Training in progress, step 157890, checkpoint
Browse files- last-checkpoint/adapter_model.safetensors +1 -1
- last-checkpoint/optimizer.pt +1 -1
- last-checkpoint/rng_state_0.pth +1 -1
- last-checkpoint/rng_state_1.pth +1 -1
- last-checkpoint/rng_state_2.pth +1 -1
- last-checkpoint/rng_state_3.pth +1 -1
- last-checkpoint/rng_state_4.pth +1 -1
- last-checkpoint/rng_state_5.pth +1 -1
- last-checkpoint/rng_state_6.pth +1 -1
- last-checkpoint/rng_state_7.pth +1 -1
- last-checkpoint/scheduler.pt +1 -1
- last-checkpoint/trainer_state.json +2923 -3
last-checkpoint/adapter_model.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 627606952
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1088d85bad76e35cb6c18e7734c2c2df37b0f09d0d513cd23193692801b8e919
|
| 3 |
size 627606952
|
last-checkpoint/optimizer.pt
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 318986436
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:74fea6dfe87ac00874c7a8077e9de8baf19330d55fbcf21369d50ffdd7952746
|
| 3 |
size 318986436
|
last-checkpoint/rng_state_0.pth
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 15984
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:908173892e21f0448410767027e18a92997b1ed2af990e2d2c4310d10a259e8d
|
| 3 |
size 15984
|
last-checkpoint/rng_state_1.pth
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 15984
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f67794cf32b5607e3c7c900f9ec980e9ddeb5878f1ecf87202a26f9f7ec578df
|
| 3 |
size 15984
|
last-checkpoint/rng_state_2.pth
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 15984
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:5d73dca6b24dc149aa09b998461a4ad6f5693f3022488555c563abdd18b0bd07
|
| 3 |
size 15984
|
last-checkpoint/rng_state_3.pth
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 15984
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:cd682e7c95609633443dbf166e5225e8744766ccd1c3ac9a8a206552b0d19437
|
| 3 |
size 15984
|
last-checkpoint/rng_state_4.pth
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 15984
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:69fb71396e38b5278906060269440d2471ddeaa97d12505747b0e67c0969d3bd
|
| 3 |
size 15984
|
last-checkpoint/rng_state_5.pth
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 15984
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:0ac676368096395c9abcdc2eb24514144adc29fa422f2ccb1a132f9486edc674
|
| 3 |
size 15984
|
last-checkpoint/rng_state_6.pth
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 15984
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:51c1b684147f6e3b2a7f0206497e7491505db91cfdd5d1908cfea9ae634ec7ed
|
| 3 |
size 15984
|
last-checkpoint/rng_state_7.pth
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 15984
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:993e06522ed60c0b873ce20890d06629f64e5c63c610b8979cd6428c8ce358d5
|
| 3 |
size 15984
|
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:cd46dcde3e0e95c0d471f441b283d1d9f5c9aaae4854865d35095a1f5c249f2e
|
| 3 |
size 1064
|
last-checkpoint/trainer_state.json
CHANGED
|
@@ -1,9 +1,9 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": null,
|
| 3 |
"best_model_checkpoint": null,
|
| 4 |
-
"epoch":
|
| 5 |
"eval_steps": 8310,
|
| 6 |
-
"global_step":
|
| 7 |
"is_hyper_param_search": false,
|
| 8 |
"is_local_process_zero": true,
|
| 9 |
"is_world_process_zero": true,
|
|
@@ -107770,6 +107770,2926 @@
|
|
| 107770 |
"learning_rate": 0.0001844548534288357,
|
| 107771 |
"loss": 11.2043,
|
| 107772 |
"step": 153730
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 107773 |
}
|
| 107774 |
],
|
| 107775 |
"logging_steps": 10,
|
|
@@ -107789,7 +110709,7 @@
|
|
| 107789 |
"attributes": {}
|
| 107790 |
}
|
| 107791 |
},
|
| 107792 |
-
"total_flos": 3.
|
| 107793 |
"train_batch_size": 1,
|
| 107794 |
"trial_name": null,
|
| 107795 |
"trial_params": null
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": null,
|
| 3 |
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 19.0,
|
| 5 |
"eval_steps": 8310,
|
| 6 |
+
"global_step": 157890,
|
| 7 |
"is_hyper_param_search": false,
|
| 8 |
"is_local_process_zero": true,
|
| 9 |
"is_world_process_zero": true,
|
|
|
|
| 107770 |
"learning_rate": 0.0001844548534288357,
|
| 107771 |
"loss": 11.2043,
|
| 107772 |
"step": 153730
|
| 107773 |
+
},
|
| 107774 |
+
{
|
| 107775 |
+
"epoch": 18.500601684717207,
|
| 107776 |
+
"grad_norm": Infinity,
|
| 107777 |
+
"learning_rate": 0.00018445281626107063,
|
| 107778 |
+
"loss": 11.0292,
|
| 107779 |
+
"step": 153740
|
| 107780 |
+
},
|
| 107781 |
+
{
|
| 107782 |
+
"epoch": 18.501805054151625,
|
| 107783 |
+
"grad_norm": Infinity,
|
| 107784 |
+
"learning_rate": 0.0001844507789710815,
|
| 107785 |
+
"loss": 11.0998,
|
| 107786 |
+
"step": 153750
|
| 107787 |
+
},
|
| 107788 |
+
{
|
| 107789 |
+
"epoch": 18.503008423586042,
|
| 107790 |
+
"grad_norm": Infinity,
|
| 107791 |
+
"learning_rate": 0.00018444874155887123,
|
| 107792 |
+
"loss": 11.1909,
|
| 107793 |
+
"step": 153760
|
| 107794 |
+
},
|
| 107795 |
+
{
|
| 107796 |
+
"epoch": 18.504211793020456,
|
| 107797 |
+
"grad_norm": Infinity,
|
| 107798 |
+
"learning_rate": 0.00018444670402444278,
|
| 107799 |
+
"loss": 11.1862,
|
| 107800 |
+
"step": 153770
|
| 107801 |
+
},
|
| 107802 |
+
{
|
| 107803 |
+
"epoch": 18.505415162454874,
|
| 107804 |
+
"grad_norm": Infinity,
|
| 107805 |
+
"learning_rate": 0.0001844446663677991,
|
| 107806 |
+
"loss": 11.2138,
|
| 107807 |
+
"step": 153780
|
| 107808 |
+
},
|
| 107809 |
+
{
|
| 107810 |
+
"epoch": 18.50661853188929,
|
| 107811 |
+
"grad_norm": Infinity,
|
| 107812 |
+
"learning_rate": 0.00018444262858894312,
|
| 107813 |
+
"loss": 11.1951,
|
| 107814 |
+
"step": 153790
|
| 107815 |
+
},
|
| 107816 |
+
{
|
| 107817 |
+
"epoch": 18.507821901323705,
|
| 107818 |
+
"grad_norm": Infinity,
|
| 107819 |
+
"learning_rate": 0.00018444059068787786,
|
| 107820 |
+
"loss": 11.0419,
|
| 107821 |
+
"step": 153800
|
| 107822 |
+
},
|
| 107823 |
+
{
|
| 107824 |
+
"epoch": 18.509025270758123,
|
| 107825 |
+
"grad_norm": Infinity,
|
| 107826 |
+
"learning_rate": 0.0001844385526646062,
|
| 107827 |
+
"loss": 11.0918,
|
| 107828 |
+
"step": 153810
|
| 107829 |
+
},
|
| 107830 |
+
{
|
| 107831 |
+
"epoch": 18.51022864019254,
|
| 107832 |
+
"grad_norm": Infinity,
|
| 107833 |
+
"learning_rate": 0.00018443651451913108,
|
| 107834 |
+
"loss": 11.1833,
|
| 107835 |
+
"step": 153820
|
| 107836 |
+
},
|
| 107837 |
+
{
|
| 107838 |
+
"epoch": 18.511432009626954,
|
| 107839 |
+
"grad_norm": Infinity,
|
| 107840 |
+
"learning_rate": 0.00018443447625145548,
|
| 107841 |
+
"loss": 11.2276,
|
| 107842 |
+
"step": 153830
|
| 107843 |
+
},
|
| 107844 |
+
{
|
| 107845 |
+
"epoch": 18.512635379061372,
|
| 107846 |
+
"grad_norm": Infinity,
|
| 107847 |
+
"learning_rate": 0.00018443243786158236,
|
| 107848 |
+
"loss": 11.2327,
|
| 107849 |
+
"step": 153840
|
| 107850 |
+
},
|
| 107851 |
+
{
|
| 107852 |
+
"epoch": 18.51383874849579,
|
| 107853 |
+
"grad_norm": Infinity,
|
| 107854 |
+
"learning_rate": 0.00018443039934951466,
|
| 107855 |
+
"loss": 11.0014,
|
| 107856 |
+
"step": 153850
|
| 107857 |
+
},
|
| 107858 |
+
{
|
| 107859 |
+
"epoch": 18.515042117930204,
|
| 107860 |
+
"grad_norm": Infinity,
|
| 107861 |
+
"learning_rate": 0.00018442836071525532,
|
| 107862 |
+
"loss": 11.2101,
|
| 107863 |
+
"step": 153860
|
| 107864 |
+
},
|
| 107865 |
+
{
|
| 107866 |
+
"epoch": 18.51624548736462,
|
| 107867 |
+
"grad_norm": Infinity,
|
| 107868 |
+
"learning_rate": 0.00018442632195880728,
|
| 107869 |
+
"loss": 11.1144,
|
| 107870 |
+
"step": 153870
|
| 107871 |
+
},
|
| 107872 |
+
{
|
| 107873 |
+
"epoch": 18.51744885679904,
|
| 107874 |
+
"grad_norm": Infinity,
|
| 107875 |
+
"learning_rate": 0.00018442428308017354,
|
| 107876 |
+
"loss": 11.1024,
|
| 107877 |
+
"step": 153880
|
| 107878 |
+
},
|
| 107879 |
+
{
|
| 107880 |
+
"epoch": 18.518652226233453,
|
| 107881 |
+
"grad_norm": Infinity,
|
| 107882 |
+
"learning_rate": 0.000184422244079357,
|
| 107883 |
+
"loss": 11.1752,
|
| 107884 |
+
"step": 153890
|
| 107885 |
+
},
|
| 107886 |
+
{
|
| 107887 |
+
"epoch": 18.51985559566787,
|
| 107888 |
+
"grad_norm": Infinity,
|
| 107889 |
+
"learning_rate": 0.00018442020495636067,
|
| 107890 |
+
"loss": 11.108,
|
| 107891 |
+
"step": 153900
|
| 107892 |
+
},
|
| 107893 |
+
{
|
| 107894 |
+
"epoch": 18.521058965102288,
|
| 107895 |
+
"grad_norm": Infinity,
|
| 107896 |
+
"learning_rate": 0.00018441816571118742,
|
| 107897 |
+
"loss": 11.1124,
|
| 107898 |
+
"step": 153910
|
| 107899 |
+
},
|
| 107900 |
+
{
|
| 107901 |
+
"epoch": 18.522262334536702,
|
| 107902 |
+
"grad_norm": Infinity,
|
| 107903 |
+
"learning_rate": 0.00018441612634384025,
|
| 107904 |
+
"loss": 11.1661,
|
| 107905 |
+
"step": 153920
|
| 107906 |
+
},
|
| 107907 |
+
{
|
| 107908 |
+
"epoch": 18.52346570397112,
|
| 107909 |
+
"grad_norm": Infinity,
|
| 107910 |
+
"learning_rate": 0.0001844140868543221,
|
| 107911 |
+
"loss": 11.1174,
|
| 107912 |
+
"step": 153930
|
| 107913 |
+
},
|
| 107914 |
+
{
|
| 107915 |
+
"epoch": 18.524669073405537,
|
| 107916 |
+
"grad_norm": Infinity,
|
| 107917 |
+
"learning_rate": 0.00018441204724263592,
|
| 107918 |
+
"loss": 11.0932,
|
| 107919 |
+
"step": 153940
|
| 107920 |
+
},
|
| 107921 |
+
{
|
| 107922 |
+
"epoch": 18.52587244283995,
|
| 107923 |
+
"grad_norm": Infinity,
|
| 107924 |
+
"learning_rate": 0.00018441000750878467,
|
| 107925 |
+
"loss": 11.1473,
|
| 107926 |
+
"step": 153950
|
| 107927 |
+
},
|
| 107928 |
+
{
|
| 107929 |
+
"epoch": 18.52707581227437,
|
| 107930 |
+
"grad_norm": Infinity,
|
| 107931 |
+
"learning_rate": 0.0001844079676527713,
|
| 107932 |
+
"loss": 11.1022,
|
| 107933 |
+
"step": 153960
|
| 107934 |
+
},
|
| 107935 |
+
{
|
| 107936 |
+
"epoch": 18.528279181708786,
|
| 107937 |
+
"grad_norm": Infinity,
|
| 107938 |
+
"learning_rate": 0.00018440592767459877,
|
| 107939 |
+
"loss": 11.0514,
|
| 107940 |
+
"step": 153970
|
| 107941 |
+
},
|
| 107942 |
+
{
|
| 107943 |
+
"epoch": 18.5294825511432,
|
| 107944 |
+
"grad_norm": Infinity,
|
| 107945 |
+
"learning_rate": 0.00018440388757427,
|
| 107946 |
+
"loss": 11.1552,
|
| 107947 |
+
"step": 153980
|
| 107948 |
+
},
|
| 107949 |
+
{
|
| 107950 |
+
"epoch": 18.530685920577618,
|
| 107951 |
+
"grad_norm": Infinity,
|
| 107952 |
+
"learning_rate": 0.000184401847351788,
|
| 107953 |
+
"loss": 11.0548,
|
| 107954 |
+
"step": 153990
|
| 107955 |
+
},
|
| 107956 |
+
{
|
| 107957 |
+
"epoch": 18.531889290012035,
|
| 107958 |
+
"grad_norm": Infinity,
|
| 107959 |
+
"learning_rate": 0.00018439980700715568,
|
| 107960 |
+
"loss": 11.1378,
|
| 107961 |
+
"step": 154000
|
| 107962 |
+
},
|
| 107963 |
+
{
|
| 107964 |
+
"epoch": 18.53309265944645,
|
| 107965 |
+
"grad_norm": Infinity,
|
| 107966 |
+
"learning_rate": 0.000184397766540376,
|
| 107967 |
+
"loss": 11.1906,
|
| 107968 |
+
"step": 154010
|
| 107969 |
+
},
|
| 107970 |
+
{
|
| 107971 |
+
"epoch": 18.534296028880867,
|
| 107972 |
+
"grad_norm": Infinity,
|
| 107973 |
+
"learning_rate": 0.0001843957259514519,
|
| 107974 |
+
"loss": 11.0453,
|
| 107975 |
+
"step": 154020
|
| 107976 |
+
},
|
| 107977 |
+
{
|
| 107978 |
+
"epoch": 18.535499398315284,
|
| 107979 |
+
"grad_norm": Infinity,
|
| 107980 |
+
"learning_rate": 0.00018439368524038633,
|
| 107981 |
+
"loss": 11.2196,
|
| 107982 |
+
"step": 154030
|
| 107983 |
+
},
|
| 107984 |
+
{
|
| 107985 |
+
"epoch": 18.5367027677497,
|
| 107986 |
+
"grad_norm": Infinity,
|
| 107987 |
+
"learning_rate": 0.0001843916444071823,
|
| 107988 |
+
"loss": 11.2259,
|
| 107989 |
+
"step": 154040
|
| 107990 |
+
},
|
| 107991 |
+
{
|
| 107992 |
+
"epoch": 18.537906137184116,
|
| 107993 |
+
"grad_norm": Infinity,
|
| 107994 |
+
"learning_rate": 0.00018438960345184272,
|
| 107995 |
+
"loss": 11.239,
|
| 107996 |
+
"step": 154050
|
| 107997 |
+
},
|
| 107998 |
+
{
|
| 107999 |
+
"epoch": 18.53910950661853,
|
| 108000 |
+
"grad_norm": Infinity,
|
| 108001 |
+
"learning_rate": 0.00018438756237437054,
|
| 108002 |
+
"loss": 11.0352,
|
| 108003 |
+
"step": 154060
|
| 108004 |
+
},
|
| 108005 |
+
{
|
| 108006 |
+
"epoch": 18.540312876052948,
|
| 108007 |
+
"grad_norm": Infinity,
|
| 108008 |
+
"learning_rate": 0.00018438552117476872,
|
| 108009 |
+
"loss": 11.2338,
|
| 108010 |
+
"step": 154070
|
| 108011 |
+
},
|
| 108012 |
+
{
|
| 108013 |
+
"epoch": 18.541516245487365,
|
| 108014 |
+
"grad_norm": Infinity,
|
| 108015 |
+
"learning_rate": 0.00018438347985304024,
|
| 108016 |
+
"loss": 11.1626,
|
| 108017 |
+
"step": 154080
|
| 108018 |
+
},
|
| 108019 |
+
{
|
| 108020 |
+
"epoch": 18.54271961492178,
|
| 108021 |
+
"grad_norm": Infinity,
|
| 108022 |
+
"learning_rate": 0.00018438143840918798,
|
| 108023 |
+
"loss": 11.1056,
|
| 108024 |
+
"step": 154090
|
| 108025 |
+
},
|
| 108026 |
+
{
|
| 108027 |
+
"epoch": 18.543922984356197,
