@qqq http://www.gputechconf.com/

@blueberry : btw, any chance you can complete https://github.com/uncomplicate/clojurednn ?

I got cuda to work fine in clojure, but cudnn I'm struggling with

if that library worked, it'd be amazing 🙂

@blueberry : is it okay if I send you a few direct messages I have a bunch of questions

@qqq It is on my todo list, but since deep learning is only one of the things I'm interested in (there are other ML topics with higher priority for me) and I have to juggle my limited resources, cuDNN will have to wait a bit, and it will not be a thin layer, but an opinionated clojure DL library tailored to my liking.

@qqq of course. DM messages are fine, but if it is something that does not absolutely require privacy, I'd prefer public messaging - that way other people can benefit too.

@qqq Also, conversations like that would be better in #uncomplicate, or for more general stuff #data-science . They're certainly not offtopic...

@jsa-aerial : uncomplicate currently has no cuDNN support

@jsa-aerial : cuDNN and data-science are orthogonal; you can do data-science without cuDNN, and you can do cuDNN without data-science

@jsa-aerial : lastly, blueberry just happened to post in off-topic; and I wanted to ask him a question; it's not clear why following up in off-topic is inappropriate

@qqq I tend to think those reasons are not sufficient. If you want to ask about uncomplicate work like the clojurednn you specifically mentioned, you will be better off in uncomplicate. Additionally, data-science seems to be the 'catch all' / generic channel for asking about any of this stuff. So, none of this is actually off-topic, in that those other channels are basically there for just that sort of conversation.

We disagree.

I also don't understand why you're trying to force traffic out of off-topic.

It's one thing to tell someone to take a conversation t off-topic; it's weird and another thing to sa "well, this doesn't below in off-topic, go to channel BLAH instead"

Not unreasonable, really; that way the conversation's results can be seen by those in relevant channels, as well. Otherwise information on a given topic gets scattered across many channels.

I think I know what's going on. Earlier, I asked where cuda conversations should go. jsa-aerial was suggesting uncomplicate / data-science for that. Then, I started asking cuda/cudnn questions -- and find it weird someone's pushing me to to data-science.

Anyway, this seems like a silly thing to argue over. Let's all get back to coding 🙂

Yeah, I kinda sympathize with both sides of the argument

🙂

The fact that history isn't stored makes it kind of a wash

there is an archive https://clojurians-log.clojureverse.org/

I stand corrected

I really wish there was a feature to create a wiki where the wiki could point to a list of piecwise continuous comments

searching through the logs with current UI is rather painful

"Just stuff it in Elasticsearch"

cockroach sql --insecure \
--database=bank \
--user=maxroach \
-e 'CREATE TABLE accounts (id INT PRIMARY KEY, balance INT)'