@frankjoined.bsky.social
Excellent! I will listen while I work. Keep up the good work Jake Broe.
March 9, 2025 at 4:43 AM
Excellent! I will listen while I work. Keep up the good work Jake Broe.
I remember playing a game called “King’s Quest” back then, “Space Quest” too. I don’t remember this one.
February 23, 2025 at 7:33 AM
I remember playing a game called “King’s Quest” back then, “Space Quest” too. I don’t remember this one.
🇨🇦🇨🇦🇨🇦🇨🇦🇺🇦🇺🇦🇺🇦🇺🇦
February 22, 2025 at 6:57 PM
🇨🇦🇨🇦🇨🇦🇨🇦🇺🇦🇺🇦🇺🇦🇺🇦
I am an engineer with a PhD and I program foundational AI models for medical imaging. I use LLM’s as an assistant and they are very helpful. My background does help me know what questions to ask. But I suppose if one is curious and determined, the background will fill in, in baby steps, as one asks?
February 18, 2025 at 12:04 AM
I am an engineer with a PhD and I program foundational AI models for medical imaging. I use LLM’s as an assistant and they are very helpful. My background does help me know what questions to ask. But I suppose if one is curious and determined, the background will fill in, in baby steps, as one asks?
My code is all PyTorch so it’s performant. But, I have a feeling Rust might be ideal for deep learning. Passing data from: data loader -> transforms -> batching -> neural network -> loss, add masking - that’s a lot of things chained together and a lot of opportunities to accidentally mutate stuff.
February 16, 2025 at 6:45 PM
My code is all PyTorch so it’s performant. But, I have a feeling Rust might be ideal for deep learning. Passing data from: data loader -> transforms -> batching -> neural network -> loss, add masking - that’s a lot of things chained together and a lot of opportunities to accidentally mutate stuff.
This is good advice. I’m going to start porting some of my Python modules to Rust. The idea of borrowing is especially relevant wrt training deep learning models where in my case, a bunch of different data types are being mixed together, transformed, and then passed through a neural network.
February 16, 2025 at 6:21 PM
This is good advice. I’m going to start porting some of my Python modules to Rust. The idea of borrowing is especially relevant wrt training deep learning models where in my case, a bunch of different data types are being mixed together, transformed, and then passed through a neural network.
I’m making multi-modal foundational models for medical imaging and clinical text. My current model has 5k lines of code and I’m starting to go nuts keeping track of everything. The stuff that makes Python easy is starting to make Python hard now. Hence, I’m looking at Rust.
February 16, 2025 at 3:48 PM
I’m making multi-modal foundational models for medical imaging and clinical text. My current model has 5k lines of code and I’m starting to go nuts keeping track of everything. The stuff that makes Python easy is starting to make Python hard now. Hence, I’m looking at Rust.
Interesting. I’ve only been dabbing with Rust for about a week. I’ve just made a few simple neural networks and a data loader so far. I was under the impression that immutability was the default. In Python of course, it’s the opposite.
February 16, 2025 at 3:45 PM
Interesting. I’ve only been dabbing with Rust for about a week. I’ve just made a few simple neural networks and a data loader so far. I was under the impression that immutability was the default. In Python of course, it’s the opposite.
Dual booting is not reliable either. I have had a Windows update nuke a Linux partition.
February 16, 2025 at 6:56 AM
Dual booting is not reliable either. I have had a Windows update nuke a Linux partition.
For a gamer, I totally get the frustration. I’m frustrated with this too. Because AI is so important now, the Nvidia GPU drivers are better on Linux now. Linux would be better for gaming but all the game development is for Windows unfortunately.
February 16, 2025 at 6:55 AM
For a gamer, I totally get the frustration. I’m frustrated with this too. Because AI is so important now, the Nvidia GPU drivers are better on Linux now. Linux would be better for gaming but all the game development is for Windows unfortunately.
Debian is especially stable too. I’m actually really grateful Linux exists. It’s given me so much. I hope I will be in a position to give back one day. Open source software is freedom.
February 16, 2025 at 6:20 AM
Debian is especially stable too. I’m actually really grateful Linux exists. It’s given me so much. I hope I will be in a position to give back one day. Open source software is freedom.
I have a machine with 4 Nvidia GPU’s as my local AI workstation. It only runs Linux and I use it to create AI. With Ubuntu, the propriety Nvidia drivers install automatically. I listen to Spotify while I program; “sudo snap install spotify” is all it takes. Maybe Linux is worth another try?
February 16, 2025 at 6:17 AM
I have a machine with 4 Nvidia GPU’s as my local AI workstation. It only runs Linux and I use it to create AI. With Ubuntu, the propriety Nvidia drivers install automatically. I listen to Spotify while I program; “sudo snap install spotify” is all it takes. Maybe Linux is worth another try?
No bourbon! Buy Canadian Whiskey.
February 15, 2025 at 5:45 PM
No bourbon! Buy Canadian Whiskey.
He is a total hero.
February 15, 2025 at 5:39 PM
He is a total hero.
Linux is awesome! What the heck are people talking about? Give Debian a whirl. It’s great!
February 15, 2025 at 5:23 PM
Linux is awesome! What the heck are people talking about? Give Debian a whirl. It’s great!
I think I remember playing this.
February 15, 2025 at 6:09 AM
I think I remember playing this.
I did read the first part of this article and the people interviewed were not happy about Musk. The Trump voter regretted his vote.
February 13, 2025 at 5:42 AM
I did read the first part of this article and the people interviewed were not happy about Musk. The Trump voter regretted his vote.
Also, they are really playing up the Mixture of Experts (MOE) angle. Mistral has been using MOE for a while now. Deepseek may be an excellent model but at its heart, it’s a Transformer Causal Decoder like all the rest, and a huge one at that.
January 29, 2025 at 9:32 PM
Also, they are really playing up the Mixture of Experts (MOE) angle. Mistral has been using MOE for a while now. Deepseek may be an excellent model but at its heart, it’s a Transformer Causal Decoder like all the rest, and a huge one at that.
I really like this! It brings me back in time. It’s great art.
January 29, 2025 at 8:51 PM
I really like this! It brings me back in time. It’s great art.
It’s a 671 billion parameter multimodal model. I’m not buying the story that it was so cheap to train, like 20 times cheaper than Chat GPT or Gemini.
January 27, 2025 at 9:07 PM
It’s a 671 billion parameter multimodal model. I’m not buying the story that it was so cheap to train, like 20 times cheaper than Chat GPT or Gemini.