English + Telugu LLMs built end to end: data engine, tokenizer, pretraining, serving. Start with the Telugu tokenizer leaderboard.
Kuluru Vineeth PRO
kuluruvineeth
AI & ML interests
LLMs end to end: data, pretraining, post-training, inference
Recent Activity
updated a bucket 4 days ago
kuluruvineeth/akshara-web-draft upvoted a collection 12 days ago
🧠Manas-64M: every training stage updated a bucket 13 days ago
kuluruvineeth/akshara-web-draftOrganizations
None yet
One 64M model trained from scratch, then SFT, DPO, PPO, GRPO, agent RL, distillation and MoE. Code: github.com/kuluruvineeth/manas
-
Manas-64M
ðŸ§A 64M LLM trained from scratch, served by my own engine
-
kuluruvineeth/manas-64m-pretrain
Text Generation • 63.9M • Updated • 136 -
kuluruvineeth/manas-64m-full-sft
Text Generation • 63.9M • Updated • 156 -
kuluruvineeth/manas-64m-dpo
Text Generation • 63.9M • Updated • 130
English + Telugu LLMs built end to end: data engine, tokenizer, pretraining, serving. Start with the Telugu tokenizer leaderboard.
One 64M model trained from scratch, then SFT, DPO, PPO, GRPO, agent RL, distillation and MoE. Code: github.com/kuluruvineeth/manas
-
Manas-64M
ðŸ§A 64M LLM trained from scratch, served by my own engine
-
kuluruvineeth/manas-64m-pretrain
Text Generation • 63.9M • Updated • 136 -
kuluruvineeth/manas-64m-full-sft
Text Generation • 63.9M • Updated • 156 -
kuluruvineeth/manas-64m-dpo
Text Generation • 63.9M • Updated • 130