step 12800 | val_ppl=9.47

README.md

@@ -25,9 +25,9 @@ Small language model (3.0M parameters) trained from scratch.
 | Total parameters | 3.000M |
 
 ## Training
-- Tokens seen: 12,989,644,800
-- Val loss: 2.2555
-- Val PPL: 9.54
+- Tokens seen: 13,408,665,600
+- Val loss: 2.2479
+- Val PPL: 9.47
 
 ## Usage
 ```python

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:7434b0fa877c05e5a03efa90bdac71424a901a6c66f86c222ffa798decb8235c
+oid sha256:8d47eaa983c7793bd06901c5a1baaa0405bb4d21718b33feacb0dcbd7ec8c750
 size 12006544

resume/ckpt.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:164a13236e703323e8b8d6a23d6e8667f8bdf7581c06607c12ac4bb7f8f05a54
+oid sha256:d8a204da107184eae47c99925beefbd7709c6643c3e51b0f5e4dbbb88c3fa168
 size 36728175

resume/latest_step.txt

@@ -1 +1 @@
-12400
+12800

training_meta.json

@@ -1,8 +1,8 @@
 {
-  "step": 12400,
-  "val_loss": 2.2555187286376954,
-  "val_ppl": 9.540240822952908,
+  "step": 12800,
+  "val_loss": 2.2479474616241455,
+  "val_ppl": 9.468281866294314,
   "params_M": 3.0,
-  "pushed_at": "2026-07-03T18:33:50.308467",
-  "tokens_seen": 12989644800
+  "pushed_at": "2026-07-03T19:21:06.751410",
+  "tokens_seen": 13408665600
 }