grpo-rl-training
Expert guidance for GRPO/RL fine-tuning with TRL for reasoning and task-specific model training
What this skill does
# GRPO/RL Training with TRL
Expert-level guidance for implementing Group Relative Policy Optimization (GRPO) using the Transformer Reinforcement Learning (TRL) library. This skill provides battle-tested patterns, critical insights, and production-ready workflows for fine-tuning language models with custom reward functions.
## When to Use This Skill
Use GRPO training when you need to:
- **Enforce specific output formats** (e.g., XML tags, JSON, structured reasoning)
- **Teach verifiable tasks** with objective correctness metrics (math, coding, fact-checking)
- **Improve reasoning capabilities** by rewarding chain-of-thought patterns
- **Align models to domain-specific behaviors** without labeled preference data
- **Optimize for multiple objectives** simultaneously (format + correctness + style)
**Do NOT use GRPO for:**
- Simple supervised fine-tuning tasks (use SFT instead)
- Tasks without clear reward signals
- When you already have high-quality preference pairs (use DPO/PPO instead)
---
## Core Concepts
### 1. GRPO Algorithm Fundamentals
**Key Mechanism:**
- Generates **multiple completions** for each prompt (group size: 4-16)
- Compares completions within each group using reward functions
- Updates policy to favor higher-rewarded responses relative to the group
**Critical Difference from PPO:**
- No separate reward model needed
- More sample-efficient (learns from within-group comparisons)
- Simpler to implement and debug
**Mathematical Intuition:**
```
For each prompt p:
1. Generate N completions: {c₁, c₂, ..., cₙ}
2. Compute rewards: {r₁, r₂, ..., rₙ}
3. Learn to increase probability of high-reward completions
relative to low-reward ones in the same group
```
### 2. Reward Function Design Philosophy
**Golden Rules:**
1. **Compose multiple reward functions** - Each handles one aspect (format, correctness, style)
2. **Scale rewards appropriately** - Higher weight = stronger signal
3. **Use incremental rewards** - Partial credit for partial compliance
4. **Test rewards independently** - Debug each reward function in isolation
**Reward Function Types:**
| Type | Use Case | Example Weight |
|------|----------|----------------|
| **Correctness** | Verifiable tasks (math, code) | 2.0 (highest) |
| **Format** | Strict structure enforcement | 0.5-1.0 |
| **Length** | Encourage verbosity/conciseness | 0.1-0.5 |
| **Style** | Penalize unwanted patterns | -0.5 to 0.5 |
---
## Implementation Workflow
### Step 1: Dataset Preparation
**Critical Requirements:**
- Prompts in chat format (list of dicts with 'role' and 'content')
- Include system prompts to set expectations
- For verifiable tasks, include ground truth answers as additional columns
**Example Structure:**
```python
from datasets import load_dataset, Dataset
SYSTEM_PROMPT = """
Respond in the following format:
<reasoning>
[Your step-by-step thinking]
</reasoning>
<answer>
[Final answer]
</answer>
"""
def prepare_dataset(raw_data):
"""
Transform raw data into GRPO-compatible format.
Returns: Dataset with columns:
- 'prompt': List[Dict] with role/content (system + user messages)
- 'answer': str (ground truth, optional but recommended)
"""
return raw_data.map(lambda x: {
'prompt': [
{'role': 'system', 'content': SYSTEM_PROMPT},
{'role': 'user', 'content': x['question']}
],
'answer': extract_answer(x['raw_answer'])
})
```
**Pro Tips:**
- Use one-shot or few-shot examples in system prompt for complex formats
- Keep prompts concise (max_prompt_length: 256-512 tokens)
- Validate data quality before training (garbage in = garbage out)
### Step 2: Reward Function Implementation
**Template Structure:**
```python
def reward_function_name(
prompts, # List[List[Dict]]: Original prompts
completions, # List[List[Dict]]: Model generations
answer=None, # Optional: Ground truth from dataset
**kwargs # Additional dataset columns
) -> list[float]:
"""
Evaluate completions and return rewards.
Returns: List of floats (one per completion)
"""
# Extract completion text
responses = [comp[0]['content'] for comp in completions]
# Compute rewards
rewards = []
for response in responses:
score = compute_score(response)
rewards.append(score)
return rewards
```
**Example 1: Correctness Reward (Math/Coding)**
```python
def correctness_reward(prompts, completions, answer, **kwargs):
"""Reward correct answers with high score."""
responses = [comp[0]['content'] for comp in completions]
extracted = [extract_final_answer(r) for r in responses]
return [2.0 if ans == gt else 0.0
for ans, gt in zip(extracted, answer)]
```
**Example 2: Format Reward (Structured Output)**
```python
import re
def format_reward(completions, **kwargs):
"""Reward XML-like structured format."""
pattern = r'<reasoning>.*?</reasoning>\s*<answer>.*?</answer>'
responses = [comp[0]['content'] for comp in completions]
return [1.0 if re.search(pattern, r, re.DOTALL) else 0.0
for r in responses]
```
**Example 3: Incremental Format Reward (Partial Credit)**
```python
def incremental_format_reward(completions, **kwargs):
"""Award partial credit for format compliance."""
responses = [comp[0]['content'] for comp in completions]
rewards = []
for r in responses:
score = 0.0
if '<reasoning>' in r:
score += 0.25
if '</reasoning>' in r:
score += 0.25
if '<answer>' in r:
score += 0.25
if '</answer>' in r:
score += 0.25
# Penalize extra text after closing tag
if r.count('</answer>') == 1:
extra_text = r.split('</answer>')[-1].strip()
score -= len(extra_text) * 0.001
rewards.append(score)
return rewards
```
**Critical Insight:**
Combine 3-5 reward functions for robust training. Order matters less than diversity of signals.
