checkpoint step 500
Browse files- step_500/config.json +14 -0
- step_500/model.pt +3 -0
- step_500/optimizer.pt +3 -0
- step_500/scaler.pt +3 -0
- step_500/scheduler.pt +3 -0
step_500/config.json
ADDED
|
@@ -0,0 +1,14 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"vocab_size": 256,
|
| 3 |
+
"d_model": 384,
|
| 4 |
+
"n_layers": 6,
|
| 5 |
+
"dropout": 0.1,
|
| 6 |
+
"max_len": 512,
|
| 7 |
+
"lr": 0.0003,
|
| 8 |
+
"warmup": 500,
|
| 9 |
+
"batch": 16,
|
| 10 |
+
"block": 512,
|
| 11 |
+
"steps": 20000,
|
| 12 |
+
"world_size": 1,
|
| 13 |
+
"step": 500
|
| 14 |
+
}
|
step_500/model.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:db6a78832de7d06d7144fef6d2ad4c72397a92ea109d42503e6282229e5c33fa
|
| 3 |
+
size 25761151
|
step_500/optimizer.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e514d111cd7fc51962fffeab8b45dc83f4ec15421a57acce01f7e58956fee545
|
| 3 |
+
size 43838795
|
step_500/scaler.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f77569c2e850b04af982cc8c1389f1430851448915c593b69e5da36ce05b71d7
|
| 3 |
+
size 1383
|
step_500/scheduler.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:384697a1cda217834318d12a9b7bb596cbba9323eebe93ef0194b448e2e9e05d
|
| 3 |
+
size 1465
|