|
| 108028 |
+
"grad_norm": Infinity,
|
| 108029 |
+
"learning_rate": 0.00018437939684321498,
|
| 108030 |
+
"loss": 11.297,
|
| 108031 |
+
"step": 154100
|
| 108032 |
+
},
|
| 108033 |
+
{
|
| 108034 |
+
"epoch": 18.545126353790614,
|
| 108035 |
+
"grad_norm": Infinity,
|
| 108036 |
+
"learning_rate": 0.00018437735515512417,
|
| 108037 |
+
"loss": 11.2027,
|
| 108038 |
+
"step": 154110
|
| 108039 |
+
},
|
| 108040 |
+
{
|
| 108041 |
+
"epoch": 18.54632972322503,
|
| 108042 |
+
"grad_norm": Infinity,
|
| 108043 |
+
"learning_rate": 0.00018437531334491848,
|
| 108044 |
+
"loss": 11.1862,
|
| 108045 |
+
"step": 154120
|
| 108046 |
+
},
|
| 108047 |
+
{
|
| 108048 |
+
"epoch": 18.547533092659446,
|
| 108049 |
+
"grad_norm": Infinity,
|
| 108050 |
+
"learning_rate": 0.00018437327141260088,
|
| 108051 |
+
"loss": 11.2691,
|
| 108052 |
+
"step": 154130
|
| 108053 |
+
},
|
| 108054 |
+
{
|
| 108055 |
+
"epoch": 18.548736462093864,
|
| 108056 |
+
"grad_norm": Infinity,
|
| 108057 |
+
"learning_rate": 0.00018437122935817434,
|
| 108058 |
+
"loss": 11.0923,
|
| 108059 |
+
"step": 154140
|
| 108060 |
+
},
|
| 108061 |
+
{
|
| 108062 |
+
"epoch": 18.549939831528278,
|
| 108063 |
+
"grad_norm": Infinity,
|
| 108064 |
+
"learning_rate": 0.0001843691871816418,
|
| 108065 |
+
"loss": 11.1205,
|
| 108066 |
+
"step": 154150
|
| 108067 |
+
},
|
| 108068 |
+
{
|
| 108069 |
+
"epoch": 18.551143200962695,
|
| 108070 |
+
"grad_norm": Infinity,
|
| 108071 |
+
"learning_rate": 0.0001843671448830062,
|
| 108072 |
+
"loss": 11.164,
|
| 108073 |
+
"step": 154160
|
| 108074 |
+
},
|
| 108075 |
+
{
|
| 108076 |
+
"epoch": 18.552346570397113,
|
| 108077 |
+
"grad_norm": Infinity,
|
| 108078 |
+
"learning_rate": 0.0001843651024622705,
|
| 108079 |
+
"loss": 11.1202,
|
| 108080 |
+
"step": 154170
|
| 108081 |
+
},
|
| 108082 |
+
{
|
| 108083 |
+
"epoch": 18.553549939831527,
|
| 108084 |
+
"grad_norm": Infinity,
|
| 108085 |
+
"learning_rate": 0.0001843630599194377,
|
| 108086 |
+
"loss": 11.1886,
|
| 108087 |
+
"step": 154180
|
| 108088 |
+
},
|
| 108089 |
+
{
|
| 108090 |
+
"epoch": 18.554753309265944,
|
| 108091 |
+
"grad_norm": Infinity,
|
| 108092 |
+
"learning_rate": 0.0001843610172545107,
|
| 108093 |
+
"loss": 11.1563,
|
| 108094 |
+
"step": 154190
|
| 108095 |
+
},
|
| 108096 |
+
{
|
| 108097 |
+
"epoch": 18.555956678700362,
|
| 108098 |
+
"grad_norm": Infinity,
|
| 108099 |
+
"learning_rate": 0.00018435897446749248,
|
| 108100 |
+
"loss": 11.1706,
|
| 108101 |
+
"step": 154200
|
| 108102 |
+
},
|
| 108103 |
+
{
|
| 108104 |
+
"epoch": 18.557160048134776,
|
| 108105 |
+
"grad_norm": Infinity,
|
| 108106 |
+
"learning_rate": 0.000184356931558386,
|
| 108107 |
+
"loss": 11.0745,
|
| 108108 |
+
"step": 154210
|
| 108109 |
+
},
|
| 108110 |
+
{
|
| 108111 |
+
"epoch": 18.558363417569193,
|
| 108112 |
+
"grad_norm": Infinity,
|
| 108113 |
+
"learning_rate": 0.00018435488852719422,
|
| 108114 |
+
"loss": 11.1546,
|
| 108115 |
+
"step": 154220
|
| 108116 |
+
},
|
| 108117 |
+
{
|
| 108118 |
+
"epoch": 18.55956678700361,
|
| 108119 |
+
"grad_norm": Infinity,
|
| 108120 |
+
"learning_rate": 0.0001843528453739201,
|
| 108121 |
+
"loss": 11.2963,
|
| 108122 |
+
"step": 154230
|
| 108123 |
+
},
|
| 108124 |
+
{
|
| 108125 |
+
"epoch": 18.560770156438025,
|
| 108126 |
+
"grad_norm": Infinity,
|
| 108127 |
+
"learning_rate": 0.00018435080209856656,
|
| 108128 |
+
"loss": 11.1517,
|
| 108129 |
+
"step": 154240
|
| 108130 |
+
},
|
| 108131 |
+
{
|
| 108132 |
+
"epoch": 18.561973525872443,
|
| 108133 |
+
"grad_norm": Infinity,
|
| 108134 |
+
"learning_rate": 0.0001843487587011366,
|
| 108135 |
+
"loss": 11.1994,
|
| 108136 |
+
"step": 154250
|
| 108137 |
+
},
|
| 108138 |
+
{
|
| 108139 |
+
"epoch": 18.56317689530686,
|
| 108140 |
+
"grad_norm": Infinity,
|
| 108141 |
+
"learning_rate": 0.00018434671518163315,
|
| 108142 |
+
"loss": 11.0373,
|
| 108143 |
+
"step": 154260
|
| 108144 |
+
},
|
| 108145 |
+
{
|
| 108146 |
+
"epoch": 18.564380264741274,
|
| 108147 |
+
"grad_norm": Infinity,
|
| 108148 |
+
"learning_rate": 0.00018434467154005918,
|
| 108149 |
+
"loss": 11.1424,
|
| 108150 |
+
"step": 154270
|
| 108151 |
+
},
|
| 108152 |
+
{
|
| 108153 |
+
"epoch": 18.56558363417569,
|
| 108154 |
+
"grad_norm": Infinity,
|
| 108155 |
+
"learning_rate": 0.00018434262777641767,
|
| 108156 |
+
"loss": 11.2854,
|
| 108157 |
+
"step": 154280
|
| 108158 |
+
},
|
| 108159 |
+
{
|
| 108160 |
+
"epoch": 18.56678700361011,
|
| 108161 |
+
"grad_norm": Infinity,
|
| 108162 |
+
"learning_rate": 0.00018434058389071154,
|
| 108163 |
+
"loss": 11.1538,
|
| 108164 |
+
"step": 154290
|
| 108165 |
+
},
|
| 108166 |
+
{
|
| 108167 |
+
"epoch": 18.567990373044523,
|
| 108168 |
+
"grad_norm": Infinity,
|
| 108169 |
+
"learning_rate": 0.00018433853988294375,
|
| 108170 |
+
"loss": 11.1253,
|
| 108171 |
+
"step": 154300
|
| 108172 |
+
},
|
| 108173 |
+
{
|
| 108174 |
+
"epoch": 18.56919374247894,
|
| 108175 |
+
"grad_norm": Infinity,
|
| 108176 |
+
"learning_rate": 0.0001843364957531173,
|
| 108177 |
+
"loss": 11.1735,
|
| 108178 |
+
"step": 154310
|
| 108179 |
+
},
|
| 108180 |
+
{
|
| 108181 |
+
"epoch": 18.57039711191336,
|
| 108182 |
+
"grad_norm": Infinity,
|
| 108183 |
+
"learning_rate": 0.00018433445150123508,
|
| 108184 |
+
"loss": 11.0829,
|
| 108185 |
+
"step": 154320
|
| 108186 |
+
},
|
| 108187 |
+
{
|
| 108188 |
+
"epoch": 18.571600481347772,
|
| 108189 |
+
"grad_norm": Infinity,
|
| 108190 |
+
"learning_rate": 0.0001843324071273001,
|
| 108191 |
+
"loss": 11.2586,
|
| 108192 |
+
"step": 154330
|
| 108193 |
+
},
|
| 108194 |
+
{
|
| 108195 |
+
"epoch": 18.57280385078219,
|
| 108196 |
+
"grad_norm": Infinity,
|
| 108197 |
+
"learning_rate": 0.00018433036263131532,
|
| 108198 |
+
"loss": 11.1248,
|
| 108199 |
+
"step": 154340
|
| 108200 |
+
},
|
| 108201 |
+
{
|
| 108202 |
+
"epoch": 18.574007220216608,
|
| 108203 |
+
"grad_norm": Infinity,
|
| 108204 |
+
"learning_rate": 0.00018432831801328366,
|
| 108205 |
+
"loss": 11.111,
|
| 108206 |
+
"step": 154350
|
| 108207 |
+
},
|
| 108208 |
+
{
|
| 108209 |
+
"epoch": 18.57521058965102,
|
| 108210 |
+
"grad_norm": Infinity,
|
| 108211 |
+
"learning_rate": 0.00018432627327320816,
|
| 108212 |
+
"loss": 11.0498,
|
| 108213 |
+
"step": 154360
|
| 108214 |
+
},
|
| 108215 |
+
{
|
| 108216 |
+
"epoch": 18.57641395908544,
|
| 108217 |
+
"grad_norm": Infinity,
|
| 108218 |
+
"learning_rate": 0.00018432422841109167,
|
| 108219 |
+
"loss": 11.2345,
|
| 108220 |
+
"step": 154370
|
| 108221 |
+
},
|
| 108222 |
+
{
|
| 108223 |
+
"epoch": 18.577617328519857,
|
| 108224 |
+
"grad_norm": Infinity,
|
| 108225 |
+
"learning_rate": 0.00018432218342693721,
|
| 108226 |
+
"loss": 11.1254,
|
| 108227 |
+
"step": 154380
|
| 108228 |
+
},
|
| 108229 |
+
{
|
| 108230 |
+
"epoch": 18.57882069795427,
|
| 108231 |
+
"grad_norm": Infinity,
|
| 108232 |
+
"learning_rate": 0.00018432013832074776,
|
| 108233 |
+
"loss": 11.2464,
|
| 108234 |
+
"step": 154390
|
| 108235 |
+
},
|
| 108236 |
+
{
|
| 108237 |
+
"epoch": 18.58002406738869,
|
| 108238 |
+
"grad_norm": Infinity,
|
| 108239 |
+
"learning_rate": 0.00018431809309252622,
|
| 108240 |
+
"loss": 11.1221,
|
| 108241 |
+
"step": 154400
|
| 108242 |
+
},
|
| 108243 |
+
{
|
| 108244 |
+
"epoch": 18.581227436823106,
|
| 108245 |
+
"grad_norm": Infinity,
|
| 108246 |
+
"learning_rate": 0.0001843160477422756,
|
| 108247 |
+
"loss": 11.194,
|
| 108248 |
+
"step": 154410
|
| 108249 |
+
},
|
| 108250 |
+
{
|
| 108251 |
+
"epoch": 18.58243080625752,
|
| 108252 |
+
"grad_norm": Infinity,
|
| 108253 |
+
"learning_rate": 0.00018431400226999882,
|
| 108254 |
+
"loss": 11.2833,
|
| 108255 |
+
"step": 154420
|
| 108256 |
+
},
|
| 108257 |
+
{
|
| 108258 |
+
"epoch": 18.583634175691937,
|
| 108259 |
+
"grad_norm": Infinity,
|
| 108260 |
+
"learning_rate": 0.00018431195667569888,
|
| 108261 |
+
"loss": 11.1216,
|
| 108262 |
+
"step": 154430
|
| 108263 |
+
},
|
| 108264 |
+
{
|
| 108265 |
+
"epoch": 18.584837545126355,
|
| 108266 |
+
"grad_norm": Infinity,
|
| 108267 |
+
"learning_rate": 0.00018430991095937872,
|
| 108268 |
+
"loss": 11.0644,
|
| 108269 |
+
"step": 154440
|
| 108270 |
+
},
|
| 108271 |
+
{
|
| 108272 |
+
"epoch": 18.58604091456077,
|
| 108273 |
+
"grad_norm": Infinity,
|
| 108274 |
+
"learning_rate": 0.00018430786512104127,
|
| 108275 |
+
"loss": 11.1347,
|
| 108276 |
+
"step": 154450
|
| 108277 |
+
},
|
| 108278 |
+
{
|
| 108279 |
+
"epoch": 18.587244283995187,
|
| 108280 |
+
"grad_norm": Infinity,
|
| 108281 |
+
"learning_rate": 0.0001843058191606896,
|
| 108282 |
+
"loss": 11.1396,
|
| 108283 |
+
"step": 154460
|
| 108284 |
+
},
|
| 108285 |
+
{
|
| 108286 |
+
"epoch": 18.588447653429604,
|
| 108287 |
+
"grad_norm": Infinity,
|
| 108288 |
+
"learning_rate": 0.00018430377307832653,
|
| 108289 |
+
"loss": 11.1106,
|
| 108290 |
+
"step": 154470
|
| 108291 |
+
},
|
| 108292 |
+
{
|
| 108293 |
+
"epoch": 18.589651022864018,
|
| 108294 |
+
"grad_norm": Infinity,
|
| 108295 |
+
"learning_rate": 0.0001843017268739551,
|
| 108296 |
+
"loss": 11.1095,
|
| 108297 |
+
"step": 154480
|
| 108298 |
+
},
|
| 108299 |
+
{
|
| 108300 |
+
"epoch": 18.590854392298436,
|
| 108301 |
+
"grad_norm": Infinity,
|
| 108302 |
+
"learning_rate": 0.00018429968054757827,
|
| 108303 |
+
"loss": 11.1176,
|
| 108304 |
+
"step": 154490
|
| 108305 |
+
},
|
| 108306 |
+
{
|
| 108307 |
+
"epoch": 18.592057761732853,
|
| 108308 |
+
"grad_norm": Infinity,
|
| 108309 |
+
"learning_rate": 0.00018429763409919897,
|
| 108310 |
+
"loss": 11.0243,
|
| 108311 |
+
"step": 154500
|
| 108312 |
+
},
|
| 108313 |
+
{
|
| 108314 |
+
"epoch": 18.593261131167267,
|
| 108315 |
+
"grad_norm": Infinity,
|
| 108316 |
+
"learning_rate": 0.0001842955875288202,
|
| 108317 |
+
"loss": 11.0968,
|
| 108318 |
+
"step": 154510
|
| 108319 |
+
},
|
| 108320 |
+
{
|
| 108321 |
+
"epoch": 18.594464500601685,
|
| 108322 |
+
"grad_norm": Infinity,
|
| 108323 |
+
"learning_rate": 0.0001842935408364449,
|
| 108324 |
+
"loss": 11.139,
|
| 108325 |
+
"step": 154520
|
| 108326 |
+
},
|
| 108327 |
+
{
|
| 108328 |
+
"epoch": 18.595667870036102,
|
| 108329 |
+
"grad_norm": Infinity,
|
| 108330 |
+
"learning_rate": 0.00018429149402207604,
|
| 108331 |
+
"loss": 11.2446,
|
| 108332 |
+
"step": 154530
|
| 108333 |
+
},
|
| 108334 |
+
{
|
| 108335 |
+
"epoch": 18.596871239470516,
|
| 108336 |
+
"grad_norm": Infinity,
|
| 108337 |
+
"learning_rate": 0.00018428944708571656,
|
| 108338 |
+
"loss": 11.1045,
|
| 108339 |
+
"step": 154540
|
| 108340 |
+
},
|
| 108341 |
+
{
|
| 108342 |
+
"epoch": 18.598074608904934,
|
| 108343 |
+
"grad_norm": Infinity,
|
| 108344 |
+
"learning_rate": 0.00018428740002736946,
|
| 108345 |
+
"loss": 11.2959,
|
| 108346 |
+
"step": 154550
|
| 108347 |
+
},
|
| 108348 |
+
{
|
| 108349 |
+
"epoch": 18.59927797833935,
|
| 108350 |
+
"grad_norm": Infinity,
|
| 108351 |
+
"learning_rate": 0.00018428535284703768,
|
| 108352 |
+
"loss": 11.1471,
|
| 108353 |
+
"step": 154560
|
| 108354 |
+
},
|
| 108355 |
+
{
|
| 108356 |
+
"epoch": 18.600481347773766,
|
| 108357 |
+
"grad_norm": Infinity,
|
| 108358 |
+
"learning_rate": 0.00018428330554472416,
|
| 108359 |
+
"loss": 11.1321,
|
| 108360 |
+
"step": 154570
|
| 108361 |
+
},
|
| 108362 |
+
{
|
| 108363 |
+
"epoch": 18.601684717208183,
|
| 108364 |
+
"grad_norm": Infinity,
|
| 108365 |
+
"learning_rate": 0.00018428125812043194,
|
| 108366 |
+
"loss": 11.1653,
|
| 108367 |
+
"step": 154580
|
| 108368 |
+
},
|
| 108369 |
+
{
|
| 108370 |
+
"epoch": 18.6028880866426,
|
| 108371 |
+
"grad_norm": Infinity,
|
| 108372 |
+
"learning_rate": 0.00018427921057416388,
|
| 108373 |
+
"loss": 11.0967,
|
| 108374 |
+
"step": 154590
|
| 108375 |
+
},
|
| 108376 |
+
{
|
| 108377 |
+
"epoch": 18.604091456077015,
|
| 108378 |
+
"grad_norm": Infinity,
|
| 108379 |
+
"learning_rate": 0.00018427716290592302,
|
| 108380 |
+
"loss": 11.0111,
|
| 108381 |
+
"step": 154600
|
| 108382 |
+
},
|
| 108383 |
+
{
|
| 108384 |
+
"epoch": 18.605294825511432,
|
| 108385 |
+
"grad_norm": Infinity,
|
| 108386 |
+
"learning_rate": 0.00018427511511571228,
|
| 108387 |
+
"loss": 11.1689,
|
| 108388 |
+
"step": 154610
|
| 108389 |
+
},
|
| 108390 |
+
{
|
| 108391 |
+
"epoch": 18.60649819494585,
|
| 108392 |
+
"grad_norm": Infinity,
|
| 108393 |
+
"learning_rate": 0.00018427306720353466,
|
| 108394 |
+
"loss": 11.1187,
|
| 108395 |
+
"step": 154620
|
| 108396 |
+
},
|
| 108397 |
+
{
|
| 108398 |
+
"epoch": 18.607701564380264,
|
| 108399 |
+
"grad_norm": Infinity,
|
| 108400 |
+
"learning_rate": 0.00018427101916939308,
|
| 108401 |
+
"loss": 11.2486,
|
| 108402 |
+
"step": 154630
|
| 108403 |
+
},
|
| 108404 |
+
{
|
| 108405 |
+
"epoch": 18.60890493381468,
|
| 108406 |
+
"grad_norm": Infinity,
|
| 108407 |
+
"learning_rate": 0.00018426897101329054,
|
| 108408 |
+
"loss": 11.0974,
|
