HTGM v2 Hindi LLM trained for 220+ hours on 41GB dataset. Real model outputs, limitations, and insights from base pre-training before SFT.
-
Updated
May 4, 2026
HTGM v2 Hindi LLM trained for 220+ hours on 41GB dataset. Real model outputs, limitations, and insights from base pre-training before SFT.
Building HTGM v2 — a Hindi LLM from scratch using GPT architecture. Real training logs, checkpoints, experiments, and progress shared publicly.
Building HTGM v2 — a Hindi LLM from scratch using GPT architecture. Real training logs, checkpoints, experiments, and progress shared publicly.
Building HTGM v2 — a Hindi LLM from scratch using GPT architecture. Real training logs, checkpoints, experiments, and progress shared publicly.
HTGM v2 crosses 150+ training hours while Mahesh Editor starts Smell AI, a new AI research project focused on smell and human emotional reactions.
HTGM v2 Hindi LLM reaches 180+ hours and 48K+ steps. Session 21 shows high learning rate instability and real LLM training dynamics.
Building HTGM v2 — a Hindi LLM from scratch using GPT architecture. Real training logs, checkpoints, experiments, and progress shared publicly.
HTGM v2 Hindi LLM reaches 210+ hours and 54K+ steps. Session 23 shows stable validation despite high training fluctuations and real LLM training dynamics.
Building HTGM v2 — a Hindi LLM from scratch using GPT architecture. Session 19 crosses 165+ hours training and 42K+ optimizer steps milestone.
Real test results of HTGM v2 Hindi LLM after 150+ training hours. Base model output analysis, Hindi Q&A testing, and learning progress.
Building HTGM v2 — a Hindi LLM from scratch using GPT architecture. Session 18 crosses 155+ hours training and 40K+ optimizer steps milestone.
Building HTGM v2 — a Hindi LLM from scratch using GPT architecture. Real training logs, experiments, and progress shared in public.
HTGM v2 Hindi LLM crosses 200+ hours and 51K+ steps. Session 22 shows recovery from instability with improved validation and stable training pipeline.
Building HTGM v2 — a Hindi LLM from scratch using GPT architecture. Real training logs, checkpoints, experiments, and progress shared publicly.
HTGM v2 Hindi LLM reaches 170+ training hours and 46K+ steps. Real training logs, fluctuation phase, and GPT-based Hindi AI development journey.
Add a description, image, and links to the htgm-v2 topic page so that developers can more easily learn about it.
To associate your repository with the htgm-v2 topic, visit your repo's landing page and select "manage topics."