Sachin Gururangan

icon indicating a website link suchin.io

icon location Seattle, WA icon location University of Washington

Results 14 comments of Sachin Gururangan

integration of type-level integers

Another area i'm interested in is type-level markers encoding 'squareness' and 'singularity'. Could be as simple as equality constraints on the indices of Dimension. Singularity may be out of scope...

Any plans to replace rustc-serialize with serde?

Hi there! Sorry for the delay. As @insanitybit mentioned, this crate was mostly just a proof of concept, haven't really had time to maintain it. Pull requests welcome for migrating...

Release the intermediate checkpoints?

We are working on releasing intermediate checkpoints and will let you know

Benchmark tok/sec with other libs

just a few datapoints from OpenLM, with default hparams: we get ~2.5K tokens/sec/GPU on 256 A100s for OpenLM-7B, and ~9.5K tokens/sec/GPU on 128 A100s for OpenLM-1B. ~11.5K tokens/sec/GPU on 32...