| 108409 |
+
"step": 154640
|
| 108410 |
+
},
|
| 108411 |
+
{
|
| 108412 |
+
"epoch": 18.6101083032491,
|
| 108413 |
+
"grad_norm": Infinity,
|
| 108414 |
+
"learning_rate": 0.00018426692273523,
|
| 108415 |
+
"loss": 11.228,
|
| 108416 |
+
"step": 154650
|
| 108417 |
+
},
|
| 108418 |
+
{
|
| 108419 |
+
"epoch": 18.611311672683513,
|
| 108420 |
+
"grad_norm": Infinity,
|
| 108421 |
+
"learning_rate": 0.00018426487433521445,
|
| 108422 |
+
"loss": 11.1379,
|
| 108423 |
+
"step": 154660
|
| 108424 |
+
},
|
| 108425 |
+
{
|
| 108426 |
+
"epoch": 18.61251504211793,
|
| 108427 |
+
"grad_norm": Infinity,
|
| 108428 |
+
"learning_rate": 0.00018426282581324678,
|
| 108429 |
+
"loss": 11.1701,
|
| 108430 |
+
"step": 154670
|
| 108431 |
+
},
|
| 108432 |
+
{
|
| 108433 |
+
"epoch": 18.613718411552348,
|
| 108434 |
+
"grad_norm": Infinity,
|
| 108435 |
+
"learning_rate": 0.00018426077716932998,
|
| 108436 |
+
"loss": 11.1545,
|
| 108437 |
+
"step": 154680
|
| 108438 |
+
},
|
| 108439 |
+
{
|
| 108440 |
+
"epoch": 18.614921780986762,
|
| 108441 |
+
"grad_norm": Infinity,
|
| 108442 |
+
"learning_rate": 0.00018425872840346707,
|
| 108443 |
+
"loss": 11.1638,
|
| 108444 |
+
"step": 154690
|
| 108445 |
+
},
|
| 108446 |
+
{
|
| 108447 |
+
"epoch": 18.61612515042118,
|
| 108448 |
+
"grad_norm": Infinity,
|
| 108449 |
+
"learning_rate": 0.00018425667951566096,
|
| 108450 |
+
"loss": 11.1699,
|
| 108451 |
+
"step": 154700
|
| 108452 |
+
},
|
| 108453 |
+
{
|
| 108454 |
+
"epoch": 18.617328519855597,
|
| 108455 |
+
"grad_norm": Infinity,
|
| 108456 |
+
"learning_rate": 0.00018425463050591464,
|
| 108457 |
+
"loss": 11.2327,
|
| 108458 |
+
"step": 154710
|
| 108459 |
+
},
|
| 108460 |
+
{
|
| 108461 |
+
"epoch": 18.61853188929001,
|
| 108462 |
+
"grad_norm": Infinity,
|
| 108463 |
+
"learning_rate": 0.00018425258137423108,
|
| 108464 |
+
"loss": 11.1129,
|
| 108465 |
+
"step": 154720
|
| 108466 |
+
},
|
| 108467 |
+
{
|
| 108468 |
+
"epoch": 18.61973525872443,
|
| 108469 |
+
"grad_norm": Infinity,
|
| 108470 |
+
"learning_rate": 0.0001842505321206132,
|
| 108471 |
+
"loss": 11.1449,
|
| 108472 |
+
"step": 154730
|
| 108473 |
+
},
|
| 108474 |
+
{
|
| 108475 |
+
"epoch": 18.620938628158846,
|
| 108476 |
+
"grad_norm": Infinity,
|
| 108477 |
+
"learning_rate": 0.00018424848274506403,
|
| 108478 |
+
"loss": 11.019,
|
| 108479 |
+
"step": 154740
|
| 108480 |
+
},
|
| 108481 |
+
{
|
| 108482 |
+
"epoch": 18.62214199759326,
|
| 108483 |
+
"grad_norm": Infinity,
|
| 108484 |
+
"learning_rate": 0.0001842464332475865,
|
| 108485 |
+
"loss": 11.1304,
|
| 108486 |
+
"step": 154750
|
| 108487 |
+
},
|
| 108488 |
+
{
|
| 108489 |
+
"epoch": 18.623345367027678,
|
| 108490 |
+
"grad_norm": Infinity,
|
| 108491 |
+
"learning_rate": 0.00018424438362818358,
|
| 108492 |
+
"loss": 11.1931,
|
| 108493 |
+
"step": 154760
|
| 108494 |
+
},
|
| 108495 |
+
{
|
| 108496 |
+
"epoch": 18.624548736462096,
|
| 108497 |
+
"grad_norm": Infinity,
|
| 108498 |
+
"learning_rate": 0.00018424233388685823,
|
| 108499 |
+
"loss": 11.0813,
|
| 108500 |
+
"step": 154770
|
| 108501 |
+
},
|
| 108502 |
+
{
|
| 108503 |
+
"epoch": 18.62575210589651,
|
| 108504 |
+
"grad_norm": Infinity,
|
| 108505 |
+
"learning_rate": 0.00018424028402361345,
|
| 108506 |
+
"loss": 11.1179,
|
| 108507 |
+
"step": 154780
|
| 108508 |
+
},
|
| 108509 |
+
{
|
| 108510 |
+
"epoch": 18.626955475330927,
|
| 108511 |
+
"grad_norm": Infinity,
|
| 108512 |
+
"learning_rate": 0.00018423823403845212,
|
| 108513 |
+
"loss": 11.2252,
|
| 108514 |
+
"step": 154790
|
| 108515 |
+
},
|
| 108516 |
+
{
|
| 108517 |
+
"epoch": 18.628158844765345,
|
| 108518 |
+
"grad_norm": Infinity,
|
| 108519 |
+
"learning_rate": 0.00018423618393137732,
|
| 108520 |
+
"loss": 11.2128,
|
| 108521 |
+
"step": 154800
|
| 108522 |
+
},
|
| 108523 |
+
{
|
| 108524 |
+
"epoch": 18.62936221419976,
|
| 108525 |
+
"grad_norm": Infinity,
|
| 108526 |
+
"learning_rate": 0.00018423413370239198,
|
| 108527 |
+
"loss": 11.1383,
|
| 108528 |
+
"step": 154810
|
| 108529 |
+
},
|
| 108530 |
+
{
|
| 108531 |
+
"epoch": 18.630565583634176,
|
| 108532 |
+
"grad_norm": Infinity,
|
| 108533 |
+
"learning_rate": 0.000184232083351499,
|
| 108534 |
+
"loss": 11.1558,
|
| 108535 |
+
"step": 154820
|
| 108536 |
+
},
|
| 108537 |
+
{
|
| 108538 |
+
"epoch": 18.63176895306859,
|
| 108539 |
+
"grad_norm": Infinity,
|
| 108540 |
+
"learning_rate": 0.00018423003287870143,
|
| 108541 |
+
"loss": 11.1106,
|
| 108542 |
+
"step": 154830
|
| 108543 |
+
},
|
| 108544 |
+
{
|
| 108545 |
+
"epoch": 18.632972322503008,
|
| 108546 |
+
"grad_norm": Infinity,
|
| 108547 |
+
"learning_rate": 0.0001842279822840022,
|
| 108548 |
+
"loss": 11.2044,
|
| 108549 |
+
"step": 154840
|
| 108550 |
+
},
|
| 108551 |
+
{
|
| 108552 |
+
"epoch": 18.634175691937426,
|
| 108553 |
+
"grad_norm": Infinity,
|
| 108554 |
+
"learning_rate": 0.0001842259315674043,
|
| 108555 |
+
"loss": 11.0732,
|
| 108556 |
+
"step": 154850
|
| 108557 |
+
},
|
| 108558 |
+
{
|
| 108559 |
+
"epoch": 18.63537906137184,
|
| 108560 |
+
"grad_norm": Infinity,
|
| 108561 |
+
"learning_rate": 0.00018422388072891063,
|
| 108562 |
+
"loss": 11.2463,
|
| 108563 |
+
"step": 154860
|
| 108564 |
+
},
|
| 108565 |
+
{
|
| 108566 |
+
"epoch": 18.636582430806257,
|
| 108567 |
+
"grad_norm": Infinity,
|
| 108568 |
+
"learning_rate": 0.00018422182976852427,
|
| 108569 |
+
"loss": 11.208,
|
| 108570 |
+
"step": 154870
|
| 108571 |
+
},
|
| 108572 |
+
{
|
| 108573 |
+
"epoch": 18.637785800240675,
|
| 108574 |
+
"grad_norm": Infinity,
|
| 108575 |
+
"learning_rate": 0.00018421977868624808,
|
| 108576 |
+
"loss": 11.0892,
|
| 108577 |
+
"step": 154880
|
| 108578 |
+
},
|
| 108579 |
+
{
|
| 108580 |
+
"epoch": 18.63898916967509,
|
| 108581 |
+
"grad_norm": Infinity,
|
| 108582 |
+
"learning_rate": 0.0001842177274820851,
|
| 108583 |
+
"loss": 11.0983,
|
| 108584 |
+
"step": 154890
|
| 108585 |
+
},
|
| 108586 |
+
{
|
| 108587 |
+
"epoch": 18.640192539109506,
|
| 108588 |
+
"grad_norm": Infinity,
|
| 108589 |
+
"learning_rate": 0.0001842156761560383,
|
| 108590 |
+
"loss": 11.0362,
|
| 108591 |
+
"step": 154900
|
| 108592 |
+
},
|
| 108593 |
+
{
|
| 108594 |
+
"epoch": 18.641395908543924,
|
| 108595 |
+
"grad_norm": Infinity,
|
| 108596 |
+
"learning_rate": 0.0001842136247081106,
|
| 108597 |
+
"loss": 11.2536,
|
| 108598 |
+
"step": 154910
|
| 108599 |
+
},
|
| 108600 |
+
{
|
| 108601 |
+
"epoch": 18.642599277978338,
|
| 108602 |
+
"grad_norm": Infinity,
|
| 108603 |
+
"learning_rate": 0.000184211573138305,
|
| 108604 |
+
"loss": 11.1086,
|
| 108605 |
+
"step": 154920
|
| 108606 |
+
},
|
| 108607 |
+
{
|
| 108608 |
+
"epoch": 18.643802647412755,
|
| 108609 |
+
"grad_norm": Infinity,
|
| 108610 |
+
"learning_rate": 0.00018420952144662444,
|
| 108611 |
+
"loss": 11.1801,
|
| 108612 |
+
"step": 154930
|
| 108613 |
+
},
|
| 108614 |
+
{
|
| 108615 |
+
"epoch": 18.645006016847173,
|
| 108616 |
+
"grad_norm": Infinity,
|
| 108617 |
+
"learning_rate": 0.0001842074696330719,
|
| 108618 |
+
"loss": 11.1866,
|
| 108619 |
+
"step": 154940
|
| 108620 |
+
},
|
| 108621 |
+
{
|
| 108622 |
+
"epoch": 18.646209386281587,
|
| 108623 |
+
"grad_norm": Infinity,
|
| 108624 |
+
"learning_rate": 0.00018420541769765042,
|
| 108625 |
+
"loss": 11.1017,
|
| 108626 |
+
"step": 154950
|
| 108627 |
+
},
|
| 108628 |
+
{
|
| 108629 |
+
"epoch": 18.647412755716005,
|
| 108630 |
+
"grad_norm": Infinity,
|
| 108631 |
+
"learning_rate": 0.00018420336564036287,
|
| 108632 |
+
"loss": 11.2785,
|
| 108633 |
+
"step": 154960
|
| 108634 |
+
},
|
| 108635 |
+
{
|
| 108636 |
+
"epoch": 18.648616125150422,
|
| 108637 |
+
"grad_norm": Infinity,
|
| 108638 |
+
"learning_rate": 0.0001842013134612123,
|
| 108639 |
+
"loss": 11.0446,
|
| 108640 |
+
"step": 154970
|
| 108641 |
+
},
|
| 108642 |
+
{
|
| 108643 |
+
"epoch": 18.649819494584836,
|
| 108644 |
+
"grad_norm": Infinity,
|
| 108645 |
+
"learning_rate": 0.0001841992611602016,
|
| 108646 |
+
"loss": 11.1732,
|
| 108647 |
+
"step": 154980
|
| 108648 |
+
},
|
| 108649 |
+
{
|
| 108650 |
+
"epoch": 18.651022864019254,
|
| 108651 |
+
"grad_norm": Infinity,
|
| 108652 |
+
"learning_rate": 0.0001841972087373338,
|
| 108653 |
+
"loss": 11.136,
|
| 108654 |
+
"step": 154990
|
| 108655 |
+
},
|
| 108656 |
+
{
|
| 108657 |
+
"epoch": 18.65222623345367,
|
| 108658 |
+
"grad_norm": Infinity,
|
| 108659 |
+
"learning_rate": 0.00018419515619261186,
|
| 108660 |
+
"loss": 11.0919,
|
| 108661 |
+
"step": 155000
|
| 108662 |
+
},
|
| 108663 |
+
{
|
| 108664 |
+
"epoch": 18.653429602888085,
|
| 108665 |
+
"grad_norm": Infinity,
|
| 108666 |
+
"learning_rate": 0.00018419310352603868,
|
| 108667 |
+
"loss": 11.1219,
|
| 108668 |
+
"step": 155010
|
| 108669 |
+
},
|
| 108670 |
+
{
|
| 108671 |
+
"epoch": 18.654632972322503,
|
| 108672 |
+
"grad_norm": Infinity,
|
| 108673 |
+
"learning_rate": 0.00018419105073761737,
|
| 108674 |
+
"loss": 11.1603,
|
| 108675 |
+
"step": 155020
|
| 108676 |
+
},
|
| 108677 |
+
{
|
| 108678 |
+
"epoch": 18.65583634175692,
|
| 108679 |
+
"grad_norm": Infinity,
|
| 108680 |
+
"learning_rate": 0.00018418899782735078,
|
| 108681 |
+
"loss": 11.1119,
|
| 108682 |
+
"step": 155030
|
| 108683 |
+
},
|
| 108684 |
+
{
|
| 108685 |
+
"epoch": 18.657039711191334,
|
| 108686 |
+
"grad_norm": Infinity,
|
| 108687 |
+
"learning_rate": 0.00018418694479524193,
|
| 108688 |
+
"loss": 11.2415,
|
| 108689 |
+
"step": 155040
|
| 108690 |
+
},
|
| 108691 |
+
{
|
| 108692 |
+
"epoch": 18.658243080625752,
|
| 108693 |
+
"grad_norm": Infinity,
|
| 108694 |
+
"learning_rate": 0.0001841848916412938,
|
| 108695 |
+
"loss": 11.1529,
|
| 108696 |
+
"step": 155050
|
| 108697 |
+
},
|
| 108698 |
+
{
|
| 108699 |
+
"epoch": 18.65944645006017,
|
| 108700 |
+
"grad_norm": Infinity,
|
| 108701 |
+
"learning_rate": 0.00018418283836550936,
|
| 108702 |
+
"loss": 11.0478,
|
| 108703 |
+
"step": 155060
|
| 108704 |
+
},
|
| 108705 |
+
{
|
| 108706 |
+
"epoch": 18.660649819494584,
|
| 108707 |
+
"grad_norm": Infinity,
|
| 108708 |
+
"learning_rate": 0.00018418078496789154,
|
| 108709 |
+
"loss": 11.1712,
|
| 108710 |
+
"step": 155070
|
| 108711 |
+
},
|
| 108712 |
+
{
|
| 108713 |
+
"epoch": 18.661853188929,
|
| 108714 |
+
"grad_norm": Infinity,
|
| 108715 |
+
"learning_rate": 0.00018417873144844333,
|
| 108716 |
+
"loss": 11.0886,
|
| 108717 |
+
"step": 155080
|
| 108718 |
+
},
|
| 108719 |
+
{
|
| 108720 |
+
"epoch": 18.66305655836342,
|
| 108721 |
+
"grad_norm": Infinity,
|
| 108722 |
+
"learning_rate": 0.00018417667780716777,
|
| 108723 |
+
"loss": 11.0901,
|
| 108724 |
+
"step": 155090
|
| 108725 |
+
},
|
| 108726 |
+
{
|
| 108727 |
+
"epoch": 18.664259927797833,
|
| 108728 |
+
"grad_norm": Infinity,
|
| 108729 |
+
"learning_rate": 0.00018417462404406773,
|
| 108730 |
+
"loss": 11.1937,
|
| 108731 |
+
"step": 155100
|
| 108732 |
+
},
|
| 108733 |
+
{
|
| 108734 |
+
"epoch": 18.66546329723225,
|
| 108735 |
+
"grad_norm": Infinity,
|
| 108736 |
+
"learning_rate": 0.0001841725701591462,
|
| 108737 |
+
"loss": 11.0931,
|
| 108738 |
+
"step": 155110
|
| 108739 |
+
},
|
| 108740 |
+
{
|
| 108741 |
+
"epoch": 18.666666666666668,
|
| 108742 |
+
"grad_norm": Infinity,
|
| 108743 |
+
"learning_rate": 0.00018417051615240624,
|
| 108744 |
+
"loss": 11.0515,
|
| 108745 |
+
"step": 155120
|
| 108746 |
+
},
|
| 108747 |
+
{
|
| 108748 |
+
"epoch": 18.667870036101082,
|
| 108749 |
+
"grad_norm": Infinity,
|
| 108750 |
+
"learning_rate": 0.00018416846202385073,
|
| 108751 |
+
"loss": 11.2758,
|
| 108752 |
+
"step": 155130
|
| 108753 |
+
},
|
| 108754 |
+
{
|
| 108755 |
+
"epoch": 18.6690734055355,
|
| 108756 |
+
"grad_norm": Infinity,
|
| 108757 |
+
"learning_rate": 0.0001841664077734827,
|
| 108758 |
+
"loss": 11.1772,
|
| 108759 |
+
"step": 155140
|
| 108760 |
+
},
|
| 108761 |
+
{
|
| 108762 |
+
"epoch": 18.670276774969917,
|
| 108763 |
+
"grad_norm": Infinity,
|
| 108764 |
+
"learning_rate": 0.00018416435340130508,
|
| 108765 |
+
"loss": 11.0956,
|
| 108766 |
+
"step": 155150
|
| 108767 |
+
},
|
| 108768 |
+
{
|
| 108769 |
+
"epoch": 18.67148014440433,
|
| 108770 |
+
"grad_norm": Infinity,
|
| 108771 |
+
"learning_rate": 0.00018416229890732083,
|
| 108772 |
+
"loss": 11.0273,
|
| 108773 |
+
"step": 155160
|
| 108774 |
+
},
|
| 108775 |
+
{
|
| 108776 |
+
"epoch": 18.67268351383875,
|
| 108777 |
+
"grad_norm": Infinity,
|
| 108778 |
+
"learning_rate": 0.000184160244291533,
|
| 108779 |
+
"loss": 11.207,
|
| 108780 |
+
"step": 155170
|
| 108781 |
+
},
|
| 108782 |
+
{
|
| 108783 |
+
"epoch": 18.673886883273166,
|
| 108784 |
+
"grad_norm": Infinity,
|
| 108785 |
+
"learning_rate": 0.0001841581895539445,
|
| 108786 |
+
"loss": 11.1246,
|
| 108787 |
+
"step": 155180
|
| 108788 |
+
},
|
| 108789 |
+
{
|
| 108790 |
+
"epoch": 18.67509025270758,
|
| 108791 |
+
"grad_norm": Infinity,
|
