udgamlabs/Vemana-Base
Text Generation • 0.3B • Updated • 2 • 5
Vemana: a 278M Telugu base model trained from scratch, live in public, for about $69. Weights, checkpoints, GGUF and demo.
Note Start here. The base model: it continues Telugu text (not a chat model). Apache-2.0.
Note 12 checkpoints from the live run: every 10%, the 25% gate model and the final.
Note Q8_0 and F16 for llama.cpp and for running in a browser.
278M Telugu base model that continues Telugu text
Note Demo on a shared CPU. To run it on your own device, in your browser: udgamlabs.com/models/vemana