Tagged gpu

Cut peak GPU memory in Megatron Bridge and fix common out-of-memory errors

Get starting MoE training configs for each GPU type in Megatron Bridge

Recommend and resize Megatron Bridge recipes for your model and GPU count

Find the right NVIDIA skill for a GPU, Jetson, CUDA, NIM or Isaac request

Write and debug GPU-accelerated RDKit fingerprint and similarity code with nvMolKit

Prepare NVSHMEM CUDA kernels for TMA shared memory registration and transfers

Plan NVIDIA Parabricks pbrun commands and map CPU options to GPU equivalents

Fine-tune any Hugging Face vision, VLM or LLM model on local NVIDIA GPUs

Run TAO training and inference containers on NVIDIA Brev GPU instances

Run TAO jobs in Docker on a local daemon or a remote GPU host over SSH

Submit TAO training jobs as Kubernetes Jobs with NVIDIA GPU scheduling

Submit TAO training and inference jobs to a SLURM GPU cluster with sbatch

Check and install compatible NVIDIA drivers, CUDA and container toolkit on GPU hosts

Add a cuTile GPU kernel to TileGym with dispatch registration, tests and benchmarks

Port cuTile Python GPU kernels to cuTile.jl, handling indexing and layout changes

Convert cuTile GPU kernels to Triton and debug illegal address and shape errors

Write validated high-performance GPU kernels with the cuTile tile programming model

Patch TileGym GPU kernels into Hugging Face transformers models before loading

Basecall raw Nanopore POD5 signal into reads with Dorado on a GPU

Organize PyTorch training with LightningModules, Trainers and multi-GPU strategies

Run GPU jobs on your own Modal account with submit, wait and harvest steps

Orchestrate GPU provisioning and training or inference jobs across GPU clouds

Prototype model serving with vLLM or SGLang on real GPUs before scaling

Install DeePMD-kit with pip, conda, Docker or source for CPU, CUDA or ROCm

Pick placements that keep usable leftover capacity for bin packing and GPU sharing

Run Python training jobs on cloud GPUs with the Modal serverless platform

Train a GPT-2 scale model on one GPU with mixed precision and modern optimizers

Crack captured password hashes offline with GPU-accelerated hashcat rules and masks

Diagnose GPU and KAITO incidents on AKS using read-only cluster evidence

Process large datasets for ML with streaming Ray Data pipelines across CPUs and GPUs

Run ML jobs across clouds with SkyPilot, spot recovery and GPU cost optimization

Speed up LLM fine-tuning with Unsloth and cut GPU memory use with LoRA

Generate C, C++ or CUDA code from PyTorch and LiteRT models with MATLAB Coder

Generate, verify and speed up C, C++ or CUDA code from MATLAB functions

Profile and rewrite MATLAB designs for faster CUDA code with GPU Coder

Detect and validate your GPU for MATLAB GPU computing and fix setup errors

Run GPU Kubernetes clusters with scheduling, autoscaling, MIG and cost controls

Install NVIDIA drivers, CUDA and the container toolkit on GPU servers

Autoscale LLM inference on Kubernetes with KEDA, GPU metrics and spot nodes

Check GPU and compute availability before experiments and stop runs that cannot start

Turn an experiment plan into implemented code, GPU runs and first results

Deploy and run ML experiments on local, remote or serverless GPU machines

Run training, fine-tuning and inference on serverless GPUs with Modal

Profile scripts, GPUs and memory to find bottlenecks and recommend fixes

Watch WandB training metrics for NaN values, loss divergence and idle GPUs

Rent, manage and destroy on-demand GPU instances on vast.ai

Add layer labels and compact GPU lanes to a Torch Profiler Chrome trace

Build and debug pixi environments for scientific Python, CUDA and Jupyter kernels

Inspect a registered server for GPUs, schedulers, sudo and Python before planning jobs