| 108792 |
+
"learning_rate": 0.0001841561346945583,
|
| 108793 |
+
"loss": 11.2164,
|
| 108794 |
+
"step": 155190
|
| 108795 |
+
},
|
| 108796 |
+
{
|
| 108797 |
+
"epoch": 18.676293622141998,
|
| 108798 |
+
"grad_norm": Infinity,
|
| 108799 |
+
"learning_rate": 0.00018415407971337743,
|
| 108800 |
+
"loss": 11.1303,
|
| 108801 |
+
"step": 155200
|
| 108802 |
+
},
|
| 108803 |
+
{
|
| 108804 |
+
"epoch": 18.677496991576415,
|
| 108805 |
+
"grad_norm": Infinity,
|
| 108806 |
+
"learning_rate": 0.00018415202461040484,
|
| 108807 |
+
"loss": 11.0508,
|
| 108808 |
+
"step": 155210
|
| 108809 |
+
},
|
| 108810 |
+
{
|
| 108811 |
+
"epoch": 18.67870036101083,
|
| 108812 |
+
"grad_norm": Infinity,
|
| 108813 |
+
"learning_rate": 0.00018414996938564347,
|
| 108814 |
+
"loss": 11.1942,
|
| 108815 |
+
"step": 155220
|
| 108816 |
+
},
|
| 108817 |
+
{
|
| 108818 |
+
"epoch": 18.679903730445247,
|
| 108819 |
+
"grad_norm": Infinity,
|
| 108820 |
+
"learning_rate": 0.0001841479140390963,
|
| 108821 |
+
"loss": 11.1507,
|
| 108822 |
+
"step": 155230
|
| 108823 |
+
},
|
| 108824 |
+
{
|
| 108825 |
+
"epoch": 18.681107099879664,
|
| 108826 |
+
"grad_norm": Infinity,
|
| 108827 |
+
"learning_rate": 0.00018414585857076635,
|
| 108828 |
+
"loss": 11.094,
|
| 108829 |
+
"step": 155240
|
| 108830 |
+
},
|
| 108831 |
+
{
|
| 108832 |
+
"epoch": 18.68231046931408,
|
| 108833 |
+
"grad_norm": Infinity,
|
| 108834 |
+
"learning_rate": 0.00018414380298065658,
|
| 108835 |
+
"loss": 11.3475,
|
| 108836 |
+
"step": 155250
|
| 108837 |
+
},
|
| 108838 |
+
{
|
| 108839 |
+
"epoch": 18.683513838748496,
|
| 108840 |
+
"grad_norm": Infinity,
|
| 108841 |
+
"learning_rate": 0.00018414174726876992,
|
| 108842 |
+
"loss": 11.2109,
|
| 108843 |
+
"step": 155260
|
| 108844 |
+
},
|
| 108845 |
+
{
|
| 108846 |
+
"epoch": 18.684717208182914,
|
| 108847 |
+
"grad_norm": Infinity,
|
| 108848 |
+
"learning_rate": 0.0001841396914351094,
|
| 108849 |
+
"loss": 11.328,
|
| 108850 |
+
"step": 155270
|
| 108851 |
+
},
|
| 108852 |
+
{
|
| 108853 |
+
"epoch": 18.685920577617328,
|
| 108854 |
+
"grad_norm": Infinity,
|
| 108855 |
+
"learning_rate": 0.00018413763547967796,
|
| 108856 |
+
"loss": 11.2317,
|
| 108857 |
+
"step": 155280
|
| 108858 |
+
},
|
| 108859 |
+
{
|
| 108860 |
+
"epoch": 18.687123947051745,
|
| 108861 |
+
"grad_norm": Infinity,
|
| 108862 |
+
"learning_rate": 0.00018413557940247862,
|
| 108863 |
+
"loss": 11.0034,
|
| 108864 |
+
"step": 155290
|
| 108865 |
+
},
|
| 108866 |
+
{
|
| 108867 |
+
"epoch": 18.688327316486163,
|
| 108868 |
+
"grad_norm": Infinity,
|
| 108869 |
+
"learning_rate": 0.00018413352320351427,
|
| 108870 |
+
"loss": 11.2417,
|
| 108871 |
+
"step": 155300
|
| 108872 |
+
},
|
| 108873 |
+
{
|
| 108874 |
+
"epoch": 18.689530685920577,
|
| 108875 |
+
"grad_norm": Infinity,
|
| 108876 |
+
"learning_rate": 0.00018413146688278797,
|
| 108877 |
+
"loss": 10.9798,
|
| 108878 |
+
"step": 155310
|
| 108879 |
+
},
|
| 108880 |
+
{
|
| 108881 |
+
"epoch": 18.690734055354994,
|
| 108882 |
+
"grad_norm": Infinity,
|
| 108883 |
+
"learning_rate": 0.00018412941044030267,
|
| 108884 |
+
"loss": 11.2268,
|
| 108885 |
+
"step": 155320
|
| 108886 |
+
},
|
| 108887 |
+
{
|
| 108888 |
+
"epoch": 18.691937424789412,
|
| 108889 |
+
"grad_norm": Infinity,
|
| 108890 |
+
"learning_rate": 0.00018412735387606131,
|
| 108891 |
+
"loss": 11.1689,
|
| 108892 |
+
"step": 155330
|
| 108893 |
+
},
|
| 108894 |
+
{
|
| 108895 |
+
"epoch": 18.693140794223826,
|
| 108896 |
+
"grad_norm": Infinity,
|
| 108897 |
+
"learning_rate": 0.00018412529719006695,
|
| 108898 |
+
"loss": 11.1137,
|
| 108899 |
+
"step": 155340
|
| 108900 |
+
},
|
| 108901 |
+
{
|
| 108902 |
+
"epoch": 18.694344163658243,
|
| 108903 |
+
"grad_norm": Infinity,
|
| 108904 |
+
"learning_rate": 0.00018412324038232249,
|
| 108905 |
+
"loss": 11.1095,
|
| 108906 |
+
"step": 155350
|
| 108907 |
+
},
|
| 108908 |
+
{
|
| 108909 |
+
"epoch": 18.69554753309266,
|
| 108910 |
+
"grad_norm": Infinity,
|
| 108911 |
+
"learning_rate": 0.00018412118345283092,
|
| 108912 |
+
"loss": 10.9851,
|
| 108913 |
+
"step": 155360
|
| 108914 |
+
},
|
| 108915 |
+
{
|
| 108916 |
+
"epoch": 18.696750902527075,
|
| 108917 |
+
"grad_norm": Infinity,
|
| 108918 |
+
"learning_rate": 0.00018411912640159525,
|
| 108919 |
+
"loss": 11.0877,
|
| 108920 |
+
"step": 155370
|
| 108921 |
+
},
|
| 108922 |
+
{
|
| 108923 |
+
"epoch": 18.697954271961493,
|
| 108924 |
+
"grad_norm": Infinity,
|
| 108925 |
+
"learning_rate": 0.00018411706922861844,
|
| 108926 |
+
"loss": 11.1739,
|
| 108927 |
+
"step": 155380
|
| 108928 |
+
},
|
| 108929 |
+
{
|
| 108930 |
+
"epoch": 18.69915764139591,
|
| 108931 |
+
"grad_norm": Infinity,
|
| 108932 |
+
"learning_rate": 0.00018411501193390346,
|
| 108933 |
+
"loss": 11.1279,
|
| 108934 |
+
"step": 155390
|
| 108935 |
+
},
|
| 108936 |
+
{
|
| 108937 |
+
"epoch": 18.700361010830324,
|
| 108938 |
+
"grad_norm": Infinity,
|
| 108939 |
+
"learning_rate": 0.00018411295451745328,
|
| 108940 |
+
"loss": 11.1035,
|
| 108941 |
+
"step": 155400
|
| 108942 |
+
},
|
| 108943 |
+
{
|
| 108944 |
+
"epoch": 18.70156438026474,
|
| 108945 |
+
"grad_norm": Infinity,
|
| 108946 |
+
"learning_rate": 0.00018411089697927086,
|
| 108947 |
+
"loss": 11.1095,
|
| 108948 |
+
"step": 155410
|
| 108949 |
+
},
|
| 108950 |
+
{
|
| 108951 |
+
"epoch": 18.70276774969916,
|
| 108952 |
+
"grad_norm": Infinity,
|
| 108953 |
+
"learning_rate": 0.00018410883931935926,
|
| 108954 |
+
"loss": 11.1272,
|
| 108955 |
+
"step": 155420
|
| 108956 |
+
},
|
| 108957 |
+
{
|
| 108958 |
+
"epoch": 18.703971119133573,
|
| 108959 |
+
"grad_norm": Infinity,
|
| 108960 |
+
"learning_rate": 0.00018410678153772136,
|
| 108961 |
+
"loss": 11.1291,
|
| 108962 |
+
"step": 155430
|
| 108963 |
+
},
|
| 108964 |
+
{
|
| 108965 |
+
"epoch": 18.70517448856799,
|
| 108966 |
+
"grad_norm": Infinity,
|
| 108967 |
+
"learning_rate": 0.0001841047236343602,
|
| 108968 |
+
"loss": 11.1143,
|
| 108969 |
+
"step": 155440
|
| 108970 |
+
},
|
| 108971 |
+
{
|
| 108972 |
+
"epoch": 18.706377858002405,
|
| 108973 |
+
"grad_norm": Infinity,
|
| 108974 |
+
"learning_rate": 0.00018410266560927875,
|
| 108975 |
+
"loss": 11.081,
|
| 108976 |
+
"step": 155450
|
| 108977 |
+
},
|
| 108978 |
+
{
|
| 108979 |
+
"epoch": 18.707581227436823,
|
| 108980 |
+
"grad_norm": Infinity,
|
| 108981 |
+
"learning_rate": 0.00018410060746247995,
|
| 108982 |
+
"loss": 11.1918,
|
| 108983 |
+
"step": 155460
|
| 108984 |
+
},
|
| 108985 |
+
{
|
| 108986 |
+
"epoch": 18.70878459687124,
|
| 108987 |
+
"grad_norm": Infinity,
|
| 108988 |
+
"learning_rate": 0.00018409854919396686,
|
| 108989 |
+
"loss": 11.1112,
|
| 108990 |
+
"step": 155470
|
| 108991 |
+
},
|
| 108992 |
+
{
|
| 108993 |
+
"epoch": 18.709987966305654,
|
| 108994 |
+
"grad_norm": Infinity,
|
| 108995 |
+
"learning_rate": 0.00018409649080374237,
|
| 108996 |
+
"loss": 11.1154,
|
| 108997 |
+
"step": 155480
|
| 108998 |
+
},
|
| 108999 |
+
{
|
| 109000 |
+
"epoch": 18.71119133574007,
|
| 109001 |
+
"grad_norm": Infinity,
|
| 109002 |
+
"learning_rate": 0.00018409443229180946,
|
| 109003 |
+
"loss": 11.1244,
|
| 109004 |
+
"step": 155490
|
| 109005 |
+
},
|
| 109006 |
+
{
|
| 109007 |
+
"epoch": 18.71239470517449,
|
| 109008 |
+
"grad_norm": Infinity,
|
| 109009 |
+
"learning_rate": 0.00018409237365817118,
|
| 109010 |
+
"loss": 11.1785,
|
| 109011 |
+
"step": 155500
|
| 109012 |
+
},
|
| 109013 |
+
{
|
| 109014 |
+
"epoch": 18.713598074608903,
|
| 109015 |
+
"grad_norm": Infinity,
|
| 109016 |
+
"learning_rate": 0.0001840903149028305,
|
| 109017 |
+
"loss": 11.0254,
|
| 109018 |
+
"step": 155510
|
| 109019 |
+
},
|
| 109020 |
+
{
|
| 109021 |
+
"epoch": 18.71480144404332,
|
| 109022 |
+
"grad_norm": Infinity,
|
| 109023 |
+
"learning_rate": 0.00018408825602579033,
|
| 109024 |
+
"loss": 11.096,
|
| 109025 |
+
"step": 155520
|
| 109026 |
+
},
|
| 109027 |
+
{
|
| 109028 |
+
"epoch": 18.71600481347774,
|
| 109029 |
+
"grad_norm": Infinity,
|
| 109030 |
+
"learning_rate": 0.00018408619702705372,
|
| 109031 |
+
"loss": 11.1088,
|
| 109032 |
+
"step": 155530
|
| 109033 |
+
},
|
| 109034 |
+
{
|
| 109035 |
+
"epoch": 18.717208182912152,
|
| 109036 |
+
"grad_norm": Infinity,
|
| 109037 |
+
"learning_rate": 0.0001840841379066236,
|
| 109038 |
+
"loss": 11.1197,
|
| 109039 |
+
"step": 155540
|
| 109040 |
+
},
|
| 109041 |
+
{
|
| 109042 |
+
"epoch": 18.71841155234657,
|
| 109043 |
+
"grad_norm": Infinity,
|
| 109044 |
+
"learning_rate": 0.000184082078664503,
|
| 109045 |
+
"loss": 11.043,
|
| 109046 |
+
"step": 155550
|
| 109047 |
+
},
|
| 109048 |
+
{
|
| 109049 |
+
"epoch": 18.719614921780988,
|
| 109050 |
+
"grad_norm": Infinity,
|
| 109051 |
+
"learning_rate": 0.00018408001930069484,
|
| 109052 |
+
"loss": 11.0468,
|
| 109053 |
+
"step": 155560
|
| 109054 |
+
},
|
| 109055 |
+
{
|
| 109056 |
+
"epoch": 18.7208182912154,
|
| 109057 |
+
"grad_norm": Infinity,
|
| 109058 |
+
"learning_rate": 0.00018407795981520216,
|
| 109059 |
+
"loss": 11.0612,
|
| 109060 |
+
"step": 155570
|
| 109061 |
+
},
|
| 109062 |
+
{
|
| 109063 |
+
"epoch": 18.72202166064982,
|
| 109064 |
+
"grad_norm": Infinity,
|
| 109065 |
+
"learning_rate": 0.00018407590020802793,
|
| 109066 |
+
"loss": 11.2637,
|
| 109067 |
+
"step": 155580
|
| 109068 |
+
},
|
| 109069 |
+
{
|
| 109070 |
+
"epoch": 18.723225030084237,
|
| 109071 |
+
"grad_norm": Infinity,
|
| 109072 |
+
"learning_rate": 0.0001840738404791751,
|
| 109073 |
+
"loss": 11.1286,
|
| 109074 |
+
"step": 155590
|
| 109075 |
+
},
|
| 109076 |
+
{
|
| 109077 |
+
"epoch": 18.72442839951865,
|
| 109078 |
+
"grad_norm": Infinity,
|
| 109079 |
+
"learning_rate": 0.00018407178062864664,
|
| 109080 |
+
"loss": 10.9937,
|
| 109081 |
+
"step": 155600
|
| 109082 |
+
},
|
| 109083 |
+
{
|
| 109084 |
+
"epoch": 18.72563176895307,
|
| 109085 |
+
"grad_norm": Infinity,
|
| 109086 |
+
"learning_rate": 0.00018406972065644557,
|
| 109087 |
+
"loss": 11.0324,
|
| 109088 |
+
"step": 155610
|
| 109089 |
+
},
|
| 109090 |
+
{
|
| 109091 |
+
"epoch": 18.726835138387486,
|
| 109092 |
+
"grad_norm": Infinity,
|
| 109093 |
+
"learning_rate": 0.00018406766056257485,
|
| 109094 |
+
"loss": 11.0811,
|
| 109095 |
+
"step": 155620
|
| 109096 |
+
},
|
| 109097 |
+
{
|
| 109098 |
+
"epoch": 18.7280385078219,
|
| 109099 |
+
"grad_norm": Infinity,
|
| 109100 |
+
"learning_rate": 0.00018406560034703748,
|
| 109101 |
+
"loss": 11.1048,
|
| 109102 |
+
"step": 155630
|
| 109103 |
+
},
|
| 109104 |
+
{
|
| 109105 |
+
"epoch": 18.729241877256317,
|
| 109106 |
+
"grad_norm": Infinity,
|
| 109107 |
+
"learning_rate": 0.00018406354000983642,
|
| 109108 |
+
"loss": 10.9592,
|
| 109109 |
+
"step": 155640
|
| 109110 |
+
},
|
| 109111 |
+
{
|
| 109112 |
+
"epoch": 18.730445246690735,
|
| 109113 |
+
"grad_norm": Infinity,
|
| 109114 |
+
"learning_rate": 0.0001840614795509747,
|
| 109115 |
+
"loss": 11.0839,
|
| 109116 |
+
"step": 155650
|
| 109117 |
+
},
|
| 109118 |
+
{
|
| 109119 |
+
"epoch": 18.73164861612515,
|
| 109120 |
+
"grad_norm": Infinity,
|
| 109121 |
+
"learning_rate": 0.00018405941897045523,
|
| 109122 |
+
"loss": 11.1915,
|
| 109123 |
+
"step": 155660
|
| 109124 |
+
},
|
| 109125 |
+
{
|
| 109126 |
+
"epoch": 18.732851985559567,
|
| 109127 |
+
"grad_norm": Infinity,
|
| 109128 |
+
"learning_rate": 0.00018405735826828107,
|
| 109129 |
+
"loss": 11.0202,
|
| 109130 |
+
"step": 155670
|
| 109131 |
+
},
|
| 109132 |
+
{
|
| 109133 |
+
"epoch": 18.734055354993984,
|
| 109134 |
+
"grad_norm": Infinity,
|
| 109135 |
+
"learning_rate": 0.0001840552974444551,
|
| 109136 |
+
"loss": 11.1132,
|
| 109137 |
+
"step": 155680
|
| 109138 |
+
},
|
| 109139 |
+
{
|
| 109140 |
+
"epoch": 18.735258724428398,
|
| 109141 |
+
"grad_norm": Infinity,
|
| 109142 |
+
"learning_rate": 0.0001840532364989804,
|
| 109143 |
+
"loss": 11.2307,
|
| 109144 |
+
"step": 155690
|
| 109145 |
+
},
|
| 109146 |
+
{
|
| 109147 |
+
"epoch": 18.736462093862816,
|
| 109148 |
+
"grad_norm": Infinity,
|
| 109149 |
+
"learning_rate": 0.0001840511754318599,
|
| 109150 |
+
"loss": 11.1407,
|
| 109151 |
+
"step": 155700
|
| 109152 |
+
},
|
| 109153 |
+
{
|
| 109154 |
+
"epoch": 18.737665463297233,
|
| 109155 |
+
"grad_norm": Infinity,
|
| 109156 |
+
"learning_rate": 0.00018404911424309658,
|
| 109157 |
+
"loss": 11.1428,
|
| 109158 |
+
"step": 155710
|
| 109159 |
+
},
|
| 109160 |
+
{
|
| 109161 |
+
"epoch": 18.738868832731647,
|
| 109162 |
+
"grad_norm": Infinity,
|
| 109163 |
+
"learning_rate": 0.00018404705293269347,
|
| 109164 |
+
"loss": 11.1013,
|
| 109165 |
+
"step": 155720
|
| 109166 |
+
},
|
| 109167 |
+
{
|
| 109168 |
+
"epoch": 18.740072202166065,
|
| 109169 |
+
"grad_norm": Infinity,
|
| 109170 |
+
"learning_rate": 0.00018404499150065351,
|
| 109171 |
+
"loss": 11.1795,
|
| 109172 |
+
"step": 155730
|
| 109173 |
+
},
