Mahesh Sunuwar

[Mahesh Sunuwar @lowkeymahesh]

There are two mistakes one can make along the road to truth... not going all the way, and not starting. - Buddha

[Mahesh Sunuwar @lowkeymahesh]

Got mine M4, 24gb.

Whats next?

#macmini #mac

[Mahesh Sunuwar @lowkeymahesh]

@edchucation If this is real, then m4 mac would be really helpful for the school + to find tune and deploy llms and for the inference for the projects (to learn). Ngl, it would exponentially increase the learning pace.

[Mahesh Sunuwar @lowkeymahesh]

Q: what is the use case of random_state=42
A:

[Mahesh Sunuwar @lowkeymahesh]

Drop the papers you read that is essential for ML👇
Needed them to be upto date.

[Mahesh Sunuwar @lowkeymahesh]

@adamghaida Raycast, iterm, notion, tmux, dia/zen/arc 👍

[Mahesh Sunuwar @lowkeymahesh]

@peer_rich @Orrick @grok Watch Silicon Valley Season 2.

[Mahesh Sunuwar @lowkeymahesh]

@buildthatidea Interested.

[Mahesh Sunuwar @lowkeymahesh]

You'll not believe what #OpenAI just did. 🥳 🎉

#ChatGPT Plus is now FREE for college students in the US and Canada until May.
No joke!

Claim here: https://t.co/UC8KFipBv4
Tell your friends and enjoy! 😎

[Mahesh Sunuwar @lowkeymahesh]

If you want to learn Python, here is a paid Udemy course for 100%o off.
Found it, and thought I should share it for others as well.

Get it while it’s available.
Good luck.

https://t.co/l4tVUq7NnD

[Mahesh Sunuwar @lowkeymahesh]

Deepseek r1 meticulously broke down the characters in strawberry and counted the number of r’s.

And it is right.

Isn’t this how we think as well?

Also, ai am running this 32b model locally.

[Mahesh Sunuwar @lowkeymahesh]

Imagine the power of having unlimited usage of powerful model.

Just fascinating.

[Mahesh Sunuwar @lowkeymahesh]

Most fascinating part of #DeepSeekR1 is that it is an open source and we can easily #selfhost it in our own system without worrying about any data going anywhere.

Need at least 20-24gb GPU for 32b model, 42-48gb for 70b model.

And 32b runs fast as well.

[Mahesh Sunuwar @lowkeymahesh]

@_blackcat1402 @ollama @OpenWebUI @deepseek_ai Here:

It is giving around 14 tokens/s evaluation rate.

It is just fast enough for our work.

[Mahesh Sunuwar @lowkeymahesh]

Running deepseek-r1 at home.

Using RTX 8000 - 48gb memory.

But looks like it utilizes less than 22 gb memory to run the 32b model.

Thanks to @ollama ,@OpenWebUI , and obviously @deepseek_ai .

#DeepSeekR1 #ollama #openwebui

[Mahesh Sunuwar @lowkeymahesh]

Is @huggingface Space down?
It’s been couple of hours and still 500 error.

[Mahesh Sunuwar @lowkeymahesh]

Visual Studio Code @code in @github Repository.
#vscode #github #programming

[Mahesh Sunuwar @lowkeymahesh]

@gowthami_s Congrats !!!!

[Mahesh Sunuwar @lowkeymahesh]

I wrote an article on AdaBoost. Its a beginner friendly article and I believer it is a must concept to understand more about Boosting and XGBoost(i will write about it later).

I hope you find it helpful
https://t.co/gzgnMOJV3r

[Mahesh Sunuwar @lowkeymahesh]

@adityakusupati @uwcse @GoogleDeepMind Congrats @adityakusupati.
Its good to see peope with PhD getting what they deserve. 🙂