Training in progress, epoch 11, checkpoint
Browse files
last-checkpoint/model.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 1575259780
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:40101c1559c0e8b31c73e81cc56dc1705da4957549fc7bd61f943d1924c4d786
|
| 3 |
size 1575259780
|
last-checkpoint/optimizer.pt
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 3150397656
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:536042680e764d5473ffed47c2a7949c1d2603a96a1970b48df10636c86cdcb5
|
| 3 |
size 3150397656
|
last-checkpoint/rng_state.pth
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 14244
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c43173be6b32e53765cce6f0605e936f3336963827ebf73a77ddc0d38d5e9853
|
| 3 |
size 14244
|
last-checkpoint/scheduler.pt
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 1064
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:840d85fbedd27b38bfcd367e046a848a780eccbdb0c8cbf7e42f589bab045a79
|
| 3 |
size 1064
|
last-checkpoint/trainer_state.json
CHANGED
|
@@ -1,9 +1,9 @@
|
|
| 1 |
{
|
| 2 |
-
"best_metric": 12.
|
| 3 |
-
"best_model_checkpoint": "/kaggle/tmp/amr-tst-indo/AMRBART-id/fine-tune/../outputs/mbart-en-id-smaller-fted/checkpoint-
|
| 4 |
-
"epoch": 11.
|
| 5 |
"eval_steps": 500,
|
| 6 |
-
"global_step":
|
| 7 |
"is_hyper_param_search": false,
|
| 8 |
"is_local_process_zero": true,
|
| 9 |
"is_world_process_zero": true,
|
|
@@ -12891,6 +12891,1174 @@
|
|
| 12891 |
"eval_samples_per_second": 2.165,
|
| 12892 |
"eval_steps_per_second": 1.083,
|
| 12893 |
"step": 42562
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 12894 |
}
|
| 12895 |
],
|
| 12896 |
"logging_steps": 20,
|
|
@@ -12910,7 +14078,7 @@
|
|
| 12910 |
"attributes": {}
|
| 12911 |
}
|
| 12912 |
},
|
| 12913 |
-
"total_flos":
|
| 12914 |
"train_batch_size": 2,
|
| 12915 |
"trial_name": null,
|
| 12916 |
"trial_params": null
|
|
|
|
| 1 |
{
|
| 2 |
+
"best_metric": 12.51,
|
| 3 |
+
"best_model_checkpoint": "/kaggle/tmp/amr-tst-indo/AMRBART-id/fine-tune/../outputs/mbart-en-id-smaller-fted/checkpoint-46428",
|
| 4 |
+
"epoch": 11.999870784339063,
|
| 5 |
"eval_steps": 500,
|
| 6 |
+
"global_step": 46428,
|
| 7 |
"is_hyper_param_search": false,
|
| 8 |
"is_local_process_zero": true,
|
| 9 |
"is_world_process_zero": true,
|
|
|
|
| 12891 |
"eval_samples_per_second": 2.165,
|
| 12892 |
"eval_steps_per_second": 1.083,
|
| 12893 |
"step": 42562
|
| 12894 |
+
},
|
| 12895 |
+
{
|
| 12896 |
+
"epoch": 11.0054270577594,
|
| 12897 |
+
"learning_rate": 1.3131725657980034e-07,
|
| 12898 |
+
"loss": 2.4441,
|
| 12899 |
+
"step": 42580
|
| 12900 |
+
},
|
| 12901 |
+
{
|
| 12902 |
+
"epoch": 11.010595684196925,
|
| 12903 |
+
"learning_rate": 1.31284843770258e-07,
|
| 12904 |
+
"loss": 2.5027,
|
| 12905 |
+
"step": 42600
|
| 12906 |
+
},
|
| 12907 |
+
{
|
| 12908 |
+
"epoch": 11.015764310634449,
|
| 12909 |
+
"learning_rate": 1.3125243096071566e-07,
|
| 12910 |
+
"loss": 2.4709,
|
| 12911 |
+
"step": 42620
|
| 12912 |
+
},
|
| 12913 |
+
{
|
| 12914 |
+
"epoch": 11.020932937071974,
|
| 12915 |
+
"learning_rate": 1.3122001815117335e-07,
|
| 12916 |
+
"loss": 2.4576,
|
| 12917 |
+
"step": 42640
|
| 12918 |
+
},
|
| 12919 |
+
{
|
| 12920 |
+
"epoch": 11.026101563509497,
|
| 12921 |
+
"learning_rate": 1.31187605341631e-07,
|
| 12922 |
+
"loss": 2.4395,
|
| 12923 |
+
"step": 42660
|
| 12924 |
+
},
|
| 12925 |
+
{
|
| 12926 |
+
"epoch": 11.031270189947021,
|
| 12927 |
+
"learning_rate": 1.3115519253208867e-07,
|
| 12928 |
+
"loss": 2.504,
|
| 12929 |
+
"step": 42680
|
| 12930 |
+
},
|
| 12931 |
+
{
|
| 12932 |
+
"epoch": 11.036438816384546,
|
| 12933 |
+
"learning_rate": 1.3112277972254634e-07,
|
| 12934 |
+
"loss": 2.5521,
|
| 12935 |
+
"step": 42700
|
| 12936 |
+
},
|
| 12937 |
+
{
|
| 12938 |
+
"epoch": 11.04160744282207,
|
| 12939 |
+
"learning_rate": 1.31090366913004e-07,
|
| 12940 |
+
"loss": 2.5074,
|
| 12941 |
+
"step": 42720
|
| 12942 |
+
},
|
| 12943 |
+
{
|
| 12944 |
+
"epoch": 11.046776069259595,
|
| 12945 |
+
"learning_rate": 1.3105795410346169e-07,
|
| 12946 |
+
"loss": 2.5079,
|
| 12947 |
+
"step": 42740
|
| 12948 |
+
},
|
| 12949 |
+
{
|
| 12950 |
+
"epoch": 11.051944695697118,
|
| 12951 |
+
"learning_rate": 1.3102554129391935e-07,
|
| 12952 |
+
"loss": 2.5337,
|
| 12953 |
+
"step": 42760
|
| 12954 |
+
},
|
| 12955 |
+
{
|
| 12956 |
+
"epoch": 11.057113322134644,
|
| 12957 |
+
"learning_rate": 1.30993128484377e-07,
|
| 12958 |
+
"loss": 2.4815,
|
| 12959 |
+
"step": 42780
|
| 12960 |
+
},
|
| 12961 |
+
{
|
| 12962 |
+
"epoch": 11.062281948572167,
|
| 12963 |
+
"learning_rate": 1.309607156748347e-07,
|
| 12964 |
+
"loss": 2.4897,
|
| 12965 |
+
"step": 42800
|
| 12966 |
+
},
|
| 12967 |
+
{
|
| 12968 |
+
"epoch": 11.06745057500969,
|
| 12969 |
+
"learning_rate": 1.3092830286529236e-07,
|
| 12970 |
+
"loss": 2.5184,
|
| 12971 |
+
"step": 42820
|
| 12972 |
+
},
|
| 12973 |
+
{
|
| 12974 |
+
"epoch": 11.072619201447216,