|
| 109174 |
+
{
|
| 109175 |
+
"epoch": 18.741275571600482,
|
| 109176 |
+
"grad_norm": Infinity,
|
| 109177 |
+
"learning_rate": 0.0001840429299469797,
|
| 109178 |
+
"loss": 11.1118,
|
| 109179 |
+
"step": 155740
|
| 109180 |
+
},
|
| 109181 |
+
{
|
| 109182 |
+
"epoch": 18.742478941034896,
|
| 109183 |
+
"grad_norm": Infinity,
|
| 109184 |
+
"learning_rate": 0.00018404086827167502,
|
| 109185 |
+
"loss": 11.0965,
|
| 109186 |
+
"step": 155750
|
| 109187 |
+
},
|
| 109188 |
+
{
|
| 109189 |
+
"epoch": 18.743682310469314,
|
| 109190 |
+
"grad_norm": Infinity,
|
| 109191 |
+
"learning_rate": 0.00018403880647474249,
|
| 109192 |
+
"loss": 11.1054,
|
| 109193 |
+
"step": 155760
|
| 109194 |
+
},
|
| 109195 |
+
{
|
| 109196 |
+
"epoch": 18.74488567990373,
|
| 109197 |
+
"grad_norm": Infinity,
|
| 109198 |
+
"learning_rate": 0.00018403674455618501,
|
| 109199 |
+
"loss": 11.0522,
|
| 109200 |
+
"step": 155770
|
| 109201 |
+
},
|
| 109202 |
+
{
|
| 109203 |
+
"epoch": 18.746089049338146,
|
| 109204 |
+
"grad_norm": Infinity,
|
| 109205 |
+
"learning_rate": 0.00018403468251600562,
|
| 109206 |
+
"loss": 10.9737,
|
| 109207 |
+
"step": 155780
|
| 109208 |
+
},
|
| 109209 |
+
{
|
| 109210 |
+
"epoch": 18.747292418772563,
|
| 109211 |
+
"grad_norm": Infinity,
|
| 109212 |
+
"learning_rate": 0.0001840326203542073,
|
| 109213 |
+
"loss": 11.2099,
|
| 109214 |
+
"step": 155790
|
| 109215 |
+
},
|
| 109216 |
+
{
|
| 109217 |
+
"epoch": 18.74849578820698,
|
| 109218 |
+
"grad_norm": Infinity,
|
| 109219 |
+
"learning_rate": 0.00018403055807079307,
|
| 109220 |
+
"loss": 11.174,
|
| 109221 |
+
"step": 155800
|
| 109222 |
+
},
|
| 109223 |
+
{
|
| 109224 |
+
"epoch": 18.749699157641395,
|
| 109225 |
+
"grad_norm": Infinity,
|
| 109226 |
+
"learning_rate": 0.00018402849566576582,
|
| 109227 |
+
"loss": 11.0626,
|
| 109228 |
+
"step": 155810
|
| 109229 |
+
},
|
| 109230 |
+
{
|
| 109231 |
+
"epoch": 18.750902527075812,
|
| 109232 |
+
"grad_norm": Infinity,
|
| 109233 |
+
"learning_rate": 0.00018402643313912862,
|
| 109234 |
+
"loss": 11.1572,
|
| 109235 |
+
"step": 155820
|
| 109236 |
+
},
|
| 109237 |
+
{
|
| 109238 |
+
"epoch": 18.75210589651023,
|
| 109239 |
+
"grad_norm": Infinity,
|
| 109240 |
+
"learning_rate": 0.00018402437049088444,
|
| 109241 |
+
"loss": 11.0243,
|
| 109242 |
+
"step": 155830
|
| 109243 |
+
},
|
| 109244 |
+
{
|
| 109245 |
+
"epoch": 18.753309265944644,
|
| 109246 |
+
"grad_norm": Infinity,
|
| 109247 |
+
"learning_rate": 0.0001840223077210362,
|
| 109248 |
+
"loss": 11.0665,
|
| 109249 |
+
"step": 155840
|
| 109250 |
+
},
|
| 109251 |
+
{
|
| 109252 |
+
"epoch": 18.75451263537906,
|
| 109253 |
+
"grad_norm": Infinity,
|
| 109254 |
+
"learning_rate": 0.000184020244829587,
|
| 109255 |
+
"loss": 11.1575,
|
| 109256 |
+
"step": 155850
|
| 109257 |
+
},
|
| 109258 |
+
{
|
| 109259 |
+
"epoch": 18.75571600481348,
|
| 109260 |
+
"grad_norm": Infinity,
|
| 109261 |
+
"learning_rate": 0.0001840181818165397,
|
| 109262 |
+
"loss": 11.2205,
|
| 109263 |
+
"step": 155860
|
| 109264 |
+
},
|
| 109265 |
+
{
|
| 109266 |
+
"epoch": 18.756919374247893,
|
| 109267 |
+
"grad_norm": Infinity,
|
| 109268 |
+
"learning_rate": 0.00018401611868189738,
|
| 109269 |
+
"loss": 11.2051,
|
| 109270 |
+
"step": 155870
|
| 109271 |
+
},
|
| 109272 |
+
{
|
| 109273 |
+
"epoch": 18.75812274368231,
|
| 109274 |
+
"grad_norm": Infinity,
|
| 109275 |
+
"learning_rate": 0.00018401405542566298,
|
| 109276 |
+
"loss": 11.1057,
|
| 109277 |
+
"step": 155880
|
| 109278 |
+
},
|
| 109279 |
+
{
|
| 109280 |
+
"epoch": 18.759326113116728,
|
| 109281 |
+
"grad_norm": Infinity,
|
| 109282 |
+
"learning_rate": 0.00018401199204783953,
|
| 109283 |
+
"loss": 11.094,
|
| 109284 |
+
"step": 155890
|
| 109285 |
+
},
|
| 109286 |
+
{
|
| 109287 |
+
"epoch": 18.760529482551142,
|
| 109288 |
+
"grad_norm": Infinity,
|
| 109289 |
+
"learning_rate": 0.00018400992854842997,
|
| 109290 |
+
"loss": 11.0583,
|
| 109291 |
+
"step": 155900
|
| 109292 |
+
},
|
| 109293 |
+
{
|
| 109294 |
+
"epoch": 18.76173285198556,
|
| 109295 |
+
"grad_norm": Infinity,
|
| 109296 |
+
"learning_rate": 0.0001840078649274373,
|
| 109297 |
+
"loss": 11.1038,
|
| 109298 |
+
"step": 155910
|
| 109299 |
+
},
|
| 109300 |
+
{
|
| 109301 |
+
"epoch": 18.762936221419977,
|
| 109302 |
+
"grad_norm": Infinity,
|
| 109303 |
+
"learning_rate": 0.0001840058011848645,
|
| 109304 |
+
"loss": 11.3338,
|
| 109305 |
+
"step": 155920
|
| 109306 |
+
},
|
| 109307 |
+
{
|
| 109308 |
+
"epoch": 18.76413959085439,
|
| 109309 |
+
"grad_norm": Infinity,
|
| 109310 |
+
"learning_rate": 0.00018400373732071457,
|
| 109311 |
+
"loss": 11.0459,
|
| 109312 |
+
"step": 155930
|
| 109313 |
+
},
|
| 109314 |
+
{
|
| 109315 |
+
"epoch": 18.76534296028881,
|
| 109316 |
+
"grad_norm": Infinity,
|
| 109317 |
+
"learning_rate": 0.00018400167333499048,
|
| 109318 |
+
"loss": 11.2301,
|
| 109319 |
+
"step": 155940
|
| 109320 |
+
},
|
| 109321 |
+
{
|
| 109322 |
+
"epoch": 18.766546329723226,
|
| 109323 |
+
"grad_norm": Infinity,
|
| 109324 |
+
"learning_rate": 0.00018399960922769527,
|
| 109325 |
+
"loss": 11.1998,
|
| 109326 |
+
"step": 155950
|
| 109327 |
+
},
|
| 109328 |
+
{
|
| 109329 |
+
"epoch": 18.76774969915764,
|
| 109330 |
+
"grad_norm": Infinity,
|
| 109331 |
+
"learning_rate": 0.00018399754499883185,
|
| 109332 |
+
"loss": 11.096,
|
| 109333 |
+
"step": 155960
|
| 109334 |
+
},
|
| 109335 |
+
{
|
| 109336 |
+
"epoch": 18.768953068592058,
|
| 109337 |
+
"grad_norm": Infinity,
|
| 109338 |
+
"learning_rate": 0.00018399548064840322,
|
| 109339 |
+
"loss": 11.1265,
|
| 109340 |
+
"step": 155970
|
| 109341 |
+
},
|
| 109342 |
+
{
|
| 109343 |
+
"epoch": 18.770156438026476,
|
| 109344 |
+
"grad_norm": Infinity,
|
| 109345 |
+
"learning_rate": 0.00018399341617641245,
|
| 109346 |
+
"loss": 11.1074,
|
| 109347 |
+
"step": 155980
|
| 109348 |
+
},
|
| 109349 |
+
{
|
| 109350 |
+
"epoch": 18.77135980746089,
|
| 109351 |
+
"grad_norm": Infinity,
|
| 109352 |
+
"learning_rate": 0.00018399135158286245,
|
| 109353 |
+
"loss": 11.1348,
|
| 109354 |
+
"step": 155990
|
| 109355 |
+
},
|
| 109356 |
+
{
|
| 109357 |
+
"epoch": 18.772563176895307,
|
| 109358 |
+
"grad_norm": Infinity,
|
| 109359 |
+
"learning_rate": 0.0001839892868677562,
|
| 109360 |
+
"loss": 11.0085,
|
| 109361 |
+
"step": 156000
|
| 109362 |
+
},
|
| 109363 |
+
{
|
| 109364 |
+
"epoch": 18.773766546329725,
|
| 109365 |
+
"grad_norm": Infinity,
|
| 109366 |
+
"learning_rate": 0.00018398722203109675,
|
| 109367 |
+
"loss": 11.0944,
|
| 109368 |
+
"step": 156010
|
| 109369 |
+
},
|
| 109370 |
+
{
|
| 109371 |
+
"epoch": 18.77496991576414,
|
| 109372 |
+
"grad_norm": Infinity,
|
| 109373 |
+
"learning_rate": 0.00018398515707288704,
|
| 109374 |
+
"loss": 11.1064,
|
| 109375 |
+
"step": 156020
|
| 109376 |
+
},
|
| 109377 |
+
{
|
| 109378 |
+
"epoch": 18.776173285198556,
|
| 109379 |
+
"grad_norm": Infinity,
|
| 109380 |
+
"learning_rate": 0.00018398309199313008,
|
| 109381 |
+
"loss": 11.202,
|
| 109382 |
+
"step": 156030
|
| 109383 |
+
},
|
| 109384 |
+
{
|
| 109385 |
+
"epoch": 18.777376654632974,
|
| 109386 |
+
"grad_norm": Infinity,
|
| 109387 |
+
"learning_rate": 0.00018398102679182885,
|
| 109388 |
+
"loss": 11.0866,
|
| 109389 |
+
"step": 156040
|
| 109390 |
+
},
|
| 109391 |
+
{
|
| 109392 |
+
"epoch": 18.778580024067388,
|
| 109393 |
+
"grad_norm": Infinity,
|
| 109394 |
+
"learning_rate": 0.00018397896146898633,
|
| 109395 |
+
"loss": 11.1692,
|
| 109396 |
+
"step": 156050
|
| 109397 |
+
},
|
| 109398 |
+
{
|
| 109399 |
+
"epoch": 18.779783393501805,
|
| 109400 |
+
"grad_norm": Infinity,
|
| 109401 |
+
"learning_rate": 0.0001839768960246055,
|
| 109402 |
+
"loss": 11.1189,
|
| 109403 |
+
"step": 156060
|
| 109404 |
+
},
|
| 109405 |
+
{
|
| 109406 |
+
"epoch": 18.780986762936223,
|
| 109407 |
+
"grad_norm": Infinity,
|
| 109408 |
+
"learning_rate": 0.00018397483045868942,
|
| 109409 |
+
"loss": 11.1767,
|
| 109410 |
+
"step": 156070
|
| 109411 |
+
},
|
| 109412 |
+
{
|
| 109413 |
+
"epoch": 18.782190132370637,
|
| 109414 |
+
"grad_norm": Infinity,
|
| 109415 |
+
"learning_rate": 0.00018397276477124098,
|
| 109416 |
+
"loss": 11.1687,
|
| 109417 |
+
"step": 156080
|
| 109418 |
+
},
|
| 109419 |
+
{
|
| 109420 |
+
"epoch": 18.783393501805055,
|
| 109421 |
+
"grad_norm": Infinity,
|
| 109422 |
+
"learning_rate": 0.00018397069896226325,
|
| 109423 |
+
"loss": 11.1573,
|
| 109424 |
+
"step": 156090
|
| 109425 |
+
},
|
| 109426 |
+
{
|
| 109427 |
+
"epoch": 18.784596871239472,
|
| 109428 |
+
"grad_norm": Infinity,
|
| 109429 |
+
"learning_rate": 0.00018396863303175918,
|
| 109430 |
+
"loss": 11.0969,
|
| 109431 |
+
"step": 156100
|
| 109432 |
+
},
|
| 109433 |
+
{
|
| 109434 |
+
"epoch": 18.785800240673886,
|
| 109435 |
+
"grad_norm": Infinity,
|
| 109436 |
+
"learning_rate": 0.00018396656697973175,
|
| 109437 |
+
"loss": 11.0836,
|
| 109438 |
+
"step": 156110
|
| 109439 |
+
},
|
| 109440 |
+
{
|
| 109441 |
+
"epoch": 18.787003610108304,
|
| 109442 |
+
"grad_norm": Infinity,
|
| 109443 |
+
"learning_rate": 0.00018396450080618396,
|
| 109444 |
+
"loss": 11.1231,
|
| 109445 |
+
"step": 156120
|
| 109446 |
+
},
|
| 109447 |
+
{
|
| 109448 |
+
"epoch": 18.78820697954272,
|
| 109449 |
+
"grad_norm": Infinity,
|
| 109450 |
+
"learning_rate": 0.00018396243451111884,
|
| 109451 |
+
"loss": 11.0268,
|
| 109452 |
+
"step": 156130
|
| 109453 |
+
},
|
| 109454 |
+
{
|
| 109455 |
+
"epoch": 18.789410348977135,
|
| 109456 |
+
"grad_norm": Infinity,
|
| 109457 |
+
"learning_rate": 0.00018396036809453933,
|
| 109458 |
+
"loss": 11.2009,
|
| 109459 |
+
"step": 156140
|
| 109460 |
+
},
|
| 109461 |
+
{
|
| 109462 |
+
"epoch": 18.790613718411553,
|
| 109463 |
+
"grad_norm": Infinity,
|
| 109464 |
+
"learning_rate": 0.00018395830155644845,
|
| 109465 |
+
"loss": 11.106,
|
| 109466 |
+
"step": 156150
|
| 109467 |
+
},
|
| 109468 |
+
{
|
| 109469 |
+
"epoch": 18.79181708784597,
|
| 109470 |
+
"grad_norm": Infinity,
|
| 109471 |
+
"learning_rate": 0.00018395623489684916,
|
| 109472 |
+
"loss": 11.1531,
|
| 109473 |
+
"step": 156160
|
| 109474 |
+
},
|
| 109475 |
+
{
|
| 109476 |
+
"epoch": 18.793020457280385,
|
| 109477 |
+
"grad_norm": Infinity,
|
| 109478 |
+
"learning_rate": 0.00018395416811574448,
|
| 109479 |
+
"loss": 11.1239,
|
| 109480 |
+
"step": 156170
|
| 109481 |
+
},
|
| 109482 |
+
{
|
| 109483 |
+
"epoch": 18.794223826714802,
|
| 109484 |
+
"grad_norm": Infinity,
|
| 109485 |
+
"learning_rate": 0.0001839521012131374,
|
| 109486 |
+
"loss": 11.0848,
|
| 109487 |
+
"step": 156180
|
| 109488 |
+
},
|
| 109489 |
+
{
|
| 109490 |
+
"epoch": 18.79542719614922,
|
| 109491 |
+
"grad_norm": Infinity,
|
| 109492 |
+
"learning_rate": 0.00018395003418903086,
|
| 109493 |
+
"loss": 11.0738,
|
| 109494 |
+
"step": 156190
|
| 109495 |
+
},
|
| 109496 |
+
{
|
| 109497 |
+
"epoch": 18.796630565583634,
|
| 109498 |
+
"grad_norm": Infinity,
|
| 109499 |
+
"learning_rate": 0.00018394796704342795,
|
| 109500 |
+
"loss": 11.1085,
|
| 109501 |
+
"step": 156200
|
| 109502 |
+
},
|
| 109503 |
+
{
|
| 109504 |
+
"epoch": 18.79783393501805,
|
| 109505 |
+
"grad_norm": Infinity,
|
| 109506 |
+
"learning_rate": 0.0001839458997763316,
|
| 109507 |
+
"loss": 11.1598,
|
| 109508 |
+
"step": 156210
|
| 109509 |
+
},
|
| 109510 |
+
{
|
| 109511 |
+
"epoch": 18.799037304452465,
|
| 109512 |
+
"grad_norm": Infinity,
|
| 109513 |
+
"learning_rate": 0.00018394383238774477,
|
| 109514 |
+
"loss": 11.1524,
|
| 109515 |
+
"step": 156220
|
| 109516 |
+
},
|
| 109517 |
+
{
|
| 109518 |
+
"epoch": 18.800240673886883,
|
| 109519 |
+
"grad_norm": Infinity,
|
| 109520 |
+
"learning_rate": 0.00018394176487767052,
|
| 109521 |
+
"loss": 11.0619,
|
| 109522 |
+
"step": 156230
|
| 109523 |
+
},
|
| 109524 |
+
{
|
| 109525 |
+
"epoch": 18.8014440433213,
|
| 109526 |
+
"grad_norm": Infinity,
|
| 109527 |
+
"learning_rate": 0.0001839396972461118,
|
| 109528 |
+
"loss": 11.0887,
|
| 109529 |
+
"step": 156240
|
| 109530 |
+
},
|
| 109531 |
+
{
|
| 109532 |
+
"epoch": 18.802647412755714,
|
| 109533 |
+
"grad_norm": Infinity,
|
| 109534 |
+
"learning_rate": 0.0001839376294930716,
|
| 109535 |
+
"loss": 11.1339,
|
| 109536 |
+
"step": 156250
|
| 109537 |
+
},
|
| 109538 |
+
{
|
| 109539 |
+
"epoch": 18.803850782190132,
|
| 109540 |
+
"grad_norm": Infinity,
|
| 109541 |
+
"learning_rate": 0.00018393556161855297,
|
| 109542 |
+
"loss": 11.2568,
|
| 109543 |
+
"step": 156260
|
| 109544 |
+
},
|
| 109545 |
+
{
|
| 109546 |
+
"epoch": 18.80505415162455,
|
| 109547 |
+
"grad_norm": Infinity,
|
| 109548 |
+
"learning_rate": 0.00018393349362255882,
|
| 109549 |
+
"loss": 11.2225,
|
| 109550 |
+
"step": 156270
|
| 109551 |
+
},
|
| 109552 |
+
{
|
| 109553 |
+
"epoch": 18.806257521058964,
|
| 109554 |
+
"grad_norm": Infinity,
|
| 109555 |
+
