{"id":46937560,"url":"https://github.com/rilltome/veldora","last_synced_at":"2026-03-11T06:01:59.624Z","repository":{"id":317259760,"uuid":"1066663626","full_name":"rillToMe/Veldora","owner":"rillToMe","description":null,"archived":false,"fork":false,"pushed_at":"2025-09-30T05:52:03.000Z","size":26,"stargazers_count":0,"open_issues_count":0,"forks_count":0,"subscribers_count":0,"default_branch":"main","last_synced_at":"2025-09-30T07:24:30.964Z","etag":null,"topics":[],"latest_commit_sha":null,"homepage":null,"language":"Python","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"apache-2.0","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/rillToMe.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":"LICENSE","code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null,"zenodo":null,"notice":null,"maintainers":null,"copyright":null,"agents":null,"dco":null,"cla":null}},"created_at":"2025-09-29T19:45:01.000Z","updated_at":"2025-09-30T05:52:06.000Z","dependencies_parsed_at":"2025-09-30T07:24:34.778Z","dependency_job_id":"aeb84489-9cc5-4216-abfe-2147ff6df43e","html_url":"https://github.com/rillToMe/Veldora","commit_stats":null,"previous_names":["rilltome/velora","rilltome/veldora"],"tags_count":null,"template":false,"template_full_name":null,"purl":"pkg:github/rillToMe/Veldora","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/rillToMe%2FVeldora","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/rillToMe%2FVeldora/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/rillToMe%2FVeldora/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/rillToMe%2FVeldora/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/rillToMe","download_url":"https://codeload.github.com/rillToMe/Veldora/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/rillToMe%2FVeldora/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":30372537,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2026-03-10T21:41:54.280Z","status":"online","status_checked_at":"2026-03-11T02:00:07.027Z","response_time":84,"last_error":null,"robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":true,"can_crawl_api":true,"host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":[],"created_at":"2026-03-11T06:01:54.847Z","updated_at":"2026-03-11T06:01:59.614Z","avatar_url":"https://github.com/rillToMe.png","language":"Python","funding_links":[],"categories":[],"sub_categories":[],"readme":"# Veldora - README\n\n## Siapa yang bikin\n- **Author:** rillToMe\n\n## Ringkas tujuan\nPipeline end‑to‑end buat belajar bikin **model bahasa kecil** dari nol:\n\n1. Ekstrak dump Wikipedia → `wiki.jsonl` + `data.txt`  \n2. Latih **SentencePiece** tokenizer (`velora.model/.vocab`)  \n3. Encode korpus → `tokens.bin`  \n4. **Train Velora Mini (PyTorch)**  \n5. Coba **generate** teks\n\n---\n\n## Prasyarat\n- **Python 3.11** (disarankan)\n- **GPU**: RTX 3050 6GB (atau setara)\n- **Windows + VS Code** (atau OS lain oke)\n\n### Buat \u0026 aktifkan venv (Windows PowerShell)\n```powershell\npy -3.11 -m venv .venv\n.\\.venv\\Scripts\\Activate.ps1\n```\n\n### Install library\n```powershell\npip install -U rich ftfy sentencepiece tqdm\n# PyTorch CUDA (pilih sesuai CUDA kamu; contoh CUDA 12.1):\npip install torch --index-url https://download.pytorch.org/whl/cu126\n```\n\n---\n\n## Struktur folder \n```\nVelora/\n├─ data/                         # taruh dump\n├─ outputs/\n│  ├─ wiki-mini/                 # wiki.jsonl + data.txt \n│  ├─ packed/                    # tokens.bin + meta.json \n│  └─ checkpoints/               # model .pt per run \n├─ wiki_tokenizer.py    # extract + train tokenizer\n├─ prep_tokens.py                # encode data.txt -\u003e tokens.bin\n├─ train_veldora_small.py            # training Velora Mini\n└─ gen_sample.py                 # sampling/generate teks\n```\n\n---\n\n## Urutan run (dari nol)\n\n### 1) Extract + Train Tokenizer\n(Otomatis bikin `outputs/wiki-mini/data.txt` dan `outputs/velora.{model,vocab}` — atur `--prefix` kalau perlu)\n```powershell\npython pipeline_wiki2tokenizer.py --prefix \"outputs/velora\"\n# tes cepat (mis. hanya 2000 artikel):\n# python pipeline_wiki2tokenizer.py --sample 2000 --prefix \"outputs/velora_test\"\n```\n\n### 2) Pack korpus → tokens.bin\n(Bikin `outputs/packed/{tokens.bin, meta.json}`)\n```powershell\npython prep_tokens.py\n```\n\n### 3) Train Velora (mini)\n(Checkpoint tersimpan di `outputs/checkpoints/velora-YYYYMMDD-HHMM/`)\n```powershell\npython train_veldora_small.py\n```\n\n### 4) Generate teks dari checkpoint terbaik\n```powershell\npython gen_sample.py\n```\n\n---\n\n## Catatan cepat\n- Kalau hanya mau **ulang training** (data/tokenizer sama), **lewati** langkah 1–2.\n- Jika OOM: turunkan `MICRO_BSZ` atau `CTX_LEN` di `train_gpt_small.py`.\n- `gen_sample.py` memakai `outputs/velora.model` dan checkpoint terbaru di `outputs/checkpoints/`.\n\n---\n\n## Library utama\n- **PyTorch** (training \u0026 inference)\n- **SentencePiece** (tokenizer)\n- **Rich** (UI progress \u0026 warna)\n- **ftfy** (perbaikan teks)\n- **tqdm** (progress bar encode)\n\n---\n\n## Versi \u0026 penamaan\n- Model: **Velora-42M**, **Velora-125M** (ke depan)\n- Rilis: **Velora v0.1 “Alpha”**, **v0.2 “Beta”**, **v1.0 “Stable”**\n- Folder run: `outputs/checkpoints/velora-YYYYMMDD-HHMM/`\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Frilltome%2Fveldora","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Frilltome%2Fveldora","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Frilltome%2Fveldora/lists"}