|
| 12975 |
+
"learning_rate": 1.3089589005575002e-07,
|
| 12976 |
+
"loss": 2.459,
|
| 12977 |
+
"step": 42840
|
| 12978 |
+
},
|
| 12979 |
+
{
|
| 12980 |
+
"epoch": 11.07778782788474,
|
| 12981 |
+
"learning_rate": 1.308634772462077e-07,
|
| 12982 |
+
"loss": 2.4623,
|
| 12983 |
+
"step": 42860
|
| 12984 |
+
},
|
| 12985 |
+
{
|
| 12986 |
+
"epoch": 11.082956454322265,
|
| 12987 |
+
"learning_rate": 1.3083106443666535e-07,
|
| 12988 |
+
"loss": 2.4422,
|
| 12989 |
+
"step": 42880
|
| 12990 |
+
},
|
| 12991 |
+
{
|
| 12992 |
+
"epoch": 11.088125080759788,
|
| 12993 |
+
"learning_rate": 1.3079865162712303e-07,
|
| 12994 |
+
"loss": 2.5212,
|
| 12995 |
+
"step": 42900
|
| 12996 |
+
},
|
| 12997 |
+
{
|
| 12998 |
+
"epoch": 11.093293707197311,
|
| 12999 |
+
"learning_rate": 1.307662388175807e-07,
|
| 13000 |
+
"loss": 2.4923,
|
| 13001 |
+
"step": 42920
|
| 13002 |
+
},
|
| 13003 |
+
{
|
| 13004 |
+
"epoch": 11.098462333634837,
|
| 13005 |
+
"learning_rate": 1.3073382600803836e-07,
|
| 13006 |
+
"loss": 2.5076,
|
| 13007 |
+
"step": 42940
|
| 13008 |
+
},
|
| 13009 |
+
{
|
| 13010 |
+
"epoch": 11.10363096007236,
|
| 13011 |
+
"learning_rate": 1.3070141319849605e-07,
|
| 13012 |
+
"loss": 2.5291,
|
| 13013 |
+
"step": 42960
|
| 13014 |
+
},
|
| 13015 |
+
{
|
| 13016 |
+
"epoch": 11.108799586509885,
|
| 13017 |
+
"learning_rate": 1.306690003889537e-07,
|
| 13018 |
+
"loss": 2.4459,
|
| 13019 |
+
"step": 42980
|
| 13020 |
+
},
|
| 13021 |
+
{
|
| 13022 |
+
"epoch": 11.113968212947409,
|
| 13023 |
+
"learning_rate": 1.3063658757941137e-07,
|
| 13024 |
+
"loss": 2.4522,
|
| 13025 |
+
"step": 43000
|
| 13026 |
+
},
|
| 13027 |
+
{
|
| 13028 |
+
"epoch": 11.119136839384934,
|
| 13029 |
+
"learning_rate": 1.3060417476986906e-07,
|
| 13030 |
+
"loss": 2.4752,
|
| 13031 |
+
"step": 43020
|
| 13032 |
+
},
|
| 13033 |
+
{
|
| 13034 |
+
"epoch": 11.124305465822458,
|
| 13035 |
+
"learning_rate": 1.3057176196032672e-07,
|
| 13036 |
+
"loss": 2.4877,
|
| 13037 |
+
"step": 43040
|
| 13038 |
+
},
|
| 13039 |
+
{
|
| 13040 |
+
"epoch": 11.129474092259981,
|
| 13041 |
+
"learning_rate": 1.3053934915078438e-07,
|
| 13042 |
+
"loss": 2.4969,
|
| 13043 |
+
"step": 43060
|
| 13044 |
+
},
|
| 13045 |
+
{
|
| 13046 |
+
"epoch": 11.134642718697506,
|
| 13047 |
+
"learning_rate": 1.3050693634124207e-07,
|
| 13048 |
+
"loss": 2.5331,
|
| 13049 |
+
"step": 43080
|
| 13050 |
+
},
|
| 13051 |
+
{
|
| 13052 |
+
"epoch": 11.13981134513503,
|
| 13053 |
+
"learning_rate": 1.304745235316997e-07,
|
| 13054 |
+
"loss": 2.4214,
|
| 13055 |
+
"step": 43100
|
| 13056 |
+
},
|
| 13057 |
+
{
|
| 13058 |
+
"epoch": 11.144979971572555,
|
| 13059 |
+
"learning_rate": 1.304421107221574e-07,
|
| 13060 |
+
"loss": 2.4795,
|
| 13061 |
+
"step": 43120
|
| 13062 |
+
},
|
| 13063 |
+
{
|
| 13064 |
+
"epoch": 11.150148598010079,
|
| 13065 |
+
"learning_rate": 1.3040969791261506e-07,
|
| 13066 |
+
"loss": 2.491,
|
| 13067 |
+
"step": 43140
|
| 13068 |
+
},
|
| 13069 |
+
{
|
| 13070 |
+
"epoch": 11.155317224447604,
|
| 13071 |
+
"learning_rate": 1.3037728510307272e-07,
|
| 13072 |
+
"loss": 2.4823,
|
| 13073 |
+
"step": 43160
|
| 13074 |
+
},
|
| 13075 |
+
{
|
| 13076 |
+
"epoch": 11.160485850885127,
|
| 13077 |
+
"learning_rate": 1.303448722935304e-07,
|
| 13078 |
+
"loss": 2.5129,
|
| 13079 |
+
"step": 43180
|
| 13080 |
+
},
|
| 13081 |
+
{
|
| 13082 |
+
"epoch": 11.16565447732265,
|
| 13083 |
+
"learning_rate": 1.3031245948398807e-07,
|
| 13084 |
+
"loss": 2.4543,
|
| 13085 |
+
"step": 43200
|
| 13086 |
+
},
|
| 13087 |
+
{
|
| 13088 |
+
"epoch": 11.170823103760176,
|
| 13089 |
+
"learning_rate": 1.3028004667444573e-07,
|
| 13090 |
+
"loss": 2.4735,
|
| 13091 |
+
"step": 43220
|
| 13092 |
+
},
|
| 13093 |
+
{
|
| 13094 |
+
"epoch": 11.1759917301977,
|
| 13095 |
+
"learning_rate": 1.3024763386490342e-07,
|
| 13096 |
+
"loss": 2.5127,
|
| 13097 |
+
"step": 43240
|
| 13098 |
+
},
|
| 13099 |
+
{
|
| 13100 |
+
"epoch": 11.181160356635225,
|
| 13101 |
+
"learning_rate": 1.3021522105536105e-07,
|
| 13102 |
+
"loss": 2.4045,
|
| 13103 |
+
"step": 43260
|
| 13104 |
+
},
|
| 13105 |
+
{
|
| 13106 |
+
"epoch": 11.186328983072748,
|
| 13107 |
+
"learning_rate": 1.3018280824581874e-07,
|
| 13108 |
+
"loss": 2.4599,
|
| 13109 |
+
"step": 43280
|
| 13110 |
+
},
|
| 13111 |
+
{
|
| 13112 |
+
"epoch": 11.191497609510273,
|
| 13113 |
+
"learning_rate": 1.301503954362764e-07,
|
| 13114 |
+
"loss": 2.4875,
|
| 13115 |
+
"step": 43300
|
| 13116 |
+
},
|
| 13117 |
+
{
|
| 13118 |
+
"epoch": 11.196666235947797,
|
| 13119 |
+
"learning_rate": 1.3011798262673407e-07,
|
| 13120 |
+
"loss": 2.4523,
|
| 13121 |
+
"step": 43320
|
| 13122 |
+
},
|
| 13123 |
+
{
|
| 13124 |
+
"epoch": 11.20183486238532,
|
| 13125 |
+
"learning_rate": 1.3008556981719175e-07,
|
| 13126 |
+
"loss": 2.47,
|
| 13127 |
+
"step": 43340
|
| 13128 |
+
},
|
| 13129 |
+
{
|
| 13130 |
+
"epoch": 11.207003488822846,
|
| 13131 |
+
