recipe_wd_t05_mix_d125m
d125m (97.2M parameters) on plus_all v2 subset (4 passes) (28.3M words, 3.51 passes), trained on L4 24GB (AWS, spot). Started 6 Oct 2026, ended 6 Oct 2026.
Hypothesis
The upweighted mix helps once weight decay is on (trained on an AWS spot L4).
What we learned
ex-Gita 0.6958; write-up pending.
Scores
Bits per byte, lower is better; change against the parent run, RECIPE WD τ0.5.
ex-Gītā (headline)
0.6958
bits per byte
+0.7%ex-Gītā, clean_v1
0.7214
bits per byte
+0.9%Pooled, all five sets
0.687
bits per byte
+0.7%Validation split
0.6725
bits per byte
−0.2%| Test set | RECIPE WD τ0.5+MIX | RECIPE WD τ0.5 (parent) | Change |
|---|---|---|---|
| DCS gold (classical) | 0.6854 | 0.6811 | +0.6% |
| Bhagavad-gītā (memorisation) | 0.4578 | 0.4493 | +1.9% |
| Out of domain | 0.6967 | 0.6943 | +0.3% |
| Prose | 0.6613 | 0.6583 | +0.5% |
| Vedic (Ṛgveda) | 1.1599 | 1.0443 | +11.1% |
Curves
Training loss
Cross-entropy per token, by step. The first few percent of the run, far higher, run off the top; hover or the table has every value.
Held-out bits per byte
The validation split, evaluated during training, by step. Lower is better.
Throughput
Tokens per second, by step.
Model
- Preset
- d125m
- Parameters
- 97,241,856
- Outside embeddings
- 84,953,856
- Layers · heads · width
- 12 · 12 · 768
- Tokenizer
- SLP1 unigram 8k
Data
- Slice
- plus_all v2 subset (4 passes)
- Words
- 28,300,000
- Training tokens
- 103,935,151
- Passes
- 3.51 passes
- Tokens seen
- 364,953,600
Compute
- GPU
- L4 24GB
- Where
- AWS (spot)
- Steps
- 7,425 / 7,425
- GPU hours
- 2.2
- Cost
- $2.03
- Spot restarts
- 0
Lineage
awsl4spotd125m