Multilingual code LLM trained entirely on a smartphone with zero budget. GPT-2 architecture. Snapdragon 685. CPU only. $0.00. Yuuki 3.7 is a multilingual code generation language model being trained entirely on a Redmi 12 smartphone (Snapdragon 685, CPU only) with zero cloud budget. This project proves that meaningful AI training doesn’t require expensive GPU infrastructure — just patience, determination, and time. The model is based on GPT-2 architecture (82M parameters) and is being trained on The Stack dataset with 75,000 carefully selected code examples across multiple programming languages. — Forward pass: 40 seconds — Backward pass: 40 seconds — Optimizer step: 20 seconds — Total: ~100 seconds — [x] Setup training pipeline — [x] Start training (Step 0) — [x] Reach Step 1,000 — [x] Break loss 2.0 barrier — [x] Break loss 1.8 barrier ⭐ — [ ] Checkpoint 2,500 (7%) — [ ] Checkpoint 5,000 (13%) — [ ] Checkpoint 10,000 (27%) — [ ] Checkpoint 18,750 (50% — Epoch 1 complete) — [ ] Checkpoint 37,500 (100% — DONE) — [x] Quantize to GGUF formats — [ ] Convert to ONNX — [
Модальности:
Генерация текста
Области применения:
Генерация кода
Задача: Генерация текста
Автор: OpceanAI
Теги: gguf, llamafile, gpt2, code, en, es, text-generation-inference, endpoints_compatible
Лайков: 3 | Загрузок: 225
Описание основано на материалах HuggingFace. Перевод выполнен автоматически.
