Popa Adrian Marius @mapopa

You can use MLX LM as a back-end with the OpenAI Python package to run locally. Just three steps to get going: gist.github.com/awni/a4613c7b2…

France dodges far right bullet amid the highest voter turnout in decades. Thank you to all voters who recognized what was at stake, and mobilized to get out the vote. You held the line not just for France, but for Europe 🇫🇷🇪🇺

@NICKIMINAJ We protest : Sound is too loud in the festival 😍

"Getting Started In Computer Graphics", good advice on how to learn and make a start in the industry, I'd advise to consider starting with a simpler API than DX12/VK, some might find learning a complex API and gfx techniques at the same time intimidating jeremyong.com/graphics/2024/…

Russian strike on a new terminal of the civilian airport in Zaporizhzhia today. Ukraine is not allowed to strike military airfields in Russia with Western weapons.

The relatives are still looking for their loved ones. The Russian strike on the hypermarket was so massive that rescuers couldn’t find the bodies. They simply don’t exist.

Olga Tokariuk @olgatokariuk

2 years ago

Imagine it's your ordinary Saturday and you are going to your local home and garden centre to do some shopping. Then, suddenly a Russian missile strike hits it and ruins everything. That's what happened yesterday in Kharkiv. At least 14 people have been killed. 🎥 Serhiy Bolvinov

42 361 766 197K 29

Transformer by Hand✍️ To study the transformer architecture, it is like opening up the hood of a car and seeing all sorts of engine parts: embeddings, positional encoding, feed-forward network, attention weighting, self-attention, cross-attention, multi-head attention, layer norm, skip connections, softmax, linear, Nx, shifted right, query, key, value, masking. This list of jargons feels overwhelming! What are the key parts that really make the transformer (🚗) run? In my opinion, the 🔑 key is the combination of: [attention weighting] and [feed-forward network]. All the other parts are enhancements to make the transformer (🚗) run faster and longer, which is still important because those enhancements are what lead us to "large" language models. 🚗 -> 🚚 Walkthrough [1] Given ↳ Input features from the previous block (5 positions) [2] Attention ↳ Feed all 5 features to a query-key attention module (QK) to obtain an attention weight matrix (A). I will skip the details of this module. In a follow-up post I will unpack this module. [3] Attention Weighting ↳ Multiply the input features with the attention weight matrix to obtain attention weighted features (Z). Note that there are still 5 positions. ↳ The effect is to combine features across positions (horizontally), in this case, X1 := X1 + X2, X2 := X2 + X3....etc. [4] FFN: First Layer ↳ Feed all 5 attention weighted features into the first layer. ↳ Multiply these features with the weights and biases. ↳ The effect is to combine features across feature dimensions (vertically). ↳ The dimensionality of each feature is increased from 3 to 4. ↳ Note that each position is processed by the same weight matrix. This is what the term "position-wise" is referring to. ↳ Note that the FFN is essentially a multi layer perceptron. [5] ReLU ↳ Negative values are set to zeros by ReLU. [6] FFN: Second Layer ↳ Feed all 5 features (d=3) into the second layer. ↳ The dimensionality of each feature is decreased from 4 back to 3. ↳ The output is fed to the next block to repeat this process. ↳ Note that the next block would have a completely separate set of parameters. Together, the two key parts: attention and FFN, transform features both across positions and across feature dimensions. This is what makes the transformer (🚗) run!

The magic trick with the disappearance of a russian tank. 📷: @United24media

For the entire day, Kharkiv has been under Russian terrorist strikes. The air raid in the Kharkiv region has been ongoing for more than 12 hours. Guided aerial bombs and missiles are targeting the ordinary buildings of Kharkiv, the ordinary civilian life of the city. A construction hypermarket has been burnt to the ground, and firefighting is still underway. There were strikes on the central park, on ordinary houses... The exact number of people killed is still unknown. In total, almost 60 people have been injured on this day alone and in Kharkiv alone. Everyone is being provided with the necessary assistance. Almost two hundred rescuers from the State Emergency Service of Ukraine and dozens of vehicles are involved in eliminating the consequences of the terror. About 400 police officers are working at the site. All emergency services are engaged. I am grateful to everyone who is taking care of people and trying to save as many lives as possible. I thank everyone in the world who supports Ukraine, and our Kharkiv now, and condemns Russian terror. It is important that we all stand together to ensure that Russian terror is overcome and that everyone responsible for these attacks receives a just response. Everyone who supports us with air defense systems, everyone who supplies Ukraine with weapons to defend against Russian terrorists, and everyone who stands with us in the world and has no lack of determination is a real lifesaver.

keyboard_arrow_left Previous keyboard_arrow_right Next

Instead of a thousand words. "This photo was taken by one of the firefighters of the Dnipropetrovsk region as they extinguished a fire after a Russian strike. A huge hole in the ceiling was left by a Russian drone. This "hole" is now in the hearts of each of us. This is the wound that hurts everyone. For every city and village, for our people. For Kharkiv Oblast, Kherson Oblast, Sumy Oblast, Chernihiv Oblast, Zaporizhzhia, Donetsk Oblast, Odesa Oblast, Mykolaiv Oblast... For every piece of our land." - State Emergency Services wrote. "While you are reading these lines, the Ukrainian defenders are firmly holding the defense. Our emergency workers of the State Emergency Service eliminate fires, rescue and evacuate people, and dismantle debris," they added. 📷: Roman Zdor/SESU of Dnipropetrovsk region

As those who follow me here already know, this past week I drove in a convoy that delivered ambulances to Ukraine. In this thread, I’d like to share some impressions from the very short time I spent in Lviv.

These were Ukrainians in Kharkiv, going to the equivalent of Lowe’s or Home Depot, thinking about fixing something up or building something new… And now they are dead because Russia murdered them.

Iuliia Mendel @IuliiaMendel

2 years ago

Kharkiv. Again. At least 11 dead, 15 missing... The numbers will increase #stoprussianow

25 194 511 371K 16

STAR WARS was released 47 years ago today. Hollywood’s biggest movie phenomenon, and the film that launched the career of star Harrison Ford, the behind the scenes story is as spectacular as blowing up the Death Star… 1/37

keyboard_arrow_left Previous keyboard_arrow_right Next

I know everyone is laughing at Google for its AI search results suggesting we should eat rocks. The funniest part is that the rocks thing also appeared on regular featured snippets, so it’s not just AI results. Perhaps Google is more broken than we think

SitRep - 24/05/24 - New US aid & ATACMS strike confirmed An overview of the daily events in Russia's invasion of Ukraine. Consequences of an ATACMS strike emerged while more Western aid to Ukraine got confirmed. Repost=appreciated! 1/X

keyboard_arrow_left Previous keyboard_arrow_right Next

@nixcraft @Xubuntu it is when the EOL for windows machines is reached #Windows #ubuntu #xubuntu

One month until release - 6/21 Windows/Mac/Linux on Steam store.steampowered.com/app/1228360/Me… #WishlistWednesday #indiegame #shmup #arpg #steamdeck #steam #release

Man, this is SO good! I get stronger emotions from an expertly crafted first stage in a shooter than I do from actual narrative driven games ha ha. When the music kicks in and the action heats...it's sublime. What a feeling!

SparseVhd - returns free space in WSL VHD image to Windows, can be enabled at creation and on existing installs #msbuild

AMD, Intel & nVidia quarterly revenues 2007 - Q1/2024 3dcenter.org/news/die-nvidi…