O6

Hi Twitter! I’m actively seeking an internship/job in software development/Back end engineering . Skilled in (Java,Golang and python). Any leads or RTs would be greatly appreciated to help me connect with opportunities. Thank you! 🙏

here it is. i implemented the complete "Google File System" paper in rust. [repo in replies] raw TCP, custom framing, bincode serialization. pipelined writes through replica chains with serial-ordered commits. record append with overflow detection and cross-replica padding. COW snapshots that fork chunks on first write, not on snapshot. operation log with rotation, checkpointing, and crash recovery. shadow masters that replicate the oplog live and catch up from disk when they reconnect. namespace locking so concurrent file ops don't step on each other. lazy GC that hides deleted files, sweeps after a retention window, then cleans chunkservers via heartbeat. re-replication when a server goes down. two-phase chunk rebalancing, copy first, confirm via heartbeat, then delete. reading the paper was the easy part. making all the pieces not break each other was THE project.

@_babe_rae Hey 👋

My core😝

@TheGeekingMoon Same

Actually For a month atleast

@Aaliyah_Lillli Bro I don't think I can ever be confrontational Sad reality of mine

be really ready to lose people once you start being confrontational. it’s so hard mann 😹

There's a clear hierarchy in this game: 1. Finance/capital: USA 2. Knowledge/expertise: Europe 3. Manufacturing: Asia 4. Resources: Africa The US military is there to stop anyone stepping outta line and trying to do more than what is required from them

@iamdwunn Honestly speaking 1% tbh But why I do prayers is cause when I doubt my ability Faith/prayers are the only way I can assure myself of the possibility it might come through

@Techton6 he dey hear una prayers??

i told nana this thing yesterday😂😭

@iamdwunn Honestly I did mine to ask God for something cause I wasn't sure of myself if I was going to get it And I already did

@Techton6 ngl i was only fasting to stop drinking alcohol cos i wasn’t even praying

@iamdwunn Atlest I don Barb my hair

Eid ladies checklist Eid Abaya ❌ Eid New Notes ❌ Eid scarf ❌ Eid hair ❌ Eid chicken ❌ Eid jewelry ❌ Eid bag ❌ Eid henna ✅

Last paragraph He no lie

@ubah_484 No no It depends on how their deal works out

Assuming this group successfully overthrows the Iranian government, they'd later be branded as Islamic terrorists by the same people that armed and funded them

@iamdwunn Why I go fear I trust you gan

@Techton6 if i send am you no go fear?

i just did lashes for like 3 women on my street for free, gave out a kitten too i dey good mood gan today

Interesting

Fun fact: All countries starting with letter "I" are in a straight line

@Aaliyah_Lillli Ofc You really are

@Techton6 you think so? 😭 i love youuuu mannn

You are to me You write really well

holy fucking shit man, im so done 🤣🤣🤣 check this out, last 3 days im pulling 8-10hr days trying work on my framework: the training on my framework was unstable, the loss was either exploding from the first few steps or randomly spiking in the middle of training or it would never even start learning started with optimizers switched btwn SGD, Adam, AdamW, tuned learning rate up and down, changed scheduler milestones, different gamma values, added gradient clipping, removed it, changed clip norms, adam looked stable, sgd made everything shit, it was worse than adam then moved to normalization, thought BatchNorm was the main problem because of eval different than train, checked eps handling, verified mean and variance updates, thousand debug prints, compared behaviour across epochs, forced model.train() during eval just to see outputs tried so many things, thought this is not training thing, maybe its my fucking framework the cause of this lmao, checked autograd again, gradient flow, shapes matching everywhere, went through my conv, linear, pooling layers, everything looked fine but still, training was unstable. the training takes good couple of mins, i prob ran it easily over 40 times for sure to check my changes/fixes, i even thought to just ditch batchnorm and implement groupnorm as later in transformers layernorm is mostly used the most, nothing too much to do with batchnorm, yet i wasnt moving ahead cuz i cant go ahead looking my framework unstable, u get me? yesterday, i had few list of bugs that i had to fix, fixed it around 1am, thought the fixes might make a difference and ran tests again, tried batch again failed twice, tried LAYERNORM this time, framework wasnt learning, the fix was so simple man, the model was being moved across devices after parts of the training pipeline were already initialized, so the optimizer ended up tracking parameters from a different device than the one actually used during forward and backward passes so basically the optimizer was stepping over stale cpu references while all the real computation was happening on cuda tensors, meaning updates were either not applied correctly or just completely disconnected from the actual model state. fixed it. no explosions, no spikes, no issue with learning, nothing. crazy man, this is simple shit but happens to best of us but anyways yes, moving to transformers and the framework got even better now, much better training

my kind of love is really being obsessed with my partner, and i’ve never met anyone who likes that.

@momistiredz Do you have any recommendations especially on history and sci fi

When someone asks me the type of movies that I watch, thinking I’m gonna say romance, but I’m more of a history, sci-fi, fantasy,and action babe 🤣