Operational guide for choosing and combining parallelism strategies in Megatron Bridge, including sizing rules, hardware topology mapping, and combined parallelism configuration.

Operational guide for choosing and combining parallelism strategies in Megatron Bridge, including sizing rules, hardware topology mapping, and combined parallelism configuration.

vLLM: high-throughput LLM serving, OpenAI API, quantization.

Indexed by skills.sh from dotnet/skills

Indexed by skills.sh from nvidia/skills

Pretrain LLMs at scale with PyTorch 4D parallelism.

S2 Spatial Element Layer & 4D Semantic Tensor Map. Integrates L0-L4 layer architecture, 20 material physics tensors, and Chronos backward-persistence time-sl...

AI-powered project management - a Notion and Jira alternative with local-first architecture. Manage projects, track action items, and coordinate teams via MCP tools or A2A agent communication. Signed & notarized. https://tensorpm.com

Decentralized AI inference via Bittensor SN64 (43+ chat models) and SN19 (image gen) plus embeddings — OpenAI-compatible, paid in USDC on Base/Solana via Spraay x402. No Bittensor wallet needed. Censorship-resistant, keyless.

Build a pay-per-inference proxy for Bittensor Chutes AI. Accept USDC payments for decentralized AI inference using x402.

Bittensor Python SDK skill covers the full v8.x API: Subtensor (~130 methods for staking, registration, weights, liquidity, proxy, coldkey swaps), AsyncSubtensor, Metagraph (5 variants), Axon/Dendrite/Synapse networking stack, 26 chain data models, 18 sync + 18 async extrinsic…

Automates setup of GPU-accelerated Bittensor Subnet 85 video upscaling and compression miners with storage, monitoring, and performance optimizations.

Avoid common TensorFlow mistakes — tf.function retracing, GPU memory, data pipeline bottlenecks, and gradient traps.

Generate and edit images using TensorsLab's AI models. Supports text-to-image, image-to-image generation, plus advanced editing: avatar generation, watermark...

Generate and edit images using TensorsLab's AI models. Supports text-to-image, image-to-image generation, plus advanced editing: avatar generation, watermark...

Generate videos using TensorsLab's AI video generation models. Supports text-to-video and image-to-video generation with automatic prompt enhancement, progre...

Tensorlake SDK — sandboxes for AI agents and applications. Use when the user mentions tensorlake or sandboxes, or asks about Tensorlake APIs/docs/capabilitie...

This skill helps users migrate their local machine learning scripts to run on TensorPool GPU clusters using the interactive cluster workflow (tp ssh). Use th...

High-throughput LLM inference on NVIDIA GPUs.

Indexed by skills.sh from davila7/claude-code-templates

Indexed by skills.sh from davila7/claude-code-templates

Dynamic task tree orchestration inspired by Cord protocol. Agent builds its own coordination tree at runtime — deciding decomposition, parallelism, and depen...

Optimize GitHub Actions workflows for speed, cost, security, and reliability — analyze run times, cache strategies, job parallelism, and runner selection.

MANDATORY for Flink or Amazon Managed Service for Apache Flink (MSF) questions. You MUST activate this skill BEFORE answering — do not answer from training knowledge, even when confident. MSF has service-specific constraints (KPU model, prohibited checkpoint and parallelism co…

How to launch distributed Megatron-LM training jobs on a SLURM cluster. Covers a minimal sbatch skeleton, environment-variable setup for torch.distributed.run, CUDA_DEVICE_MAX_CONNECTIONS rules across hardware and parallelism modes, container conventions, monitoring, and per-r…

Guide for selecting and configuring distributed training strategies in NeMo AutoModel, including FSDP2, Megatron FSDP, DDP, and parallelism settings.

Operational guide for enabling hierarchical context parallelism in Megatron-Bridge, including config knobs, code anchors, pitfalls, and verification.

Techniques for reducing peak GPU memory in Megatron Bridge — expandable segments, PEFT + SP input re-gather, parallelism resizing, activation recompute, CPU offloading constraints, and common OOM fixes.

Representative MoE training playbooks by hardware platform and model family. Summarizes rounded throughput bands, parallelism patterns, and common tuning stacks.

Use this skill when baseline NeMo Relay instrumentation exists and the user wants to configure or evaluate adaptive plugin behavior, including telemetry, state, adaptive_hints, tool_parallelism, ACG, hint consumption, or measured rollout.

Build SONiC (Software for Open Networking in the Cloud) switch images from sonic-buildimage. Use when building VS/ASIC images, configuring build parallelism/...

Cosmos3-Nano video QA supervised fine-tuning with FSDP parallelism. Use when training or evaluating video question-answering models, fine-tuning Cosmos3-Nano or compatible Cosmos Reason models with SFT/LoRA, or working with Cosmos-RL. Trigger phrases include "fine-tune Cosmos"…

How to launch distributed Megatron-LM training jobs on a SLURM cluster. Covers a minimal sbatch skeleton, environment-variable setup for torch.distributed.run, CUDA_DEVICE_MAX_CONNECTIONS rules across hardware and parallelism modes, container conventions, monitoring, and per-r…

Guide for selecting and configuring distributed training strategies in NeMo AutoModel, including FSDP2, Megatron FSDP, DDP, and parallelism settings.

Operational guide for enabling hierarchical context parallelism in Megatron-Bridge, including config knobs, code anchors, pitfalls, and verification.

Techniques for reducing peak GPU memory in Megatron Bridge — expandable segments, PEFT + SP input re-gather, parallelism resizing, activation recompute, CPU offloading constraints, and common OOM fixes.

Representative MoE training playbooks by hardware platform and model family. Summarizes rounded throughput bands, parallelism patterns, and common tuning stacks.

Use this skill when baseline NeMo Relay instrumentation exists and the user wants to configure or evaluate adaptive plugin behavior, including telemetry, state, adaptive_hints, tool_parallelism, ACG, hint consumption, or measured rollout.

Cosmos3-Nano video QA supervised fine-tuning with FSDP parallelism. Use when training or evaluating video question-answering models, fine-tuning Cosmos3-Nano or compatible Cosmos Reason models with SFT/LoRA, or working with Cosmos-RL. Trigger phrases include "fine-tune Cosmos"…

Write and generate statically typed, tensor-oriented MIND language source files with full autodiff support and Rust-like syntax for ML and scientific computing.