# medium v2

- **preset:** medium
- **training corpus:** data/medium (see corpus_index.md / corpus_stats.txt)
- **trained:** 2026-07-03 16:04 (checkpoint mtime)
- **training wall-clock:** unknown
- **final val loss:** 1.2278 (iter 7999)
- **parameters:** 25.44M | vocab 97

## Training hyperparameters

| param | value |
|---|---|
| block_size | 256 |
| n_embd | 512 |
| n_head | 8 |
| n_layer | 8 |
| dropout | 0.0 |
| batch_size | 40 |
| max_iters | 8000 |
| eval_interval | 500 |
| eval_iters | 80 |
| learning_rate | 0.0003 |
| vocab_size | 97 |

## Validation curve

| step | train | val |
|---:|---:|---:|

## Reproducing the corpus
`corpus_index.md` lists every author and work in the training set; re-run `python -m corpus add-author` / `add-topic` per that index and `make finalize` to rebuild an equivalent corpus.