Name: Mahesh Editor Age: 17 Developed by: Mahesh Editor (with AI assistance) Email: theindiaaiofficial@gmail.com This model was built end-to-end by Mahesh Editor, including: — Data collection — Data cleaning — Tokenizer training — Model architecture design — Training pipeline creation — Checkpoint management — Evaluation system AI was used as a research and development assistant during the process. — Source: AI4Bharat Hindi datasets — Raw Size: ~9GB — Cleaned Size: ~8.98GB — Training Duration (this release): ~50 GPU hours — Training Type: From scratch (not fine-tuned) — Hindi text generation — Prompt completion — Educational experiments — Research projects — Hindi NLP development — Building chat systems — Fine-tuning for downstream tasks HTGM is an independent Hindi LLM project built from scratch with the goal of contributing to open Hindi AI development. — Larger training runs — Better structured datasets — Instruction tuning — Retrieval Augmented Generation (RAG) — Improved alignment
Модальности:
Генерация текста
Задача: Генерация текста
Автор: indiaaiofficial
Теги: hindi, gpt2, transformer, causal-lm, hi, endpoints_compatible
Лайков: 3 | Загрузок: 0
Описание основано на материалах HuggingFace. Перевод выполнен автоматически.
