+
{
|
| 3908 |
+
"epoch": 0.26,
|
| 3909 |
+
"learning_rate": 4.5739348370927324e-05,
|
| 3910 |
+
"loss": 3.2538,
|
| 3911 |
+
"step": 3170
|
| 3912 |
+
},
|
| 3913 |
+
{
|
| 3914 |
+
"epoch": 0.26,
|
| 3915 |
+
"learning_rate": 4.5732611097636566e-05,
|
| 3916 |
+
"loss": 3.2111,
|
| 3917 |
+
"step": 3175
|
| 3918 |
+
},
|
| 3919 |
+
{
|
| 3920 |
+
"epoch": 0.26,
|
| 3921 |
+
"learning_rate": 4.572587382434581e-05,
|
| 3922 |
+
"loss": 3.0219,
|
| 3923 |
+
"step": 3180
|
| 3924 |
+
},
|
| 3925 |
+
{
|
| 3926 |
+
"epoch": 0.26,
|
| 3927 |
+
"learning_rate": 4.571913655105506e-05,
|
| 3928 |
+
"loss": 3.2369,
|
| 3929 |
+
"step": 3185
|
| 3930 |
+
},
|
| 3931 |
+
{
|
| 3932 |
+
"epoch": 0.26,
|
| 3933 |
+
"learning_rate": 4.5712399277764306e-05,
|
| 3934 |
+
"loss": 3.1409,
|
| 3935 |
+
"step": 3190
|
| 3936 |
+
},
|
| 3937 |
+
{
|
| 3938 |
+
"epoch": 0.26,
|
| 3939 |
+
"learning_rate": 4.5705662004473555e-05,
|
| 3940 |
+
"loss": 3.1233,
|
| 3941 |
+
"step": 3195
|
| 3942 |
+
},
|
| 3943 |
+
{
|
| 3944 |
+
"epoch": 0.26,
|
| 3945 |
+
"learning_rate": 4.56989247311828e-05,
|
| 3946 |
+
"loss": 3.2719,
|
| 3947 |
+
"step": 3200
|
| 3948 |
+
},
|
| 3949 |
+
{
|
| 3950 |
+
"epoch": 0.26,
|
| 3951 |
+
"learning_rate": 4.5692187457892046e-05,
|
| 3952 |
+
"loss": 2.9358,
|
| 3953 |
+
"step": 3205
|
| 3954 |
+
},
|
| 3955 |
+
{
|
| 3956 |
+
"epoch": 0.26,
|
| 3957 |
+
"learning_rate": 4.5685450184601295e-05,
|
| 3958 |
+
"loss": 3.1265,
|
| 3959 |
+
"step": 3210
|
| 3960 |
+
},
|
| 3961 |
+
{
|
| 3962 |
+
"epoch": 0.26,
|
| 3963 |
+
"learning_rate": 4.567871291131054e-05,
|
| 3964 |
+
"loss": 3.0724,
|
| 3965 |
+
"step": 3215
|
| 3966 |
+
},
|
| 3967 |
+
{
|
| 3968 |
+
"epoch": 0.26,
|
| 3969 |
+
"learning_rate": 4.567197563801978e-05,
|
| 3970 |
+
"loss": 3.0934,
|
| 3971 |
+
"step": 3220
|
| 3972 |
+
},
|
| 3973 |
+
{
|
| 3974 |
+
"epoch": 0.26,
|
| 3975 |
+
"learning_rate": 4.566523836472903e-05,
|
| 3976 |
+
"loss": 3.3711,
|
| 3977 |
+
"step": 3225
|
| 3978 |
+
},
|
| 3979 |
+
{
|
| 3980 |
+
"epoch": 0.26,
|
| 3981 |
+
"learning_rate": 4.565850109143828e-05,
|
| 3982 |
+
"loss": 2.8675,
|
| 3983 |
+
"step": 3230
|
| 3984 |
+
},
|
| 3985 |
+
{
|
| 3986 |
+
"epoch": 0.26,
|
| 3987 |
+
"learning_rate": 4.565176381814752e-05,
|
| 3988 |
+
"loss": 3.0045,
|
| 3989 |
+
"step": 3235
|
| 3990 |
+
},
|
| 3991 |
+
{
|
| 3992 |
+
"epoch": 0.26,
|
| 3993 |
+
"learning_rate": 4.564502654485677e-05,
|
| 3994 |
+
"loss": 3.0661,
|
| 3995 |
+
"step": 3240
|
| 3996 |
+
},
|
| 3997 |
+
{
|
| 3998 |
+
"epoch": 0.26,
|
| 3999 |
+
"learning_rate": 4.5638289271566017e-05,
|
| 4000 |
+
"loss": 3.1129,
|
| 4001 |
+
"step": 3245
|
| 4002 |
+
},
|
| 4003 |
+
{
|
| 4004 |
+
"epoch": 0.26,
|
| 4005 |
+
"learning_rate": 4.5631551998275265e-05,
|
| 4006 |
+
"loss": 3.0902,
|
| 4007 |
+
"step": 3250
|
| 4008 |
+
},
|
| 4009 |
+
{
|
| 4010 |
+
"epoch": 0.26,
|
| 4011 |
+
"learning_rate": 4.562481472498451e-05,
|
| 4012 |
+
"loss": 3.1345,
|
| 4013 |
+
"step": 3255
|
| 4014 |
+
},
|
| 4015 |
+
{
|
| 4016 |
+
"epoch": 0.26,
|
| 4017 |
+
"learning_rate": 4.561807745169375e-05,
|
| 4018 |
+
"loss": 2.9902,
|
| 4019 |
+
"step": 3260
|
| 4020 |
+
},
|
| 4021 |
+
{
|
| 4022 |
+
"epoch": 0.26,
|
| 4023 |
+
"learning_rate": 4.5611340178403e-05,
|
| 4024 |
+
"loss": 3.191,
|
| 4025 |
+
"step": 3265
|
| 4026 |
+
},
|
| 4027 |
+
{
|
| 4028 |
+
"epoch": 0.26,
|
| 4029 |
+
"learning_rate": 4.560460290511224e-05,
|
| 4030 |
+
"loss": 3.1233,
|
| 4031 |
+
"step": 3270
|
| 4032 |
+
},
|
| 4033 |
+
{
|
| 4034 |
+
"epoch": 0.26,
|
| 4035 |
+
"learning_rate": 4.559786563182149e-05,
|
| 4036 |
+
"loss": 3.203,
|
| 4037 |
+
"step": 3275
|
| 4038 |
+
