Community Plugin

research-skills

ML research experiment management. Use when: (1) Planning training runs or hyperparameter sweeps, (2) Documenting model experiments and ablations, (3) Tracking what worked/failed across training sessions

1.0.0

Updated 11 hours ago

Capabilities

Commands

Agents

Skills

Hooks

MCP Servers

LSP Servers

Install

Add the repository(one-time)

/plugin marketplace add Erland366/transformers_playground

Install the plugin

/plugin install erland366-research-skills@Erland366/transformers_playground

Component Details

domain-advisor

inherits

Plan experiments or development tasks using past knowledge. Adapts behavior based on project domain (research, unsloth, cuda) by reading registry.json. Triggers on <advise> or questions like "What should we try next?"

domain-retrospective

inherits

Turn experiment reports and development notes into summaries and reusable skills. Adapts behavior based on project domain (research, unsloth, cuda) by reading registry.json. Triggers on <retrospective> or requests for lessons learned.

gradient-accumulation-deterministic

inherits

Implement gradient accumulation that produces bit-identical results to standard batching. Use when: comparing GA vs non-GA runs, debugging training reproducibility.