"learning_rate": 1.3005315700764942e-07,
|
| 13132 |
+
"loss": 2.464,
|
| 13133 |
+
"step": 43360
|
| 13134 |
+
},
|
| 13135 |
+
{
|
| 13136 |
+
"epoch": 11.212172115260369,
|
| 13137 |
+
"learning_rate": 1.3002074419810708e-07,
|
| 13138 |
+
"loss": 2.5223,
|
| 13139 |
+
"step": 43380
|
| 13140 |
+
},
|
| 13141 |
+
{
|
| 13142 |
+
"epoch": 11.217340741697894,
|
| 13143 |
+
"learning_rate": 1.2998833138856477e-07,
|
| 13144 |
+
"loss": 2.4749,
|
| 13145 |
+
"step": 43400
|
| 13146 |
+
},
|
| 13147 |
+
{
|
| 13148 |
+
"epoch": 11.222509368135418,
|
| 13149 |
+
"learning_rate": 1.2995591857902243e-07,
|
| 13150 |
+
"loss": 2.5268,
|
| 13151 |
+
"step": 43420
|
| 13152 |
+
},
|
| 13153 |
+
{
|
| 13154 |
+
"epoch": 11.227677994572943,
|
| 13155 |
+
"learning_rate": 1.299235057694801e-07,
|
| 13156 |
+
"loss": 2.4837,
|
| 13157 |
+
"step": 43440
|
| 13158 |
+
},
|
| 13159 |
+
{
|
| 13160 |
+
"epoch": 11.232846621010466,
|
| 13161 |
+
"learning_rate": 1.2989109295993778e-07,
|
| 13162 |
+
"loss": 2.4285,
|
| 13163 |
+
"step": 43460
|
| 13164 |
+
},
|
| 13165 |
+
{
|
| 13166 |
+
"epoch": 11.23801524744799,
|
| 13167 |
+
"learning_rate": 1.2985868015039541e-07,
|
| 13168 |
+
"loss": 2.4918,
|
| 13169 |
+
"step": 43480
|
| 13170 |
+
},
|
| 13171 |
+
{
|
| 13172 |
+
"epoch": 11.243183873885515,
|
| 13173 |
+
"learning_rate": 1.298262673408531e-07,
|
| 13174 |
+
"loss": 2.4585,
|
| 13175 |
+
"step": 43500
|
| 13176 |
+
},
|
| 13177 |
+
{
|
| 13178 |
+
"epoch": 11.248352500323039,
|
| 13179 |
+
"learning_rate": 1.2979385453131076e-07,
|
| 13180 |
+
"loss": 2.5156,
|
| 13181 |
+
"step": 43520
|
| 13182 |
+
},
|
| 13183 |
+
{
|
| 13184 |
+
"epoch": 11.253521126760564,
|
| 13185 |
+
"learning_rate": 1.2976144172176843e-07,
|
| 13186 |
+
"loss": 2.5031,
|
| 13187 |
+
"step": 43540
|
| 13188 |
+
},
|
| 13189 |
+
{
|
| 13190 |
+
"epoch": 11.258689753198087,
|
| 13191 |
+
"learning_rate": 1.2972902891222611e-07,
|
| 13192 |
+
"loss": 2.456,
|
| 13193 |
+
"step": 43560
|
| 13194 |
+
},
|
| 13195 |
+
{
|
| 13196 |
+
"epoch": 11.263858379635613,
|
| 13197 |
+
"learning_rate": 1.2969661610268378e-07,
|
| 13198 |
+
"loss": 2.5083,
|
| 13199 |
+
"step": 43580
|
| 13200 |
+
},
|
| 13201 |
+
{
|
| 13202 |
+
"epoch": 11.269027006073136,
|
| 13203 |
+
"learning_rate": 1.2966420329314144e-07,
|
| 13204 |
+
"loss": 2.5241,
|
| 13205 |
+
"step": 43600
|
| 13206 |
+
},
|
| 13207 |
+
{
|
| 13208 |
+
"epoch": 11.27419563251066,
|
| 13209 |
+
"learning_rate": 1.2963179048359913e-07,
|
| 13210 |
+
"loss": 2.4751,
|
| 13211 |
+
"step": 43620
|
| 13212 |
+
},
|
| 13213 |
+
{
|
| 13214 |
+
"epoch": 11.279364258948185,
|
| 13215 |
+
"learning_rate": 1.295993776740568e-07,
|
| 13216 |
+
"loss": 2.4695,
|
| 13217 |
+
"step": 43640
|
| 13218 |
+
},
|
| 13219 |
+
{
|
| 13220 |
+
"epoch": 11.284532885385708,
|
| 13221 |
+
"learning_rate": 1.2956696486451445e-07,
|
| 13222 |
+
"loss": 2.5578,
|
| 13223 |
+
"step": 43660
|
| 13224 |
+
},
|
| 13225 |
+
{
|
| 13226 |
+
"epoch": 11.289701511823234,
|
| 13227 |
+
"learning_rate": 1.2953455205497214e-07,
|
| 13228 |
+
"loss": 2.4649,
|
| 13229 |
+
"step": 43680
|
| 13230 |
+
},
|
| 13231 |
+
{
|
| 13232 |
+
"epoch": 11.294870138260757,
|
| 13233 |
+
"learning_rate": 1.2950213924542977e-07,
|
| 13234 |
+
"loss": 2.4747,
|
| 13235 |
+
"step": 43700
|
| 13236 |
+
},
|
| 13237 |
+
{
|
| 13238 |
+
"epoch": 11.300038764698282,
|
| 13239 |
+
"learning_rate": 1.2946972643588746e-07,
|
| 13240 |
+
"loss": 2.4679,
|
| 13241 |
+
"step": 43720
|
| 13242 |
+
},
|
| 13243 |
+
{
|
| 13244 |
+
"epoch": 11.305207391135806,
|
| 13245 |
+
"learning_rate": 1.2943731362634512e-07,
|
| 13246 |
+
"loss": 2.4683,
|
| 13247 |
+
"step": 43740
|
| 13248 |
+
},
|
| 13249 |
+
{
|
| 13250 |
+
"epoch": 11.31037601757333,
|
| 13251 |
+
"learning_rate": 1.2940490081680279e-07,
|
| 13252 |
+
"loss": 2.4857,
|
| 13253 |
+
"step": 43760
|
| 13254 |
+
},
|
| 13255 |
+
{
|
| 13256 |
+
"epoch": 11.315544644010854,
|
| 13257 |
+
"learning_rate": 1.2937248800726047e-07,
|
| 13258 |
+
"loss": 2.426,
|
| 13259 |
+
"step": 43780
|
| 13260 |
+
},
|
| 13261 |
+
{
|
| 13262 |
+
"epoch": 11.320713270448378,
|
| 13263 |
+
"learning_rate": 1.2934007519771814e-07,
|
| 13264 |
+
"loss": 2.4733,
|
| 13265 |
+
"step": 43800
|
| 13266 |
+
},
|
| 13267 |
+
{
|
| 13268 |
+
"epoch": 11.325881896885903,
|
| 13269 |
+
"learning_rate": 1.293076623881758e-07,
|
| 13270 |
+
"loss": 2.4812,
|
| 13271 |
+
"step": 43820
|
| 13272 |
+
},
|
| 13273 |
+
{
|
| 13274 |
+
"epoch": 11.331050523323427,
|
| 13275 |
+
"learning_rate": 1.2927524957863349e-07,
|
| 13276 |
+
"loss": 2.4528,
|
| 13277 |
+
"step": 43840
|
| 13278 |
+
},
|
| 13279 |
+
{
|
| 13280 |
+
"epoch": 11.33621914976095,
|
| 13281 |
+
"learning_rate": 1.2924283676909112e-07,
|
| 13282 |
+
"loss": 2.494,
|
| 13283 |
+
"step": 43860
|
| 13284 |
+
},
|
| 13285 |
+
{
|
| 13286 |
+
"epoch": 11.341387776198475,
|
| 13287 |
+
"learning_rate": 1.292104239595488e-07,
|
