3 ms·
I'm surprised no one has done this for a GPU cluster yet - we used to do this for RNNs on GPUs & FPGAs at Baidu: https://proceedings.mlr.press/v48/diamos16.pdf
by gdiamos 2y ago
I'm surprised no one has done this for a GPU cluster yet - we used to do this for RNNs on GPUs & FPGAs at Baidu:
https://proceedings.mlr.press/v48/diamos16.pdf https://proceedings.mlr.press/v48/diamos16.pdf
Or better yet - on Cerebras
Kudos to groq for writing that kernel