Smart India Hackathon: Horrible experience
Background:- Our task was to do load forecasting for a given area. Hourly energy consumption data for past 5 years was given to us.

One government official asks the following questions:-
1. Why are you using deep learning for the project? Why are you not doing data analysis?
2. Which neural network "algorithm" you are using? He wanted to ask which model we are using, but he didn't have a single clue about Neural Networks.

3. Why are you using libraries? Why not your own code?

Here comes the biggest one,

4. Why haven't you developed your own "algorithm" (again, he meant model)? All you have done is used sone library. Where is "novelty" in your project?

I just want to say that if you don't know anything about ML/AI, then don't comment anything about it. And worst thing was, he was not ready to accept the fact that for capturing temporal dependencies where underlying probability distribution ia unknown, deep learning performs much better than traditional data analysis techniques.
After hearing his first question, second one was not a surprise for us. We were expecting something like that. For a few moments, we were speechless. Then one of us started by showing neural network architecture. But after some time, he rudely repeated the same question, "where is the algorithm". We told him every fucking thing used in the project, ranging from RMSprop optimizer to Backpropagation through time algorithm to mean squared loss error function.

Then very calmly, he asked third question, why are you using libraries? That moron wanted us to write a whole fucking optimized library. We were speechless at this question. Finally, one of us told him the "obvious" answer. We were completely demotivated. But it didnt end here. The real question was waiting. At the end, after listening to all of us, he dropped the final bomb, WHY HAVE YOU USED A NEURAL NETWORK "ALGORITHM" WHICH HAS ALREADY BEEN IMPLEMENTED? WHY DIDN'T YOU MAKE YOU OWN "ALGORITHM"? We again stated the obvious answer that it takes atleast an year or two of continuous hardwork to develop a state of art algorithm, that too when gou build it on top of some existing "algorithm". After listening to this, he left. His final response was "Try to make a new "algorithm"".

Needless to say, we were completely demotivated after this evaluation. We all had worked too hard for this. And we had ability to explain each and every part of the project intuitively and mathematically, but he was not even ready to listen.
Now, all of us are sitting aimlessly, waiting for Hackathon to end.😢😢😢😢😢

After a lot of work I figured out how to build the graph component of my LLM. Figured out the basic architecture, how to connect it in, and how to train it. The design and how-to is 100%.
Ironically generating the embeddings is slower than I expect the training itself to take.

A few extensions of the design will also allow bootstrapped and transfer learning, and as a reach, unsupervised learning but I still need to work out the fine details on that.

Right now because of the design of the embeddings (different from standard transformers in a key aspect), they're slow. Like 10 tokens per minute on an i5 (python, no multithreading, no optimization at all, no training on gpu). I've came up with a modification that takes the token embeddings and turns them into hash keys, which should be significantly faster for a variety of reasons. Essentially I generate a tree of all weights, where the parent nodes are the mean of their immediate child nodes, split the tree on lesser-than-greater-than values, and then convert the node values to keys in a hashmap to make lookup very fast.

Weight comparison can be done either directly through tree traversal, or using normalized hamming distance between parent/child weight keys and the lookup weight.

That last bit is designed already and just needs implemented but it is completely doable.

The design itself is 100% attention free incidentally.

I'm outlining the step by step, only the essentials to train a word boundary detector, noun detector, verb detector, as I already considered prior. But now I'm actually able to implement it.

The hard part was figuring out the *graph* part of the model, not the NN part (if you could even call it an NN, which it doesn't fit the definition of, but I don't know what else to call it). Determining what the design would look like, the necessary graph token types, what function they should have, *how* they use the context, how thats calculated, how loss is to be calculated, and how to train it.

I'm happy to report all that is now settled.

I'm hoping to get more work done on it on my day off, but thats seven days away, 9-10 hour shifts, working fucking BurgerKing and all I want to do is program.

And all because no one takes me seriously due to not having a degree.

Fucking aye. What is life.

If I had a laptop and insurance and taxes weren't a thing, I'd go live in my car and code in a fucking mcdonalds or a park all day and not have to give a shit about any of these other externalities like earning minimum wage to pay 25% of it in rent a month and 20% in taxes and other government bullshit.