"learning_rate": 0.0001839314255050922,
|
| 109556 |
+
"loss": 11.1804,
|
| 109557 |
+
"step": 156280
|
| 109558 |
+
},
|
| 109559 |
+
{
|
| 109560 |
+
"epoch": 18.80746089049338,
|
| 109561 |
+
"grad_norm": Infinity,
|
| 109562 |
+
"learning_rate": 0.00018392935726615608,
|
| 109563 |
+
"loss": 11.0418,
|
| 109564 |
+
"step": 156290
|
| 109565 |
+
},
|
| 109566 |
+
{
|
| 109567 |
+
"epoch": 18.8086642599278,
|
| 109568 |
+
"grad_norm": Infinity,
|
| 109569 |
+
"learning_rate": 0.00018392728890575345,
|
| 109570 |
+
"loss": 11.169,
|
| 109571 |
+
"step": 156300
|
| 109572 |
+
},
|
| 109573 |
+
{
|
| 109574 |
+
"epoch": 18.809867629362213,
|
| 109575 |
+
"grad_norm": Infinity,
|
| 109576 |
+
"learning_rate": 0.00018392522042388733,
|
| 109577 |
+
"loss": 11.1228,
|
| 109578 |
+
"step": 156310
|
| 109579 |
+
},
|
| 109580 |
+
{
|
| 109581 |
+
"epoch": 18.81107099879663,
|
| 109582 |
+
"grad_norm": Infinity,
|
| 109583 |
+
"learning_rate": 0.0001839231518205607,
|
| 109584 |
+
"loss": 11.2595,
|
| 109585 |
+
"step": 156320
|
| 109586 |
+
},
|
| 109587 |
+
{
|
| 109588 |
+
"epoch": 18.812274368231048,
|
| 109589 |
+
"grad_norm": Infinity,
|
| 109590 |
+
"learning_rate": 0.00018392108309577652,
|
| 109591 |
+
"loss": 11.2344,
|
| 109592 |
+
"step": 156330
|
| 109593 |
+
},
|
| 109594 |
+
{
|
| 109595 |
+
"epoch": 18.813477737665462,
|
| 109596 |
+
"grad_norm": Infinity,
|
| 109597 |
+
"learning_rate": 0.0001839190142495378,
|
| 109598 |
+
"loss": 11.0632,
|
| 109599 |
+
"step": 156340
|
| 109600 |
+
},
|
| 109601 |
+
{
|
| 109602 |
+
"epoch": 18.81468110709988,
|
| 109603 |
+
"grad_norm": Infinity,
|
| 109604 |
+
"learning_rate": 0.00018391694528184757,
|
| 109605 |
+
"loss": 11.1457,
|
| 109606 |
+
"step": 156350
|
| 109607 |
+
},
|
| 109608 |
+
{
|
| 109609 |
+
"epoch": 18.815884476534297,
|
| 109610 |
+
"grad_norm": Infinity,
|
| 109611 |
+
"learning_rate": 0.00018391487619270883,
|
| 109612 |
+
"loss": 11.2568,
|
| 109613 |
+
"step": 156360
|
| 109614 |
+
},
|
| 109615 |
+
{
|
| 109616 |
+
"epoch": 18.81708784596871,
|
| 109617 |
+
"grad_norm": Infinity,
|
| 109618 |
+
"learning_rate": 0.0001839128069821245,
|
| 109619 |
+
"loss": 11.1566,
|
| 109620 |
+
"step": 156370
|
| 109621 |
+
},
|
| 109622 |
+
{
|
| 109623 |
+
"epoch": 18.81829121540313,
|
| 109624 |
+
"grad_norm": Infinity,
|
| 109625 |
+
"learning_rate": 0.00018391073765009765,
|
| 109626 |
+
"loss": 11.3208,
|
| 109627 |
+
"step": 156380
|
| 109628 |
+
},
|
| 109629 |
+
{
|
| 109630 |
+
"epoch": 18.819494584837546,
|
| 109631 |
+
"grad_norm": Infinity,
|
| 109632 |
+
"learning_rate": 0.00018390866819663123,
|
| 109633 |
+
"loss": 11.101,
|
| 109634 |
+
"step": 156390
|
| 109635 |
+
},
|
| 109636 |
+
{
|
| 109637 |
+
"epoch": 18.82069795427196,
|
| 109638 |
+
"grad_norm": Infinity,
|
| 109639 |
+
"learning_rate": 0.0001839065986217283,
|
| 109640 |
+
"loss": 11.1122,
|
| 109641 |
+
"step": 156400
|
| 109642 |
+
},
|
| 109643 |
+
{
|
| 109644 |
+
"epoch": 18.821901323706378,
|
| 109645 |
+
"grad_norm": Infinity,
|
| 109646 |
+
"learning_rate": 0.00018390452892539175,
|
| 109647 |
+
"loss": 11.1221,
|
| 109648 |
+
"step": 156410
|
| 109649 |
+
},
|
| 109650 |
+
{
|
| 109651 |
+
"epoch": 18.823104693140795,
|
| 109652 |
+
"grad_norm": Infinity,
|
| 109653 |
+
"learning_rate": 0.00018390245910762464,
|
| 109654 |
+
"loss": 11.084,
|
| 109655 |
+
"step": 156420
|
| 109656 |
+
},
|
| 109657 |
+
{
|
| 109658 |
+
"epoch": 18.82430806257521,
|
| 109659 |
+
"grad_norm": Infinity,
|
| 109660 |
+
"learning_rate": 0.00018390038916842998,
|
| 109661 |
+
"loss": 11.1787,
|
| 109662 |
+
"step": 156430
|
| 109663 |
+
},
|
| 109664 |
+
{
|
| 109665 |
+
"epoch": 18.825511432009627,
|
| 109666 |
+
"grad_norm": Infinity,
|
| 109667 |
+
"learning_rate": 0.00018389831910781073,
|
| 109668 |
+
"loss": 11.1843,
|
| 109669 |
+
"step": 156440
|
| 109670 |
+
},
|
| 109671 |
+
{
|
| 109672 |
+
"epoch": 18.826714801444044,
|
| 109673 |
+
"grad_norm": Infinity,
|
| 109674 |
+
"learning_rate": 0.00018389624892576988,
|
| 109675 |
+
"loss": 11.1098,
|
| 109676 |
+
"step": 156450
|
| 109677 |
+
},
|
| 109678 |
+
{
|
| 109679 |
+
"epoch": 18.82791817087846,
|
| 109680 |
+
"grad_norm": Infinity,
|
| 109681 |
+
"learning_rate": 0.00018389417862231048,
|
| 109682 |
+
"loss": 11.1299,
|
| 109683 |
+
"step": 156460
|
| 109684 |
+
},
|
| 109685 |
+
{
|
| 109686 |
+
"epoch": 18.829121540312876,
|
| 109687 |
+
"grad_norm": Infinity,
|
| 109688 |
+
"learning_rate": 0.00018389210819743545,
|
| 109689 |
+
"loss": 11.0784,
|
| 109690 |
+
"step": 156470
|
| 109691 |
+
},
|
| 109692 |
+
{
|
| 109693 |
+
"epoch": 18.830324909747294,
|
| 109694 |
+
"grad_norm": Infinity,
|
| 109695 |
+
"learning_rate": 0.00018389003765114785,
|
| 109696 |
+
"loss": 11.1469,
|
| 109697 |
+
"step": 156480
|
| 109698 |
+
},
|
| 109699 |
+
{
|
| 109700 |
+
"epoch": 18.831528279181708,
|
| 109701 |
+
"grad_norm": Infinity,
|
| 109702 |
+
"learning_rate": 0.00018388796698345065,
|
| 109703 |
+
"loss": 11.1581,
|
| 109704 |
+
"step": 156490
|
| 109705 |
+
},
|
| 109706 |
+
{
|
| 109707 |
+
"epoch": 18.832731648616125,
|
| 109708 |
+
"grad_norm": Infinity,
|
| 109709 |
+
"learning_rate": 0.00018388589619434687,
|
| 109710 |
+
"loss": 11.0997,
|
| 109711 |
+
"step": 156500
|
| 109712 |
+
},
|
| 109713 |
+
{
|
| 109714 |
+
"epoch": 18.833935018050543,
|
| 109715 |
+
"grad_norm": Infinity,
|
| 109716 |
+
"learning_rate": 0.00018388382528383948,
|
| 109717 |
+
"loss": 11.1977,
|
| 109718 |
+
"step": 156510
|
| 109719 |
+
},
|
| 109720 |
+
{
|
| 109721 |
+
"epoch": 18.835138387484957,
|
| 109722 |
+
"grad_norm": Infinity,
|
| 109723 |
+
"learning_rate": 0.00018388175425193147,
|
| 109724 |
+
"loss": 11.1397,
|
| 109725 |
+
"step": 156520
|
| 109726 |
+
},
|
| 109727 |
+
{
|
| 109728 |
+
"epoch": 18.836341756919374,
|
| 109729 |
+
"grad_norm": Infinity,
|
| 109730 |
+
"learning_rate": 0.00018387968309862583,
|
| 109731 |
+
"loss": 11.3134,
|
| 109732 |
+
"step": 156530
|
| 109733 |
+
},
|
| 109734 |
+
{
|
| 109735 |
+
"epoch": 18.837545126353792,
|
| 109736 |
+
"grad_norm": Infinity,
|
| 109737 |
+
"learning_rate": 0.0001838776118239256,
|
| 109738 |
+
"loss": 11.0607,
|
| 109739 |
+
"step": 156540
|
| 109740 |
+
},
|
| 109741 |
+
{
|
| 109742 |
+
"epoch": 18.838748495788206,
|
| 109743 |
+
"grad_norm": Infinity,
|
| 109744 |
+
"learning_rate": 0.00018387554042783376,
|
| 109745 |
+
"loss": 11.0607,
|
| 109746 |
+
"step": 156550
|
| 109747 |
+
},
|
| 109748 |
+
{
|
| 109749 |
+
"epoch": 18.839951865222623,
|
| 109750 |
+
"grad_norm": Infinity,
|
| 109751 |
+
"learning_rate": 0.00018387346891035332,
|
| 109752 |
+
"loss": 11.1197,
|
| 109753 |
+
"step": 156560
|
| 109754 |
+
},
|
| 109755 |
+
{
|
| 109756 |
+
"epoch": 18.84115523465704,
|
| 109757 |
+
"grad_norm": Infinity,
|
| 109758 |
+
"learning_rate": 0.00018387139727148725,
|
| 109759 |
+
"loss": 11.1981,
|
| 109760 |
+
"step": 156570
|
| 109761 |
+
},
|
| 109762 |
+
{
|
| 109763 |
+
"epoch": 18.842358604091455,
|
| 109764 |
+
"grad_norm": Infinity,
|
| 109765 |
+
"learning_rate": 0.00018386932551123857,
|
| 109766 |
+
"loss": 11.165,
|
| 109767 |
+
"step": 156580
|
| 109768 |
+
},
|
| 109769 |
+
{
|
| 109770 |
+
"epoch": 18.843561973525873,
|
| 109771 |
+
"grad_norm": Infinity,
|
| 109772 |
+
"learning_rate": 0.00018386725362961025,
|
| 109773 |
+
"loss": 11.1773,
|
| 109774 |
+
"step": 156590
|
| 109775 |
+
},
|
| 109776 |
+
{
|
| 109777 |
+
"epoch": 18.84476534296029,
|
| 109778 |
+
"grad_norm": Infinity,
|
| 109779 |
+
"learning_rate": 0.00018386518162660532,
|
| 109780 |
+
"loss": 11.1637,
|
| 109781 |
+
"step": 156600
|
| 109782 |
+
},
|
| 109783 |
+
{
|
| 109784 |
+
"epoch": 18.845968712394704,
|
| 109785 |
+
"grad_norm": Infinity,
|
| 109786 |
+
"learning_rate": 0.00018386310950222677,
|
| 109787 |
+
"loss": 11.1841,
|
| 109788 |
+
"step": 156610
|
| 109789 |
+
},
|
| 109790 |
+
{
|
| 109791 |
+
"epoch": 18.84717208182912,
|
| 109792 |
+
"grad_norm": Infinity,
|
| 109793 |
+
"learning_rate": 0.0001838610372564776,
|
| 109794 |
+
"loss": 11.2402,
|
| 109795 |
+
"step": 156620
|
| 109796 |
+
},
|
| 109797 |
+
{
|
| 109798 |
+
"epoch": 18.84837545126354,
|
| 109799 |
+
"grad_norm": Infinity,
|
| 109800 |
+
"learning_rate": 0.00018385896488936075,
|
| 109801 |
+
"loss": 11.1267,
|
| 109802 |
+
"step": 156630
|
| 109803 |
+
},
|
| 109804 |
+
{
|
| 109805 |
+
"epoch": 18.849578820697953,
|
| 109806 |
+
"grad_norm": Infinity,
|
| 109807 |
+
"learning_rate": 0.00018385689240087936,
|
| 109808 |
+
"loss": 11.2748,
|
| 109809 |
+
"step": 156640
|
| 109810 |
+
},
|
| 109811 |
+
{
|
| 109812 |
+
"epoch": 18.85078219013237,
|
| 109813 |
+
"grad_norm": Infinity,
|
| 109814 |
+
"learning_rate": 0.00018385481979103628,
|
| 109815 |
+
"loss": 11.2116,
|
| 109816 |
+
"step": 156650
|
| 109817 |
+
},
|
| 109818 |
+
{
|
| 109819 |
+
"epoch": 18.85198555956679,
|
| 109820 |
+
"grad_norm": Infinity,
|
| 109821 |
+
"learning_rate": 0.0001838527470598346,
|
| 109822 |
+
"loss": 11.1882,
|
| 109823 |
+
"step": 156660
|
| 109824 |
+
},
|
| 109825 |
+
{
|
| 109826 |
+
"epoch": 18.853188929001202,
|
| 109827 |
+
"grad_norm": Infinity,
|
| 109828 |
+
"learning_rate": 0.0001838506742072773,
|
| 109829 |
+
"loss": 11.1638,
|
| 109830 |
+
"step": 156670
|
| 109831 |
+
},
|
| 109832 |
+
{
|
| 109833 |
+
"epoch": 18.85439229843562,
|
| 109834 |
+
"grad_norm": Infinity,
|
| 109835 |
+
"learning_rate": 0.00018384860123336734,
|
| 109836 |
+
"loss": 11.208,
|
| 109837 |
+
"step": 156680
|
| 109838 |
+
},
|
| 109839 |
+
{
|
| 109840 |
+
"epoch": 18.855595667870038,
|
| 109841 |
+
"grad_norm": Infinity,
|
| 109842 |
+
"learning_rate": 0.00018384652813810778,
|
| 109843 |
+
"loss": 11.1221,
|
| 109844 |
+
"step": 156690
|
| 109845 |
+
},
|
| 109846 |
+
{
|
| 109847 |
+
"epoch": 18.85679903730445,
|
| 109848 |
+
"grad_norm": Infinity,
|
| 109849 |
+
"learning_rate": 0.00018384445492150158,
|
| 109850 |
+
"loss": 11.0525,
|
| 109851 |
+
"step": 156700
|
| 109852 |
+
},
|
| 109853 |
+
{
|
| 109854 |
+
"epoch": 18.85800240673887,
|
| 109855 |
+
"grad_norm": Infinity,
|
| 109856 |
+
"learning_rate": 0.00018384238158355176,
|
| 109857 |
+
"loss": 11.1971,
|
| 109858 |
+
"step": 156710
|
| 109859 |
+
},
|
| 109860 |
+
{
|
| 109861 |
+
"epoch": 18.859205776173287,
|
| 109862 |
+
"grad_norm": Infinity,
|
| 109863 |
+
"learning_rate": 0.00018384030812426132,
|
| 109864 |
+
"loss": 11.1792,
|
| 109865 |
+
"step": 156720
|
| 109866 |
+
},
|
| 109867 |
+
{
|
| 109868 |
+
"epoch": 18.8604091456077,
|
| 109869 |
+
"grad_norm": Infinity,
|
| 109870 |
+
"learning_rate": 0.00018383823454363324,
|
| 109871 |
+
"loss": 11.1946,
|
| 109872 |
+
"step": 156730
|
| 109873 |
+
},
|
| 109874 |
+
{
|
| 109875 |
+
"epoch": 18.86161251504212,
|
| 109876 |
+
"grad_norm": Infinity,
|
| 109877 |
+
"learning_rate": 0.00018383616084167054,
|
| 109878 |
+
"loss": 11.064,
|
| 109879 |
+
"step": 156740
|
| 109880 |
+
},
|
| 109881 |
+
{
|
| 109882 |
+
"epoch": 18.862815884476536,
|
| 109883 |
+
"grad_norm": Infinity,
|
| 109884 |
+
"learning_rate": 0.00018383408701837622,
|
| 109885 |
+
"loss": 11.0675,
|
| 109886 |
+
"step": 156750
|
| 109887 |
+
},
|
| 109888 |
+
{
|
| 109889 |
+
"epoch": 18.86401925391095,
|
| 109890 |
+
"grad_norm": Infinity,
|
| 109891 |
+
"learning_rate": 0.00018383201307375327,
|
| 109892 |
+
"loss": 11.1382,
|
| 109893 |
+
"step": 156760
|
| 109894 |
+
},
|
| 109895 |
+
{
|
| 109896 |
+
"epoch": 18.865222623345367,
|
| 109897 |
+
"grad_norm": Infinity,
|
| 109898 |
+
"learning_rate": 0.0001838299390078047,
|
| 109899 |
+
"loss": 11.054,
|
| 109900 |
+
"step": 156770
|
| 109901 |
+
},
|
| 109902 |
+
{
|
| 109903 |
+
"epoch": 18.866425992779785,
|
| 109904 |
+
"grad_norm": Infinity,
|
| 109905 |
+
"learning_rate": 0.0001838278648205335,
|
| 109906 |
+
"loss": 11.105,
|
| 109907 |
+
"step": 156780
|
| 109908 |
+
},
|
| 109909 |
+
{
|
| 109910 |
+
"epoch": 18.8676293622142,
|
| 109911 |
+
"grad_norm": Infinity,
|
| 109912 |
+
"learning_rate": 0.00018382579051194273,
|
| 109913 |
+
"loss": 11.1738,
|
| 109914 |
+
"step": 156790
|
| 109915 |
+
},
|
| 109916 |
+
{
|
| 109917 |
+
"epoch": 18.868832731648617,
|
| 109918 |
+
"grad_norm": Infinity,
|
| 109919 |
+
"learning_rate": 0.0001838237160820353,
|
| 109920 |
+
"loss": 11.0556,
|
| 109921 |
+
"step": 156800
|
| 109922 |
+
},
|
| 109923 |
+
{
|
| 109924 |
+
"epoch": 18.870036101083034,
|
| 109925 |
+
"grad_norm": Infinity,
|
| 109926 |
+
"learning_rate": 0.00018382164153081426,
|
| 109927 |
+
"loss": 11.1819,
|
| 109928 |
+
"step": 156810
|
| 109929 |
+
},
|
| 109930 |
+
{
|
| 109931 |
+
"epoch": 18.871239470517448,
|
| 109932 |
+
"grad_norm": Infinity,
|
| 109933 |
+
"learning_rate": 0.00018381956685828264,
|
| 109934 |
+
"loss": 11.1954,
|
| 109935 |
+
"step": 156820
|
| 109936 |
+
},
|
| 109937 |
+
{
|
| 109938 |
+
