Text-Only component of gemma-3-27b-it

  • The tokenizer's eos token has been explicitly set to <end_of_turn> instead of <eos>, in alignment with the original chat format used by Gemma.

์ด ๋ชจ๋ธ์€ Gemma-3-27b-it ๋ชจ๋ธ์—์„œ ๋น„์ „ ๋ถ€๋ถ„(400m ํŒŒ๋ผ๋ฏธํ„ฐ)์„ ์ œ๊ฑฐํ•œ ๋ชจ๋ธ์ž…๋‹ˆ๋‹ค.
ํ† ํฌ๋‚˜์ด์ €์˜ eos ํ† ํฐ ์„ค์ • ๋˜ํ•œ <end_of_turn>์œผ๋กœ ๋ณ€๊ฒฝํ•˜์˜€์Šต๋‹ˆ๋‹ค.

Downloads last month
5
Safetensors
Model size
27B params
Tensor type
BF16
ยท
Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐Ÿ™‹ Ask for provider support