records/track_10min_16mb/2026-03-19_10L_MixedPrecision/submission.json


1
2
3
4
5
6
7
8
9
10
11

{
  "author": "Nan Liu",
  "github_id": "nanlliu",
  "name": "10L Mixed Precision",
  "blurb": "10-layer 512-dim model with lower LR (MATRIX_LR=0.02) and mixed int8/int6 compression: full int8 for first/last 3 layers, int6 (step=4 rounding) for middle layers 3-6. Fits 16MB via better compression while gaining an extra transformer layer over baseline.",
  "date": "2026-03-19T03:30:00Z",
  "val_loss": 2.05104604,
  "val_bpb": 1.21474500,
  "bytes_total": 15928974,
  "bytes_code": 48917
}