Agent Skills by hhhi21g
Showing 95 vetted skills indexed across 1 GitHub repositories.
hhhi21gdspy
Optimize prompts and build modular AI systems with dspy.
hhhi21gopenrlhf-training
Train large language models with RLHF using Ray and vLLM.
hhhi21gnemo-guardrails
Implement runtime safety checks for LLM applications.
hhhi21gevaluating-cosmos-policy
Evaluate NVIDIA Cosmos Policy in LIBERO and RoboCasa simulation environments.
hhhi21gllamaindex
Create RAG applications with document indexing and querying via LlamaIndex.
hhhi21gdeepspeed
Optimize distributed training with DeepSpeed ZeRO, pipeline parallelism, and mixed precision.
hhhi21gcrewai-multi-agent
Coordinate autonomous AI agents with the crewai library for collaborative tasks.
hhhi21gmoe-training
Train large-scale Mixture of Experts models with DeepSpeed or HuggingFace.
hhhi21gnnsight-remote-interpretability
Run nnsight interpretability experiments on large models via NDIF remote execution.
hhhi21glangsmith-observability
Trace and monitor LLM interactions in AI applications with LangSmith.
hhhi21gaudiocraft-audio-generation
Generate music and sound effects from text using audiocraft and torch.
hhhi21gnemo-curator
Automates GPU-accelerated deduplication, quality filtering, and PII redaction for LLM training datasets.
hhhi21ghuggingface-accelerate
Simplifies PyTorch distributed training across multiple GPUs and machines via unified API and automatic device placement, mixed precision, scaling.
hhhi21gchroma
Store embeddings and metadata in a vector database for semantic search and RAG applications.
hhhi21gprompt-guard
Detect prompt injections and jailbreak attempts in LLM applications.
hhhi21gevaluating-llms-harness
Automate language model evaluation across 60+ academic benchmarks.
hhhi21gllama-factory
Fine-tune large language models through LLaMA-Factory's web interface.
hhhi21gacademic-plotting
Generate publication-quality diagrams and data-driven figures for machine learning papers.
hhhi21gknowledge-distillation
Compress large language models using knowledge distillation with Transformers, Torch, and Datasets.
hhhi21gsimpo-training
Optimizes language models for alignment and performance without a reference model.
hhhi21gfine-tuning-openvla-oft
Fine-tune and evaluate OpenVLA-OFT policies for robot action generation.
hhhi21gslime-rl-training
Scale RL post-training for LLMs with Megatron-LM and SGLang.
hhhi21gblip-2-vision-language
Generate textual descriptions for images using a pre-trained vision-language model.
hhhi21gpytorch-fsdp2
Implement PyTorch FSDP2 for distributed training with parameter sharding and checkpointing.