model-setup
CommunityLoad, inspect, and surgically modify models.
Software Engineering#hooks#ablation#transformers#interpretability#architecture-discovery#model-loading
AuthorEmaRimoldi
Version1.0.0
Installs0
System Documentation
What problem does it solve?
Streamslining loading, configuring, and surgically modifying models for experiment-ready setups by reducing boilerplate around loaders, hooks, and ablation utilities.
Core Features & Use Cases
- Model Loading: Load models from HuggingFace Hub or local checkpoints with correct config and dtype handling.
- Architecture Discovery: Generate architecture reports (model-architecture.json) and enumerate layers, parameters, and components.
- Hook Infrastructure: Attach forward hooks to extract intermediate activations and support controlled ablation.
- Ablation & Fine-tuning: Provide zero/mean activation ablation, activation patching, and reversible state changes for safe experiments.
- Comparison Tools: Enable representation similarity analyses (CKA/RSA) across models and checkpoints.
- Hydra Configs: Output Hydra YAMLs for model loading and experiment pipelines.
Quick Start
Load a pretrained model, attach a hook to an attention layer, run a test input to collect activations, and generate the architecture report.
Dependency Matrix
Required Modules
None requiredComponents
references
💻 Claude Code Installation
Recommended: Let Claude install automatically. Simply copy and paste the text below to Claude Code.
Please help me install this Skill: Name: model-setup Download link: https://github.com/EmaRimoldi/Claude-scholar-extended/archive/main.zip#model-setup Please download this .zip file, extract it, and install it in the .claude/skills/ directory.
Agent Skills Search Helper
Install a tiny helper to your Agent, search and equip skill from 471,000+ vetted skills library on demand.