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.
7.0
Rating
0
Installs
Machine Learning
Category
Excellent skill for neural network interpretability using nnsight. The description clearly identifies when to use this skill (massive model interpretability, remote execution). SKILL.md provides comprehensive task knowledge with 5 detailed workflows covering activation analysis, patching, remote execution, cross-prompt sharing, and gradient analysis. Code examples are practical and include error-handling patterns. Structure is well-organized with clear sections, comparison tables, and references to additional documentation files. The skill addresses a genuine pain point (accessing 70B+ models without local GPUs) that would require significant tokens for a CLI agent to solve independently. Minor deductions: slightly verbose for a skill file (could consolidate some examples into referenced tutorials), and novelty is moderate since the underlying nnsight library does the heavy lifting, though the skill provides valuable guidance and workflow patterns that meaningfully reduce complexity.
zechenzhangAGI
Skill Author
Loading SKILL.md…
