Skill UI

This guide details advanced methods for merging multiple specialized LLMs (e.g., combining math, coding, and chat capabilities) without the need for expensive retraining. Learn various techniques such as Linear Merging, SLERP, Task Arithmetic, TIES-Merging, and DARE. Use mergekit to create high-performing, specialized model variants quickly, significantly improving performance and reducing computational costs.

This skill provides comprehensive guidance on training Mixture of Experts (MoE) models, including architectures like Mixtral and DeepSeek-V3. It details techniques for achieving large-scale model capacity with significantly reduced computational costs (e.g., 5x reduction vs. dense models). Topics covered include top-k routing, load balancing, expert parallelism, and integration with frameworks like DeepSpeed and HuggingFace for efficient, resource-constrained training.

A comprehensive guide to leveraging Stable Diffusion models via the HuggingFace Diffusers library. This skill facilitates state-of-the-art image creation, including Text-to-Image generation from prompts, Image-to-Image translation, inpainting, outpainting, and advanced workflow building. Supports various models like SDXL and SD 3.0, ideal for professional creative and computational vision tasks.

This comprehensive runbook provides rapid, structured procedures for incident response on Vast.ai cloud GPU instances. It covers common failure scenarios like spot preemption, training job crashes, and GPU hardware issues. Users can follow the steps for triage, mitigation, recovery, and postmortem analysis to minimize downtime and ensure computational continuity.

Biopython provides Python APIs for sequence manipulation, database access, BLAST automation, structural bioinformatics, and phylogenetics, making it ideal for scripting custom pipelines and batch processing in computational molecular biology workflows.

Denario orchestrates a multi-agent AI to automate research workflows, covering idea generation, methodology development, computational experiments, and LaTeX publication output.

PathML is a Python toolkit for computational pathology workflows, loading 160+ slide formats, preprocessing H&E images, constructing spatial graphs, training HoVer-Net/HACTNet models, and analyzing multiplex imaging for ML-ready pathology data.

RDKit delivers a Python-based cheminformatics toolkit for parsing SMILES/SDF, computing descriptors, generating 2D/3D coordinates, fingerprinting, searching substructures, managing reactions, and visualizing molecules, ideal for drug discovery and computational chemistry workflows.

Provides a detailed, multi-platform workflow to monitor long-running computational experiments. It checks the status of jobs across various environments (SSH, vast.ai, Modal), collects screen logs, fetches structured JSON results, and pulls detailed training metrics (loss curves, accuracy) from Weights & Biases (W&B). Essential for MLOps and automated result review.

LaminDB is an open-source lineage-native lakehouse designed for biological datasets and models. It enables queryable, traceable, and validated data management across various storage formats like S3 and Postgres. Users can track computational workflows, curate data with ontologies via Bionty, and ensure FAIR compliance. Ideal for managing scRNA-seq, spatial transcriptomics, and integrating ML pipelines while maintaining reproducibility and collaboration.

Build local, research-only computational pathology workflows with PathML: load and tile whole-slide images, run preprocessing and QC pipelines, manage h5path data, quantify multiplex images, construct spatial graphs, and plan bounded model inference.

Python bindings to OpenMS for computational mass spectrometry. Analyze proteomics and metabolomics LC-MS/MS data: feature detection, peptide and protein identification, label-free and isobaric quantification, adduct and accurate-mass annotation, plus ready-to-run workflow scripts.