Philip Meier
Philip Meier
Oh yeah, SBU is also failing. I'll have a look.
> How will we remember to revisit this before the next release? Given that we are not using GH milestone feature or any other project management tool that I'm aware...
> Oh yeah, SBU is also failing. See #6390. Do you want me to disable this in this PR as well or should I send another one?
> Is it possible to turn the test into an xfail one so that if it gets fixed we know it? Done.
I agree that there is a theoretical benefit, but does this an impact on a practical case? Benchmarking it on my machine gives me roughly 3 ns per category in...
Makes no significant difference for me: ```py import random from time import perf_counter_ns import string num_categories = 1_000 # maximum that we currently have num_runs = 1_000_000 categories = ["".join(random.choices(string.ascii_lowercase,...
Right now we create the `Label` with a sequence of categories: https://github.com/pytorch/vision/blob/2e70ee1af3bb3c5ba0f9a8bcf16cd1611831690f/torchvision/prototype/features/_label.py#L18 https://github.com/pytorch/vision/blob/2e70ee1af3bb3c5ba0f9a8bcf16cd1611831690f/torchvision/prototype/features/_label.py#L40 Do we want to keep this and only internally use dictionaries or should we switch to dictionaries...
This seems like a cool idea. Given that they currently only spamming me, I don't think this is urgent. But if you (or someone else) wants to look into this,...
I think if we determine a server to be down there are two ways forward: 1. Disable the download test like I did in #6389. 2. Remove the download functionality...
From the [website](http://places2.csail.mit.edu/download.html): > Dataset is under maintenance. If you have urgent use of the dataset, please contact Bolei Zhou at [email protected]. Please describe your affiliation and your purpose of...