--- id: "/model-training/training-infrastructure/page/4/" title: "Training Infrastructure: 123 curated resources · page 4" creator: null creator_url: null creator_x: null url: "https://skillissue.dev/model-training/training-infrastructure/page/4/" level: null level_max: null category: null subcategory: null region: null license: null license_scope: null access: null models: null source_date: null date_added: null resource_kind: null verification_basis: null access_evidence: null domain_top: null domain_sub: null last_verified: null --- # Training Infrastructure: 123 curated resources Curated material on training infrastructure within model training. [Beginner](/model-training/beginner/index.md) [Intermediate](/model-training/intermediate/index.md) [Advanced](/model-training/advanced/index.md) [Expert](/model-training/expert/index.md) 123 resources · page 4 of 4 - ### [Building an efficient harness for advanced enterprise work (Glean)](/model-training/training-infrastructure/glean-building-an-efficient-harness-for-advanced-enterprise-work/index.md) Creator not supplied; see original source. Glean post on its coding-agent harness using code execution, tool-output truncation and progressive disclosure to cut tokens. Advanced · unknown · unknown resource · en - ### [Grouped-Query Attention](/model-training/training-infrastructure/grouped-query-attention/index.md) Creator not supplied; see original source. arXiv paper (2023) introducing grouped-query attention (GQA) and uptraining multi-head models to it. Advanced · unknown · unknown resource · en - ### [MindSpore 中文 API 文档](/model-training/training-infrastructure/mindspore-api/index.md) Creator not supplied; see original source. MindSpore API documentation in Chinese for Huawei's deep learning training and inference framework. Advanced · unknown · unknown resource · zh - ### [Modal guides](/model-training/training-infrastructure/modal-guides/index.md) Creator not supplied; see original source. Modal guides: deploying model serving, GPU jobs and scalable applications on Modal's serverless platform. Advanced · unknown · unknown resource · en - ### [Scaling to 1 million concurrent sandboxes in seconds (Modal)](/model-training/training-infrastructure/modal-scaling-to-1m-concurrent-sandboxes/index.md) Creator not supplied; see original source. Modal engineering post on building a scheduler that runs one million concurrent code sandboxes per workspace. Advanced · unknown · unknown resource · en - ### [Moore Threads docs center](/model-training/training-infrastructure/moore-threads-docs-center/index.md) Creator not supplied; see original source. Moore Threads documentation centre for its Chinese GPUs and the MUSA SDK (in Chinese). Advanced · unknown · unknown resource · zh - ### [PEFT documentation](/model-training/training-infrastructure/peft-documentation/index.md) Creator not supplied; see original source. Hugging Face PEFT documentation: parameter-efficient fine-tuning methods such as LoRA, with recipes. Advanced · unknown · unknown resource · en - ### [Qiskit Machine Learning](/model-training/training-infrastructure/qiskit-machine-learning/index.md) Creator not supplied; see original source. Qiskit Machine Learning documentation: quantum kernels and quantum neural networks for ML experiments. Advanced · unknown · unknown resource · en - ### [QLoRA](/model-training/training-infrastructure/qlora/index.md) Creator not supplied; see original source. arXiv paper (2023) introducing QLoRA, fine-tuning 4-bit quantized LLMs with low-rank adapters. Advanced · unknown · unknown resource · en - ### [Ray Serve documentation](/model-training/training-infrastructure/ray-serve-documentation/index.md) Creator not supplied; see original source. Ray Serve documentation: scalable library for building online inference APIs and composing model deployments. Advanced · unknown · unknown resource · en - ### [SGLang docs: HiCache best practices](/model-training/training-infrastructure/sglang-hicache-best-practices/index.md) Creator not supplied; see original source. SGLang HiCache best practices: configuring hierarchical KV caching across GPU, CPU and storage. Advanced · unknown · unknown resource · en - ### [Tinker documentation](/model-training/training-infrastructure/tinker-documentation/index.md) Creator not supplied; see original source. Tinker documentation from Thinking Machines Lab: training API for fine-tuning open-weight LLMs, with SDK reference and recipes. Advanced · unknown · unknown resource · en - ### [AdapTive-LeArning Speculator System (ATLAS) (Together AI)](/model-training/training-infrastructure/together-ai-atlas-runtime-learning-speculators/index.md) Creator not