step 2400 | val_ppl=12.71

README.md

@@ -25,9 +25,9 @@ Small language model (9.9M parameters) trained from scratch.
 | Total parameters | 9.853M |
 
 ## Training
-- Tokens seen: 2,095,104,000
-- Val loss: 2.5937
-- Val PPL: 13.38
+- Tokens seen: 2,514,124,800
+- Val loss: 2.5423
+- Val PPL: 12.71
 
 ## Usage
 ```python

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ce31762a215b9af8d1be6e67b35ff25caaca6f4e7c4e88930d34527bbc38088b
+oid sha256:ccd484c9a3bd3ed27da0daaaf6717864e2dfcabc40ebef6de58c7e83a2482ff8
 size 39421616

resume/ckpt.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d36c694f3dac51d63f3162c316b7666e7cca2b78a0b81ac3ebf7f526dd7a90e2
+oid sha256:fb766580d4298eddc88430c5df213f4b99214965ce11e4624e5a832636bd4269
 size 119811943

resume/latest_step.txt

@@ -1 +1 @@
-2000
+2400

training_meta.json

@@ -1,8 +1,8 @@
 {
-  "step": 2000,
-  "val_loss": 2.593691735458374,
-  "val_ppl": 13.3790725406917,
+  "step": 2400,
+  "val_loss": 2.542284320831299,
+  "val_ppl": 12.70866851460382,
   "params_M": 9.853,
-  "pushed_at": "2026-06-17T16:36:36.179467",
-  "tokens_seen": 2095104000
+  "pushed_at": "2026-06-17T18:20:35.191660",
+  "tokens_seen": 2514124800
 }