Results for “rlhf”

Train AI models to be harmless through self-critique and AI feedback using a set of constitutional principles, without requiring human labels for harmful outputs.

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.

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.

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.

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.

Enforce code quality with Ruff in a controlled, iterative workflow: run checks with optional scope and rule overrides, apply safe and unsafe autofixes, review diffs, and resolve remaining findings.

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

Foundational SPARQL patterns covering RDF querying, triple patterns, graph patterns, and semantic web fundamentals. Use when querying RDF data or working with knowledge graphs. This is the entry point for SPARQL development.

Fine-tune and align language models using reinforcement learning with TRL, including SFT, DPO, PPO, GRPO, and reward model training.

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.

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).

CRLF injection playbook. Use when user input reaches HTTP response headers, Location redirects, Set-Cookie values, or log files where carriage-return/line-feed characters can split or inject content.

Train or fine-tune language and vision models using TRL or Unsloth on Hugging Face Jobs cloud infrastructure, with support for SFT, DPO, GRPO, and reward modeling, plus GGUF conversion for local deployment.

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.

Train and fine-tune transformer language models using TRL (Transformers Reinforcement Learning) with support for SFT, DPO, GRPO, KTO, RLOO, and reward model training via CLI commands.

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.

Structures a Request for Comments (RFC) document for proposing significant technical changes that require cross-team input and approval, guiding authors through problem definition, proposed solution, alternatives analysis, and rollout planning.

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

Use RWKV, a linear-time RNN-Transformer hybrid, for efficient long-context inference and training with constant memory usage.

Executes Python code iteratively via an MCP bridge to produce verified results for calculations, data analysis, and task decomposition.

Lint and format Python with Ruff. Use when a user asks to set up Python linting, replace flake8/black/isort, configure code quality rules, or speed up Python code formatting.

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.

Executes Python code iteratively via an MCP bridge to produce verified results for calculations, data analysis, and task decomposition.

Use when turning `bcftools roh` output plus a VCF/BCF into an interactive HTML visualization of ROH segments and homozygosity rates.

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.

Create Request for Information templates for construction. TRIGGERS - Use when user needs help with rfi-template related tasks.

Django REST Framework scaffolding best practices, and gotchas.

Hugging Face CLI to estimate the required memory to load Safetensors or GGUF model weights for inference from the Hugging Face Hub

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

Evaluates LLMs across 60+ academic benchmarks (MMLU, HumanEval, GSM8K, TruthfulQA, HellaSwag). Use when benchmarking model quality, comparing models, reporting academic results, or tracking training progress. Industry standard used by EleutherAI, HuggingFace, and major labs. Supports HuggingFace, vLLM, APIs.