},
|
| 4039 |
+
{
|
| 4040 |
+
"epoch": 0.27,
|
| 4041 |
+
"learning_rate": 4.559112835853074e-05,
|
| 4042 |
+
"loss": 3.112,
|
| 4043 |
+
"step": 3280
|
| 4044 |
+
},
|
| 4045 |
+
{
|
| 4046 |
+
"epoch": 0.27,
|
| 4047 |
+
"learning_rate": 4.558439108523999e-05,
|
| 4048 |
+
"loss": 3.0281,
|
| 4049 |
+
"step": 3285
|
| 4050 |
+
},
|
| 4051 |
+
{
|
| 4052 |
+
"epoch": 0.27,
|
| 4053 |
+
"learning_rate": 4.557765381194923e-05,
|
| 4054 |
+
"loss": 3.3552,
|
| 4055 |
+
"step": 3290
|
| 4056 |
+
},
|
| 4057 |
+
{
|
| 4058 |
+
"epoch": 0.27,
|
| 4059 |
+
"learning_rate": 4.557091653865848e-05,
|
| 4060 |
+
"loss": 3.0418,
|
| 4061 |
+
"step": 3295
|
| 4062 |
+
},
|
| 4063 |
+
{
|
| 4064 |
+
"epoch": 0.27,
|
| 4065 |
+
"learning_rate": 4.556417926536772e-05,
|
| 4066 |
+
"loss": 3.0119,
|
| 4067 |
+
"step": 3300
|
| 4068 |
+
},
|
| 4069 |
+
{
|
| 4070 |
+
"epoch": 0.27,
|
| 4071 |
+
"learning_rate": 4.555744199207697e-05,
|
| 4072 |
+
"loss": 3.003,
|
| 4073 |
+
"step": 3305
|
| 4074 |
+
},
|
| 4075 |
+
{
|
| 4076 |
+
"epoch": 0.27,
|
| 4077 |
+
"learning_rate": 4.555070471878621e-05,
|
| 4078 |
+
"loss": 3.1365,
|
| 4079 |
+
"step": 3310
|
| 4080 |
+
},
|
| 4081 |
+
{
|
| 4082 |
+
"epoch": 0.27,
|
| 4083 |
+
"learning_rate": 4.554396744549546e-05,
|
| 4084 |
+
"loss": 3.0694,
|
| 4085 |
+
"step": 3315
|
| 4086 |
+
},
|
| 4087 |
+
{
|
| 4088 |
+
"epoch": 0.27,
|
| 4089 |
+
"learning_rate": 4.553723017220471e-05,
|
| 4090 |
+
"loss": 2.8513,
|
| 4091 |
+
"step": 3320
|
| 4092 |
+
},
|
| 4093 |
+
{
|
| 4094 |
+
"epoch": 0.27,
|
| 4095 |
+
"learning_rate": 4.553049289891395e-05,
|
| 4096 |
+
"loss": 3.1681,
|
| 4097 |
+
"step": 3325
|
| 4098 |
+
},
|
| 4099 |
+
{
|
| 4100 |
+
"epoch": 0.27,
|
| 4101 |
+
"learning_rate": 4.55237556256232e-05,
|
| 4102 |
+
"loss": 3.1393,
|
| 4103 |
+
"step": 3330
|
| 4104 |
+
},
|
| 4105 |
+
{
|
| 4106 |
+
"epoch": 0.27,
|
| 4107 |
+
"learning_rate": 4.551701835233245e-05,
|
| 4108 |
+
"loss": 3.0198,
|
| 4109 |
+
"step": 3335
|
| 4110 |
+
},
|
| 4111 |
+
{
|
| 4112 |
+
"epoch": 0.27,
|
| 4113 |
+
"learning_rate": 4.55102810790417e-05,
|
| 4114 |
+
"loss": 3.0511,
|
| 4115 |
+
"step": 3340
|
| 4116 |
+
},
|
| 4117 |
+
{
|
| 4118 |
+
"epoch": 0.27,
|
| 4119 |
+
"learning_rate": 4.550354380575094e-05,
|
| 4120 |
+
"loss": 3.1593,
|
| 4121 |
+
"step": 3345
|
| 4122 |
+
},
|
| 4123 |
+
{
|
| 4124 |
+
"epoch": 0.27,
|
| 4125 |
+
"learning_rate": 4.549680653246018e-05,
|
| 4126 |
+
"loss": 2.9654,
|
| 4127 |
+
"step": 3350
|
| 4128 |
+
},
|
| 4129 |
+
{
|
| 4130 |
+
"epoch": 0.27,
|
| 4131 |
+
"learning_rate": 4.549006925916943e-05,
|
| 4132 |
+
"loss": 3.1572,
|
| 4133 |
+
"step": 3355
|
| 4134 |
+
},
|
| 4135 |
+
{
|
| 4136 |
+
"epoch": 0.27,
|
| 4137 |
+
"learning_rate": 4.548333198587868e-05,
|
| 4138 |
+
"loss": 3.283,
|
| 4139 |
+
"step": 3360
|
| 4140 |
+
},
|
| 4141 |
+
{
|
| 4142 |
+
"epoch": 0.27,
|
| 4143 |
+
"learning_rate": 4.547659471258792e-05,
|
| 4144 |
+
"loss": 3.1824,
|
| 4145 |
+
"step": 3365
|
| 4146 |
+
},
|
| 4147 |
+
{
|
| 4148 |
+
"epoch": 0.27,
|
| 4149 |
+
"learning_rate": 4.546985743929717e-05,
|
| 4150 |
+
"loss": 3.086,
|
| 4151 |
+
"step": 3370
|
| 4152 |
+
},
|
| 4153 |
+
{
|
| 4154 |
+
"epoch": 0.27,
|
| 4155 |
+
"learning_rate": 4.546312016600642e-05,
|
| 4156 |
+
"loss": 3.1806,
|
| 4157 |
+
"step": 3375
|
| 4158 |
+
},
|
| 4159 |
+
{
|
| 4160 |
+
"epoch": 0.27,
|
| 4161 |
+
"learning_rate": 4.545638289271566e-05,
|
| 4162 |
+
"loss": 3.2051,
|
| 4163 |
+
"step": 3380
|
| 4164 |
+
},
|
| 4165 |
+
{
|
| 4166 |
+
"epoch": 0.27,
|
| 4167 |
+
"learning_rate": 4.544964561942491e-05,
|
| 4168 |
+
"loss": 3.0539,
|
| 4169 |
+
"step": 3385
|
| 4170 |
