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.

Biopython is a robust, comprehensive toolkit for computational molecular biology written in Python. It enables users to manipulate biological sequences (DNA, RNA, protein), parse standard formats (FASTA, GenBank, PDB, etc.), and programmatically access major bioinformatics databases like NCBI via Entrez. This skill is ideal for developing complex bioinformatics pipelines, including sequence alignment, structural analysis, and phylogenetic tree construction.

DeepChem is a comprehensive Python library designed for applying machine learning to complex fields like chemistry, materials science, and biology. It enables users to perform molecular property prediction (e.g., ADMET, toxicity), train models using specialized graph neural networks (GNNs), and process diverse molecular formats (SMILES, SDF, FASTA). It provides extensive featurization options, making it foundational for drug discovery and computational chemistry research.

The ESM skill provides comprehensive tools for computational biology, leveraging ESM3 and ESM-C models. It enables sophisticated protein sequence generation, structure prediction (sequence to 3D structure), inverse folding (structure to sequence design), and generating function-conditioned proteins. Use it for de novo protein design, feature embedding, and advanced structural analysis in bioinformatics workflows.