Figured I'd post for some advice here and see if anybody has had previous experience or success with a situation like this.

My team is generally comprised of full-stack developers completing front-end custom work on sites, writing back-end tools, and fixing broken sites. We are a rapid-response DEV team, and we typically turn around any custom requests in less than 5 days and fix any broken sites on the same day as they were reported. We manage almost 15,000 sites across multiple countries, and deal with very large corporations that many of you interact with every day (I'm trying to be cryptic here hahaha.) There are 16 of us on our team, and we are the only DEV team within our department of 500+ people. We are also the only DEV team taking requests from these 500+ people. The way the department works, we are the final say on whether a specific piece of custom work will get completed or not, and we are the go-to people when anybody has a question about our system infrastructure or if our system can accommodate a request, along with how to fix any broken pieces of our platform. We typically get about 150 requests per day. Lately, the entire team has become unhappy with our compensation for the work we do. We're quite underpaid, and they keep giving us more responsibilities without any sort of extra compensation. We've discovered that there are a large amount of non-developers below us that are getting paid more than we are. We've found that we get paid about $15,000 less than a comparable DEV team in a different department (let's call that team DEV_2,) just because of which department our team exists within, and how our department defined our job back when this position was created a few years ago. Ever since the position was created, our team's responsibilities have exponentially increased. We believe that there is absolutely no reason that an entry-level position below us should get paid just as much, or even more in some cases, than a developer. Of course, we're not asking to pay them less. Instead, we've decided that we're going to bring this up with our manager and schedule a meeting with him, our Department Director, and Human Resources, and voice that we believe that we should be on the same payscale as the comparable DEV_2 in the other department.

To be a good developer on our team, you need to not only have coding expertise, but also an encyclopedic knowledge of what you can do within our platform without any coding. You need this knowledge so you can pass it along to any people in positions below you, in case they didn't know that something could be done without custom code.

We're going to argue that if it weren't for our team, the company would be losing millions of dollars in clients, because people wouldn't have anybody to go to for platform infrastructure questions, broken websites, or custom work. Instead, they would need to send these requests to the DEV_2 team, which currently take about 6 months to turnaround requests. Like I said, we are a rapid-response DEV team, and these particular clients think that a 5 day turnaround time is ridiculous. If they had to wait 6 months for their request to be completed, they would cancel their contracts.

Not to mention the general loss of knowledge if the members of our team went to a different department, which would be catastrophic for our current department. Believe me, this department could not function without this DEV team. If we all went on vacation for a week, the place would be on fire by the time we got back, and many clients would be lost.

Do any of you have any experience with a situation like this, and if so, how did it turn out? Thank you!

This kind of stuff is what causes developer's to loss their mind. If you have to write a comment saying do not use a function, REMOVE IT!

So I have a problem and I was hoping for some insight.

I figured out how to get
(surd(n, x)-surd(n, y))

without knowing x or y, (only n), through a convergent series of approximate identities.
n is the product of x and y, where x<y

My only issue is I don't know where to go from here. I've basically hit the limit of my insight into the problem.

surd() here is just a function that returns the results of two arguments, a, b, such that (a^2)-b.
Both are guaranteed to be positive integers, greater than 1.

But, having come this far, with a couple pages of intermediate identities, I'm at a loss.

They keep training bigger language models (GPT et al). All the resear4chers appear to be doing this as a first step, and then running self-learning. The way they do this is train a smaller network, using the bigger network as a teacher. Another way of doing this is dropping some parameters and nodes and testing the performance of the network to see if the smaller version performs roughly the same, on the theory that there are some initialization and configurations that start out, just by happenstance, to be efficient (like finding a "winning lottery ticket").

My question is why aren't they running these two procedures *during* training and validation?

If [x] is a good initialization or larger network and [y] is a smaller network, then
after each training and validation, we run it against a potential [y]. If the result is acceptable and [y] is a good substitute, y becomes x, and we repeat the entire procedure.

The idea is not to look to optimize mere training and validation loss, but to bootstrap a sort of meta-loss that exists across the whole span of training, amortizing the loss function.

Anyone seen this in the wild yet?

