step 4400 | val_ppl=11.01

README.md

@@ -25,9 +25,9 @@ Small language model (3.0M parameters) trained from scratch.
 | Total parameters | 3.000M |
 
 ## Training
-- Tokens seen: 4,190,208,000
-- Val loss: 2.4698
-- Val PPL: 11.82
+- Tokens seen: 4,609,228,800
+- Val loss: 2.3991
+- Val PPL: 11.01
 
 ## Usage
 ```python

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:6da6fb15f486529d924fd47602ba5e0c2682dc5c43fc50d99db1f7a5e50a347c
+oid sha256:7cfc51ddc13e6059777c01a6493d182ebf0d99d5821156e24b608f06ca392508
 size 12006544

resume/ckpt.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:7a7862a9e0e2a43da0071c119c5e2ed1cfe965bcab8287a9446804fda90b89cf
+oid sha256:e041851242ed1c2d602f8afdb7a53718c97208c13619e2f33dd982c684ef6193
 size 36728047

resume/latest_step.txt

@@ -1 +1 @@
-4000
+4400

training_meta.json

@@ -1,8 +1,8 @@
 {
-  "step": 4000,
-  "val_loss": 2.4697529681396486,
-  "val_ppl": 11.819526691307864,
+  "step": 4400,
+  "val_loss": 2.3991224730682372,
+  "val_ppl": 11.013507489470683,
   "params_M": 3.0,
-  "pushed_at": "2026-07-01T17:40:06.755448",
-  "tokens_seen": 4190208000
+  "pushed_at": "2026-07-01T18:30:45.022102",
+  "tokens_seen": 4609228800
 }