supplied; see original source. Together AI post on ATLAS, a speculative decoding system whose speculators keep learning from the live workload. Advanced · unknown · unknown resource · en - ### [Cache-aware prefill-decode disaggregation (Together AI)](/model-training/training-infrastructure/together-ai-cache-aware-prefill-decode-disaggregation/index.md) Creator not supplied; see original source. Together AI post on cache-aware prefill-decode disaggregation for faster long-context inference serving. Advanced · unknown · unknown resource · en - ### [Together AI documentation](/model-training/training-infrastructure/together-ai-documentation/index.md) Creator not supplied; see original source. Together AI documentation for running, fine-tuning and serving open-source models. Advanced · unknown · unknown resource · en - ### [vLLM docs: Automatic prefix caching](/model-training/training-infrastructure/vllm-automatic-prefix-caching-design/index.md) Creator not supplied; see original source. vLLM design document on automatic prefix caching and how shared prompt prefixes reuse the KV cache. Advanced · unknown · unknown resource · en - ### [vLLM docs](/model-training/training-infrastructure/vllm-docs/index.md) Creator not supplied; see original source. vLLM latest documentation: installation, serving, supported models and quantisation for the LLM inference engine. Advanced · unknown · unknown resource · en - ### [CMU 11-868 LLM Systems](/model-training/training-infrastructure/cmu-11-868-llm-systems/index.md) Creator not supplied; see original source. CMU 11-868 LLM Systems course site: systems for training and serving large language models, with schedule and readings. Expert · unknown · unknown resource · en - ### [CUDA C++ Programming Guide](/model-training/training-infrastructure/cuda-c-programming-guide/index.md) Creator not supplied; see original source. NVIDIA CUDA C++ Programming Guide: reference for the CUDA programming model, memory hierarchy and interfaces. Expert · unknown · unknown resource · en - ### [EAGLE](/model-training/training-infrastructure/eagle/index.md) Creator not supplied; see original source. arXiv paper (2024) introducing EAGLE, speculative sampling with feature-level draft heads. Expert · unknown · unknown resource · en - ### [Fast Inference from Transformers via Speculative Decoding](/model-training/training-infrastructure/fast-inference-from-transformers-via-speculative-decoding/index.md) Creator not supplied; see original source. arXiv paper (2022) introducing speculative decoding to speed up transformer inference without changing outputs. Expert · unknown · unknown resource · en - ### [FlashAttention](/model-training/training-infrastructure/flashattention/index.md) Creator not supplied; see original source. arXiv paper (2022) introducing FlashAttention, an IO-aware exact attention algorithm for GPUs. Expert · unknown · unknown resource · en - ### [How to Scale Your Model (JAX scaling book)](/model-training/training-infrastructure/how-to-scale-your-model-jax-scaling-book/index.md) Creator not supplied; see original source. How to Scale Your Model: free online book on TPU/GPU rooflines, transformer math, parallelism and serving for LLMs. Expert · unknown · unknown resource · en - ### [The Ultra-Scale Playbook](/model-training/training-infrastructure/hugging-face-the-ultra-scale-playbook/index.md) by [Hugging Face (nanotron)](https://nanotron-ultrascale-playbook.static.hf.space/index.html) Understand how parallelism changes memory use and training throughput. Expert · other · closed resource · en - ### [PagedAttention / vLLM paper](/model-training/training-infrastructure/pagedattention-vllm-paper/index.md) Creator not supplied; see original source. arXiv paper (2023) introducing PagedAttention and the vLLM serving system for efficient KV-cache memory management. Expert · unknown · unknown resource · en - ### [SGLang docs](/model-training/training-infrastructure/sglang-docs/index.md) Creator not supplied; see original source. SGLang documentation: LLM serving framework with RadixAttention prefix caching and fast structured outputs. Expert · unknown · unknown resource · en - ### [How to Optimize a CUDA Matmul Kernel for cuBLAS-like Performance (Simon Boehm)](/model-training/training-infrastructure/simon-boehm-cuda-matmul-worklog/index.md) Creator not supplied; see original source. Simon Boehm's worklog optimising a CUDA matrix-multiplication kernel step by step towards cuBLAS performance. Expert · unknown · unknown resource · en [Previous page](/model-training/training-infrastructure/page/3/index.md) [Plain-text Markdown copy](/model-training/training-infrastructure/page/4/index.md)