A no-cost infrastructure benchmark measuring the VRAM and throughput impact of NF4 (4-bit) quantization on LLMs
-
Updated
Dec 7, 2025 - Python
#
gpu-optimization
Here are 149 public repositories matching this topic...
Enterprise-grade Power BI dashboard for tracking AI infrastructure unit economics. Features real-time GPU profitability ratios, idle waste detection (Zombie GPUs), and ESG carbon footprint reporting.
power-bi dax data-modeling cost-optimization gpu-optimization finops platform-engineering a100 esg-reporting cloud-economics unit-economics llmops h100 sustainable-it ai-infrastructure
-
Updated
Feb 14, 2026
Archived history — active development moved to troycheng/cuda-kernel-optimizer
performance gpu cuda triton cutlass gpu-optimization nsight-compute kernel-optimization codex-skills codex-skill
-
Updated
Jul 21, 2026 - Python
A massively parallel implementation of the Synthetic Aperture Radar (SAR) Time-Domain Backprojection algorithm optimized for NVIDIA GPUs using CUDA C++.
-
Updated
Apr 25, 2026 - Cuda
High-throughput, low-latency LLM inference platform for LLaMA-3 & Mistral — dynamic batching, KV-cache optimization, FP16/BF16 mixed precision, tensor parallelism, with PyTorch profiling, Prometheus/Grafana observability, Docker & Kubernetes (HPA) deployment.
kubernetes inference prometheus pytorch mistral model-serving gpu-optimization mlops fastapi kv-cache deepspeed dynamic-batching tensor-parallelism llm llama3
-
Updated
Jun 24, 2026 - Python
"Fix for Docker Desktop 4.79.0 stuck on 'Starting the Docker Engine…' due to NVIDIA Container Toolkit conflicts in WSL2. No data loss, no downgrade needed."
-
Updated
Jun 27, 2026
GPU Stability Mastery 2026: Optimize Load, Fix Crashes, and Cut Heat
windows steam gpu-optimization crash-fix windows-11 stutter-fix frame-time how-to-fish how-to-fish-performance how-to-fish-display-fix how-to-fish-crash-fix how-to-fish-optimizer how-to-fish-gpu-fix how-to-fish-gpu-usage how-to-fish-high-gpu-usage how-to-fish-lag-fix how-to-fish-stutter-fix how-to-fish-black-screen how-to-fish-gray-screen gpu-load
-
Updated
Sep 23, 2026 - HTML
GPU Infrastructure Optimization
-
Updated
Feb 12, 2026 - HTML
Optimized LSTM-based character-level text generator trained on Shakespeare, achieving 3.5x faster training with mixed precision.
python nlp machine-learning natural-language-processing deep-learning text-generation pytorch lstm neural-networks rnn academic-project shakespeare character-level gpu-optimization mixed-precision
-
Updated
Sep 25, 2025 - Python
Collection of Triton operators for transformer models.
pytorch transformer triton attention normalization gpu-optimization pytorch-implementation triton-kernels pytorch-optimization
-
Updated
Jun 25, 2025 - Python
Handwritten Flash Attention 2 CUDA kernel for Blackwell (SM120) with TMA, swizzle, double buffering & warp specialization
-
Updated
May 25, 2026 - Cuda
TrainSight: Sub-millisecond pre-flight LLM dataset profiler predicting HBM OOM crashes, padding waste, and attention entropy dispersion before GPU allocation. Built on the Two-Factor Law of LLM Compute.
python benchmarking transformers data-quality gpu-optimization sft dpo huggingface data-centric-ai llm vllm llm-training kubernetes-init-container grpo pre-flight-checks
-
Updated
Aug 4, 2026 - Python
Progressive CUDA SGEMM kernel(6) optimization (naive → vectorized) with cross-architecture profiling on RTX 3090 vs RTX 4050, benchmarked using Nsight Compute.
benchmark hpc cuda cublas matrix-multiplication cuda-kernels performance-analysis nvidia-gpu sgemm gpu-architecture gpu-optimization nsight-compute
-
Updated
Jul 2, 2026 - Cuda
LLM inference engine implementing continuous batching, PagedAttention, prefix caching, prefill-decode disaggregation, and KV-cache-aware routing.
-
Updated
Aug 7, 2026 - Python
Fused Triton kernels for LIF spiking neuron dynamics. Bit-exact with snnTorch, up to 130x faster, 22% less memory.
performance gpu cuda pytorch triton spiking-neural-networks neuromorphic gpu-optimization kernel-fusion snntorch
-
Updated
Sep 3, 2026 - Python
GPU-accelerated multi-agent incident-triage system: LangGraph orchestration, TensorRT-LLM/Triton inference, Azure-native, fully observable.
multi-agent-systems gpu-optimization triton-inference-server mlops-workflow llm-inference tensorrt-llm azure-ai-search langgraph rag-pipeline agentic-ai
-
Updated
May 31, 2026 - Python
Rule-evolving GPU kernel optimization loop — measurement feedback updates the rule table (matmul 6.4x, batched GEMM 4.5x on A100, KernelBench-validated)
-
Updated
Jul 24, 2026 - Python
Diagnose LLM inference cost and latency, then autotune serving configs with verified experiments. CLI + browser report viewer (v0.1), private autotune engine (v0.2), residual-driven search (v0.3).
autotuning bayesian-optimization gpu-optimization mlops inference-optimization llm-serving vllm llm-inference tensorrt-llm
-
Updated
Sep 14, 2026 - Python
DGX Spark (GB10/SM121) platform support for Meta's KernelAgent — auto-detect, hardware constraints, safe Triton configs
-
Updated
Mar 14, 2026 - Python
Add this topic to your repo
To associate your repository with the gpu-optimization topic, visit your repo's landing page and select "manage topics."