alexandretl
commited on
Commit
•
f5cc918
1
Parent(s):
9d5a057
Upload folder using huggingface_hub
Browse files- runs/good-glade-64/ckpt_0/model.pth +3 -0
- runs/good-glade-64/ckpt_20000/model.pth +3 -0
- runs/good-glade-64/ckpt_40000/model.pth +3 -0
- runs/good-glade-64/ckpt_60000/model.pth +3 -0
- runs/good-glade-64/ckpt_80000_before_cooldown/model.pth +3 -0
- runs/good-glade-64/config.json +1 -0
- runs/good-glade-64/model.pth +3 -0
runs/good-glade-64/ckpt_0/model.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:4b9079873ee04ea2cdbe1374f7bfa3a1afe988a7a34a21140b6ba4512df0cd15
|
3 |
+
size 2191206134
|
runs/good-glade-64/ckpt_20000/model.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:7a54bf06182de7ea3ad8ebe99486a5d7dfe2760b316eec7b3cc3f760ac55f8f0
|
3 |
+
size 2191206134
|
runs/good-glade-64/ckpt_40000/model.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:1969421b80ad2dafe68df3b248a5f359359af7f9559ebfea9cfce7f01f044577
|
3 |
+
size 2191206134
|
runs/good-glade-64/ckpt_60000/model.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:a239401a6f595a7b6a3dc4f052ca7591a0d80aa2296d2f15f9fafc5efc147c65
|
3 |
+
size 2191206134
|
runs/good-glade-64/ckpt_80000_before_cooldown/model.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:cc9ea7be9b725771d1ad2b9d3c634c7a2f53ce35b3283fc222e0bec11ba27bf5
|
3 |
+
size 2191206134
|
runs/good-glade-64/config.json
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
{"d_model": 1024, "n_layers": 12, "n_heads": 16, "max_len": 512, "dropout": 0.0, "bias": false, "norm_eps": 1e-05, "base_std": 0.02, "d_ff": 3584, "n_kv_heads": 16, "optimised_attn": false, "efficient_attn": false, "super_attn": false, "pos_emb": "rope", "rope_theta": 10000, "mup": false, "mup_base_width": 288, "flash": true, "architecture": "Transformer"}
|
runs/good-glade-64/model.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:6e7172dbefb354ecf55996367fb43724ce56d9e9a8187cc7940975458fb5d315
|
3 |
+
size 2191206134
|