+
},
|
| 4171 |
+
{
|
| 4172 |
+
"epoch": 0.27,
|
| 4173 |
+
"learning_rate": 4.544290834613415e-05,
|
| 4174 |
+
"loss": 3.0205,
|
| 4175 |
+
"step": 3390
|
| 4176 |
+
},
|
| 4177 |
+
{
|
| 4178 |
+
"epoch": 0.27,
|
| 4179 |
+
"learning_rate": 4.54361710728434e-05,
|
| 4180 |
+
"loss": 3.2201,
|
| 4181 |
+
"step": 3395
|
| 4182 |
+
},
|
| 4183 |
+
{
|
| 4184 |
+
"epoch": 0.27,
|
| 4185 |
+
"learning_rate": 4.5429433799552644e-05,
|
| 4186 |
+
"loss": 3.2691,
|
| 4187 |
+
"step": 3400
|
| 4188 |
+
},
|
| 4189 |
+
{
|
| 4190 |
+
"epoch": 0.28,
|
| 4191 |
+
"learning_rate": 4.542269652626189e-05,
|
| 4192 |
+
"loss": 3.1167,
|
| 4193 |
+
"step": 3405
|
| 4194 |
+
},
|
| 4195 |
+
{
|
| 4196 |
+
"epoch": 0.28,
|
| 4197 |
+
"learning_rate": 4.541595925297114e-05,
|
| 4198 |
+
"loss": 2.9539,
|
| 4199 |
+
"step": 3410
|
| 4200 |
+
},
|
| 4201 |
+
{
|
| 4202 |
+
"epoch": 0.28,
|
| 4203 |
+
"learning_rate": 4.540922197968039e-05,
|
| 4204 |
+
"loss": 3.1993,
|
| 4205 |
+
"step": 3415
|
| 4206 |
+
},
|
| 4207 |
+
{
|
| 4208 |
+
"epoch": 0.28,
|
| 4209 |
+
"learning_rate": 4.540248470638963e-05,
|
| 4210 |
+
"loss": 3.1305,
|
| 4211 |
+
"step": 3420
|
| 4212 |
+
},
|
| 4213 |
+
{
|
| 4214 |
+
"epoch": 0.28,
|
| 4215 |
+
"learning_rate": 4.539574743309888e-05,
|
| 4216 |
+
"loss": 3.1626,
|
| 4217 |
+
"step": 3425
|
| 4218 |
+
},
|
| 4219 |
+
{
|
| 4220 |
+
"epoch": 0.28,
|
| 4221 |
+
"learning_rate": 4.538901015980812e-05,
|
| 4222 |
+
"loss": 3.2176,
|
| 4223 |
+
"step": 3430
|
| 4224 |
+
},
|
| 4225 |
+
{
|
| 4226 |
+
"epoch": 0.28,
|
| 4227 |
+
"learning_rate": 4.5382272886517365e-05,
|
| 4228 |
+
"loss": 3.0691,
|
| 4229 |
+
"step": 3435
|
| 4230 |
+
},
|
| 4231 |
+
{
|
| 4232 |
+
"epoch": 0.28,
|
| 4233 |
+
"learning_rate": 4.5375535613226614e-05,
|
| 4234 |
+
"loss": 3.0611,
|
| 4235 |
+
"step": 3440
|
| 4236 |
+
},
|
| 4237 |
+
{
|
| 4238 |
+
"epoch": 0.28,
|
| 4239 |
+
"learning_rate": 4.536879833993586e-05,
|
| 4240 |
+
"loss": 3.1824,
|
| 4241 |
+
"step": 3445
|
| 4242 |
+
},
|
| 4243 |
+
{
|
| 4244 |
+
"epoch": 0.28,
|
| 4245 |
+
"learning_rate": 4.536206106664511e-05,
|
| 4246 |
+
"loss": 3.1394,
|
| 4247 |
+
"step": 3450
|
| 4248 |
+
},
|
| 4249 |
+
{
|
| 4250 |
+
"epoch": 0.28,
|
| 4251 |
+
"learning_rate": 4.5355323793354354e-05,
|
| 4252 |
+
"loss": 3.1272,
|
| 4253 |
+
"step": 3455
|
| 4254 |
+
},
|
| 4255 |
+
{
|
| 4256 |
+
"epoch": 0.28,
|
| 4257 |
+
"learning_rate": 4.53485865200636e-05,
|
| 4258 |
+
"loss": 2.9905,
|
| 4259 |
+
"step": 3460
|
| 4260 |
+
},
|
| 4261 |
+
{
|
| 4262 |
+
"epoch": 0.28,
|
| 4263 |
+
"learning_rate": 4.534184924677285e-05,
|
| 4264 |
+
"loss": 2.8781,
|
| 4265 |
+
"step": 3465
|
| 4266 |
+
},
|
| 4267 |
+
{
|
| 4268 |
+
"epoch": 0.28,
|
| 4269 |
+
"learning_rate": 4.53351119734821e-05,
|
| 4270 |
+
"loss": 3.1255,
|
| 4271 |
+
"step": 3470
|
| 4272 |
+
},
|
| 4273 |
+
{
|
| 4274 |
+
"epoch": 0.28,
|
| 4275 |
+
"learning_rate": 4.532837470019134e-05,
|
| 4276 |
+
"loss": 3.0489,
|
| 4277 |
+
"step": 3475
|
| 4278 |
+
},
|
| 4279 |
+
{
|
| 4280 |
+
"epoch": 0.28,
|
| 4281 |
+
"learning_rate": 4.5321637426900585e-05,
|
| 4282 |
+
"loss": 3.1483,
|
| 4283 |
+
"step": 3480
|
| 4284 |
+
},
|
| 4285 |
+
{
|
| 4286 |
+
"epoch": 0.28,
|
| 4287 |
+
"learning_rate": 4.5314900153609834e-05,
|
| 4288 |
+
"loss": 2.923,
|
| 4289 |
+
"step": 3485
|
| 4290 |
+
},
|
| 4291 |
+
{
|
| 4292 |
+
"epoch": 0.28,
|
| 4293 |
+
"learning_rate": 4.5308162880319076e-05,
|
| 4294 |
+
"loss": 3.1935,
|
| 4295 |
+
"step": 3490
|
| 4296 |
+
},
|
| 4297 |
+
{
|
| 4298 |
+
"epoch": 0.28,
|
| 4299 |
+
"learning_rate": 4.5301425607028325e-05,
|
| 4300 |
+
"loss": 2.8445,
|
| 4301 |
+
"step": 3495
