Subscribe to Oleg Parashchenko, fullstack developer in lead roles. LLMs, AWS, Kubernetes, TypeScript, Python, Rust’s news feed.
Click on “Follow” and decide if you want to get news from Oleg Parashchenko, fullstack developer in lead roles. LLMs, AWS, Kubernetes, TypeScript, Python, Rust via RSS, as email newsletter, via mobile or on your personal news page.
Subscription to Oleg Parashchenko, fullstack developer in lead roles. LLMs, AWS, Kubernetes, TypeScript, Python, Rust comes without risk as you can unsubscribe instantly at any time.
You can also filter the feed to your needs via topics and keywords so that you only receive the news from Oleg Parashchenko, fullstack developer in lead roles. LLMs, AWS, Kubernetes, TypeScript, Python, Rust which you are really interested in. Click on the blue “Filter” button below to get started.
Website title: Hello! - Oleg Parashchenko, fullstack developer in lead roles. LLMs, AWS, Kubernetes, TypeScript, Python, Rust
Is this your feed? Claim it!
The habit of double-checking the basics has bitten me again. This time it was GPU acceleration.
To see that GPU makes the calculations faster, I’ve run a code snippet to benchmark matrix multiplication. The result was: 10 times slower than it should be.
Issue 1: the GTX 1650 doesn’t have Tensor Cores for matrix multiplications. cuBLAS, and then PyTorch, doesn’...
Re-implementing the transformer architecture and reasoning about its time and space complexity is hard. This activity requires that the brain loads extra details such as traditional naming of parameters and dimensions. It’s a side activity for me; each time I have a time slot to do something, I realize I spend the whole slot remembering the details.
As a solution atte...