No singular library has fascinated me more than pyro. I never could use it because I thought people implemented the models entirely by hand typing the code. Pieced together how these models could be described by patches in a graphical composer which then could generate model description code. Im noticing that these frameworks are written entirely modularly and translated through python from the underlying torch. I just found torch studio which seems to be close to the notion of a torch design suite. Was hoping to identify the ide Uber developers use to interact with their software, could it be torch studio? I haven’t seen any sign of this sort of utility until just the other night when surmising it would be a cool. Vs code studio seems to almost humorously not support visual interaction found graph composer but I can’t find explicative demos and don’t have nvidia gear. Couldn’t one just do this in a web portal via low power tech such as iOS (iPad air ideally).
Piecing together auto guides as a textual conundrum “auto gui ide” assuming it’s not just a joke similar to the non-visual visual studio code, is there a possibility that the model examples and code naming conventions are derived from self explainable ai systems implemented in a graphical style along with an automating function whether systematically active during execution or post processed? Out of my depths in regards to software reverse engineering. Is there a standard live model designer and multi module/model composer? What ide/design studio like this is used at Uber? The Uber ai ecosystem is the coolest thing I came across as I learned about deep learning. From my background in stats it would make sense we would want probabilistic deep neural models yet they’re not as publicly accessible as one might imagine.
Any other developer tips and tricks for no code intuitive interaction? Any advice for how to learn this package to its fullest extent? How do you guys like to build pyro models?