|
| 4302 |
+
},
|
| 4303 |
+
{
|
| 4304 |
+
"epoch": 0.28,
|
| 4305 |
+
"learning_rate": 4.5294688333737574e-05,
|
| 4306 |
+
"loss": 3.1171,
|
| 4307 |
+
"step": 3500
|
| 4308 |
+
},
|
| 4309 |
+
{
|
| 4310 |
+
"epoch": 0.28,
|
| 4311 |
+
"learning_rate": 4.528795106044682e-05,
|
| 4312 |
+
"loss": 3.2084,
|
| 4313 |
+
"step": 3505
|
| 4314 |
+
},
|
| 4315 |
+
{
|
| 4316 |
+
"epoch": 0.28,
|
| 4317 |
+
"learning_rate": 4.5281213787156065e-05,
|
| 4318 |
+
"loss": 3.1549,
|
| 4319 |
+
"step": 3510
|
| 4320 |
+
},
|
| 4321 |
+
{
|
| 4322 |
+
"epoch": 0.28,
|
| 4323 |
+
"learning_rate": 4.527447651386531e-05,
|
| 4324 |
+
"loss": 3.0279,
|
| 4325 |
+
"step": 3515
|
| 4326 |
+
},
|
| 4327 |
+
{
|
| 4328 |
+
"epoch": 0.28,
|
| 4329 |
+
"learning_rate": 4.5267739240574555e-05,
|
| 4330 |
+
"loss": 2.7719,
|
| 4331 |
+
"step": 3520
|
| 4332 |
+
},
|
| 4333 |
+
{
|
| 4334 |
+
"epoch": 0.28,
|
| 4335 |
+
"learning_rate": 4.52610019672838e-05,
|
| 4336 |
+
"loss": 3.1109,
|
| 4337 |
+
"step": 3525
|
| 4338 |
+
},
|
| 4339 |
+
{
|
| 4340 |
+
"epoch": 0.29,
|
| 4341 |
+
"learning_rate": 4.5254264693993046e-05,
|
| 4342 |
+
"loss": 2.9881,
|
| 4343 |
+
"step": 3530
|
| 4344 |
+
},
|
| 4345 |
+
{
|
| 4346 |
+
"epoch": 0.29,
|
| 4347 |
+
"learning_rate": 4.5247527420702295e-05,
|
| 4348 |
+
"loss": 3.2017,
|
| 4349 |
+
"step": 3535
|
| 4350 |
+
},
|
| 4351 |
+
{
|
| 4352 |
+
"epoch": 0.29,
|
| 4353 |
+
"learning_rate": 4.5240790147411544e-05,
|
| 4354 |
+
"loss": 2.9567,
|
| 4355 |
+
"step": 3540
|
| 4356 |
+
},
|
| 4357 |
+
{
|
| 4358 |
+
"epoch": 0.29,
|
| 4359 |
+
"learning_rate": 4.5234052874120786e-05,
|
| 4360 |
+
"loss": 3.2274,
|
| 4361 |
+
"step": 3545
|
| 4362 |
+
},
|
| 4363 |
+
{
|
| 4364 |
+
"epoch": 0.29,
|
| 4365 |
+
"learning_rate": 4.5227315600830035e-05,
|
| 4366 |
+
"loss": 3.1517,
|
| 4367 |
+
"step": 3550
|
| 4368 |
+
},
|
| 4369 |
+
{
|
| 4370 |
+
"epoch": 0.29,
|
| 4371 |
+
"learning_rate": 4.5220578327539284e-05,
|
| 4372 |
+
"loss": 3.0133,
|
| 4373 |
+
"step": 3555
|
| 4374 |
+
},
|
| 4375 |
+
{
|
| 4376 |
+
"epoch": 0.29,
|
| 4377 |
+
"learning_rate": 4.5213841054248526e-05,
|
| 4378 |
+
"loss": 3.2494,
|
| 4379 |
+
"step": 3560
|
| 4380 |
+
},
|
| 4381 |
+
{
|
| 4382 |
+
"epoch": 0.29,
|
| 4383 |
+
"learning_rate": 4.520710378095777e-05,
|
| 4384 |
+
"loss": 3.1607,
|
| 4385 |
+
"step": 3565
|
| 4386 |
+
},
|
| 4387 |
+
{
|
| 4388 |
+
"epoch": 0.29,
|
| 4389 |
+
"learning_rate": 4.520036650766702e-05,
|
| 4390 |
+
"loss": 3.1622,
|
| 4391 |
+
"step": 3570
|
| 4392 |
+
},
|
| 4393 |
+
{
|
| 4394 |
+
"epoch": 0.29,
|
| 4395 |
+
"learning_rate": 4.5193629234376266e-05,
|
| 4396 |
+
"loss": 3.0597,
|
| 4397 |
+
"step": 3575
|
| 4398 |
+
},
|
| 4399 |
+
{
|
| 4400 |
+
"epoch": 0.29,
|
| 4401 |
+
"learning_rate": 4.518689196108551e-05,
|
| 4402 |
+
"loss": 3.1157,
|
| 4403 |
+
"step": 3580
|
| 4404 |
+
},
|
| 4405 |
+
{
|
| 4406 |
+
"epoch": 0.29,
|
| 4407 |
+
"learning_rate": 4.518015468779476e-05,
|
| 4408 |
+
"loss": 3.1916,
|
| 4409 |
+
"step": 3585
|
| 4410 |
+
},
|
| 4411 |
+
{
|
| 4412 |
+
"epoch": 0.29,
|
| 4413 |
+
"learning_rate": 4.5173417414504006e-05,
|
| 4414 |
+
"loss": 3.3892,
|
| 4415 |
+
"step": 3590
|
| 4416 |
+
},
|
| 4417 |
+
{
|
| 4418 |
+
"epoch": 0.29,
|
| 4419 |
+
"learning_rate": 4.5166680141213255e-05,
|
| 4420 |
+
"loss": 3.2352,
|
| 4421 |
+
"step": 3595
|
| 4422 |
+
},
|
| 4423 |
+
{
|
| 4424 |
+
"epoch": 0.29,
|
| 4425 |
+
"learning_rate": 4.51599428679225e-05,
|
| 4426 |
+
"loss": 2.9455,
|
| 4427 |
+
"step": 3600
|
| 4428 |
+
},
|
| 4429 |
+
{
|
| 4430 |
+
"epoch": 0.29,
|
| 4431 |
+
"learning_rate": 4.515455304928989e-05,
|
| 4432 |
+
"loss": 3.2882,