| 13288 |
+
"loss": 2.4795,
|
| 13289 |
+
"step": 43880
|
| 13290 |
+
},
|
| 13291 |
+
{
|
| 13292 |
+
"epoch": 11.346556402635999,
|
| 13293 |
+
"learning_rate": 1.2917801115000647e-07,
|
| 13294 |
+
"loss": 2.4816,
|
| 13295 |
+
"step": 43900
|
| 13296 |
+
},
|
| 13297 |
+
{
|
| 13298 |
+
"epoch": 11.351725029073524,
|
| 13299 |
+
"learning_rate": 1.2914559834046413e-07,
|
| 13300 |
+
"loss": 2.5282,
|
| 13301 |
+
"step": 43920
|
| 13302 |
+
},
|
| 13303 |
+
{
|
| 13304 |
+
"epoch": 11.356893655511048,
|
| 13305 |
+
"learning_rate": 1.2911318553092182e-07,
|
| 13306 |
+
"loss": 2.4849,
|
| 13307 |
+
"step": 43940
|
| 13308 |
+
},
|
| 13309 |
+
{
|
| 13310 |
+
"epoch": 11.362062281948573,
|
| 13311 |
+
"learning_rate": 1.2908077272137948e-07,
|
| 13312 |
+
"loss": 2.4246,
|
| 13313 |
+
"step": 43960
|
| 13314 |
+
},
|
| 13315 |
+
{
|
| 13316 |
+
"epoch": 11.367230908386096,
|
| 13317 |
+
"learning_rate": 1.2904835991183715e-07,
|
| 13318 |
+
"loss": 2.5491,
|
| 13319 |
+
"step": 43980
|
| 13320 |
+
},
|
| 13321 |
+
{
|
| 13322 |
+
"epoch": 11.372399534823622,
|
| 13323 |
+
"learning_rate": 1.2901594710229483e-07,
|
| 13324 |
+
"loss": 2.4584,
|
| 13325 |
+
"step": 44000
|
| 13326 |
+
},
|
| 13327 |
+
{
|
| 13328 |
+
"epoch": 11.377568161261145,
|
| 13329 |
+
"learning_rate": 1.289835342927525e-07,
|
| 13330 |
+
"loss": 2.5351,
|
| 13331 |
+
"step": 44020
|
| 13332 |
+
},
|
| 13333 |
+
{
|
| 13334 |
+
"epoch": 11.382736787698668,
|
| 13335 |
+
"learning_rate": 1.2895112148321016e-07,
|
| 13336 |
+
"loss": 2.5108,
|
| 13337 |
+
"step": 44040
|
| 13338 |
+
},
|
| 13339 |
+
{
|
| 13340 |
+
"epoch": 11.387905414136194,
|
| 13341 |
+
"learning_rate": 1.2891870867366782e-07,
|
| 13342 |
+
"loss": 2.5328,
|
| 13343 |
+
"step": 44060
|
| 13344 |
+
},
|
| 13345 |
+
{
|
| 13346 |
+
"epoch": 11.393074040573717,
|
| 13347 |
+
"learning_rate": 1.2888629586412548e-07,
|
| 13348 |
+
"loss": 2.4501,
|
| 13349 |
+
"step": 44080
|
| 13350 |
+
},
|
| 13351 |
+
{
|
| 13352 |
+
"epoch": 11.398242667011242,
|
| 13353 |
+
"learning_rate": 1.2885388305458317e-07,
|
| 13354 |
+
"loss": 2.434,
|
| 13355 |
+
"step": 44100
|
| 13356 |
+
},
|
| 13357 |
+
{
|
| 13358 |
+
"epoch": 11.403411293448766,
|
| 13359 |
+
"learning_rate": 1.2882147024504083e-07,
|
| 13360 |
+
"loss": 2.4713,
|
| 13361 |
+
"step": 44120
|
| 13362 |
+
},
|
| 13363 |
+
{
|
| 13364 |
+
"epoch": 11.40857991988629,
|
| 13365 |
+
"learning_rate": 1.287890574354985e-07,
|
| 13366 |
+
"loss": 2.5093,
|
| 13367 |
+
"step": 44140
|
| 13368 |
+
},
|
| 13369 |
+
{
|
| 13370 |
+
"epoch": 11.413748546323815,
|
| 13371 |
+
"learning_rate": 1.2875664462595618e-07,
|
| 13372 |
+
"loss": 2.5031,
|
| 13373 |
+
"step": 44160
|
| 13374 |
+
},
|
| 13375 |
+
{
|
| 13376 |
+
"epoch": 11.418917172761338,
|
| 13377 |
+
"learning_rate": 1.2872423181641384e-07,
|
| 13378 |
+
"loss": 2.4712,
|
| 13379 |
+
"step": 44180
|
| 13380 |
+
},
|
| 13381 |
+
{
|
| 13382 |
+
"epoch": 11.424085799198863,
|
| 13383 |
+
"learning_rate": 1.286918190068715e-07,
|
| 13384 |
+
"loss": 2.4963,
|
| 13385 |
+
"step": 44200
|
| 13386 |
+
},
|
| 13387 |
+
{
|
| 13388 |
+
"epoch": 11.429254425636387,
|
| 13389 |
+
"learning_rate": 1.2865940619732917e-07,
|
| 13390 |
+
"loss": 2.4866,
|
| 13391 |
+
"step": 44220
|
| 13392 |
+
},
|
| 13393 |
+
{
|
| 13394 |
+
"epoch": 11.434423052073912,
|
| 13395 |
+
"learning_rate": 1.2862699338778686e-07,
|
| 13396 |
+
"loss": 2.4665,
|
| 13397 |
+
"step": 44240
|
| 13398 |
+
},
|
| 13399 |
+
{
|
| 13400 |
+
"epoch": 11.439591678511436,
|
| 13401 |
+
"learning_rate": 1.2859458057824452e-07,
|
| 13402 |
+
"loss": 2.4525,
|
| 13403 |
+
"step": 44260
|
| 13404 |
+
},
|
| 13405 |
+
{
|
| 13406 |
+
"epoch": 11.444760304948959,
|
| 13407 |
+
"learning_rate": 1.2856216776870218e-07,
|
| 13408 |
+
"loss": 2.5271,
|
| 13409 |
+
"step": 44280
|
| 13410 |
+
},
|
| 13411 |
+
{
|
| 13412 |
+
"epoch": 11.449928931386484,
|
| 13413 |
+
"learning_rate": 1.2852975495915984e-07,
|
| 13414 |
+
"loss": 2.4676,
|
| 13415 |
+
"step": 44300
|
| 13416 |
+
},
|
| 13417 |
+
{
|
| 13418 |
+
"epoch": 11.455097557824008,
|
| 13419 |
+
"learning_rate": 1.2849734214961753e-07,
|
| 13420 |
+
"loss": 2.4996,
|
| 13421 |
+
"step": 44320
|
| 13422 |
+
},
|
| 13423 |
+
{
|
| 13424 |
+
"epoch": 11.460266184261533,
|
| 13425 |
+
"learning_rate": 1.284649293400752e-07,
|
| 13426 |
+
"loss": 2.4901,
|
| 13427 |
+
"step": 44340
|
| 13428 |
+
},
|
| 13429 |
+
{
|
| 13430 |
+
"epoch": 11.465434810699056,
|
| 13431 |
+
"learning_rate": 1.2843251653053285e-07,
|
| 13432 |
+
"loss": 2.411,
|
| 13433 |
+
"step": 44360
|
| 13434 |
+
},
|
| 13435 |
+
{
|
| 13436 |
+
"epoch": 11.470603437136582,
|
| 13437 |
+
"learning_rate": 1.2840010372099054e-07,
|
| 13438 |
+
"loss": 2.4688,
|
| 13439 |
+
"step": 44380
|
| 13440 |
+
},
|
| 13441 |
+
{
|
| 13442 |
+
"epoch": 11.475772063574105,
|
| 13443 |
+
"learning_rate": 1.283676909114482e-07,
|
| 13444 |
+
"loss": 2.4458,
