Skill UI

Automates adapting pretrained models via transfer learning by analyzing requirements, generating PyTorch or TensorFlow fine-tuning code, validating data, tracking KPIs, and saving artifacts so you can quickly reuse existing models for new datasets and tasks.

Expert guidance for fine-tuning LLMs with Axolotl, covering YAML configs, LoRA/QLoRA/DPO variants, multimodal support, and practical troubleshooting patterns for development and training workflows.

Expert guidance for fine-tuning LLMs with Axolotl, covering YAML configs, LoRA/QLoRA/DPO variants, multimodal support, and practical troubleshooting patterns for development and training workflows.

Bridges frozen vision encoders and LLMs for high-quality image captioning, VQA, retrieval, and multimodal chat, enabling zero-shot reasoning without fine-tuning and offering efficient Q-Former training.

CLIP enables general-purpose vision-language tasks through zero-shot image classification, similarity scoring, semantic search, and content moderation using pretrained contrastive embeddings without fine-tuning.

This skill guides users on fine-tuning CodeRabbit's review quality. It teaches the AI agent team preferences through PR feedback, integrates custom code guidelines (e.g., security policies), allows customization of the review tone, and implements path filters to drastically reduce false positives and noise. This ensures that all code feedback is highly relevant, actionable, and aligns with specific team standards.

This skill provides comprehensive methods to optimize Deepgram API usage and manage costs. It implements budget-aware transcription pipelines that preemptively check projected spending against set limits, utilizes audio preprocessing techniques (like silence removal via ffmpeg) to minimize billable minutes, and includes API calls to monitor current usage and spending against defined budgets. Ideal for cost-sensitive, production-grade deployments.

Guides ML engineers through dataset prep, PEFT strategy selection, LoRA/QLoRA adapter setup, training monitoring, and deployment checks for practical LLM fine-tuning, quantization, and adapter merge workflows.

Fine-tune LLMs with TRL’s RLHF toolkit, covering SFT, DPO, PPO/GRPO and reward model training so you can align HuggingFace Transformers with human preferences or feedback in post-training workflows.

Provides expert guidance for GRPO training with TRL, covering dataset prep, reward engineering, structured outputs, and multi-objective alignment when preference pairs are scarce.

Provides calibration-free weight quantization for LLMs down to 1-bit precision with multiple optimized backends, ideal for fast deployments with vLLM/HuggingFace and LoRA/PEFT fine-tuning without collecting calibration data.

LitGPT provides clean, readable, and production-ready implementations for over 20 state-of-the-art LLMs (including Llama, Gemma, Phi, Mistral). It supports comprehensive workflows, including full fine-tuning, efficient LoRA/QLoRA adaptation, and pretraining from scratch. Ideal for developers needing academic understanding or robust, deployable model architectures.