๊ฐ€์ž… ํ›„ ์ดˆ๋Œ€ ๋งํฌ๋ฅผ ๊ณต์œ ํ•˜๋ฉด ๋™์˜์ƒ ์žฌ์ƒ ๋ฐ ์ดˆ๋Œ€ ๋ณด์ƒ์„ ๋ฐ›์„ ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค.

Sayak Paul
@RisingSayak
ML at Hugging Face ๐Ÿค—
๊ฐ€์ž… May 2012
125 ํŒ”๋กœ์ž‰ ์ค‘    24.8K ํŒฌ
Tensor parallel loading in ๐Ÿค— Diffusers got a massive upgrade ๐Ÿ”ฅ On Flux.2-Dev DiT, with a TP degree of 4 (A10G): โ€ข 30.4s โ†’ 12.5s loading โ€ข 64.1 โ†’ 6.8 GB peak CPU memory/rank ~2.4ร— faster loading. ~89% less CPU memory. Check it out:
๋” ๋ณด๊ธฐ