Why is https://t.co/o0wNDrIKHz down?
All of CRAN is down? No one can download R or update their docker images right now. And no status pages to help?
#rstats
@ctjlewis Easy to get sample data/code like that crossed up when it's not test{ed,able}. Almost makes one miss the days of R CMD check barfing on issues in \example{...}
@ctjlewis and since their rng.sample takes messages from self.train-samples in self.train_samples_per_prompt-sized chunks, using that to resample is fine.
@ctjlewis Their code implies those objects are dicts, each with a question/answer pair in keys ‘problem’ and ‘answer’ respectively? So they need to be transformed a bit to fit the list-of-messages format they’re looking for
Maybe my most controversial opinion: Email is currently the optimal form of business communication. It's slightly more cumbersome but that encourages much more thoughtful writing.
@abhi1thakur@lcjr86 I’d argue it’s even simpler than that. The basics of file handling are important regardless of resource constraints; opening and (naively) parsing a csv is a dead simple test of that knowledge
I'm of the increasing belief that the main technical challenge for companies to successfully adopt ML isn’t the lack of functionality, but legacy systems.
The bigger a company is, the more existing tools it uses, and the slower it will be in adopting new tools.