|
| 4433 |
+
"step": 3605
|
| 4434 |
+
},
|
| 4435 |
+
{
|
| 4436 |
+
"epoch": 0.29,
|
| 4437 |
+
"learning_rate": 4.514781577599914e-05,
|
| 4438 |
+
"loss": 3.1644,
|
| 4439 |
+
"step": 3610
|
| 4440 |
+
},
|
| 4441 |
+
{
|
| 4442 |
+
"epoch": 0.29,
|
| 4443 |
+
"learning_rate": 4.514107850270838e-05,
|
| 4444 |
+
"loss": 3.1653,
|
| 4445 |
+
"step": 3615
|
| 4446 |
+
},
|
| 4447 |
+
{
|
| 4448 |
+
"epoch": 0.29,
|
| 4449 |
+
"learning_rate": 4.513434122941763e-05,
|
| 4450 |
+
"loss": 3.286,
|
| 4451 |
+
"step": 3620
|
| 4452 |
+
},
|
| 4453 |
+
{
|
| 4454 |
+
"epoch": 0.29,
|
| 4455 |
+
"learning_rate": 4.512760395612688e-05,
|
| 4456 |
+
"loss": 3.2771,
|
| 4457 |
+
"step": 3625
|
| 4458 |
+
},
|
| 4459 |
+
{
|
| 4460 |
+
"epoch": 0.29,
|
| 4461 |
+
"learning_rate": 4.512086668283613e-05,
|
| 4462 |
+
"loss": 3.4175,
|
| 4463 |
+
"step": 3630
|
| 4464 |
+
},
|
| 4465 |
+
{
|
| 4466 |
+
"epoch": 0.29,
|
| 4467 |
+
"learning_rate": 4.511412940954537e-05,
|
| 4468 |
+
"loss": 3.0911,
|
| 4469 |
+
"step": 3635
|
| 4470 |
+
},
|
| 4471 |
+
{
|
| 4472 |
+
"epoch": 0.29,
|
| 4473 |
+
"learning_rate": 4.510739213625462e-05,
|
| 4474 |
+
"loss": 3.0502,
|
| 4475 |
+
"step": 3640
|
| 4476 |
+
},
|
| 4477 |
+
{
|
| 4478 |
+
"epoch": 0.29,
|
| 4479 |
+
"learning_rate": 4.510065486296386e-05,
|
| 4480 |
+
"loss": 3.0543,
|
| 4481 |
+
"step": 3645
|
| 4482 |
+
},
|
| 4483 |
+
{
|
| 4484 |
+
"epoch": 0.3,
|
| 4485 |
+
"learning_rate": 4.5093917589673105e-05,
|
| 4486 |
+
"loss": 3.125,
|
| 4487 |
+
"step": 3650
|
| 4488 |
+
},
|
| 4489 |
+
{
|
| 4490 |
+
"epoch": 0.3,
|
| 4491 |
+
"learning_rate": 4.5087180316382353e-05,
|
| 4492 |
+
"loss": 3.1286,
|
| 4493 |
+
"step": 3655
|
| 4494 |
+
},
|
| 4495 |
+
{
|
| 4496 |
+
"epoch": 0.3,
|
| 4497 |
+
"learning_rate": 4.50804430430916e-05,
|
| 4498 |
+
"loss": 3.1895,
|
| 4499 |
+
"step": 3660
|
| 4500 |
+
},
|
| 4501 |
+
{
|
| 4502 |
+
"epoch": 0.3,
|
| 4503 |
+
"learning_rate": 4.507370576980085e-05,
|
| 4504 |
+
"loss": 3.0078,
|
| 4505 |
+
"step": 3665
|
| 4506 |
+
},
|
| 4507 |
+
{
|
| 4508 |
+
"epoch": 0.3,
|
| 4509 |
+
"learning_rate": 4.506696849651009e-05,
|
| 4510 |
+
"loss": 3.0766,
|
| 4511 |
+
"step": 3670
|
| 4512 |
+
},
|
| 4513 |
+
{
|
| 4514 |
+
"epoch": 0.3,
|
| 4515 |
+
"learning_rate": 4.506023122321934e-05,
|
| 4516 |
+
"loss": 3.4027,
|
| 4517 |
+
"step": 3675
|
| 4518 |
+
},
|
| 4519 |
+
{
|
| 4520 |
+
"epoch": 0.3,
|
| 4521 |
+
"learning_rate": 4.505349394992859e-05,
|
| 4522 |
+
"loss": 2.8891,
|
| 4523 |
+
"step": 3680
|
| 4524 |
+
},
|
| 4525 |
+
{
|
| 4526 |
+
"epoch": 0.3,
|
| 4527 |
+
"learning_rate": 4.504675667663783e-05,
|
| 4528 |
+
"loss": 3.0093,
|
| 4529 |
+
"step": 3685
|
| 4530 |
+
},
|
| 4531 |
+
{
|
| 4532 |
+
"epoch": 0.3,
|
| 4533 |
+
"learning_rate": 4.5040019403347075e-05,
|
| 4534 |
+
"loss": 3.1543,
|
| 4535 |
+
"step": 3690
|
| 4536 |
+
},
|
| 4537 |
+
{
|
| 4538 |
+
"epoch": 0.3,
|
| 4539 |
+
"learning_rate": 4.5033282130056324e-05,
|
| 4540 |
+
"loss": 3.0005,
|
| 4541 |
+
"step": 3695
|
| 4542 |
+
},
|
| 4543 |
+
{
|
| 4544 |
+
"epoch": 0.3,
|
| 4545 |
+
"learning_rate": 4.502654485676557e-05,
|
| 4546 |
+
"loss": 3.2393,
|
| 4547 |
+
"step": 3700
|
| 4548 |
+
},
|
| 4549 |
+
{
|
| 4550 |
+
"epoch": 0.3,
|
| 4551 |
+
"learning_rate": 4.5019807583474815e-05,
|
| 4552 |
+
"loss": 3.2052,
|
| 4553 |
+
"step": 3705
|
| 4554 |
+
},
|
| 4555 |
+
{
|
| 4556 |
+
"epoch": 0.3,
|
| 4557 |
+
"learning_rate": 4.5013070310184064e-05,
|
| 4558 |
+
"loss": 3.1979,
|
| 4559 |
+
"step": 3710
|
| 4560 |
+
},
|
| 4561 |
+
{
|
| 4562 |
+
"epoch": 0.3,
|
| 4563 |
+
"learning_rate": 4.500633303689331e-05,
|
| 4564 |
+
"loss": 3.1785,
|