|
| 13445 |
+
"step": 44400
|
| 13446 |
+
},
|
| 13447 |
+
{
|
| 13448 |
+
"epoch": 11.480940690011629,
|
| 13449 |
+
"learning_rate": 1.2833527810190587e-07,
|
| 13450 |
+
"loss": 2.4871,
|
| 13451 |
+
"step": 44420
|
| 13452 |
+
},
|
| 13453 |
+
{
|
| 13454 |
+
"epoch": 11.486109316449154,
|
| 13455 |
+
"learning_rate": 1.2830286529236353e-07,
|
| 13456 |
+
"loss": 2.4476,
|
| 13457 |
+
"step": 44440
|
| 13458 |
+
},
|
| 13459 |
+
{
|
| 13460 |
+
"epoch": 11.491277942886677,
|
| 13461 |
+
"learning_rate": 1.282704524828212e-07,
|
| 13462 |
+
"loss": 2.4647,
|
| 13463 |
+
"step": 44460
|
| 13464 |
+
},
|
| 13465 |
+
{
|
| 13466 |
+
"epoch": 11.496446569324203,
|
| 13467 |
+
"learning_rate": 1.2823803967327888e-07,
|
| 13468 |
+
"loss": 2.4887,
|
| 13469 |
+
"step": 44480
|
| 13470 |
+
},
|
| 13471 |
+
{
|
| 13472 |
+
"epoch": 11.501615195761726,
|
| 13473 |
+
"learning_rate": 1.2820562686373654e-07,
|
| 13474 |
+
"loss": 2.4835,
|
| 13475 |
+
"step": 44500
|
| 13476 |
+
},
|
| 13477 |
+
{
|
| 13478 |
+
"epoch": 11.506783822199251,
|
| 13479 |
+
"learning_rate": 1.281732140541942e-07,
|
| 13480 |
+
"loss": 2.4435,
|
| 13481 |
+
"step": 44520
|
| 13482 |
+
},
|
| 13483 |
+
{
|
| 13484 |
+
"epoch": 11.511952448636775,
|
| 13485 |
+
"learning_rate": 1.281408012446519e-07,
|
| 13486 |
+
"loss": 2.4716,
|
| 13487 |
+
"step": 44540
|
| 13488 |
+
},
|
| 13489 |
+
{
|
| 13490 |
+
"epoch": 11.517121075074298,
|
| 13491 |
+
"learning_rate": 1.2810838843510955e-07,
|
| 13492 |
+
"loss": 2.457,
|
| 13493 |
+
"step": 44560
|
| 13494 |
+
},
|
| 13495 |
+
{
|
| 13496 |
+
"epoch": 11.522289701511824,
|
| 13497 |
+
"learning_rate": 1.2807597562556721e-07,
|
| 13498 |
+
"loss": 2.4465,
|
| 13499 |
+
"step": 44580
|
| 13500 |
+
},
|
| 13501 |
+
{
|
| 13502 |
+
"epoch": 11.527458327949347,
|
| 13503 |
+
"learning_rate": 1.2804356281602488e-07,
|
| 13504 |
+
"loss": 2.4804,
|
| 13505 |
+
"step": 44600
|
| 13506 |
+
},
|
| 13507 |
+
{
|
| 13508 |
+
"epoch": 11.532626954386872,
|
| 13509 |
+
"learning_rate": 1.2801115000648256e-07,
|
| 13510 |
+
"loss": 2.5168,
|
| 13511 |
+
"step": 44620
|
| 13512 |
+
},
|
| 13513 |
+
{
|
| 13514 |
+
"epoch": 11.537795580824396,
|
| 13515 |
+
"learning_rate": 1.2797873719694023e-07,
|
| 13516 |
+
"loss": 2.4759,
|
| 13517 |
+
"step": 44640
|
| 13518 |
+
},
|
| 13519 |
+
{
|
| 13520 |
+
"epoch": 11.542964207261921,
|
| 13521 |
+
"learning_rate": 1.279463243873979e-07,
|
| 13522 |
+
"loss": 2.4929,
|
| 13523 |
+
"step": 44660
|
| 13524 |
+
},
|
| 13525 |
+
{
|
| 13526 |
+
"epoch": 11.548132833699444,
|
| 13527 |
+
"learning_rate": 1.2791391157785555e-07,
|
| 13528 |
+
"loss": 2.4136,
|
| 13529 |
+
"step": 44680
|
| 13530 |
+
},
|
| 13531 |
+
{
|
| 13532 |
+
"epoch": 11.553301460136968,
|
| 13533 |
+
"learning_rate": 1.2788149876831324e-07,
|
| 13534 |
+
"loss": 2.4497,
|
| 13535 |
+
"step": 44700
|
| 13536 |
+
},
|
| 13537 |
+
{
|
| 13538 |
+
"epoch": 11.558470086574493,
|
| 13539 |
+
"learning_rate": 1.278490859587709e-07,
|
| 13540 |
+
"loss": 2.4912,
|
| 13541 |
+
"step": 44720
|
| 13542 |
+
},
|
| 13543 |
+
{
|
| 13544 |
+
"epoch": 11.563638713012017,
|
| 13545 |
+
"learning_rate": 1.2781667314922856e-07,
|
| 13546 |
+
"loss": 2.5089,
|
| 13547 |
+
"step": 44740
|
| 13548 |
+
},
|
| 13549 |
+
{
|
| 13550 |
+
"epoch": 11.568807339449542,
|
| 13551 |
+
"learning_rate": 1.2778426033968622e-07,
|
| 13552 |
+
"loss": 2.4516,
|
| 13553 |
+
"step": 44760
|
| 13554 |
+
},
|
| 13555 |
+
{
|
| 13556 |
+
"epoch": 11.573975965887065,
|
| 13557 |
+
"learning_rate": 1.277518475301439e-07,
|
| 13558 |
+
"loss": 2.5023,
|
| 13559 |
+
"step": 44780
|
| 13560 |
+
},
|
| 13561 |
+
{
|
| 13562 |
+
"epoch": 11.579144592324589,
|
| 13563 |
+
"learning_rate": 1.2771943472060157e-07,
|
| 13564 |
+
"loss": 2.4883,
|
| 13565 |
+
"step": 44800
|
| 13566 |
+
},
|
| 13567 |
+
{
|
| 13568 |
+
"epoch": 11.584313218762114,
|
| 13569 |
+
"learning_rate": 1.2768702191105924e-07,
|
| 13570 |
+
"loss": 2.4885,
|
| 13571 |
+
"step": 44820
|
| 13572 |
+
},
|
| 13573 |
+
{
|
| 13574 |
+
"epoch": 11.589481845199638,
|
| 13575 |
+
"learning_rate": 1.2765460910151692e-07,
|
| 13576 |
+
"loss": 2.5421,
|
| 13577 |
+
"step": 44840
|
| 13578 |
+
},
|
| 13579 |
+
{
|
| 13580 |
+
"epoch": 11.594650471637163,
|
| 13581 |
+
"learning_rate": 1.2762219629197459e-07,
|
| 13582 |
+
"loss": 2.4938,
|
| 13583 |
+
"step": 44860
|
| 13584 |
+
},
|
| 13585 |
+
{
|
| 13586 |
+
"epoch": 11.599819098074686,
|
| 13587 |
+
"learning_rate": 1.2758978348243225e-07,
|
| 13588 |
+
"loss": 2.4712,
|
| 13589 |
+
"step": 44880
|
| 13590 |
+
},
|
| 13591 |
+
{
|
| 13592 |
+
"epoch": 11.604987724512211,
|
| 13593 |
+
"learning_rate": 1.275573706728899e-07,
|
| 13594 |
+
"loss": 2.5227,
|
| 13595 |
+
"step": 44900
|
| 13596 |
+
},
|
| 13597 |
+
{
|
| 13598 |
+
"epoch": 11.610156350949735,
|
| 13599 |
+
"learning_rate": 1.275249578633476e-07,
|
| 13600 |
+
"loss": 2.4463,
|
| 13601 |
+