"epoch": 18.872442839951866,
|
| 109939 |
+
"grad_norm": Infinity,
|
| 109940 |
+
"learning_rate": 0.0001838174920644434,
|
| 109941 |
+
"loss": 11.17,
|
| 109942 |
+
"step": 156830
|
| 109943 |
+
},
|
| 109944 |
+
{
|
| 109945 |
+
"epoch": 18.87364620938628,
|
| 109946 |
+
"grad_norm": Infinity,
|
| 109947 |
+
"learning_rate": 0.00018381541714929955,
|
| 109948 |
+
"loss": 11.2456,
|
| 109949 |
+
"step": 156840
|
| 109950 |
+
},
|
| 109951 |
+
{
|
| 109952 |
+
"epoch": 18.874849578820697,
|
| 109953 |
+
"grad_norm": Infinity,
|
| 109954 |
+
"learning_rate": 0.00018381334211285412,
|
| 109955 |
+
"loss": 11.1573,
|
| 109956 |
+
"step": 156850
|
| 109957 |
+
},
|
| 109958 |
+
{
|
| 109959 |
+
"epoch": 18.876052948255115,
|
| 109960 |
+
"grad_norm": Infinity,
|
| 109961 |
+
"learning_rate": 0.00018381126695511004,
|
| 109962 |
+
"loss": 11.1414,
|
| 109963 |
+
"step": 156860
|
| 109964 |
+
},
|
| 109965 |
+
{
|
| 109966 |
+
"epoch": 18.87725631768953,
|
| 109967 |
+
"grad_norm": Infinity,
|
| 109968 |
+
"learning_rate": 0.0001838091916760704,
|
| 109969 |
+
"loss": 11.2085,
|
| 109970 |
+
"step": 156870
|
| 109971 |
+
},
|
| 109972 |
+
{
|
| 109973 |
+
"epoch": 18.878459687123947,
|
| 109974 |
+
"grad_norm": Infinity,
|
| 109975 |
+
"learning_rate": 0.00018380711627573817,
|
| 109976 |
+
"loss": 11.1137,
|
| 109977 |
+
"step": 156880
|
| 109978 |
+
},
|
| 109979 |
+
{
|
| 109980 |
+
"epoch": 18.879663056558364,
|
| 109981 |
+
"grad_norm": Infinity,
|
| 109982 |
+
"learning_rate": 0.00018380504075411633,
|
| 109983 |
+
"loss": 11.1565,
|
| 109984 |
+
"step": 156890
|
| 109985 |
+
},
|
| 109986 |
+
{
|
| 109987 |
+
"epoch": 18.880866425992778,
|
| 109988 |
+
"grad_norm": Infinity,
|
| 109989 |
+
"learning_rate": 0.0001838029651112079,
|
| 109990 |
+
"loss": 11.1592,
|
| 109991 |
+
"step": 156900
|
| 109992 |
+
},
|
| 109993 |
+
{
|
| 109994 |
+
"epoch": 18.882069795427196,
|
| 109995 |
+
"grad_norm": Infinity,
|
| 109996 |
+
"learning_rate": 0.00018380088934701592,
|
| 109997 |
+
"loss": 11.0634,
|
| 109998 |
+
"step": 156910
|
| 109999 |
+
},
|
| 110000 |
+
{
|
| 110001 |
+
"epoch": 18.883273164861613,
|
| 110002 |
+
"grad_norm": Infinity,
|
| 110003 |
+
"learning_rate": 0.00018379881346154337,
|
| 110004 |
+
"loss": 11.1236,
|
| 110005 |
+
"step": 156920
|
| 110006 |
+
},
|
| 110007 |
+
{
|
| 110008 |
+
"epoch": 18.884476534296027,
|
| 110009 |
+
"grad_norm": Infinity,
|
| 110010 |
+
"learning_rate": 0.0001837967374547932,
|
| 110011 |
+
"loss": 11.0946,
|
| 110012 |
+
"step": 156930
|
| 110013 |
+
},
|
| 110014 |
+
{
|
| 110015 |
+
"epoch": 18.885679903730445,
|
| 110016 |
+
"grad_norm": Infinity,
|
| 110017 |
+
"learning_rate": 0.00018379466132676853,
|
| 110018 |
+
"loss": 11.1062,
|
| 110019 |
+
"step": 156940
|
| 110020 |
+
},
|
| 110021 |
+
{
|
| 110022 |
+
"epoch": 18.886883273164862,
|
| 110023 |
+
"grad_norm": Infinity,
|
| 110024 |
+
"learning_rate": 0.00018379258507747223,
|
| 110025 |
+
"loss": 11.0293,
|
| 110026 |
+
"step": 156950
|
| 110027 |
+
},
|
| 110028 |
+
{
|
| 110029 |
+
"epoch": 18.888086642599276,
|
| 110030 |
+
"grad_norm": Infinity,
|
| 110031 |
+
"learning_rate": 0.00018379050870690738,
|
| 110032 |
+
"loss": 11.0878,
|
| 110033 |
+
"step": 156960
|
| 110034 |
+
},
|
| 110035 |
+
{
|
| 110036 |
+
"epoch": 18.889290012033694,
|
| 110037 |
+
"grad_norm": Infinity,
|
| 110038 |
+
"learning_rate": 0.000183788432215077,
|
| 110039 |
+
"loss": 10.9764,
|
| 110040 |
+
"step": 156970
|
| 110041 |
+
},
|
| 110042 |
+
{
|
| 110043 |
+
"epoch": 18.89049338146811,
|
| 110044 |
+
"grad_norm": Infinity,
|
| 110045 |
+
"learning_rate": 0.0001837863556019841,
|
| 110046 |
+
"loss": 11.0772,
|
| 110047 |
+
"step": 156980
|
| 110048 |
+
},
|
| 110049 |
+
{
|
| 110050 |
+
"epoch": 18.891696750902526,
|
| 110051 |
+
"grad_norm": Infinity,
|
| 110052 |
+
"learning_rate": 0.0001837842788676316,
|
| 110053 |
+
"loss": 11.1573,
|
| 110054 |
+
"step": 156990
|
| 110055 |
+
},
|
| 110056 |
+
{
|
| 110057 |
+
"epoch": 18.892900120336943,
|
| 110058 |
+
"grad_norm": Infinity,
|
| 110059 |
+
"learning_rate": 0.0001837822020120226,
|
| 110060 |
+
"loss": 11.0387,
|
| 110061 |
+
"step": 157000
|
| 110062 |
+
},
|
| 110063 |
+
{
|
| 110064 |
+
"epoch": 18.89410348977136,
|
| 110065 |
+
"grad_norm": Infinity,
|
| 110066 |
+
"learning_rate": 0.00018378012503516007,
|
| 110067 |
+
"loss": 11.1678,
|
| 110068 |
+
"step": 157010
|
| 110069 |
+
},
|
| 110070 |
+
{
|
| 110071 |
+
"epoch": 18.895306859205775,
|
| 110072 |
+
"grad_norm": Infinity,
|
| 110073 |
+
"learning_rate": 0.000183778047937047,
|
| 110074 |
+
"loss": 11.1658,
|
| 110075 |
+
"step": 157020
|
| 110076 |
+
},
|
| 110077 |
+
{
|
| 110078 |
+
"epoch": 18.896510228640192,
|
| 110079 |
+
"grad_norm": Infinity,
|
| 110080 |
+
"learning_rate": 0.0001837759707176864,
|
| 110081 |
+
"loss": 11.1512,
|
| 110082 |
+
"step": 157030
|
| 110083 |
+
},
|
| 110084 |
+
{
|
| 110085 |
+
"epoch": 18.89771359807461,
|
| 110086 |
+
"grad_norm": Infinity,
|
| 110087 |
+
"learning_rate": 0.0001837738933770813,
|
| 110088 |
+
"loss": 11.0356,
|
| 110089 |
+
"step": 157040
|
| 110090 |
+
},
|
| 110091 |
+
{
|
| 110092 |
+
"epoch": 18.898916967509024,
|
| 110093 |
+
"grad_norm": Infinity,
|
| 110094 |
+
"learning_rate": 0.00018377181591523468,
|
| 110095 |
+
"loss": 11.2211,
|
| 110096 |
+
"step": 157050
|
| 110097 |
+
},
|
| 110098 |
+
{
|
| 110099 |
+
"epoch": 18.90012033694344,
|
| 110100 |
+
"grad_norm": Infinity,
|
| 110101 |
+
"learning_rate": 0.0001837697383321496,
|
| 110102 |
+
"loss": 11.0747,
|
| 110103 |
+
"step": 157060
|
| 110104 |
+
},
|
| 110105 |
+
{
|
| 110106 |
+
"epoch": 18.90132370637786,
|
| 110107 |
+
"grad_norm": Infinity,
|
| 110108 |
+
"learning_rate": 0.00018376766062782896,
|
| 110109 |
+
"loss": 11.206,
|
| 110110 |
+
"step": 157070
|
| 110111 |
+
},
|
| 110112 |
+
{
|
| 110113 |
+
"epoch": 18.902527075812273,
|
| 110114 |
+
"grad_norm": Infinity,
|
| 110115 |
+
"learning_rate": 0.00018376558280227588,
|
| 110116 |
+
"loss": 11.1095,
|
| 110117 |
+
"step": 157080
|
| 110118 |
+
},
|
| 110119 |
+
{
|
| 110120 |
+
"epoch": 18.90373044524669,
|
| 110121 |
+
"grad_norm": Infinity,
|
| 110122 |
+
"learning_rate": 0.00018376350485549333,
|
| 110123 |
+
"loss": 11.1608,
|
| 110124 |
+
"step": 157090
|
| 110125 |
+
},
|
| 110126 |
+
{
|
| 110127 |
+
"epoch": 18.904933814681108,
|
| 110128 |
+
"grad_norm": Infinity,
|
| 110129 |
+
"learning_rate": 0.00018376142678748426,
|
| 110130 |
+
"loss": 11.0815,
|
| 110131 |
+
"step": 157100
|
| 110132 |
+
},
|
| 110133 |
+
{
|
| 110134 |
+
"epoch": 18.906137184115522,
|
| 110135 |
+
"grad_norm": Infinity,
|
| 110136 |
+
"learning_rate": 0.00018375934859825177,
|
| 110137 |
+
"loss": 11.0471,
|
| 110138 |
+
"step": 157110
|
| 110139 |
+
},
|
| 110140 |
+
{
|
| 110141 |
+
"epoch": 18.90734055354994,
|
| 110142 |
+
"grad_norm": Infinity,
|
| 110143 |
+
"learning_rate": 0.0001837572702877988,
|
| 110144 |
+
"loss": 11.2041,
|
| 110145 |
+
"step": 157120
|
| 110146 |
+
},
|
| 110147 |
+
{
|
| 110148 |
+
"epoch": 18.908543922984357,
|
| 110149 |
+
"grad_norm": Infinity,
|
| 110150 |
+
"learning_rate": 0.0001837551918561284,
|
| 110151 |
+
"loss": 11.1298,
|
| 110152 |
+
"step": 157130
|
| 110153 |
+
},
|
| 110154 |
+
{
|
| 110155 |
+
"epoch": 18.90974729241877,
|
| 110156 |
+
"grad_norm": Infinity,
|
| 110157 |
+
"learning_rate": 0.00018375311330324356,
|
| 110158 |
+
"loss": 11.1328,
|
| 110159 |
+
"step": 157140
|
| 110160 |
+
},
|
| 110161 |
+
{
|
| 110162 |
+
"epoch": 18.91095066185319,
|
| 110163 |
+
"grad_norm": Infinity,
|
| 110164 |
+
"learning_rate": 0.00018375103462914727,
|
| 110165 |
+
"loss": 11.069,
|
| 110166 |
+
"step": 157150
|
| 110167 |
+
},
|
| 110168 |
+
{
|
| 110169 |
+
"epoch": 18.912154031287606,
|
| 110170 |
+
"grad_norm": Infinity,
|
| 110171 |
+
"learning_rate": 0.00018374895583384254,
|
| 110172 |
+
"loss": 11.1397,
|
| 110173 |
+
"step": 157160
|
| 110174 |
+
},
|
| 110175 |
+
{
|
| 110176 |
+
"epoch": 18.91335740072202,
|
| 110177 |
+
"grad_norm": Infinity,
|
| 110178 |
+
"learning_rate": 0.00018374687691733244,
|
| 110179 |
+
"loss": 11.1763,
|
| 110180 |
+
"step": 157170
|
| 110181 |
+
},
|
| 110182 |
+
{
|
| 110183 |
+
"epoch": 18.914560770156438,
|
| 110184 |
+
"grad_norm": Infinity,
|
| 110185 |
+
"learning_rate": 0.00018374479787961989,
|
| 110186 |
+
"loss": 11.2508,
|
| 110187 |
+
"step": 157180
|
| 110188 |
+
},
|
| 110189 |
+
{
|
| 110190 |
+
"epoch": 18.915764139590856,
|
| 110191 |
+
"grad_norm": Infinity,
|
| 110192 |
+
"learning_rate": 0.00018374271872070795,
|
| 110193 |
+
"loss": 11.2083,
|
| 110194 |
+
"step": 157190
|
| 110195 |
+
},
|
| 110196 |
+
{
|
| 110197 |
+
"epoch": 18.91696750902527,
|
| 110198 |
+
"grad_norm": Infinity,
|
| 110199 |
+
"learning_rate": 0.00018374063944059964,
|
| 110200 |
+
"loss": 11.1163,
|
| 110201 |
+
"step": 157200
|
| 110202 |
+
},
|
| 110203 |
+
{
|
| 110204 |
+
"epoch": 18.918170878459687,
|
| 110205 |
+
"grad_norm": Infinity,
|
| 110206 |
+
"learning_rate": 0.00018373856003929794,
|
| 110207 |
+
"loss": 11.1841,
|
| 110208 |
+
"step": 157210
|
| 110209 |
+
},
|
| 110210 |
+
{
|
| 110211 |
+
"epoch": 18.919374247894105,
|
| 110212 |
+
"grad_norm": Infinity,
|
| 110213 |
+
"learning_rate": 0.0001837364805168059,
|
| 110214 |
+
"loss": 11.0735,
|
| 110215 |
+
"step": 157220
|
| 110216 |
+
},
|
| 110217 |
+
{
|
| 110218 |
+
"epoch": 18.92057761732852,
|
| 110219 |
+
"grad_norm": Infinity,
|
| 110220 |
+
"learning_rate": 0.00018373440087312643,
|
| 110221 |
+
"loss": 11.1616,
|
| 110222 |
+
"step": 157230
|
| 110223 |
+
},
|
| 110224 |
+
{
|
| 110225 |
+
"epoch": 18.921780986762936,
|
| 110226 |
+
"grad_norm": Infinity,
|
| 110227 |
+
"learning_rate": 0.00018373232110826266,
|
| 110228 |
+
"loss": 11.1764,
|
| 110229 |
+
"step": 157240
|
| 110230 |
+
},
|
| 110231 |
+
{
|
| 110232 |
+
"epoch": 18.922984356197354,
|
| 110233 |
+
"grad_norm": Infinity,
|
| 110234 |
+
"learning_rate": 0.00018373024122221753,
|
| 110235 |
+
"loss": 11.16,
|
| 110236 |
+
"step": 157250
|
| 110237 |
+
},
|
| 110238 |
+
{
|
| 110239 |
+
"epoch": 18.924187725631768,
|
| 110240 |
+
"grad_norm": Infinity,
|
| 110241 |
+
"learning_rate": 0.00018372816121499405,
|
| 110242 |
+
"loss": 11.0687,
|
| 110243 |
+
"step": 157260
|
| 110244 |
+
},
|
| 110245 |
+
{
|
| 110246 |
+
"epoch": 18.925391095066185,
|
| 110247 |
+
"grad_norm": Infinity,
|
| 110248 |
+
"learning_rate": 0.00018372608108659528,
|
| 110249 |
+
"loss": 11.1236,
|
| 110250 |
+
"step": 157270
|
| 110251 |
+
},
|
| 110252 |
+
{
|
| 110253 |
+
"epoch": 18.926594464500603,
|
| 110254 |
+
"grad_norm": Infinity,
|
| 110255 |
+
"learning_rate": 0.00018372400083702418,
|
| 110256 |
+
"loss": 11.1659,
|
| 110257 |
+
"step": 157280
|
| 110258 |
+
},
|
| 110259 |
+
{
|
| 110260 |
+
"epoch": 18.927797833935017,
|
| 110261 |
+
"grad_norm": Infinity,
|
| 110262 |
+
"learning_rate": 0.0001837219204662838,
|
| 110263 |
+
"loss": 11.0458,
|
| 110264 |
+
"step": 157290
|
| 110265 |
+
},
|
| 110266 |
+
{
|
| 110267 |
+
"epoch": 18.929001203369435,
|
| 110268 |
+
"grad_norm": Infinity,
|
| 110269 |
+
"learning_rate": 0.00018371983997437713,
|
| 110270 |
+
"loss": 11.0637,
|
| 110271 |
+
"step": 157300
|
| 110272 |
+
},
|
| 110273 |
+
{
|
| 110274 |
+
"epoch": 18.930204572803852,
|
| 110275 |
+
"grad_norm": Infinity,
|
| 110276 |
+
"learning_rate": 0.00018371775936130717,
|
| 110277 |
+
"loss": 11.1876,
|
| 110278 |
+
"step": 157310
|
| 110279 |
+
},
|
| 110280 |
+
{
|
| 110281 |
+
"epoch": 18.931407942238266,
|
| 110282 |
+
"grad_norm": Infinity,
|
| 110283 |
+
"learning_rate": 0.00018371567862707694,
|
| 110284 |
+
"loss": 11.2099,
|
| 110285 |
+
"step": 157320
|
| 110286 |
+
},
|
| 110287 |
+
{
|
| 110288 |
+
"epoch": 18.932611311672684,
|
| 110289 |
+
"grad_norm": Infinity,
|
| 110290 |
+
"learning_rate": 0.00018371359777168947,
|
| 110291 |
+
"loss": 11.1799,
|
| 110292 |
+
"step": 157330
|
| 110293 |
+
},
|
| 110294 |
+
{
|
| 110295 |
+
"epoch": 18.9338146811071,
|
| 110296 |
+
"grad_norm": Infinity,
|
| 110297 |
+
"learning_rate": 0.00018371151679514775,
|
| 110298 |
+
"loss": 11.0395,
|
| 110299 |
+
"step": 157340
|
| 110300 |
+
},
|
| 110301 |
+
{
|
| 110302 |
+
"epoch": 18.935018050541515,
|
| 110303 |
+
"grad_norm": Infinity,
|
| 110304 |
+
"learning_rate": 0.0001837094356974548,
|
| 110305 |
+
"loss": 11.1093,
|
| 110306 |
+
"step": 157350
|
| 110307 |
+
},
|
| 110308 |
+
{
|
| 110309 |
+
"epoch": 18.936221419975933,
|
| 110310 |
+
"grad_norm": Infinity,
|
| 110311 |
+
"learning_rate": 0.0001837073544786136,
|
| 110312 |
+
"loss": 11.1713,
|
| 110313 |
+
"step": 157360
|
| 110314 |
+
},
|
| 110315 |
+
{
|
| 110316 |
+
"epoch": 18.93742478941035,
|
| 110317 |
+
"grad_norm": Infinity,
|
| 110318 |
+
"learning_rate": 0.00018370527313862724,
|
| 110319 |