| 4565 |
+
"step": 3715
|
| 4566 |
+
},
|
| 4567 |
+
{
|
| 4568 |
+
"epoch": 0.3,
|
| 4569 |
+
"learning_rate": 4.499959576360256e-05,
|
| 4570 |
+
"loss": 3.1174,
|
| 4571 |
+
"step": 3720
|
| 4572 |
+
},
|
| 4573 |
+
{
|
| 4574 |
+
"epoch": 0.3,
|
| 4575 |
+
"learning_rate": 4.4992858490311804e-05,
|
| 4576 |
+
"loss": 3.0281,
|
| 4577 |
+
"step": 3725
|
| 4578 |
+
},
|
| 4579 |
+
{
|
| 4580 |
+
"epoch": 0.3,
|
| 4581 |
+
"learning_rate": 4.4986121217021046e-05,
|
| 4582 |
+
"loss": 3.1682,
|
| 4583 |
+
"step": 3730
|
| 4584 |
+
},
|
| 4585 |
+
{
|
| 4586 |
+
"epoch": 0.3,
|
| 4587 |
+
"learning_rate": 4.4979383943730295e-05,
|
| 4588 |
+
"loss": 2.9703,
|
| 4589 |
+
"step": 3735
|
| 4590 |
+
},
|
| 4591 |
+
{
|
| 4592 |
+
"epoch": 0.3,
|
| 4593 |
+
"learning_rate": 4.4972646670439544e-05,
|
| 4594 |
+
"loss": 3.0364,
|
| 4595 |
+
"step": 3740
|
| 4596 |
+
},
|
| 4597 |
+
{
|
| 4598 |
+
"epoch": 0.3,
|
| 4599 |
+
"learning_rate": 4.4965909397148786e-05,
|
| 4600 |
+
"loss": 2.8145,
|
| 4601 |
+
"step": 3745
|
| 4602 |
+
},
|
| 4603 |
+
{
|
| 4604 |
+
"epoch": 0.3,
|
| 4605 |
+
"learning_rate": 4.4959172123858035e-05,
|
| 4606 |
+
"loss": 3.0781,
|
| 4607 |
+
"step": 3750
|
| 4608 |
+
},
|
| 4609 |
+
{
|
| 4610 |
+
"epoch": 0.3,
|
| 4611 |
+
"learning_rate": 4.4952434850567283e-05,
|
| 4612 |
+
"loss": 3.0378,
|
| 4613 |
+
"step": 3755
|
| 4614 |
+
},
|
| 4615 |
+
{
|
| 4616 |
+
"epoch": 0.3,
|
| 4617 |
+
"learning_rate": 4.4945697577276526e-05,
|
| 4618 |
+
"loss": 3.1284,
|
| 4619 |
+
"step": 3760
|
| 4620 |
+
},
|
| 4621 |
+
{
|
| 4622 |
+
"epoch": 0.3,
|
| 4623 |
+
"learning_rate": 4.4938960303985774e-05,
|
| 4624 |
+
"loss": 3.2207,
|
| 4625 |
+
"step": 3765
|
| 4626 |
+
},
|
| 4627 |
+
{
|
| 4628 |
+
"epoch": 0.3,
|
| 4629 |
+
"learning_rate": 4.4932223030695017e-05,
|
| 4630 |
+
"loss": 3.1658,
|
| 4631 |
+
"step": 3770
|
| 4632 |
+
},
|
| 4633 |
+
{
|
| 4634 |
+
"epoch": 0.31,
|
| 4635 |
+
"learning_rate": 4.4925485757404265e-05,
|
| 4636 |
+
"loss": 2.9948,
|
| 4637 |
+
"step": 3775
|
| 4638 |
+
},
|
| 4639 |
+
{
|
| 4640 |
+
"epoch": 0.31,
|
| 4641 |
+
"learning_rate": 4.491874848411351e-05,
|
| 4642 |
+
"loss": 3.1124,
|
| 4643 |
+
"step": 3780
|
| 4644 |
+
},
|
| 4645 |
+
{
|
| 4646 |
+
"epoch": 0.31,
|
| 4647 |
+
"learning_rate": 4.4912011210822756e-05,
|
| 4648 |
+
"loss": 2.9172,
|
| 4649 |
+
"step": 3785
|
| 4650 |
+
},
|
| 4651 |
+
{
|
| 4652 |
+
"epoch": 0.31,
|
| 4653 |
+
"learning_rate": 4.4905273937532005e-05,
|
| 4654 |
+
"loss": 3.22,
|
| 4655 |
+
"step": 3790
|
| 4656 |
+
},
|
| 4657 |
+
{
|
| 4658 |
+
"epoch": 0.31,
|
| 4659 |
+
"learning_rate": 4.4898536664241254e-05,
|
| 4660 |
+
"loss": 2.9467,
|
| 4661 |
+
"step": 3795
|
| 4662 |
+
},
|
| 4663 |
+
{
|
| 4664 |
+
"epoch": 0.31,
|
| 4665 |
+
"learning_rate": 4.4891799390950496e-05,
|
| 4666 |
+
"loss": 3.1856,
|
| 4667 |
+
"step": 3800
|
| 4668 |
+
},
|
| 4669 |
+
{
|
| 4670 |
+
"epoch": 0.31,
|
| 4671 |
+
"learning_rate": 4.4885062117659745e-05,
|
| 4672 |
+
"loss": 3.3563,
|
| 4673 |
+
"step": 3805
|
| 4674 |
+
},
|
| 4675 |
+
{
|
| 4676 |
+
"epoch": 0.31,
|
| 4677 |
+
"learning_rate": 4.4878324844368994e-05,
|
| 4678 |
+
"loss": 3.1518,
|
| 4679 |
+
"step": 3810
|
| 4680 |
+
},
|
| 4681 |
+
{
|
| 4682 |
+
"epoch": 0.31,
|
| 4683 |
+
"learning_rate": 4.4871587571078236e-05,
|
| 4684 |
+
"loss": 3.1046,
|
| 4685 |
+
"step": 3815
|
| 4686 |
+
},
|
| 4687 |
+
{
|
| 4688 |
+
"epoch": 0.31,
|
| 4689 |
+
"learning_rate": 4.486485029778748e-05,
|
| 4690 |
+
"loss": 3.1422,
|
| 4691 |
+
"step": 3820
|
| 4692 |
+
},
|
| 4693 |
+
{
|
| 4694 |
+
"epoch": 0.31,
|
| 4695 |
+
"learning_rate": 4.485811302449673e-05,
|
| 4696 |
+
