Luis Capelo

Hi @mtuncer, yes. I cannot recall the details. From what I remember, I did increase the `work_mem`, but the issue still remained. Apparently the amount of memory necessary was way...

Interesting. The issue constantly happens on a Docker deployment (locally and via CI) of Postgres 9.6 on `debian:jessie`. I am using this Docker `postgres` base image: https://github.com/docker-library/postgres/blob/5e7d985129d82e8757ac21c4c15cf51703d3477f/9.6/Dockerfile And here is...

Hello @Narsil @patrickvonplaten 👋! Awesome to see this merged. I am wondering if you have latency figures that compare loading a Stable Diffusion pipeline with and without `safetensors` on GPUs....

Nice! I'll take a look at `runsc run --detach hello`.

We tested this in A100s with 40 and 80GB and the merged fix works.

I would agree with @takavarasha's points above.

Mountpoint supports `AWS_REGION` [as a parameter](https://github.com/awslabs/mountpoint-s3/blob/bd9b1af33128f235b26bbeaee5c586d22de748e2/doc/CONFIGURATION.md#region-detection). The main use-case seems to be resolving issues that may happen with "cross-region mount with a directory bucket", among other things. Seems fair to...

> Are we just going to assume that our existing unit tests cover all the behaviors that we need? This should be the case, yes. The [tests I added](https://github.com/modal-labs/modal-client/pull/1357/files#diff-91e91e76633aefa71ffbbd8080e1176cc7f61a2d01e3d387cb1f05fc058a4f40R915) test...

> It seems a bit janky to patch sockets from inside the client – is there a way we can do it from the outside (from the worker)? We need...

> Could we make an upstream change to handle this in the runtime, or perhaps just raise an issue describing our problem? We can but they are implementing the correct...