I am building a Django app where a function needs to be called asynchronously. I need to handle any interruptions like loss of internet connectivity during that time. How do I achieve this? Can anyone help?

ohhhhh I am pissseddddddddddd

itss the fucking pytorch.module class it would seem !

I do exactly the same goddamn shit as its supposed to do in a goddamn notebook and run it step by step and the fucking model trains and the output values change !!!! and the loss decreases !!

I do this in the goddamn class derived from model with a call to model.parameters() and the fucker fails !!!

why ???
why ?????
why ??????
is it cloning the goddamn parameters so the references aren't there ????
seems to work goddamn fine when i call a layer and activation function at a goddamn time chaining the calls one after another !!!!!!

UGHHHH IT LOOKS LIKE IF YOU DEFINE THE LOSS AND OPTIMIZER OUTSIDE THE FUCKING CLASS IN A SEPERATE TRAINING FUNCTION IT DOESN'T TRAIN !!!!!!

WHY ??
A REFERENCE IS A GODDAMN REFERENCE !!!!

i need some advice on how to deal with office culture. i am a covid graduate and this is my first wfo job. it is technically hybrid but quickly turning into full time office, and there are several examples of scenarios, where i am not only feeling just frustrated, but hurt and retaliation.

my whole team is in a different city except 4 of us : pm, sr ios dev, me(android dev) and a sr android dev. in our office, there are 50 more people , but i rarely need to contact anyone except my team from another city or these 3 folks. also, we 4 are new joinees like just joined in last 2 months.

so let's discuss the problems.

1. there have been very shitty decisions that are leading to loss of everyone just because a few are unlucky. here's an example. on may 1, international labor day, we 4 had a leave showing. but it was not showing for other people. maybe because ourbleave calender was aligned to other city or maybe coz we are new, idk. but someone told the boss of manager, and he mailed to us that there is no leave :/ wtf

2. another news: our is shifting from we work to another co-working space. it is being heard that office will be now 3/5 days instead of 2/5 . when we joined, it was showing 3/5 days in our hr portal, but hr assured that it is 2/5 days. and we would still go 2/5 days only. but like that holiday scenario, people are buzzing and talking, and they might end up getting our 2/5 culture tonget fucked too. this is very stupid, since i am wasting 4 hrs everyday travelling.

3. let's talk about the snakes in the 4 ppl group. the ios dev and manager are sweet looking girl snakes. ios girl is the meek snake and pm is the wicked snake. once i discussed with ios girl about how we need to rush every morning at 8 am to reach office as our standup is at 10. i told her that i would raise this matter in standup and when i did, she was just mum as fuck. didn't even voted a fucking yes when the boss said "ok let's have a vote on it" . i mean man what the fuck are your scared of? the boss won't kill you bitch for clocking 30 mins late

4. the other snake is pm. i am pretty sure she was one of the people for which that leave was not showing and she informed the boss's boss. day before that i told her jokingly that once i leave the office, I won't be opening my laptop and since today it was decided that tomorrow is the holiday, I am unreachable and therefore enjoying the vacay due to lack of latest info.
the bitch fucking whatsapped me to say that she got a call from boss that tomorrow's a working day. it would have been the perfect fucking leave.
I am pretty sure a lot of people are hating me for leaving so early too. i oeave at 5pm , as i have to be at gym by 7. also 1 minute past 5 and i would be travelling in a jam packed metro, so yeah, no thanks. but this bitch is definitely telling my boss about this.

5 finally the biggest snake is this *cough-cough* "sr" android guy. dude's code is so shittu and hacky, i can sense that he didn't tried to understand the class and just added a function at any place he felt fit. he also is a schemy bitch, as he has somehow convinced noss to let him wotk just 1/5 days in wfo.

but i didn't cared about him much until now. yesterday i sent a link regarding latest Android dev update in the official channel as a fun read, and his reply was "probably should have seen theeynote yesterday" bitch it wasn't even mentioned in that keynote! i just checked its summary after his message, but then it was too late to retaliate.

and now that i see, he always tries to be smug and cool. not that i care, roast me all you want in front of your crush, I won't mind, but if you're trying to show people that am not an able dev, then buckle up bitch, either you or me are counting last breaths.