-
# TL;DR
Refocus Yolk on pipelined, parallel computation.
# Abstract
Yolk currently focuses on single-chip numerical computing. This proposal would allow Yolk to seamlessly spread its Yolol ou…
-
Implement parallelism using `async` keyword on function calls:
```ruby
void def iterate_and_print(num list a)
foreach a as i
print i
end
end
async iterate_and_print([1, 2, 3…
-
-
It would be good to be able to perform optimistic concurrency check to avoid data corruption when running a few instances at the same time.
I propose to add `Revision` property of type `uint` to th…
-
It's not clear that the first naive attempt at parallelising has actually done anything useful. My understanding of the way I've done it is that there should be an async task produced per pixel, and t…
-
Short term: compare/contrast Ray vs. PySpark, figure out which is less invasive and easier to deal with.
Long term: get this so it runs, and autoscales, on a big cluster
- Use of spot instances fo…
-
-
-
CLIP's objective involves a symmetric cross entropy loss between the representstions of the text and the images (spectrograms, in our case) in a batch. It benefits from very large batch sizes (OpenAI …
-
As the network size is proportional to the dataset size, and the GPU memory is limited, some parallelism has to be implemented.
A naive approach is to split the dataset into possibly-overlapped win…