bs-vision-train-v02 / README.md
shaban2024's picture
v02 kartica
16769ca verified
|
Raw History Blame Contribute Delete
1.26 kB
---
language:
- bs
license: cc-by-4.0
task_categories:
- image-text-to-text
- visual-question-answering
- document-question-answering
tags:
- bosnian
- ocr
- chart-qa
- synthetic
- bh-language-ai
---
# bs-vision-train-v02 — sintetički bs trening set (2179 uzoraka)
Vlastiti generisani sadržaj (100% čista licenca, CC-BY-4.0). Generator: `scripts/make_synth_bs.py` u
https://github.com/BH-Language-AI/bh-vision (seed 43, `OUT=data/synth2 N_DOC=800 N_CHART=400`).
- `data/synth2/train.jsonl` (1961) + `data/synth2/val.jsonl` (218), slike `data/synth2/img/` (1200 jpg, 32MB)
- Tipovi: računi (ukupan iznos, broj, broj stavki), fakture-tabele (cijena artikla, broj stavki),
obavještenja/potvrde/opomene/rješenja (naslov, ime), izvodi (stanje) — sve sa čćdžšđ, 30% sa laganom rotacijom;
bar-chartovi (max, vrijednost, razlika max-min) i line-chartovi temperatura (min, vrijednost)
- Format: image-text-to-text chat JSONL — `messages[{role, content:[{type:image},{type:text}]}]` + `images`
- Font: DejaVu Sans; chartovi: vlastiti podaci, matplotlib
- NIKAD miješati sa held-outom `bs-vision-ocr-heldout-v01` niti proxy evalom (XM3600-hr, postocr-sr)
- Trening: `configs/smo*-qlora.yaml`, start od najmanjeg (SmolVLM2-2.2B)