step 10000 | val_ppl=10.24

README.md

@@ -25,9 +25,9 @@ Small language model (3.0M parameters) trained from scratch.
 | Total parameters | 3.000M |
 
 ## Training
-- Tokens seen: 10,056,499,200
-- Val loss: 2.2615
-- Val PPL: 9.60
+- Tokens seen: 10,475,520,000
+- Val loss: 2.3259
+- Val PPL: 10.24
 
 ## Usage
 ```python

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:9b1dc3bf192434e1bdfd5ea55faddc192d1abdb6092ab2ed859687ac0babcb5c
+oid sha256:3ede254a139363308710178b96aaa4f19b503b34c20128f9d3fd8d3fe0d1bf67
 size 12006544

resume/ckpt.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:8d8f31f9b568c4b77b535bc0fa75ed71c59de72b81a744db78f33a4022c1a5be
+oid sha256:2315bb1dffc03db7e42ed61e15b4fb751163e33dbb991a3558ad684d33ed952c
 size 36728175

resume/latest_step.txt

@@ -1 +1 @@
-9600
+10000

training_meta.json

@@ -1,8 +1,8 @@
 {
-  "step": 9600,
-  "val_loss": 2.2615336409759523,
-  "val_ppl": 9.597797460724992,
+  "step": 10000,
+  "val_loss": 2.3259349578094484,
+  "val_ppl": 10.236246071401576,
   "params_M": 3.0,
-  "pushed_at": "2026-07-02T05:09:24.249085",
-  "tokens_seen": 10056499200
+  "pushed_at": "2026-07-03T13:50:29.383411",
+  "tokens_seen": 10475520000
 }