Nnsight Remote Interpretability
技能 已验证 活跃Provides guidance for interpreting and manipulating neural network internals using nnsight with optional NDIF remote execution. Use when needing to run interpretability experiments on massive models (70B+) without local GPU resources, or when working with any PyTorch architecture.
To democratize access to large language model internals for research and experimentation by enabling consistent interpretability workflows across various model sizes and execution environments.
功能
- Interpret and manipulate neural network internals
- Run experiments on massive models (70B+) remotely via NDIF
- Use the same code for local and remote execution
- Support for any PyTorch architecture
- Access activations, gradients, and logits for analysis
使用场景
- Running interpretability experiments on models too large for local GPUs
- Performing multi-token generation interventions
- Sharing activations between different prompts
- Analyzing PyTorch models of any architecture, including custom ones
非目标
- Providing a unified API across all model types (TransformerLens serves this)
- Declarative, shareable interventions (pyvene is for this)
- Training SAEs (SAELens is for this)
- Working exclusively with small models locally (TransformerLens may be simpler)
安装
请先添加 Marketplace
/plugin marketplace add Orchestra-Research/AI-Research-SKILLs/plugin install AI-Research-SKILLs@ai-research-skills质量评分
已验证类似扩展
Nnsight Remote Interpretability
99Provides guidance for interpreting and manipulating neural network internals using nnsight with optional NDIF remote execution. Use when needing to run interpretability experiments on massive models (70B+) without local GPU resources, or when working with any PyTorch architecture.
PyTorch Lightning
100Deep learning framework (PyTorch Lightning). Organize PyTorch code into LightningModules, configure Trainers for multi-GPU/TPU, implement data pipelines, callbacks, logging (W&B, TensorBoard), distributed training (DDP, FSDP, DeepSpeed), for scalable neural network training.
Implementing Llms Litgpt
100Implements and trains LLMs using Lightning AI's LitGPT with 20+ pretrained architectures (Llama, Gemma, Phi, Qwen, Mistral). Use when need clean model implementations, educational understanding of architectures, or production fine-tuning with LoRA/QLoRA. Single-file implementations, no abstraction layers.
Crabbox
100Use Crabbox for OpenClaw remote validation across Linux, macOS, Windows, and WSL2. Default to Blacksmith Testbox for broad Linux proof; includes direct Blacksmith and owned AWS/Hetzner fallback notes when Crabbox fails.
Transformer Lens Interpretability
99Provides guidance for mechanistic interpretability research using TransformerLens to inspect and manipulate transformer internals via HookPoints and activation caching. Use when reverse-engineering model algorithms, studying attention patterns, or performing activation patching experiments.
ML Training Recipes
99Battle-tested PyTorch training recipes for all domains — LLMs, vision, diffusion, medical imaging, protein/drug discovery, spatial omics, genomics. Covers training loops, optimizer selection (AdamW, Muon), LR scheduling, mixed precision, debugging, and systematic experimentation. Use when training or fine-tuning neural networks, debugging loss spikes or OOM, choosing architectures, or optimizing GPU throughput.