YHacker Newsscreenshots

hanlab.mit.edu on Hacker News

3 stories on the front page, newest first. Scroll down to watch the site change over time.

Screenshot of hanlab.mit.edu: How attention sinks keep language models stable

How attention sinks keep language models stable (hanlab.mit.edu)

191 points by pr337h4m 1 year ago | 31 comments

Screenshot of hanlab.mit.edu: SVDQuant+NVFP4: 4× Smaller, 3× Faster FLUX with 16-bit Quality on Blackwell GPUs

SVDQuant+NVFP4: 4× Smaller, 3× Faster FLUX with 16-bit Quality on Blackwell GPUs (hanlab.mit.edu)

48 points by lmxyy 1 year ago | 8 comments

Screenshot of hanlab.mit.edu: SVDQuant: 4-Bit Quantization Powers 12B Flux on a 16GB 4090 GPU with 3x Speedup

SVDQuant: 4-Bit Quantization Powers 12B Flux on a 16GB 4090 GPU with 3x Speedup (hanlab.mit.edu)

167 points by lmxyy 1 year ago | 56 comments