How much GPU energy does fp16 save? Measured on a Tesla T4
How much GPU energy does fp16 actually save over fp32? Spec sheets give TDP and peak FLOPS, not joules per job, so I measured it on a Tesla T4 using…
Latest Web news from Tech News
All topics
AI
agents
ai
api
architecture
automation
backend
beginners
career
database
devchallenge
devops
discuss
javascript
linux
llm
machinelearning
mcp
opensource
performance
productivity
programming
python
react
security
showdev
testing
tutorial
typescript
webdev
How Hugging Face's Transformers Added UltraBERT with Layer-Specific Rotary Embeddings
Why UltraBERT Integration Matters The Hugging Face Transformers library is a cornerstone for NLP research and applications, providing a unified inte…
Keras #22561 speedups: compiled Keras on PyTorch now generates text 11.6 faster
Between master on Jul 18 and master on Oct 15, 2026, 15 of our PRs landed in Keras. Same GPU, same torch and jax versions, same models; only the Ker…
Demystifying LLMs: Building a 124M-Parameter Decoder-Only Transformer in PyTorch
If you want to truly understand how ChatGPT, Llama, or Claude works, reading about them isn't enough. You have to build one from the ground up. In 2…
Зачем NVIDIA скрестила NeMo Automodel с диффузионными моделями
Еще в июле NVIDIA и Hugging Face выпустили совместный пост про то, как их опенсорсная библиотека для обучения нейросетей NeMo Automodel умеет теперь…
Discriminative Fine-Tuning: Why Your Backbone and Your Head Shouldn't Learn at the Same Speed
I was recently fine-tuning DEIMv2 — a DETR-style detector with a DINOv3 Vision Transformer backbone — starting from a COCO-pretrained checkpoint, to…
Undefined type Float8_e4m3fn on Apple Silicon: BF16 and GGUF Workarounds for FP8 Models
Two FP8 image models crashed with Undefined type Float8_e4m3fn on an Apple Silicon Mac, and each needed a different fix — a BF16 swap for one, a GGU…
Neural Networks: Weights, Activation, and Backpropagation
If you've ever looked at a neural network diagram and every wondered what is actually happening in these networks, this post is just for you. I pull…
GPU-Accelerating MSCRED with CUDA, im2col, GEMM, and a Custom PyTorch Extension
Introduction During my M.Tech in Data Science and Artificial Intelligence at IIT Bhilai (2021–2023), I conducted thesis research on multivariate tim…
Your quantized model got worse, and nothing told you
Getting a model onto a phone means exporting it, and almost always quantizing it. Both of those change the numbers. Everyone knows that part. What g…
От ANN к честному KNN на GPU: как мы пересобрали отбор кандидатов в рекомендациях Ozon
Привет! Мы команда рекомендательной системы Ozon, и сегодня мы хотим рассказать о нашем пути от приближённого поиска соседей (ANN) к точному KNN на…
I Trained a 6.4M-Parameter Transformer From Scratch to Talk About Recipes
Every LLM-powered app I'd built up to this point followed the same recipe (pun intended): call an API, write a good prompt, wrap it in a nice UI. Th…
Running Qwen3 Through the ExecuTorch MLX Delegate: Up to 4.52x Faster on M1 Max
Hello, everyone. There are now many ways to run an LLM on a Mac, but exporting a PyTorch model for Apple Silicon and executing it in a lightweight r…
Testing PyTorch 2.13 MPS FlexAttention on M1 Max: Up to 7.83x Faster for Sparse Attention
Hello, everyone. Attention becomes expensive very quickly as more text is given to an AI model. Can a Mac GPU make it faster when every token is res…
What Does `unsqueeze` Do in PyTorch? (And Why Your Model Keeps Asking For It)
You passed a single image to your model and got ValueError: expected 4D input (got 3D input) . Someone on Stack Overflow said "just add .unsqueeze(0…
Debugging a Python "Memory Leak" That Was Actually a Measurement Bug (ru_maxrss vs VmRSS)
I was running a pipeline for building a RAG knowledge base: crawl web articles, split them into chunks, create embeddings, and push them into Qdrant…
Classifier-free guidance above 7.5 oversaturated our product renders
TL;DR: Classifier-free guidance above a scale of ~7.5 pushed our SDXL product renders into oversaturation and clipped highlights. Adding CFG rescale…
Channels-last memory format cut our conv backbone latency 22%
TL;DR: Switching our convolutional segmentation backbone to PyTorch's channels-last memory format cut inference latency by about 22% on A100s, with…
Speculative decoding shifted our output distribution and evals missed it
TL;DR: We turned on speculative decoding in vLLM to cut latency on a fine-tuned 8B. Got a 1.9x throughput win. Three weeks later a customer flagged…
Winograd convolutions cost us 2 mAP and we didn't notice for a month
TL;DR: We turned on Winograd convolution to shave latency off a pedestrian detector running on a Cortex-A53, got a clean 18% speedup, and silently l…
Как я сделал Smart Select для Krita: локальное AI-выделение объектов по лассо
Я недавно начал пользоваться Krita, и после Фотошопа основной болью для меня было отсутствие удобного инструмента для умного выделения объектов выде…
PyTorch from Scratch — Part 1: Tensors, Gradients & Activations
Most people use PyTorch without really knowing what's happening underneath. This series breaks the foundations down into the simplest possible expla…
Why JAX Is a Much Better Backend for Quantum Circuit Simulation Than PyTorch
Modern quantum circuit simulation is not just “machine learning with complex tensors.” It involves irregular tensor contractions, sparse operators,…
AI Coding Tools for Machine Learning Engineers in 2026: Jupyter, PyTorch, and the CUDA Trap
This article was originally published on aicoderscope.com ML engineers aren't software engineers who happen to write some Python. They live in noteb…
Our event-camera detector lost 6 mAP to a badly chosen accumulation window
TL;DR: We spent three weeks chasing a 6 mAP regression in an event-camera object detector. The model was fine. The bug was the accumulation window w…
From Bayesian to deep knowledge tracing — upgrading NumPath's student model with a PyTorch LSTM
BKT told us how well a student knows subtraction-with-borrowing. It had no idea that a student who reverses digits on subtraction problems probably…
QAT vs PTQ on our edge vision model: 6 months of A/B data
TL;DR: We ran post-training quantisation (PTQ) and quantisation-aware training (QAT) side by side on the same defect-classification model deployed o…
LLM-as-judge variance broke our DPO training signal for 3 weeks
TL;DR: Our DPO pipeline used a single LLM as the preference judge. Training reward climbed every run. Production accuracy fell 4 points. The judge w…
The bf16 grad accumulator that killed our SDXL LoRA training
TL;DR: Our SDXL LoRA fine-tune for a Photoroom product photography model trained for six days while silently corrupting its adapter weights. The cau…
Prefix caching in vLLM under multi-tenant agent traffic
TL;DR: We turned on vLLM's prefix cache for our agent workloads at Nexus Labs and watched TTFT drop from 480ms to 110ms on one tenant and stay exact…