"loss": 2.9912,
|
| 4697 |
+
"step": 3825
|
| 4698 |
+
},
|
| 4699 |
+
{
|
| 4700 |
+
"epoch": 0.31,
|
| 4701 |
+
"learning_rate": 4.4851375751205976e-05,
|
| 4702 |
+
"loss": 3.3937,
|
| 4703 |
+
"step": 3830
|
| 4704 |
+
},
|
| 4705 |
+
{
|
| 4706 |
+
"epoch": 0.31,
|
| 4707 |
+
"learning_rate": 4.484463847791522e-05,
|
| 4708 |
+
"loss": 2.772,
|
| 4709 |
+
"step": 3835
|
| 4710 |
+
},
|
| 4711 |
+
{
|
| 4712 |
+
"epoch": 0.31,
|
| 4713 |
+
"learning_rate": 4.483790120462447e-05,
|
| 4714 |
+
"loss": 3.1256,
|
| 4715 |
+
"step": 3840
|
| 4716 |
+
},
|
| 4717 |
+
{
|
| 4718 |
+
"epoch": 0.31,
|
| 4719 |
+
"learning_rate": 4.4831163931333716e-05,
|
| 4720 |
+
"loss": 3.2421,
|
| 4721 |
+
"step": 3845
|
| 4722 |
+
},
|
| 4723 |
+
{
|
| 4724 |
+
"epoch": 0.31,
|
| 4725 |
+
"learning_rate": 4.4824426658042965e-05,
|
| 4726 |
+
"loss": 2.8323,
|
| 4727 |
+
"step": 3850
|
| 4728 |
+
},
|
| 4729 |
+
{
|
| 4730 |
+
"epoch": 0.31,
|
| 4731 |
+
"learning_rate": 4.481768938475221e-05,
|
| 4732 |
+
"loss": 3.0066,
|
| 4733 |
+
"step": 3855
|
| 4734 |
+
},
|
| 4735 |
+
{
|
| 4736 |
+
"epoch": 0.31,
|
| 4737 |
+
"learning_rate": 4.481095211146145e-05,
|
| 4738 |
+
"loss": 3.2898,
|
| 4739 |
+
"step": 3860
|
| 4740 |
+
},
|
| 4741 |
+
{
|
| 4742 |
+
"epoch": 0.31,
|
| 4743 |
+
"learning_rate": 4.48042148381707e-05,
|
| 4744 |
+
"loss": 3.0331,
|
| 4745 |
+
"step": 3865
|
| 4746 |
+
},
|
| 4747 |
+
{
|
| 4748 |
+
"epoch": 0.31,
|
| 4749 |
+
"learning_rate": 4.479747756487994e-05,
|
| 4750 |
+
"loss": 3.0689,
|
| 4751 |
+
"step": 3870
|
| 4752 |
+
},
|
| 4753 |
+
{
|
| 4754 |
+
"epoch": 0.31,
|
| 4755 |
+
"learning_rate": 4.479074029158919e-05,
|
| 4756 |
+
"loss": 2.9034,
|
| 4757 |
+
"step": 3875
|
| 4758 |
+
},
|
| 4759 |
+
{
|
| 4760 |
+
"epoch": 0.31,
|
| 4761 |
+
"learning_rate": 4.478400301829844e-05,
|
| 4762 |
+
"loss": 2.8014,
|
| 4763 |
+
"step": 3880
|
| 4764 |
+
},
|
| 4765 |
+
{
|
| 4766 |
+
"epoch": 0.31,
|
| 4767 |
+
"learning_rate": 4.4777265745007686e-05,
|
| 4768 |
+
"loss": 2.984,
|
| 4769 |
+
"step": 3885
|
| 4770 |
+
},
|
| 4771 |
+
{
|
| 4772 |
+
"epoch": 0.31,
|
| 4773 |
+
"learning_rate": 4.477052847171693e-05,
|
| 4774 |
+
"loss": 3.0642,
|
| 4775 |
+
"step": 3890
|
| 4776 |
+
},
|
| 4777 |
+
{
|
| 4778 |
+
"epoch": 0.31,
|
| 4779 |
+
"learning_rate": 4.476379119842618e-05,
|
| 4780 |
+
"loss": 2.9044,
|
| 4781 |
+
"step": 3895
|
| 4782 |
+
},
|
| 4783 |
+
{
|
| 4784 |
+
"epoch": 0.32,
|
| 4785 |
+
"learning_rate": 4.475705392513542e-05,
|
| 4786 |
+
"loss": 3.2954,
|
| 4787 |
+
"step": 3900
|
| 4788 |
+
},
|
| 4789 |
+
{
|
| 4790 |
+
"epoch": 0.32,
|
| 4791 |
+
"learning_rate": 4.475031665184466e-05,
|
| 4792 |
+
"loss": 3.1445,
|
| 4793 |
+
"step": 3905
|
| 4794 |
+
},
|
| 4795 |
+
{
|
| 4796 |
+
"epoch": 0.32,
|
| 4797 |
+
"learning_rate": 4.474357937855391e-05,
|
| 4798 |
+
"loss": 3.3147,
|
| 4799 |
+
"step": 3910
|
| 4800 |
+
},
|
| 4801 |
+
{
|
| 4802 |
+
"epoch": 0.32,
|
| 4803 |
+
"learning_rate": 4.473684210526316e-05,
|
| 4804 |
+
"loss": 3.2616,
|
| 4805 |
+
"step": 3915
|
| 4806 |
+
},
|
| 4807 |
+
{
|
| 4808 |
+
"epoch": 0.32,
|
| 4809 |
+
"learning_rate": 4.473010483197241e-05,
|
| 4810 |
+
"loss": 3.3639,
|
| 4811 |
+
"step": 3920
|
| 4812 |
+
},
|
| 4813 |
+
{
|
| 4814 |
+
"epoch": 0.32,
|
| 4815 |
+
"learning_rate": 4.472336755868165e-05,
|
| 4816 |
+
"loss": 2.9422,
|
| 4817 |
+
"step": 3925
|
| 4818 |
+
},
|
| 4819 |
+
{
|
| 4820 |
+
"epoch": 0.32,
|
| 4821 |
+
"learning_rate": 4.47166302853909e-05,
|
| 4822 |
+
"loss": 3.193,
|
| 4823 |
+
"step": 3930
|
| 4824 |
+
},
|
| 4825 |
+
{
|
| 4826 |
+
"epoch": 0.32,
|
| 4827 |
+
"learning_rate": 4.470989301210015e-05,
|
