@maxveroo: Alpamys Makhajan’s Soup tool just revolutionized AI fine-tuning by slashing VRAM needs from 16GB to 3.3GB with as-layer streaming. Now anyone can fine-tune large language models quickly and cheaply—no expensive cloud GPUs required. This open-source breakthrough is changing the AI game. #AI #AITools #MachineLearning #OpenSource #TechNews
119 t/s is rather slow for llm reasoning. What that number means for training or tuning is a complete unknown for me. A more useful measurement that I can understand is time. Does training a model take one day? One month? One year? The lifetime of the universe?
2026-10-01 17:35:01
0
To see more videos from user @maxveroo, please go to the Tikwm
homepage.