tg-medium / README.md
AILaborant's picture
Create README.md
d340acb verified
|
raw
history blame
394 Bytes
metadata
datasets:
  - AILaborant/crazy_tg
  - AILaborant/crazy_tg_tiny
language:
  - ru
pipeline_tag: text2text-generation

A small lm. (Russian only) Created to emulate a really simple one way dialogue like most telegram users do. WARNING!!! CAN SWEAR! It was trained on two T4s from scratch. Final training time: 1 hour 2 minutes. The model consists of 3 transformer blocks stacked forming 6 layers.