-
It looks like you don't have plans to keep developing this prototype, but I may be interested in integrating the PV-DM distributed memory method into the implementation (mainly because it has shown be…
-
**Inspired by the article:** https://journal.stuffwithstuff.com/2015/02/01/what-color-is-your-function/
> But if you have threads (green- or OS-level), you don’t need to do that. You can just suspe…
-
I have used `dask.delayed` to wire together some classes and when using `dask.threaded.get` everything works properly. When same code is run using `distributed.Client` memory used by process keeps gro…
iljau updated
3 years ago
-
https://dl.acm.org/doi/10.1145/3600006.3613145
-
### 🐛 Describe the bug
Symptom:
Each non-0 rank is occupying ~ 1GB memory on GPU 0.
### Versions
…
-
**Describe the bug**
The log settings defined by `logging_on()`, and therefore by any trollflow2 process, are not inherited by tasks scheduled using dask.distributed when called inside an `if __nam…
-
## Environment
- OS: [Ubuntu Ubuntu 22.04.2]
- Hardware (GPU, or instance type): [DGX with 8xH100, CUDA 12.0]
## To reproduce
Steps to reproduce the behavior:
Using the mosaic [codeba…
-
Hi,
Is there a simple way to run this code on a webdataset?
Thanks!
-
**cc**: @janeyx99
When attempting to use `OffloadActivations(use_streams=True)` on my particular use case, I get NaN gradients. Here is a script capable of reproducing this behavior (requires ~3.4…
-
PyTorch 2.5.0 is officially [released](https://github.com/pytorch/pytorch/releases/tag/v2.5.0) which includes features such as FlexAttention that is now public API and other compile features. We can n…