### Step 3: Training Configuration
**Memory-Optimized Config (Small GPU)**
```python
from trl import GRPOConfig
training_args = GRPOConfig(
output_dir="outputs/grpo-model",
# Learning rate
learning_rate=5e-6, # Lower = more stable
adam_beta1=0.9,
adam_beta2=0.99,
weight_decay=0.1,
warmup_ratio=0.1,
lr_scheduler_type='cosine',
# Batch settings
per_device_train_batch_size=1,
gradient_accumulation_steps=4, # Effective batch = 4
# GRPO-specific
num_generations=8, # Group size: 8-16 recommended
max_prompt_length=256,
max_completion_length=512,
# Training duration
num_train_epochs=1,
max_steps=None, # Or set fixed steps (e.g., 500)
# Optimization
bf16=True, # Faster on A100/H100
optim="adamw_8bit", # Memory-efficient optimizer
max_grad_norm=0.1,
# Logging
logging_steps=1,
save_steps=100,
report_to="wandb", # Or "none" for no logging
)
```
**High-Performance Config (Large GPU)**
```python
training_args = GRPOConfig(
output_dir="outputs/grpo-model",
learning_rate=1e-5,
per_device_train_batch_size=4,
gradient_accumulation_steps=2,
num_generations=16, # Larger groups = better signal
max_prompt_length=512,
max_completion_length=1024,
num_train_epochs=1,
bf16=True,
use_vllm=True, # Fast generation with vLLM
logging_steps=10,
)
```
**Critical Hyperparameters:**
| Parameter | Impact | Tuning Advice |
|-----------|--------|---------------|
| `num_generations` | Group size for comparison | Start with 8, increase to 16 if GPU allows |
| `learning_rate` | Convergence speed/stability | 5e-6 (safe), 1e-5 (faster, riskier) |
| `max_completion_length` | Output verbosity | Match your task (512 for reasoning, 256 for short answers) |
| `gradient_accumulation_steps` | Effective batch size | Increase if GPU memorRelated in Productivity
gitea-workflow
IncludedOrchestrate agile development workflows for Gitea repositories using the tea CLI. Use when working with Gitea-hosted repos and asking to 'run the workflow', 'continue working', 'what's next', 'complete the task cycle', 'start my day', 'end the sprint', 'implement the next task', or wanting guided step-by-step development assistance. Keywords: workflow, orchestrate, agile, task cycle, sprint, daily, implement, review, PR, standup, retrospective, gitea, tea.
microsoft-graph-gateway
IncludedRoute Microsoft Graph work in this workspace. Use when users want to read or write Outlook mail, calendar events, contacts, OneDrive or SharePoint files, Teams, Planner, To Do, users, groups, directory data, or arbitrary Microsoft Graph endpoints from VS Code. Prefer WorkIQ for common read scenarios. Use Microsoft Graph for write actions and gap-read scenarios that need exact Graph properties, filters, permissions, or endpoints.
copilotkit
IncludedUse when building with CopilotKit — setup, development, integrations, debugging, upgrading, or contributing. Routes to the appropriate specialized skill based on the task.
wordly-wisdom
IncludedProvides calibrated decision analysis using Charlie Munger-style multiple mental models, inversion, incentive mapping, circle-of-competence checks, misjudgment audits, second-order effects, and forecast updates. Use when the user asks for an oracle take, a hard call, a decision memo, a premortem, an outside view, a red-team, a sanity-check, what am I missing, think this through, or wants a strategy, hire, investment, plan, product, partnership, or major life choice analysed. Avoid for simple factual lookups or time-sensitive legal, medical, or market questions without fresh evidence.
swain-session
IncludedSession management and project status dashboard. Owns the full session lifecycle (start/work/close/resume), focus lane, bookmarks, worktree detection, and tab naming. Also serves as the project status dashboard — shows active epics, progress, actionable next steps, blocked items, tasks, GitHub issues, and recommendations. Worktree creation is deferred to swain-do task dispatch (SPEC-195). Triggers on: 'session', 'status', 'what's next', 'dashboard', 'overview', 'where are we', 'what should I work on', 'show me priorities', 'bookmark', 'focus on', 'session info'.
gandi
IncludedComprehensive Gandi domain registrar integration for domain and DNS management. Register and manage domains, create/update/delete DNS records (A, AAAA, CNAME, MX, TXT, SRV, and more), configure email forwarding and aliases, check SSL certificate status, create DNS snapshots for safe rollback, bulk update zone files, and monitor domain expiration. Supports multi-domain management, zone file import/export, and automated DNS backups. Includes both read-only and destructive operations with safety controls.