Menu
SkillsMP has collected 24 skills from uiuc-kang-lab/rlvr_generalization_bounds. Open a skill to review its source and details.
Showing 24 of 24 collected skills.
Guide for checkpointing — saving, loading, and resuming training with CheckpointRecord. Use when the user asks about saving weights, resuming training, checkpoint management, or the checkpoint lifecycle.
Guide for testing conventions and CI pipelines — unit tests, integration smoke tests, pytest markers, and GitHub Actions workflows. Use when the user asks about testing, CI, running tests, or adding tests for a recipe.
Guide for using completers — TokenCompleter and MessageCompleter for text generation during RL rollouts and evaluation. Use when the user asks about generating text, completing messages, or using completers in RL environments.
Guide for contributing to the tinker-cookbook repo — development setup, code style, type checking, PR process, and design conventions. Use when the user asks about how to contribute, set up the dev environment, code style, or project conventions.
Guide for dataset construction — SupervisedDatasetBuilder, RLDatasetBuilder, ChatDatasetBuilder, and custom dataset creation from JSONL, HuggingFace, or conversation data. Use when the user asks about datasets, data loading, data preparation, or custom data…
Set up and run knowledge distillation (on-policy, off-policy, or multi-teacher) from a teacher model to a student model using the Tinker API. Use when the user wants to distill knowledge, compress models, or train a student from a teacher.
Set up and run Direct Preference Optimization (DPO) training on preference datasets using the Tinker API. Use when the user wants to train with preference data, chosen/rejected pairs, or DPO.
Guide for defining RL environments — the Env protocol, EnvGroupBuilder, RLDataset, and custom environment creation. Use when the user asks about RL environments, reward functions, or how to define custom tasks for RL training.
Guide for evaluation — inline evaluators, Inspect AI integration, and custom evaluators for measuring training progress. Use when the user asks about evaluation, metrics, benchmarks, or how to measure model quality during training.
Set up and run reinforcement learning with verifiable rewards (RLVR/GRPO) for math, code, or custom environments using the Tinker API. Use when the user wants to do RL training, GRPO, reward-based optimization, or train with verifiable rewards.
Guide for hyperparameter selection — learning rate formulas, LoRA rank, batch size, group size, schedules, and model-specific tuning. Use when the user asks about learning rate, batch size, hyperparameter tuning, or how to configure training parameters.
Guide for training outputs, metrics logging, logtree reports, tracing/profiling, and debugging training runs. Use when the user asks about training logs, metrics, debugging, tracing, profiling, timing, Gantt charts, or understanding training output files.
Create, update, or organize Claude Code skills in this repo. Use when adding a new skill, reviewing existing skills for consistency, or maintaining the skill taxonomy.
Guide for choosing models in Tinker — available model families, model types (base, instruction, reasoning, hybrid, vision), architecture (dense vs MoE), and how to match renderers to models. Use when the user asks which model to use, what models are…
Set up and run multi-turn RL training for interactive environments (terminal tasks, tool use, search/RAG, games) using the Tinker API. Use when the user wants multi-turn RL, agentic training, tool-use RL, or interactive environment training.
Scaffold a new training recipe for the Tinker cookbook following repo conventions. Use when the user wants to create a new recipe, training script, or experiment.
Guide for using renderers — the bridge between chat-style messages and token sequences. Covers renderer setup, TrainOnWhat, vision inputs, model family matching, and custom renderers. Use when the user asks about renderers, tokenization, message formatting,…
Set up and run the full RLHF pipeline (SFT, reward model training, RL from reward model) using the Tinker API. Use when the user wants to do RLHF, train a reward model, or run the full preference-based RL pipeline.
Guide for installing Tinker, setting up the environment, getting an API key, and verifying everything works. Use when the user is getting started, setting up their environment, or troubleshooting installation issues.
Set up and run supervised fine-tuning (SFT) on instruction or chat datasets using the Tinker API. Use when the user wants to do instruction tuning, chat fine-tuning, or supervised learning.
Guide for the Tinker CLI — managing training runs, checkpoints, downloading weights, and publishing to HuggingFace. Use when the user asks about CLI commands, listing runs, managing checkpoints from the terminal, or uploading to HF Hub.
Guide for using the Tinker Python SDK APIs — ServiceClient, TrainingClient, SamplingClient, RestClient, forward_backward, optim_step, sampling, and async patterns. Use when the user asks about Tinker API basics, how to call training/sampling, or how the SDK…
Reference for Tinker SDK types — Datum, ModelInput, TensorData, SamplingParams, response types, error types, and helper functions. Use when the user needs to build training data, construct model inputs, understand response objects, or handle errors.
Guide for the weight lifecycle — downloading trained weights from Tinker, merging LoRA adapters into HuggingFace models, and publishing to HuggingFace Hub. Use when the user asks about exporting, downloading, merging, or publishing trained model weights.