"step": 44920
|
| 13602 |
+
},
|
| 13603 |
+
{
|
| 13604 |
+
"epoch": 11.61532497738726,
|
| 13605 |
+
"learning_rate": 1.2749254505380526e-07,
|
| 13606 |
+
"loss": 2.4526,
|
| 13607 |
+
"step": 44940
|
| 13608 |
+
},
|
| 13609 |
+
{
|
| 13610 |
+
"epoch": 11.620493603824784,
|
| 13611 |
+
"learning_rate": 1.2746013224426292e-07,
|
| 13612 |
+
"loss": 2.4763,
|
| 13613 |
+
"step": 44960
|
| 13614 |
+
},
|
| 13615 |
+
{
|
| 13616 |
+
"epoch": 11.625662230262307,
|
| 13617 |
+
"learning_rate": 1.2742771943472058e-07,
|
| 13618 |
+
"loss": 2.5575,
|
| 13619 |
+
"step": 44980
|
| 13620 |
+
},
|
| 13621 |
+
{
|
| 13622 |
+
"epoch": 11.630830856699832,
|
| 13623 |
+
"learning_rate": 1.2739530662517827e-07,
|
| 13624 |
+
"loss": 2.5079,
|
| 13625 |
+
"step": 45000
|
| 13626 |
+
},
|
| 13627 |
+
{
|
| 13628 |
+
"epoch": 11.635999483137356,
|
| 13629 |
+
"learning_rate": 1.2736289381563593e-07,
|
| 13630 |
+
"loss": 2.4868,
|
| 13631 |
+
"step": 45020
|
| 13632 |
+
},
|
| 13633 |
+
{
|
| 13634 |
+
"epoch": 11.641168109574881,
|
| 13635 |
+
"learning_rate": 1.273304810060936e-07,
|
| 13636 |
+
"loss": 2.4547,
|
| 13637 |
+
"step": 45040
|
| 13638 |
+
},
|
| 13639 |
+
{
|
| 13640 |
+
"epoch": 11.646336736012405,
|
| 13641 |
+
"learning_rate": 1.2729806819655126e-07,
|
| 13642 |
+
"loss": 2.4594,
|
| 13643 |
+
"step": 45060
|
| 13644 |
+
},
|
| 13645 |
+
{
|
| 13646 |
+
"epoch": 11.651505362449928,
|
| 13647 |
+
"learning_rate": 1.2726565538700895e-07,
|
| 13648 |
+
"loss": 2.495,
|
| 13649 |
+
"step": 45080
|
| 13650 |
+
},
|
| 13651 |
+
{
|
| 13652 |
+
"epoch": 11.656673988887453,
|
| 13653 |
+
"learning_rate": 1.272332425774666e-07,
|
| 13654 |
+
"loss": 2.5174,
|
| 13655 |
+
"step": 45100
|
| 13656 |
+
},
|
| 13657 |
+
{
|
| 13658 |
+
"epoch": 11.661842615324977,
|
| 13659 |
+
"learning_rate": 1.2720082976792427e-07,
|
| 13660 |
+
"loss": 2.4223,
|
| 13661 |
+
"step": 45120
|
| 13662 |
+
},
|
| 13663 |
+
{
|
| 13664 |
+
"epoch": 11.667011241762502,
|
| 13665 |
+
"learning_rate": 1.2716841695838196e-07,
|
| 13666 |
+
"loss": 2.4523,
|
| 13667 |
+
"step": 45140
|
| 13668 |
+
},
|
| 13669 |
+
{
|
| 13670 |
+
"epoch": 11.672179868200026,
|
| 13671 |
+
"learning_rate": 1.2713600414883962e-07,
|
| 13672 |
+
"loss": 2.4271,
|
| 13673 |
+
"step": 45160
|
| 13674 |
+
},
|
| 13675 |
+
{
|
| 13676 |
+
"epoch": 11.67734849463755,
|
| 13677 |
+
"learning_rate": 1.2710359133929728e-07,
|
| 13678 |
+
"loss": 2.5235,
|
| 13679 |
+
"step": 45180
|
| 13680 |
+
},
|
| 13681 |
+
{
|
| 13682 |
+
"epoch": 11.682517121075074,
|
| 13683 |
+
"learning_rate": 1.2707117852975494e-07,
|
| 13684 |
+
"loss": 2.4666,
|
| 13685 |
+
"step": 45200
|
| 13686 |
+
},
|
| 13687 |
+
{
|
| 13688 |
+
"epoch": 11.687685747512598,
|
| 13689 |
+
"learning_rate": 1.2703876572021263e-07,
|
| 13690 |
+
"loss": 2.4413,
|
| 13691 |
+
"step": 45220
|
| 13692 |
+
},
|
| 13693 |
+
{
|
| 13694 |
+
"epoch": 11.692854373950123,
|
| 13695 |
+
"learning_rate": 1.270063529106703e-07,
|
| 13696 |
+
"loss": 2.422,
|
| 13697 |
+
"step": 45240
|
| 13698 |
+
},
|
| 13699 |
+
{
|
| 13700 |
+
"epoch": 11.698023000387646,
|
| 13701 |
+
"learning_rate": 1.2697394010112796e-07,
|
| 13702 |
+
"loss": 2.5057,
|
| 13703 |
+
"step": 45260
|
| 13704 |
+
},
|
| 13705 |
+
{
|
| 13706 |
+
"epoch": 11.703191626825172,
|
| 13707 |
+
"learning_rate": 1.2694152729158562e-07,
|
| 13708 |
+
"loss": 2.5153,
|
| 13709 |
+
"step": 45280
|
| 13710 |
+
},
|
| 13711 |
+
{
|
| 13712 |
+
"epoch": 11.708360253262695,
|
| 13713 |
+
"learning_rate": 1.269091144820433e-07,
|
| 13714 |
+
"loss": 2.4894,
|
| 13715 |
+
"step": 45300
|
| 13716 |
+
},
|
| 13717 |
+
{
|
| 13718 |
+
"epoch": 11.71352887970022,
|
| 13719 |
+
"learning_rate": 1.2687670167250097e-07,
|
| 13720 |
+
"loss": 2.5211,
|
| 13721 |
+
"step": 45320
|
| 13722 |
+
},
|
| 13723 |
+
{
|
| 13724 |
+
"epoch": 11.718697506137744,
|
| 13725 |
+
"learning_rate": 1.2684428886295863e-07,
|
| 13726 |
+
"loss": 2.4449,
|
| 13727 |
+
"step": 45340
|
| 13728 |
+
},
|
| 13729 |
+
{
|
| 13730 |
+
"epoch": 11.723866132575267,
|
| 13731 |
+
"learning_rate": 1.268118760534163e-07,
|
| 13732 |
+
"loss": 2.4859,
|
| 13733 |
+
"step": 45360
|
| 13734 |
+
},
|
| 13735 |
+
{
|
| 13736 |
+
"epoch": 11.729034759012793,
|
| 13737 |
+
"learning_rate": 1.2677946324387398e-07,
|
| 13738 |
+
"loss": 2.4578,
|
| 13739 |
+
"step": 45380
|
| 13740 |
+
},
|
| 13741 |
+
{
|
| 13742 |
+
"epoch": 11.734203385450316,
|
| 13743 |
+
"learning_rate": 1.2674705043433164e-07,
|
| 13744 |
+
"loss": 2.4715,
|
| 13745 |
+
"step": 45400
|
| 13746 |
+
},
|
| 13747 |
+
{
|
| 13748 |
+
"epoch": 11.739372011887841,
|
| 13749 |
+
"learning_rate": 1.267146376247893e-07,
|
| 13750 |
+
"loss": 2.4258,
|
| 13751 |
+
"step": 45420
|
| 13752 |
+
},
|
| 13753 |
+
{
|
| 13754 |
+
"epoch": 11.744540638325365,
|
| 13755 |
+
"learning_rate": 1.26682224815247e-07,
|
| 13756 |
+
"loss": 2.4635,
|
| 13757 |
+
"step": 45440
|
| 13758 |