| 4828 |
+
"loss": 3.0869,
|
| 4829 |
+
"step": 3935
|
| 4830 |
+
},
|
| 4831 |
+
{
|
| 4832 |
+
"epoch": 0.32,
|
| 4833 |
+
"learning_rate": 4.47031557388094e-05,
|
| 4834 |
+
"loss": 3.097,
|
| 4835 |
+
"step": 3940
|
| 4836 |
+
},
|
| 4837 |
+
{
|
| 4838 |
+
"epoch": 0.32,
|
| 4839 |
+
"learning_rate": 4.469641846551864e-05,
|
| 4840 |
+
"loss": 3.2279,
|
| 4841 |
+
"step": 3945
|
| 4842 |
+
},
|
| 4843 |
+
{
|
| 4844 |
+
"epoch": 0.32,
|
| 4845 |
+
"learning_rate": 4.468968119222788e-05,
|
| 4846 |
+
"loss": 3.4025,
|
| 4847 |
+
"step": 3950
|
| 4848 |
+
},
|
| 4849 |
+
{
|
| 4850 |
+
"epoch": 0.32,
|
| 4851 |
+
"learning_rate": 4.468294391893713e-05,
|
| 4852 |
+
"loss": 3.0516,
|
| 4853 |
+
"step": 3955
|
| 4854 |
+
},
|
| 4855 |
+
{
|
| 4856 |
+
"epoch": 0.32,
|
| 4857 |
+
"learning_rate": 4.467620664564637e-05,
|
| 4858 |
+
"loss": 3.0323,
|
| 4859 |
+
"step": 3960
|
| 4860 |
+
},
|
| 4861 |
+
{
|
| 4862 |
+
"epoch": 0.32,
|
| 4863 |
+
"learning_rate": 4.466946937235562e-05,
|
| 4864 |
+
"loss": 2.7716,
|
| 4865 |
+
"step": 3965
|
| 4866 |
+
},
|
| 4867 |
+
{
|
| 4868 |
+
"epoch": 0.32,
|
| 4869 |
+
"learning_rate": 4.466273209906487e-05,
|
| 4870 |
+
"loss": 3.1923,
|
| 4871 |
+
"step": 3970
|
| 4872 |
+
},
|
| 4873 |
+
{
|
| 4874 |
+
"epoch": 0.32,
|
| 4875 |
+
"learning_rate": 4.465599482577412e-05,
|
| 4876 |
+
"loss": 3.1534,
|
| 4877 |
+
"step": 3975
|
| 4878 |
+
},
|
| 4879 |
+
{
|
| 4880 |
+
"epoch": 0.32,
|
| 4881 |
+
"learning_rate": 4.464925755248336e-05,
|
| 4882 |
+
"loss": 2.8173,
|
| 4883 |
+
"step": 3980
|
| 4884 |
+
},
|
| 4885 |
+
{
|
| 4886 |
+
"epoch": 0.32,
|
| 4887 |
+
"learning_rate": 4.464252027919261e-05,
|
| 4888 |
+
"loss": 3.0176,
|
| 4889 |
+
"step": 3985
|
| 4890 |
+
},
|
| 4891 |
+
{
|
| 4892 |
+
"epoch": 0.32,
|
| 4893 |
+
"learning_rate": 4.463578300590185e-05,
|
| 4894 |
+
"loss": 3.2045,
|
| 4895 |
+
"step": 3990
|
| 4896 |
+
},
|
| 4897 |
+
{
|
| 4898 |
+
"epoch": 0.32,
|
| 4899 |
+
"learning_rate": 4.46290457326111e-05,
|
| 4900 |
+
"loss": 3.0714,
|
| 4901 |
+
"step": 3995
|
| 4902 |
+
},
|
| 4903 |
+
{
|
| 4904 |
+
"epoch": 0.32,
|
| 4905 |
+
"learning_rate": 4.462230845932034e-05,
|
| 4906 |
+
"loss": 3.0958,
|
| 4907 |
+
"step": 4000
|
| 4908 |
+
},
|
| 4909 |
+
{
|
| 4910 |
+
"epoch": 0.32,
|
| 4911 |
+
"eval_loss": 3.0753977298736572,
|
| 4912 |
+
"eval_rouge2_fmeasure": 0.0055,
|
| 4913 |
+
"eval_rouge2_precision": 0.0064,
|
| 4914 |
+
"eval_rouge2_recall": 0.0063,
|
| 4915 |
+
"eval_runtime": 2992.0589,
|
| 4916 |
+
"eval_samples_per_second": 0.092,
|
| 4917 |
+
"eval_steps_per_second": 0.046,
|
| 4918 |
+
"step": 4000
|
| 4919 |
}
|
| 4920 |
],
|
| 4921 |
+
"max_steps": 37107,
|
| 4922 |
"num_train_epochs": 3,
|
| 4923 |
+
"total_flos": 1.56090117390336e+17,
|
| 4924 |
"trial_name": null,
|
| 4925 |
"trial_params": null
|
| 4926 |
}
|
last-checkpoint/training_args.bin
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 4091
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b0a9b2c35b144e3bc8d84766c46ac44d3749c9a4ed84e04c65e17520de36cdda
|
| 3 |
size 4091
|
pytorch_model.bin
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 1140034381
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c57d5e652de14e5f01a9d33e0167070f639ab8a97f101a3d3639e153530869ce
|
| 3 |
size 1140034381
|
spiece.model
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:fa8e856a5564ec886599b5ed6728e3bd0cf9e040a2a56dfb6bf87c63a32a5caf
|
| 3 |
+
size 2243614
|
training_args.bin
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 4091
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b0a9b2c35b144e3bc8d84766c46ac44d3749c9a4ed84e04c65e17520de36cdda
|
| 3 |
size 4091
|