step 800 | val_ppl=13.63

README.md

@@ -25,9 +25,9 @@ Small language model (9.9M parameters) trained from scratch.
 | Total parameters | 9.902M |
 
 ## Training
-- Tokens seen: 1,414,195,200
-- Val loss: 2.7217
-- Val PPL: 15.21
+- Tokens seen: 1,885,593,600
+- Val loss: 2.6120
+- Val PPL: 13.63
 
 ## Usage
 ```python

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:8199d836370546707616149db92e7e33b094943920bcc761483b11fc49ed4d69
+oid sha256:180ba389f6267a34f073d17be98ec447903d43ed473f3094a45a82b5c8a4827a
 size 39621248

resume/ckpt.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:2f32673203de5e65c77a75d46efc82bbd2cd027794c72b8c0e4d71686e71b976
+oid sha256:9cc60874c07ad029c305b458c1cdc1c3995c834c77f4e1386b4e5ebe87930ffa
 size 122094685

resume/latest_step.txt

@@ -1 +1 @@
-600
+800

training_meta.json

@@ -1,8 +1,8 @@
 {
-  "step": 600,
-  "val_loss": 2.7217227363586427,
-  "val_ppl": 15.206496477198597,
+  "step": 800,
+  "val_loss": 2.6119989347457886,
+  "val_ppl": 13.626261654337085,
   "params_M": 9.902,
-  "pushed_at": "2026-06-11T11:05:55.862283",
-  "tokens_seen": 1414195200
+  "pushed_at": "2026-06-11T12:47:10.876870",
+  "tokens_seen": 1885593600
 }