{"id":23190483,"url":"https://github.com/lukasdrews97/dumblellm","last_synced_at":"2025-04-05T06:42:02.206Z","repository":{"id":268617041,"uuid":"901940772","full_name":"LukasDrews97/DumbleLLM","owner":"LukasDrews97","description":"Decoder-only LLM trained on the Harry Potter books.","archived":false,"fork":false,"pushed_at":"2024-12-20T11:37:49.000Z","size":241,"stargazers_count":0,"open_issues_count":0,"forks_count":0,"subscribers_count":1,"default_branch":"main","last_synced_at":"2025-02-10T14:26:41.335Z","etag":null,"topics":["byte-pair-encoding","flash-attention","grouped-query-attention","large-language-model","rotary-position-embedding","transformer"],"latest_commit_sha":null,"homepage":"","language":"Python","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":null,"status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/LukasDrews97.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":null,"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}},"created_at":"2024-12-11T15:47:42.000Z","updated_at":"2024-12-20T11:37:52.000Z","dependencies_parsed_at":"2024-12-17T22:22:06.692Z","dependency_job_id":"fb49340e-dbcb-44f0-9f0b-677f3212dec5","html_url":"https://github.com/LukasDrews97/DumbleLLM","commit_stats":null,"previous_names":["lukasdrews97/dumblellm"],"tags_count":0,"template":false,"template_full_name":null,"repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/LukasDrews97%2FDumbleLLM","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/LukasDrews97%2FDumbleLLM/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/LukasDrews97%2FDumbleLLM/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/LukasDrews97%2FDumbleLLM/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/LukasDrews97","download_url":"https://codeload.github.com/LukasDrews97/DumbleLLM/tar.gz/refs/heads/main","host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":247299792,"owners_count":20916186,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2022-07-04T15:15:14.044Z","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":["byte-pair-encoding","flash-attention","grouped-query-attention","large-language-model","rotary-position-embedding","transformer"],"created_at":"2024-12-18T12:14:39.987Z","updated_at":"2025-04-05T06:42:02.163Z","avatar_url":"https://github.com/LukasDrews97.png","language":"Python","funding_links":[],"categories":[],"sub_categories":[],"readme":"# DumbleLLM - Custom Large Language Model\n\nThis repository contains the code for a decoder-only transformer, similar to Llama or GPT. It was trained on an English corpus built from the seven Harry Potter books and has roughly 75M trainable parameters.\n\n# Technical Features\n\n- Tokenization: Byte pair encoding (sentencepiece)\n- FlashAttention, Grouped Query Attention\n- Rotary Position Embeddings\n- Key Value Cache\n- Sampling: top-p, top-k\n\n\n# Training Configuration\n| **Parameter**          | **Value**   |\n|------------------------|-------------|\n| Layer                  | 4           |\n| Model Dimension        | 768         |\n| Context Length         | 1024        |\n| Attention Heads        | 8           |\n| Key/Value Heads        | 4           |\n| Vocabulary Size        | 32000       |\n| RoPE Theta             | 10000       |\n\n\n# Roadmap\n- [x] ~~Grouped Query Attention~~\n- [x] ~~Rotary Position Embeddings~~\n- [x] ~~Key Value Cache~~\n- [ ] Distributed training\n- [ ] Finetuning with (Q)LoRA\n- [ ] Add Mixture of Experts model\n\n# Example Prompts\nTODO\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Flukasdrews97%2Fdumblellm","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Flukasdrews97%2Fdumblellm","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Flukasdrews97%2Fdumblellm/lists"}