Skill UI

A comprehensive guide for training or fine-tuning advanced sentence-transformer models. This skill assists in selecting the correct architecture—including Bi-encoders (SentenceTransformer), CrossEncoders (for reranking), and SparseEncoders—for various Natural Language Processing (NLP) tasks. Use cases include semantic retrieval, similarity matching, document clustering, and pair-wise classification.

This skill provides expert capabilities for training and fine-tuning transformer language models using the TRL (Transformers Reinforcement Learning) library. It supports state-of-the-art post-training methods, including Supervised Fine-Tuning (SFT), Direct Preference Optimization (DPO), Group Relative Policy Optimization (GRPO), and Reward Model training. Use this skill to align and customize foundation models for advanced tasks via CLI commands.

This skill converts vague speed requests into a structured optimization process. It establishes a baseline, defines correctness gates, and iteratively tests variants against metrics like latency or cost. Each variant is benchmarked and validated before promotion, ensuring reproducible improvements without breaking functionality. Ideal for performance tuning, recursive optimization, and selecting the best implementation through measured tests.

Provides comprehensive configuration management for local LLM inference using RuVLLM. Users can select models, perform task-specific fine-tuning with MicroLoRA adapters, or implement continuous, real-time adaptation using SONA. This tool guides the entire advanced LLM workflow, covering status checks, configuration generation, and advanced adaptation loops.

Utilize zero-shot NER to extract custom entity types (e.g., Drug, Symptom, Device) from complex clinical or biomedical text without requiring any labeled training data or fine-tuning. OpenMed wraps GLiNER/GLiNER2, allowing users to define labels on the fly. The workflow involves building a model index and then running inference via CLI or Python API, making it ideal for rapid prototyping or unstable schemas.

Determines an organization's role under the EU AI Act—provider, deployer, importer, distributor, or quasi-provider—and assesses Article 25 substantial modification risk. Useful for AI Act role questions, value-chain responsibilities, German terminology (Anbieter/Betreiber), and finetuning implications.

This skill orchestrates advanced Machine Learning training workflows on dedicated compute clusters. It manages complex stages (pre-training, fine-tuning, RL) by ensuring strict adherence to immutable manifests, fresh booking entitlements, and multi-step confirmation gates. It acts as a canonical interface for job submission, emphasizing security, idempotency, and resource validation before execution.

Guide to selecting and optimizing embedding models for vector search and RAG applications, covering model comparison, chunking, fine-tuning, and dimension reduction.

This skill covers setting up infrastructure for fine-tuning Large Language Models (LLMs) using techniques like QLoRA, LoRA, and full fine-tuning. It includes configuring distributed training with DeepSpeed or FSDP, utilizing frameworks such as Hugging Face TRL and Axolotl, and implementing alignment methods like RLHF and DPO. Suitable for domain-specific model adaptation and production deployment.

Comprehensive guide to LLM Operations (LLMOps). Covers advanced topics like Retrieval-Augmented Generation (RAG), vector database management (Pinecone, Chroma), embeddings, fine-tuning, prompt engineering, and cost optimization. Essential for bridging the gap between AI prototypes and reliable, scalable, production-grade AI systems.

Guides agents through bounded parameter optimization for configurable systems with measurable objectives: define search space, budget, and baseline, run trials one at a time, log evidence, and recommend a configuration for HPO, inference tuning, simulations, or RL/control experiments.

Train and fine-tune transformer language models with Hugging Face TRL via CLI. Covers SFT, DPO, GRPO, RLOO, KTO and reward model training, plus LoRA adapters, YAML configs and distributed multi-GPU runs.