+
},
|
| 13759 |
+
{
|
| 13760 |
+
"epoch": 11.74970926476289,
|
| 13761 |
+
"learning_rate": 1.2664981200570465e-07,
|
| 13762 |
+
"loss": 2.4803,
|
| 13763 |
+
"step": 45460
|
| 13764 |
+
},
|
| 13765 |
+
{
|
| 13766 |
+
"epoch": 11.754877891200413,
|
| 13767 |
+
"learning_rate": 1.2661739919616232e-07,
|
| 13768 |
+
"loss": 2.4333,
|
| 13769 |
+
"step": 45480
|
| 13770 |
+
},
|
| 13771 |
+
{
|
| 13772 |
+
"epoch": 11.760046517637937,
|
| 13773 |
+
"learning_rate": 1.2658498638661998e-07,
|
| 13774 |
+
"loss": 2.5486,
|
| 13775 |
+
"step": 45500
|
| 13776 |
+
},
|
| 13777 |
+
{
|
| 13778 |
+
"epoch": 11.765215144075462,
|
| 13779 |
+
"learning_rate": 1.2655257357707767e-07,
|
| 13780 |
+
"loss": 2.4803,
|
| 13781 |
+
"step": 45520
|
| 13782 |
+
},
|
| 13783 |
+
{
|
| 13784 |
+
"epoch": 11.770383770512986,
|
| 13785 |
+
"learning_rate": 1.2652016076753533e-07,
|
| 13786 |
+
"loss": 2.4958,
|
| 13787 |
+
"step": 45540
|
| 13788 |
+
},
|
| 13789 |
+
{
|
| 13790 |
+
"epoch": 11.775552396950511,
|
| 13791 |
+
"learning_rate": 1.26487747957993e-07,
|
| 13792 |
+
"loss": 2.4918,
|
| 13793 |
+
"step": 45560
|
| 13794 |
+
},
|
| 13795 |
+
{
|
| 13796 |
+
"epoch": 11.780721023388034,
|
| 13797 |
+
"learning_rate": 1.2645533514845065e-07,
|
| 13798 |
+
"loss": 2.513,
|
| 13799 |
+
"step": 45580
|
| 13800 |
+
},
|
| 13801 |
+
{
|
| 13802 |
+
"epoch": 11.78588964982556,
|
| 13803 |
+
"learning_rate": 1.2642292233890834e-07,
|
| 13804 |
+
"loss": 2.4897,
|
| 13805 |
+
"step": 45600
|
| 13806 |
+
},
|
| 13807 |
+
{
|
| 13808 |
+
"epoch": 11.791058276263083,
|
| 13809 |
+
"learning_rate": 1.26390509529366e-07,
|
| 13810 |
+
"loss": 2.4468,
|
| 13811 |
+
"step": 45620
|
| 13812 |
+
},
|
| 13813 |
+
{
|
| 13814 |
+
"epoch": 11.796226902700607,
|
| 13815 |
+
"learning_rate": 1.2635809671982366e-07,
|
| 13816 |
+
"loss": 2.4928,
|
| 13817 |
+
"step": 45640
|
| 13818 |
+
},
|
| 13819 |
+
{
|
| 13820 |
+
"epoch": 11.801395529138132,
|
| 13821 |
+
"learning_rate": 1.2632568391028132e-07,
|
| 13822 |
+
"loss": 2.4793,
|
| 13823 |
+
"step": 45660
|
| 13824 |
+
},
|
| 13825 |
+
{
|
| 13826 |
+
"epoch": 11.806564155575655,
|
| 13827 |
+
"learning_rate": 1.26293271100739e-07,
|
| 13828 |
+
"loss": 2.4788,
|
| 13829 |
+
"step": 45680
|
| 13830 |
+
},
|
| 13831 |
+
{
|
| 13832 |
+
"epoch": 11.81173278201318,
|
| 13833 |
+
"learning_rate": 1.2626085829119668e-07,
|
| 13834 |
+
"loss": 2.4913,
|
| 13835 |
+
"step": 45700
|
| 13836 |
+
},
|
| 13837 |
+
{
|
| 13838 |
+
"epoch": 11.816901408450704,
|
| 13839 |
+
"learning_rate": 1.2622844548165434e-07,
|
| 13840 |
+
"loss": 2.4966,
|
| 13841 |
+
"step": 45720
|
| 13842 |
+
},
|
| 13843 |
+
{
|
| 13844 |
+
"epoch": 11.82207003488823,
|
| 13845 |
+
"learning_rate": 1.2619603267211203e-07,
|
| 13846 |
+
"loss": 2.4954,
|
| 13847 |
+
"step": 45740
|
| 13848 |
+
},
|
| 13849 |
+
{
|
| 13850 |
+
"epoch": 11.827238661325753,
|
| 13851 |
+
"learning_rate": 1.261636198625697e-07,
|
| 13852 |
+
"loss": 2.4608,
|
| 13853 |
+
"step": 45760
|
| 13854 |
+
},
|
| 13855 |
+
{
|
| 13856 |
+
"epoch": 11.832407287763276,
|
| 13857 |
+
"learning_rate": 1.2613120705302735e-07,
|
| 13858 |
+
"loss": 2.4901,
|
| 13859 |
+
"step": 45780
|
| 13860 |
+
},
|
| 13861 |
+
{
|
| 13862 |
+
"epoch": 11.837575914200801,
|
| 13863 |
+
"learning_rate": 1.26098794243485e-07,
|
| 13864 |
+
"loss": 2.5462,
|
| 13865 |
+
"step": 45800
|
| 13866 |
+
},
|
| 13867 |
+
{
|
| 13868 |
+
"epoch": 11.842744540638325,
|
| 13869 |
+
"learning_rate": 1.260663814339427e-07,
|
| 13870 |
+
"loss": 2.5217,
|
| 13871 |
+
"step": 45820
|
| 13872 |
+
},
|
| 13873 |
+
{
|
| 13874 |
+
"epoch": 11.84791316707585,
|
| 13875 |
+
"learning_rate": 1.2603396862440036e-07,
|
| 13876 |
+
"loss": 2.4889,
|
| 13877 |
+
"step": 45840
|
| 13878 |
+
},
|
| 13879 |
+
{
|
| 13880 |
+
"epoch": 11.853081793513374,
|
| 13881 |
+
"learning_rate": 1.2600155581485802e-07,
|
| 13882 |
+
"loss": 2.5021,
|
| 13883 |
+
"step": 45860
|
| 13884 |
+
},
|
| 13885 |
+
{
|
| 13886 |
+
"epoch": 11.858250419950899,
|
| 13887 |
+
"learning_rate": 1.2596914300531568e-07,
|
| 13888 |
+
"loss": 2.5027,
|
| 13889 |
+
"step": 45880
|
| 13890 |
+
},
|
| 13891 |
+
{
|
| 13892 |
+
"epoch": 11.863419046388422,
|
| 13893 |
+
"learning_rate": 1.2593673019577337e-07,
|
| 13894 |
+
"loss": 2.4681,
|
| 13895 |
+
"step": 45900
|
| 13896 |
+
},
|
| 13897 |
+
{
|
| 13898 |
+
"epoch": 11.868587672825946,
|
| 13899 |
+
"learning_rate": 1.2590431738623104e-07,
|
| 13900 |
+
"loss": 2.409,
|
| 13901 |
+
"step": 45920
|
| 13902 |
+
},
|
| 13903 |
+
{
|
| 13904 |
+
"epoch": 11.873756299263471,
|
| 13905 |
+
"learning_rate": 1.258719045766887e-07,
|
| 13906 |
+
"loss": 2.4426,
|
| 13907 |
+
"step": 45940
|
| 13908 |
+
},
|
| 13909 |
+
{
|
| 13910 |
+
"epoch": 11.878924925700995,
|
| 13911 |
+
"learning_rate": 1.2583949176714636e-07,
|
| 13912 |
+
"loss": 2.4285,
|
| 13913 |
+
"step": 45960
|
| 13914 |
+
},
|
| 13915 |
+
{
|
| 13916 |
