Transformer Programming Tutorials & Engineering Articles

The article discusses how NVIDIA's hardware-software co-design significantly enhanced the inference performance of Sarvam AI's Sovereign 30B model, achieving a 4x speedup on NVIDIA Blackwell archit...

The article discusses how NVFP4, a low-precision floating-point format developed by NVIDIA, enhances AI training and inference performance.

The article discusses how Pinterest enhances its ad candidate generation process using behavioral sequence modeling.

This article introduces Dynamic Context Parallelism (Dynamic-CP), a scheduling approach in NVIDIA Megatron Core designed to optimize training for variable-length sequences in large-scale models.

The article discusses the limitations of current large language models (LLMs) in handling long contexts and introduces Test-Time Training with an end-to-end formulation (TTT-E2E) as a solution.

The article discusses the NVIDIA Rubin platform, which introduces six new chips designed to create a powerful AI supercomputer.

This article provides a comprehensive tutorial on building a voice agent using NVIDIA's Nemotron models, focusing on retrieval-augmented generation (RAG) and safety guardrails.

The article discusses the NVIDIA Nemotron 3, a family of open models designed for agentic AI systems, emphasizing its efficiency and accuracy through innovative architectures and techniques.

The article discusses Mirakl's vision for agentic commerce, emphasizing the integration of AI across the company to enhance workflows and product offerings.

The article discusses model quantization, a technique essential for deploying complex AI models on resource-constrained hardware.

The NVIDIA Blackwell architecture has achieved the fastest training times across all MLPerf Training v5. 1 benchmarks, showcasing significant advancements in AI training performance.

The article discusses how NVIDIA's NeMo Automodel simplifies the training of large-scale mixture-of-experts (MoE) models in PyTorch, making it accessible to a broader audience.

The article discusses how to scale biology transformer models using PyTorch and NVIDIA BioNeMo Recipes, focusing on advanced parallel computing techniques and the integration of the NVIDIA Transfor...

The article reflects on a decade of AI platform development at Pinterest, detailing the evolution from fragmented machine learning stacks to a unified AI platform that supports various models.

The article introduces CodonFM, a new state-of-the-art RNA foundation model developed by NVIDIA as part of the Clara open model family.

The article discusses the optimization of large language models (LLMs) through pruning and knowledge distillation using NVIDIA TensorRT Model Optimizer.

The article discusses how id Software integrated RTX neural rendering and path tracing into DOOM: The Dark Ages, highlighting the advancements in real-time graphics and the technical challenges ove...

The article provides an in-depth exploration of the EmbeddingGemma architecture, detailing its origins, embedding generation process, and the comprehensive training methodology.

The article discusses three neural innovations from NVIDIA Research that are enhancing robot learning capabilities, specifically focusing on bridging the gap between controlled simulations and real...

The article discusses the deployment of on-device generative AI (GenAI) using LiteRT-LM in Chrome, Chromebook Plus, and Pixel Watch.

This article discusses the advantages of using FP8 precision for faster training throughput in large-scale deep learning models with NVIDIA NeMo.

The article discusses ReaSyn, a generative model developed by NVIDIA to predict molecular synthesis pathways, addressing the challenges of synthesizability in molecular design.

The article discusses how the NVIDIA HGX B200 significantly reduces embodied carbon emissions intensity compared to its predecessor, the HGX H100, while enhancing performance and energy efficiency.

The article introduces speculative decoding as a technique to reduce latency in AI inference, particularly for large language models (LLMs).

The article discusses the release of two new open-source models, Qwen3-Next 80B-A3B-Thinking and Qwen3-Next 80B-A3B-Instruct, which utilize a hybrid Mixture of Experts (MoE) architecture to enhance...

The article discusses fine-tuning the gpt-oss model for improved accuracy and performance through Quantization Aware Training (QAT) and Supervised Fine-Tuning (SFT).

The article discusses NVIDIA's NVFP4, a new 4-bit precision format for training large language models (LLMs) that enhances efficiency and scalability while maintaining accuracy.

The article introduces the NVIDIA Jetson Thor, a powerful platform designed for physical AI and humanoid robotics.

The article discusses the NVIDIA Blackwell Ultra GPU, a significant advancement in the Blackwell architecture designed to enhance AI training and reasoning capabilities.

The article discusses how NVIDIA's hardware innovations, particularly the Blackwell architecture and NVFP4 precision, along with their open source contributions, are driving advancements in AI.

This article discusses the development and implementation of forecasting models aimed at improving driver availability at airports, which are critical to Uber's ridesharing ecosystem.

The article discusses the evolving landscape of AI security, focusing on how hackers exploit the problem-solving instincts of multimodal AI systems through cognitive challenges.

NVIDIA has optimized OpenAI's gpt-oss models for accelerated inference performance on the NVIDIA GB200 NVL72 system, achieving up to 1. 5 million tokens per second (TPS).

The article introduces gpt-oss, two state-of-the-art open-weight language models, gpt-oss-120b and gpt-oss-20b, which excel in reasoning tasks and are optimized for deployment on consumer hardware.

The article discusses the advancements in multilingual human-like speech synthesis and voice cloning using NVIDIA Riva TTS.

The article introduces T5Gemma, a new collection of encoder-decoder models derived from pretrained decoder-only models.

The article discusses the optimization of the FLUX. 1 Kontext model for image editing through low-precision quantization techniques.

This article discusses FP8 scaling strategies, including per-tensor and per-block scaling, essential for maintaining numerical stability and accuracy during low-precision training.

The article introduces Gemma 3n, a mobile-first architecture designed for on-device AI, highlighting its multimodal capabilities and architectural innovations.

The article discusses advancements in AI-based 3D robot perception and mapping, focusing on NVIDIA's research efforts to create a unified 3D perception stack.

The article discusses NVIDIA's advancements in molecular AI modeling through the introduction of cuEquivariance and NIM microservices, which enhance the speed and efficiency of training and inferen...

The article discusses the advancements in AI autonomy through NVIDIA's Nemotron open reasoning models, which enhance AI agents' decision-making capabilities in complex environments.

This article discusses how Pinterest enhances its recommendation system through the TransActV2 model, which leverages over 16,000 lifelong user actions to improve personalization.

The article introduces the Nemotron-H Reasoning Model Family developed by NVIDIA, which addresses the challenges of reasoning-intensive tasks in large language models by significantly improving thr...

The article discusses the performance improvements delivered by NVIDIA's Blackwell architecture in MLPerf Training v5. 0, showcasing up to 2.

The article discusses the advancements in AI training through the introduction of floating-point 8 (FP8) precision, emphasizing its benefits in computational efficiency and memory usage.

The article discusses the advancements in large language models (LLMs) focusing on the importance of extended context lengths for processing and generating text.

This article discusses advanced optimization strategies for training large language models (LLMs) on the NVIDIA Grace Hopper Superchip.

The article discusses JUDE, LinkedIn's platform for generating high-quality embeddings for job recommendations using fine-tuned Large Language Models (LLMs).

FM-Intent is a novel recommendation model developed by Netflix that enhances user session intent prediction through hierarchical multi-task learning.