Nathan Daly

Also, some of the other MultiDict changes I've proposed: #654, #655 would also be breaking changes.. Maybe it makes sense to do them all at once, in one release?

😭 oh no! Bummer. I'm fine leaving all of these on like one big "multi-dict" branch or something that we can wait to merge until you're comfortable with another breaking...

Okay, makes sense! If you can create a new branch (maybe `MultiDict`? Or `MultiDict-changes` or something?), I'll switch the PR base over to it!

Or I guess you're saying maybe we can still support the older constructors but deprecated? Unfortunately I don't think so, without the behavior remaining ambiguous and confusing. For example, the...

Thanks friends! Sorry it took me so long to get to this. Now that i have the committer bit on this repo, i've created a new [MultiDict-reboot](https://github.com/JuliaCollections/DataStructures.jl/tree/MultiDict-reboot) branch, and adjusted...

I'm not sure if it's related or not, but I ended up getting stuck on what appears to be a deadlock? ```julia Internal error: encountered unexpected error in runtime: InterruptException()...

Just adding some more context about the way this is parsed: > for now our block finding doesn't look at the code at all and only cares about the indentation...

> (If I read the output correctly, 20% of full time on summer benchmark): Yep! To clarify, that's 0.43% of total time spent actually executing that frame (exclusive), and 20%...

The above latent_spec worked okay for me. ``` python c3_celebA_latent_spec = [ (Uniform(128), False), # Noise (Categorical(10), True), (Categorical(10), True), (Categorical(10), True), (Categorical(10), True), (Categorical(10), True), (Categorical(10), True), (Categorical(10), True),...

> how to config C.5 Chairs as below I might be misunderstanding, but it seems like > For this task, we use 1 continuous latent code, **3 discrete latent codes...