+
"loss": 11.2034,
|
| 110320 |
+
"step": 157370
|
| 110321 |
+
},
|
| 110322 |
+
{
|
| 110323 |
+
"epoch": 18.938628158844764,
|
| 110324 |
+
"grad_norm": Infinity,
|
| 110325 |
+
"learning_rate": 0.00018370319167749865,
|
| 110326 |
+
"loss": 11.1406,
|
| 110327 |
+
"step": 157380
|
| 110328 |
+
},
|
| 110329 |
+
{
|
| 110330 |
+
"epoch": 18.939831528279182,
|
| 110331 |
+
"grad_norm": Infinity,
|
| 110332 |
+
"learning_rate": 0.00018370111009523087,
|
| 110333 |
+
"loss": 11.2147,
|
| 110334 |
+
"step": 157390
|
| 110335 |
+
},
|
| 110336 |
+
{
|
| 110337 |
+
"epoch": 18.9410348977136,
|
| 110338 |
+
"grad_norm": Infinity,
|
| 110339 |
+
"learning_rate": 0.00018369902839182695,
|
| 110340 |
+
"loss": 11.2302,
|
| 110341 |
+
"step": 157400
|
| 110342 |
+
},
|
| 110343 |
+
{
|
| 110344 |
+
"epoch": 18.942238267148014,
|
| 110345 |
+
"grad_norm": Infinity,
|
| 110346 |
+
"learning_rate": 0.00018369694656728984,
|
| 110347 |
+
"loss": 11.1735,
|
| 110348 |
+
"step": 157410
|
| 110349 |
+
},
|
| 110350 |
+
{
|
| 110351 |
+
"epoch": 18.94344163658243,
|
| 110352 |
+
"grad_norm": Infinity,
|
| 110353 |
+
"learning_rate": 0.00018369486462162262,
|
| 110354 |
+
"loss": 11.1265,
|
| 110355 |
+
"step": 157420
|
| 110356 |
+
},
|
| 110357 |
+
{
|
| 110358 |
+
"epoch": 18.94464500601685,
|
| 110359 |
+
"grad_norm": Infinity,
|
| 110360 |
+
"learning_rate": 0.00018369278255482826,
|
| 110361 |
+
"loss": 11.1389,
|
| 110362 |
+
"step": 157430
|
| 110363 |
+
},
|
| 110364 |
+
{
|
| 110365 |
+
"epoch": 18.945848375451263,
|
| 110366 |
+
"grad_norm": Infinity,
|
| 110367 |
+
"learning_rate": 0.00018369070036690976,
|
| 110368 |
+
"loss": 11.1313,
|
| 110369 |
+
"step": 157440
|
| 110370 |
+
},
|
| 110371 |
+
{
|
| 110372 |
+
"epoch": 18.94705174488568,
|
| 110373 |
+
"grad_norm": Infinity,
|
| 110374 |
+
"learning_rate": 0.00018368861805787018,
|
| 110375 |
+
"loss": 11.151,
|
| 110376 |
+
"step": 157450
|
| 110377 |
+
},
|
| 110378 |
+
{
|
| 110379 |
+
"epoch": 18.948255114320098,
|
| 110380 |
+
"grad_norm": Infinity,
|
| 110381 |
+
"learning_rate": 0.00018368653562771245,
|
| 110382 |
+
"loss": 11.1124,
|
| 110383 |
+
"step": 157460
|
| 110384 |
+
},
|
| 110385 |
+
{
|
| 110386 |
+
"epoch": 18.949458483754512,
|
| 110387 |
+
"grad_norm": Infinity,
|
| 110388 |
+
"learning_rate": 0.00018368445307643969,
|
| 110389 |
+
"loss": 11.1773,
|
| 110390 |
+
"step": 157470
|
| 110391 |
+
},
|
| 110392 |
+
{
|
| 110393 |
+
"epoch": 18.95066185318893,
|
| 110394 |
+
"grad_norm": Infinity,
|
| 110395 |
+
"learning_rate": 0.00018368237040405485,
|
| 110396 |
+
"loss": 11.0857,
|
| 110397 |
+
"step": 157480
|
| 110398 |
+
},
|
| 110399 |
+
{
|
| 110400 |
+
"epoch": 18.951865222623347,
|
| 110401 |
+
"grad_norm": Infinity,
|
| 110402 |
+
"learning_rate": 0.00018368028761056094,
|
| 110403 |
+
"loss": 11.0975,
|
| 110404 |
+
"step": 157490
|
| 110405 |
+
},
|
| 110406 |
+
{
|
| 110407 |
+
"epoch": 18.95306859205776,
|
| 110408 |
+
"grad_norm": Infinity,
|
| 110409 |
+
"learning_rate": 0.00018367820469596105,
|
| 110410 |
+
"loss": 11.022,
|
| 110411 |
+
"step": 157500
|
| 110412 |
+
},
|
| 110413 |
+
{
|
| 110414 |
+
"epoch": 18.95427196149218,
|
| 110415 |
+
"grad_norm": Infinity,
|
| 110416 |
+
"learning_rate": 0.0001836761216602581,
|
| 110417 |
+
"loss": 11.2443,
|
| 110418 |
+
"step": 157510
|
| 110419 |
+
},
|
| 110420 |
+
{
|
| 110421 |
+
"epoch": 18.955475330926596,
|
| 110422 |
+
"grad_norm": Infinity,
|
| 110423 |
+
"learning_rate": 0.00018367403850345516,
|
| 110424 |
+
"loss": 11.1917,
|
| 110425 |
+
"step": 157520
|
| 110426 |
+
},
|
| 110427 |
+
{
|
| 110428 |
+
"epoch": 18.95667870036101,
|
| 110429 |
+
"grad_norm": Infinity,
|
| 110430 |
+
"learning_rate": 0.0001836719552255552,
|
| 110431 |
+
"loss": 11.022,
|
| 110432 |
+
"step": 157530
|
| 110433 |
+
},
|
| 110434 |
+
{
|
| 110435 |
+
"epoch": 18.957882069795428,
|
| 110436 |
+
"grad_norm": Infinity,
|
| 110437 |
+
"learning_rate": 0.00018366987182656127,
|
| 110438 |
+
"loss": 11.2335,
|
| 110439 |
+
"step": 157540
|
| 110440 |
+
},
|
| 110441 |
+
{
|
| 110442 |
+
"epoch": 18.959085439229845,
|
| 110443 |
+
"grad_norm": Infinity,
|
| 110444 |
+
"learning_rate": 0.0001836677883064764,
|
| 110445 |
+
"loss": 11.2388,
|
| 110446 |
+
"step": 157550
|
| 110447 |
+
},
|
| 110448 |
+
{
|
| 110449 |
+
"epoch": 18.96028880866426,
|
| 110450 |
+
"grad_norm": Infinity,
|
| 110451 |
+
"learning_rate": 0.00018366570466530358,
|
| 110452 |
+
"loss": 11.0706,
|
| 110453 |
+
"step": 157560
|
| 110454 |
+
},
|
| 110455 |
+
{
|
| 110456 |
+
"epoch": 18.961492178098677,
|
| 110457 |
+
"grad_norm": Infinity,
|
| 110458 |
+
"learning_rate": 0.0001836636209030458,
|
| 110459 |
+
"loss": 11.1347,
|
| 110460 |
+
"step": 157570
|
| 110461 |
+
},
|
| 110462 |
+
{
|
| 110463 |
+
"epoch": 18.96269554753309,
|
| 110464 |
+
"grad_norm": Infinity,
|
| 110465 |
+
"learning_rate": 0.00018366153701970613,
|
| 110466 |
+
"loss": 11.1475,
|
| 110467 |
+
"step": 157580
|
| 110468 |
+
},
|
| 110469 |
+
{
|
| 110470 |
+
"epoch": 18.96389891696751,
|
| 110471 |
+
"grad_norm": Infinity,
|
| 110472 |
+
"learning_rate": 0.00018365945301528758,
|
| 110473 |
+
"loss": 11.1984,
|
| 110474 |
+
"step": 157590
|
| 110475 |
+
},
|
| 110476 |
+
{
|
| 110477 |
+
"epoch": 18.965102286401926,
|
| 110478 |
+
"grad_norm": Infinity,
|
| 110479 |
+
"learning_rate": 0.00018365736888979314,
|
| 110480 |
+
"loss": 11.1857,
|
| 110481 |
+
"step": 157600
|
| 110482 |
+
},
|
| 110483 |
+
{
|
| 110484 |
+
"epoch": 18.96630565583634,
|
| 110485 |
+
"grad_norm": Infinity,
|
| 110486 |
+
"learning_rate": 0.00018365528464322578,
|
| 110487 |
+
"loss": 11.0506,
|
| 110488 |
+
"step": 157610
|
| 110489 |
+
},
|
| 110490 |
+
{
|
| 110491 |
+
"epoch": 18.967509025270758,
|
| 110492 |
+
"grad_norm": Infinity,
|
| 110493 |
+
"learning_rate": 0.0001836532002755886,
|
| 110494 |
+
"loss": 11.2283,
|
| 110495 |
+
"step": 157620
|
| 110496 |
+
},
|
| 110497 |
+
{
|
| 110498 |
+
"epoch": 18.968712394705175,
|
| 110499 |
+
"grad_norm": Infinity,
|
| 110500 |
+
"learning_rate": 0.0001836511157868846,
|
| 110501 |
+
"loss": 11.2945,
|
| 110502 |
+
"step": 157630
|
| 110503 |
+
},
|
| 110504 |
+
{
|
| 110505 |
+
"epoch": 18.96991576413959,
|
| 110506 |
+
"grad_norm": Infinity,
|
| 110507 |
+
"learning_rate": 0.00018364903117711677,
|
| 110508 |
+
"loss": 11.0815,
|
| 110509 |
+
"step": 157640
|
| 110510 |
+
},
|
| 110511 |
+
{
|
| 110512 |
+
"epoch": 18.971119133574007,
|
| 110513 |
+
"grad_norm": Infinity,
|
| 110514 |
+
"learning_rate": 0.00018364694644628816,
|
| 110515 |
+
"loss": 11.174,
|
| 110516 |
+
"step": 157650
|
| 110517 |
+
},
|
| 110518 |
+
{
|
| 110519 |
+
"epoch": 18.972322503008424,
|
| 110520 |
+
"grad_norm": Infinity,
|
| 110521 |
+
"learning_rate": 0.00018364486159440173,
|
| 110522 |
+
"loss": 11.0937,
|
| 110523 |
+
"step": 157660
|
| 110524 |
+
},
|
| 110525 |
+
{
|
| 110526 |
+
"epoch": 18.97352587244284,
|
| 110527 |
+
"grad_norm": Infinity,
|
| 110528 |
+
"learning_rate": 0.00018364277662146056,
|
| 110529 |
+
"loss": 11.1016,
|
| 110530 |
+
"step": 157670
|
| 110531 |
+
},
|
| 110532 |
+
{
|
| 110533 |
+
"epoch": 18.974729241877256,
|
| 110534 |
+
"grad_norm": Infinity,
|
| 110535 |
+
"learning_rate": 0.00018364069152746764,
|
| 110536 |
+
"loss": 11.2318,
|
| 110537 |
+
"step": 157680
|
| 110538 |
+
},
|
| 110539 |
+
{
|
| 110540 |
+
"epoch": 18.975932611311674,
|
| 110541 |
+
"grad_norm": Infinity,
|
| 110542 |
+
"learning_rate": 0.00018363860631242599,
|
| 110543 |
+
"loss": 11.0833,
|
| 110544 |
+
"step": 157690
|
| 110545 |
+
},
|
| 110546 |
+
{
|
| 110547 |
+
"epoch": 18.977135980746088,
|
| 110548 |
+
"grad_norm": Infinity,
|
| 110549 |
+
"learning_rate": 0.0001836365209763386,
|
| 110550 |
+
"loss": 11.0869,
|
| 110551 |
+
"step": 157700
|
| 110552 |
+
},
|
| 110553 |
+
{
|
| 110554 |
+
"epoch": 18.978339350180505,
|
| 110555 |
+
"grad_norm": Infinity,
|
| 110556 |
+
"learning_rate": 0.00018363443551920857,
|
| 110557 |
+
"loss": 11.23,
|
| 110558 |
+
"step": 157710
|
| 110559 |
+
},
|
| 110560 |
+
{
|
| 110561 |
+
"epoch": 18.979542719614923,
|
| 110562 |
+
"grad_norm": Infinity,
|
| 110563 |
+
"learning_rate": 0.0001836323499410388,
|
| 110564 |
+
"loss": 11.1504,
|
| 110565 |
+
"step": 157720
|
| 110566 |
+
},
|
| 110567 |
+
{
|
| 110568 |
+
"epoch": 18.980746089049337,
|
| 110569 |
+
"grad_norm": Infinity,
|
| 110570 |
+
"learning_rate": 0.0001836302642418324,
|
| 110571 |
+
"loss": 11.1502,
|
| 110572 |
+
"step": 157730
|
| 110573 |
+
},
|
| 110574 |
+
{
|
| 110575 |
+
"epoch": 18.981949458483754,
|
| 110576 |
+
"grad_norm": Infinity,
|
| 110577 |
+
"learning_rate": 0.00018362817842159236,
|
| 110578 |
+
"loss": 11.1727,
|
| 110579 |
+
"step": 157740
|
| 110580 |
+
},
|
| 110581 |
+
{
|
| 110582 |
+
"epoch": 18.983152827918172,
|
| 110583 |
+
"grad_norm": Infinity,
|
| 110584 |
+
"learning_rate": 0.0001836260924803217,
|
| 110585 |
+
"loss": 11.0221,
|
| 110586 |
+
"step": 157750
|
| 110587 |
+
},
|
| 110588 |
+
{
|
| 110589 |
+
"epoch": 18.984356197352586,
|
| 110590 |
+
"grad_norm": Infinity,
|
| 110591 |
+
"learning_rate": 0.0001836240064180234,
|
| 110592 |
+
"loss": 11.1395,
|
| 110593 |
+
"step": 157760
|
| 110594 |
+
},
|
| 110595 |
+
{
|
| 110596 |
+
"epoch": 18.985559566787003,
|
| 110597 |
+
"grad_norm": Infinity,
|
| 110598 |
+
"learning_rate": 0.00018362192023470058,
|
| 110599 |
+
"loss": 11.0751,
|
| 110600 |
+
"step": 157770
|
| 110601 |
+
},
|
| 110602 |
+
{
|
| 110603 |
+
"epoch": 18.98676293622142,
|
| 110604 |
+
"grad_norm": Infinity,
|
| 110605 |
+
"learning_rate": 0.00018361983393035618,
|
| 110606 |
+
"loss": 11.1112,
|
| 110607 |
+
"step": 157780
|
| 110608 |
+
},
|
| 110609 |
+
{
|
| 110610 |
+
"epoch": 18.987966305655835,
|
| 110611 |
+
"grad_norm": Infinity,
|
| 110612 |
+
"learning_rate": 0.00018361774750499322,
|
| 110613 |
+
"loss": 11.0373,
|
| 110614 |
+
"step": 157790
|
| 110615 |
+
},
|
| 110616 |
+
{
|
| 110617 |
+
"epoch": 18.989169675090253,
|
| 110618 |
+
"grad_norm": Infinity,
|
| 110619 |
+
"learning_rate": 0.00018361566095861473,
|
| 110620 |
+
"loss": 11.0098,
|
| 110621 |
+
"step": 157800
|
| 110622 |
+
},
|
| 110623 |
+
{
|
| 110624 |
+
"epoch": 18.99037304452467,
|
| 110625 |
+
"grad_norm": Infinity,
|
| 110626 |
+
"learning_rate": 0.00018361357429122377,
|
| 110627 |
+
"loss": 11.1946,
|
| 110628 |
+
"step": 157810
|
| 110629 |
+
},
|
| 110630 |
+
{
|
| 110631 |
+
"epoch": 18.991576413959084,
|
| 110632 |
+
"grad_norm": Infinity,
|
| 110633 |
+
"learning_rate": 0.0001836114875028233,
|
| 110634 |
+
"loss": 11.1975,
|
| 110635 |
+
"step": 157820
|
| 110636 |
+
},
|
| 110637 |
+
{
|
| 110638 |
+
"epoch": 18.9927797833935,
|
| 110639 |
+
"grad_norm": Infinity,
|
| 110640 |
+
"learning_rate": 0.00018360940059341635,
|
| 110641 |
+
"loss": 11.2993,
|
| 110642 |
+
"step": 157830
|
| 110643 |
+
},
|
| 110644 |
+
{
|
| 110645 |
+
"epoch": 18.99398315282792,
|
| 110646 |
+
"grad_norm": Infinity,
|
| 110647 |
+
"learning_rate": 0.00018360731356300596,
|
| 110648 |
+
"loss": 11.138,
|
| 110649 |
+
"step": 157840
|
| 110650 |
+
},
|
| 110651 |
+
{
|
| 110652 |
+
"epoch": 18.995186522262333,
|
| 110653 |
+
"grad_norm": Infinity,
|
| 110654 |
+
"learning_rate": 0.00018360522641159516,
|
| 110655 |
+
"loss": 11.2325,
|
| 110656 |
+
"step": 157850
|
| 110657 |
+
},
|
| 110658 |
+
{
|
| 110659 |
+
"epoch": 18.99638989169675,
|
| 110660 |
+
"grad_norm": Infinity,
|
| 110661 |
+
"learning_rate": 0.00018360313913918694,
|
| 110662 |
+
"loss": 11.1623,
|
| 110663 |
+
"step": 157860
|
| 110664 |
+
},
|
| 110665 |
+
{
|
| 110666 |
+
"epoch": 18.99759326113117,
|
| 110667 |
+
"grad_norm": Infinity,
|
| 110668 |
+
"learning_rate": 0.00018360105174578437,
|
| 110669 |
+
"loss": 11.0993,
|
| 110670 |
+
"step": 157870
|
| 110671 |
+
},
|
| 110672 |
+
{
|
| 110673 |
+
"epoch": 18.998796630565582,
|
| 110674 |
+
"grad_norm": Infinity,
|
| 110675 |
+
"learning_rate": 0.0001835989642313904,
|
| 110676 |
+
"loss": 11.1284,
|
| 110677 |
+
"step": 157880
|
| 110678 |
+
},
|
| 110679 |
+
{
|
| 110680 |
+
"epoch": 19.0,
|
| 110681 |
+
"grad_norm": Infinity,
|
| 110682 |
+
"learning_rate": 0.0001835968765960081,
|
| 110683 |
+
"loss": 11.1654,
|
| 110684 |
+
"step": 157890
|
| 110685 |
+
},
|
| 110686 |
+
{
|
| 110687 |
+
"epoch": 19.0,
|
| 110688 |
+
"eval_loss": 11.139603614807129,
|
| 110689 |
+
"eval_runtime": 118.4351,
|
| 110690 |
+
"eval_samples_per_second": 62.372,
|
| 110691 |
+
"eval_steps_per_second": 7.802,
|
| 110692 |
+
"step": 157890
|
| 110693 |
}
|
| 110694 |
],
|
| 110695 |
"logging_steps": 10,
|
|
|
|
| 110709 |
"attributes": {}
|
| 110710 |
}
|
| 110711 |
},
|
| 110712 |
+
"total_flos": 3.3196933925675467e+19,
|
| 110713 |
"train_batch_size": 1,
|
| 110714 |
"trial_name": null,
|
| 110715 |
"trial_params": null
|