Results for “openrlhf”

High-performance RLHF framework with Ray+vLLM acceleration. Use for PPO, GRPO, RLOO, DPO training of large models (7B-70B+). Built on Ray, vLLM, ZeRO-3. 2× faster than DeepSpeedChat with distributed architecture and GPU resource sharing.

High-performance RLHF framework with Ray+vLLM acceleration. Use for PPO, GRPO, RLOO, DPO training of large models (7B-70B+). Built on Ray, vLLM, ZeRO-3. 2× faster than DeepSpeedChat with distributed architecture and GPU resource sharing.

Comprehensive guide to rlhf. Master the concepts, implementation, best practices, and real-world applications of rlhf in professional environments.

Imported skill tektur_ofl from anthropic

Search the Hugging Face Hub for llama.cpp-compatible GGUF models, select the right quantization, and run them locally with llama-cli or llama-server.

Fine-tune HuggingFace CV, VLM, or LLM models on local NVIDIA GPUs using an NGC PyTorch container, with support for full or LoRA training, dataset handling, and optional model push to the Hub.

Import HD Map or OpenDRIVE files into a RoadRunner scene using MATLAB. Use when loading driving scenes in RoadRunner or RoadRunner Scene Builder, importing RRHD, OpenDRIVE, or other RoadRunner-supported formats for simulation, or verifying Lanelet2-to-RRHD conversion results visually. Requires rrApp handle from roadrunner-core.

HuggingFace hf CLI: search/download/upload models, datasets.

Convert Lanelet2 maps (.osm) to RoadRunner HD Map (.rrhd) format using MATLAB. Use when converting Lanelet2 maps into RoadRunner Scene Builder, building driving scenes from open-source map data, or transforming road network definitions for simulation.

Fine-tunes and evaluates OpenVLA-OFT and OpenVLA-OFT+ policies for robot action generation with continuous action heads, LoRA adaptation, and FiLM conditioning on LIBERO simulation and ALOHA real-world setups. Use when reproducing OpenVLA-OFT paper results, training custom VLA action heads (L1 or diffusion), deploying server-client inference for ALOHA, or debugging normalization, LoRA merge, and cross-GPU issues.

Fine-tune LLMs using reinforcement learning with TRL - SFT for instruction tuning, DPO for preference alignment, PPO/GRPO for reward optimization, and reward model training. Use when need RLHF, align model with preferences, or train from human feedback. Works with HuggingFace Transformers.

Hugging Face Hub CLI (hf) — search, download, and upload models and datasets, manage repos, query datasets with SQL, deploy inference endpoints, manage Spaces and buckets.

Operate Hugging Face Hub repositories, models, datasets, and Spaces via the hf CLI, including downloads, uploads, authentication, and compute jobs.

Hugging Face Hub CLI (hf) — search, download, and upload models and datasets, manage repos, query datasets with SQL, deploy inference endpoints, manage Spaces and buckets.

Fine-tune LLMs using reinforcement learning with TRL - SFT for instruction tuning, DPO for preference alignment, PPO/GRPO for reward optimization, and reward model training. Use when need RLHF, align model with preferences, or train from human feedback. Works with HuggingFace Transformers.

Identify and exploit SSRF vulnerabilities to access internal services, cloud metadata, and restricted network resources during authorized penetration tests.

Create, validate, and consume Google's Open Knowledge Format (OKF) bundles — YAML-frontmatter Markdown files with type / title / description / resource / tags / timestamp fields for portable, interoperable AI-agent knowledge sharing. OKF formalizes the LLM-Wiki pattern into a vendor-neutral open specification so any producer can write and any agent can consume without translation. Routes: use `llm-wiki` for raw source capture + vault maintenance, `obsidian` for Obsidian-vault workflows, `graphify` for durable committed graphs, `scrapling` for web-content extraction into OKF docs. Triggers on: okf, open knowledge format, knowledge bundle, okf document, llm wiki standard, knowledge atom, agent context format, okf frontmatter, okf bundle, knowledge interoperability.

Imported skill outfit_ofl from anthropic

Laravel security best practices for authn/authz, validation, CSRF, mass assignment, file uploads, secrets, rate limiting, and secure deployment.

Fine-tunes and evaluates OpenVLA-OFT and OpenVLA-OFT+ policies for robot action generation with continuous action heads, LoRA adaptation, and FiLM conditioning on LIBERO simulation and ALOHA real-world setups.

RNN+Transformer hybrid with O(n) inference. Linear time, infinite context, no KV cache. Train like GPT (parallel), infer like RNN (sequential). Linux Foundation AI project. Production at Windows, Office, NeMo. RWKV-7 (March 2025). Models up to 14B parameters.

Build RoadRunner HD Map entities in MATLAB — lanes, boundaries, markings, junctions, signs, signals, barriers, parking. Use when creating driving scenes from scratch, authoring road networks for simulation and testing automated driving systems, or assembling RRHD maps from Lanelet2 or other HD map sources.

Query, search, and download public datasets from the City of Raleigh Open Data portal. Use this whenever someone wants to explore city data — crime reports, food inspections, building permits, bike lanes, parks, zoning, traffic, budgets, or any of 200+ public datasets.

Train LLMs with reinforcement learning using verl (Volcano Engine RL), supporting RLHF, GRPO, PPO, and other algorithms for scalable post-training with flexible infrastructure backends.

Imported skill lora_ofl from anthropic

Imported skill gloock_ofl from anthropic

Tracks multi-step work across sessions using the hzl CLI, coordinating sub-agents with projects, tasks, checkpoints, and leases.

Universal Reasoning Framework implementing λο.τ calculus over holarchic structures. Provides severity-based routing (R0-R3 pipelines), modular cognitive architecture (DEC, EVL, PAT, SYN, MEA, HYP, INT), fractal execution patterns, multi-level validation (η≥4, KROG), and adaptive learning. Triggers on: (1) complex multi-step reasoning, (2) high-stakes decisions requiring validation, (3) research synthesis across domains, (4) system design and architecture, (5) crisis management, (6) performance optimization. Implements scale-invariant reasoning from micro (tool calls) through meso (skill composition) to macro (orchestrated workflows).

Fine-tune LLMs using reinforcement learning with TRL - SFT for instruction tuning, DPO for preference alignment, PPO/GRPO for reward optimization, and reward model training. Use when need RLHF, align model with preferences, or train from human feedback. Works with HuggingFace Transformers.

Imported skill jura_ofl from anthropic

Query external data from Oxla — Kafka topics via catalogs, Apache Iceberg tables, and S3/GCS/Azure parquet/ORC files — alongside native Oxla tables. Use when querying Kafka topics with CREATE KAFKA CATALOG or CREATE REDPANDA CATALOG, reading Apache Iceberg tables with the catalog=>path.table syntax, loading or.

Open redirect playbook. Use when URL parameters, form actions, or JavaScript sinks control navigation targets and may redirect users to attacker-controlled destinations.

HuggingFace hf CLI: search/download/upload models, datasets.

Fetch unread GitHub notifications for open PRs where review is requested from a specified team or opened by a team member.

Open Neural Network Exchange format for model interoperability across frameworks. Export models from PyTorch, TensorFlow, and other frameworks to ONNX, optimize with ONNX Runtime, and deploy for cross-platform inference on CPU, GPU, and edge devices.