+
"epoch": 11.88409355213852,
|
| 13917 |
+
"learning_rate": 1.2580707895760405e-07,
|
| 13918 |
+
"loss": 2.4452,
|
| 13919 |
+
"step": 45980
|
| 13920 |
+
},
|
| 13921 |
+
{
|
| 13922 |
+
"epoch": 11.889262178576043,
|
| 13923 |
+
"learning_rate": 1.257746661480617e-07,
|
| 13924 |
+
"loss": 2.4816,
|
| 13925 |
+
"step": 46000
|
| 13926 |
+
},
|
| 13927 |
+
{
|
| 13928 |
+
"epoch": 11.894430805013567,
|
| 13929 |
+
"learning_rate": 1.2574225333851937e-07,
|
| 13930 |
+
"loss": 2.4585,
|
| 13931 |
+
"step": 46020
|
| 13932 |
+
},
|
| 13933 |
+
{
|
| 13934 |
+
"epoch": 11.899599431451092,
|
| 13935 |
+
"learning_rate": 1.2570984052897706e-07,
|
| 13936 |
+
"loss": 2.4171,
|
| 13937 |
+
"step": 46040
|
| 13938 |
+
},
|
| 13939 |
+
{
|
| 13940 |
+
"epoch": 11.904768057888615,
|
| 13941 |
+
"learning_rate": 1.2567742771943472e-07,
|
| 13942 |
+
"loss": 2.4597,
|
| 13943 |
+
"step": 46060
|
| 13944 |
+
},
|
| 13945 |
+
{
|
| 13946 |
+
"epoch": 11.90993668432614,
|
| 13947 |
+
"learning_rate": 1.2564501490989238e-07,
|
| 13948 |
+
"loss": 2.4195,
|
| 13949 |
+
"step": 46080
|
| 13950 |
+
},
|
| 13951 |
+
{
|
| 13952 |
+
"epoch": 11.915105310763664,
|
| 13953 |
+
"learning_rate": 1.2561260210035004e-07,
|
| 13954 |
+
"loss": 2.5034,
|
| 13955 |
+
"step": 46100
|
| 13956 |
+
},
|
| 13957 |
+
{
|
| 13958 |
+
"epoch": 11.92027393720119,
|
| 13959 |
+
"learning_rate": 1.2558018929080773e-07,
|
| 13960 |
+
"loss": 2.4407,
|
| 13961 |
+
"step": 46120
|
| 13962 |
+
},
|
| 13963 |
+
{
|
| 13964 |
+
"epoch": 11.925442563638713,
|
| 13965 |
+
"learning_rate": 1.255477764812654e-07,
|
| 13966 |
+
"loss": 2.4813,
|
| 13967 |
+
"step": 46140
|
| 13968 |
+
},
|
| 13969 |
+
{
|
| 13970 |
+
"epoch": 11.930611190076238,
|
| 13971 |
+
"learning_rate": 1.2551536367172306e-07,
|
| 13972 |
+
"loss": 2.4599,
|
| 13973 |
+
"step": 46160
|
| 13974 |
+
},
|
| 13975 |
+
{
|
| 13976 |
+
"epoch": 11.935779816513762,
|
| 13977 |
+
"learning_rate": 1.2548295086218072e-07,
|
| 13978 |
+
"loss": 2.4641,
|
| 13979 |
+
"step": 46180
|
| 13980 |
+
},
|
| 13981 |
+
{
|
| 13982 |
+
"epoch": 11.940948442951285,
|
| 13983 |
+
"learning_rate": 1.254505380526384e-07,
|
| 13984 |
+
"loss": 2.5262,
|
| 13985 |
+
"step": 46200
|
| 13986 |
+
},
|
| 13987 |
+
{
|
| 13988 |
+
"epoch": 11.94611706938881,
|
| 13989 |
+
"learning_rate": 1.2541812524309607e-07,
|
| 13990 |
+
"loss": 2.4518,
|
| 13991 |
+
"step": 46220
|
| 13992 |
+
},
|
| 13993 |
+
{
|
| 13994 |
+
"epoch": 11.951285695826334,
|
| 13995 |
+
"learning_rate": 1.2538571243355373e-07,
|
| 13996 |
+
"loss": 2.5018,
|
| 13997 |
+
"step": 46240
|
| 13998 |
+
},
|
| 13999 |
+
{
|
| 14000 |
+
"epoch": 11.956454322263859,
|
| 14001 |
+
"learning_rate": 1.253532996240114e-07,
|
| 14002 |
+
"loss": 2.4634,
|
| 14003 |
+
"step": 46260
|
| 14004 |
+
},
|
| 14005 |
+
{
|
| 14006 |
+
"epoch": 11.961622948701383,
|
| 14007 |
+
"learning_rate": 1.2532088681446908e-07,
|
| 14008 |
+
"loss": 2.4889,
|
| 14009 |
+
"step": 46280
|
| 14010 |
+
},
|
| 14011 |
+
{
|
| 14012 |
+
"epoch": 11.966791575138906,
|
| 14013 |
+
"learning_rate": 1.2528847400492674e-07,
|
| 14014 |
+
"loss": 2.4577,
|
| 14015 |
+
"step": 46300
|
| 14016 |
+
},
|
| 14017 |
+
{
|
| 14018 |
+
"epoch": 11.971960201576431,
|
| 14019 |
+
"learning_rate": 1.252560611953844e-07,
|
| 14020 |
+
"loss": 2.4812,
|
| 14021 |
+
"step": 46320
|
| 14022 |
+
},
|
| 14023 |
+
{
|
| 14024 |
+
"epoch": 11.977128828013955,
|
| 14025 |
+
"learning_rate": 1.252236483858421e-07,
|
| 14026 |
+
"loss": 2.4448,
|
| 14027 |
+
"step": 46340
|
| 14028 |
+
},
|
| 14029 |
+
{
|
| 14030 |
+
"epoch": 11.98229745445148,
|
| 14031 |
+
"learning_rate": 1.2519123557629975e-07,
|
| 14032 |
+
"loss": 2.4594,
|
| 14033 |
+
"step": 46360
|
| 14034 |
+
},
|
| 14035 |
+
{
|
| 14036 |
+
"epoch": 11.987466080889003,
|
| 14037 |
+
"learning_rate": 1.2515882276675742e-07,
|
| 14038 |
+
"loss": 2.4473,
|
| 14039 |
+
"step": 46380
|
| 14040 |
+
},
|
| 14041 |
+
{
|
| 14042 |
+
"epoch": 11.992634707326529,
|
| 14043 |
+
"learning_rate": 1.2512640995721508e-07,
|
| 14044 |
+
"loss": 2.4831,
|
| 14045 |
+
"step": 46400
|
| 14046 |
+
},
|
| 14047 |
+
{
|
| 14048 |
+
"epoch": 11.997803333764052,
|
| 14049 |
+
"learning_rate": 1.2509399714767277e-07,
|
| 14050 |
+
"loss": 2.4706,
|
| 14051 |
+
"step": 46420
|
| 14052 |
+
},
|
| 14053 |
+
{
|
| 14054 |
+
"epoch": 11.999870784339063,
|
| 14055 |
+
"eval_bleu": 12.51,
|
| 14056 |
+
"eval_gen_len": 37.4146,
|
| 14057 |
+
"eval_loss": 2.4478776454925537,
|
| 14058 |
+
"eval_runtime": 807.7318,
|
| 14059 |
+
"eval_samples_per_second": 2.132,
|
| 14060 |
+
"eval_steps_per_second": 1.066,
|
| 14061 |
+
"step": 46428
|
| 14062 |
}
|
| 14063 |
],
|
| 14064 |
"logging_steps": 20,
|
|
|
|
| 14078 |
"attributes": {}
|
| 14079 |
}
|
| 14080 |
},
|
| 14081 |
+
"total_flos": 3.0475659174912e+17,
|
| 14082 |
"train_batch_size": 2,
|
| 14083 |
"trial_name": null,
|
| 14084 |
"trial_params": null
|