26 skills · sorted by relevance
Guides LLM compression through pruning, which removes selected model weights. Covers Wanda and SparseGPT methods without retraining, accuracy comparisons, and structured weight patterns that can accelerate inference on compatible hardware.
Reduce LLM weight usage through pruning, which removes selected model parameters. Explore Wanda and SparseGPT for compression without retraining, choose hardware-friendly patterns for faster inference, and compare the compressed model's accuracy with the original.
Build and train Mixture of Experts models, which activate selected specialist networks for each input. Use DeepSpeed or HuggingFace workflows to configure routing, balance workloads, distribute experts across GPUs, and optimize inference.
Train sentence-transformers models for search, similarity, and reranking, which reorders search results by relevance. Choose a model type, prepare data, configure training, compare results against a baseline, and publish the trained model to Hugging Face Hub.
Train Mixture of Experts models, which activate selected specialist networks for each input. The skill guides architecture, input routing, workload balancing, and GPU distribution using DeepSpeed or HuggingFace, with examples for training configuration and inference optimization.
Train sentence-transformers models for search, similarity, and reranking, which reorders search results by relevance. Choose a model type, prepare data, configure training, compare results against a baseline, and publish the trained model to Hugging Face Hub.
Use SAELens to train and analyze sparse autoencoders, which separate neural network activity into features that may represent recognizable concepts. Explore pretrained features, evaluate reconstruction quality, and investigate how individual features influence language model outputs.
Use SAELens to train and analyze sparse autoencoders, which break model activations into a small set of interpretable features. Explore pretrained features, evaluate reconstruction quality, troubleshoot training, and test how individual features influence generated text.
Train Mixture of Experts (MoE) models using DeepSpeed or HuggingFace. Use when training large-scale models with limited compute (5× cost reduction vs dense models), implementing sparse architectures like Mixtral 8x7B or DeepSeek-V3, or scaling model capacity without proportional compute increase. Covers MoE architectures, routing mechanisms, load balancing, expert parallelism, and inference optimization.
The English summary is being prepared. Showing the original description.
Build graph neural networks, models that learn from connected nodes, with PyTorch Geometric. Supports node and graph classification, link prediction, custom datasets, and training approaches for large graphs or graphs containing multiple types of nodes and connections.
Helps configure and troubleshoot DeepSpeed for training large models across multiple devices. Covers memory optimization, parallel training, and DeepNVMe storage transfers, with examples for tuning data movement and configuring models that divide work among specialized components.
Applies the engineering and research philosophies of Jeff Dean, Chief Scientist at Google DeepMind and Google Research. Reach for this skill whenever you are designing large-scale distributed systems, optimizing latency and energy efficiency, or making architectural decisions about machine learning infrastructure. It should trigger automatically for topics involving hardware-ML co-design, model distillation, sparse activation, massively multi-task models, or scaling systems by 5x to 10x. Use this skill to evaluate system bottlenecks, transition from specialized to unified models, and optimize experimental velocity. Apply his mental models to avoid premature 100x scaling and to treat AI models as reasoning engines rather than memorization databases.
The English summary is being prepared. Showing the original description.
Applies the engineering and research philosophies of Jeff Dean, Chief Scientist at Google DeepMind and Google Research. Reach for this skill whenever you are designing large-scale distributed systems, optimizing latency and energy efficiency, or making architectural decisions about machine learning infrastructure. It should trigger automatically for topics involving hardware-ML co-design, model distillation, sparse activation, massively multi-task models, or scaling systems by 5x to 10x. Use this skill to evaluate system bottlenecks, transition from specialized to unified models, and optimize experimental velocity. Apply his mental models to avoid premature 100x scaling and to treat AI models as reasoning engines rather than memorization databases.
The English summary is being prepared. Showing the original description.
Explore concepts learned by language models using sparse autoencoders, which separate internal activity into a small set of features. Analyze pretrained models, train custom autoencoders, assess reconstruction quality, and experiment with feature based output adjustments.
Provides guidance for training and analyzing Sparse Autoencoders (SAEs) using SAELens to decompose neural network activations into interpretable features. Use when discovering interpretable features, analyzing superposition, or studying monosemantic representations in language models.
The English summary is being prepared. Showing the original description.
Train or fine-tune sentence-transformers models across `SentenceTransformer` (bi-encoder, dense or static embedding model for retrieval, similarity, clustering, classification, paraphrase mining, dedup, multimodal), `CrossEncoder` (reranker, pair scoring for two-stage retrieval / pair classification), `SparseEncoder` (SPLADE, sparse embedding model for learned-sparse retrieval), and `MultiVectorEncoder` (ColBERT / late-interaction, per-token embeddings scored with MaxSim). Covers loss selection, hard-negative mining, evaluators, distillation, LoRA, Matryoshka, and Hugging Face Hub publishing. Use for any sentence-transformers training task.
The English summary is being prepared. Showing the original description.
Sparse Identification of Nonlinear Dynamics (SINDy) — discover governing equations from time-series data. Builds sparse dynamical system models dx/dt = f(x) from measurements using PySINDy. Use when you have trajectory data and want to find the underlying ODE.
The English summary is being prepared. Showing the original description.
Provides guidance for training and analyzing Sparse Autoencoders (SAEs) using SAELens to decompose neural network activations into interpretable features. Use when discovering interpretable features, analyzing superposition, or studying monosemantic representations in language models.
The English summary is being prepared. Showing the original description.
Reduce LLM size and accelerate inference using pruning techniques like Wanda and SparseGPT. Use when compressing models without retraining, achieving 50% sparsity with minimal accuracy loss, or enabling faster inference on hardware accelerators. Covers unstructured pruning, structured pruning, N:M sparsity, magnitude pruning, and one-shot methods.
The English summary is being prepared. Showing the original description.
Provides guidance for training and analyzing Sparse Autoencoders (SAEs) using SAELens to decompose neural network activations into interpretable features. Use when discovering interpretable features, analyzing superposition, or studying monosemantic representations in language models.
The English summary is being prepared. Showing the original description.
Provides guidance for training and analyzing Sparse Autoencoders (SAEs) using SAELens to decompose neural network activations into interpretable features. Use when discovering interpretable features, analyzing superposition, or studying monosemantic representations in language models.
The English summary is being prepared. Showing the original description.
Implement and troubleshoot large-model training with DeepSpeed across multiple GPUs. The skill covers memory optimization, parallel training, and numerical precision choices, alongside examples for storage transfers and models that divide work among specialized components.
Use for formulating, solving, debugging, and validating mixed-integer linear optimization models with open-source solvers, including variable indexing, sparse constraints, linearized costs, solver limits, MIP gaps, incumbent extraction, numerical tolerances, and deterministic output reporting.
The English summary is being prepared. Showing the original description.
Train or fine-tune SentenceTransformer bi-encoders, CrossEncoder rerankers, or SparseEncoder models, including losses, negatives, evaluation, distillation, LoRA, and Matryoshka. Use for sentence-transformers training tasks.
The English summary is being prepared. Showing the original description.