home / skills / orchestra-research / ai-research-skills / slime

slime skill

/06-post-training/slime

This skill helps you accelerate RL-based LLM post-training with slime's Megatron-LM and SGLang for scalable data generation and rollout.

npx playbooks add skill orchestra-research/ai-research-skills --skill slime

Review the files below or copy the command above to add this skill to your agents.

Files (3)
SKILL.md
11.3 KB
---
name: slime-rl-training
description: Provides guidance for LLM post-training with RL using slime, a Megatron+SGLang framework. Use when training GLM models, implementing custom data generation workflows, or needing tight Megatron-LM integration for RL scaling.
version: 1.0.0
author: Orchestra Research
license: MIT
tags: [Reinforcement Learning, Megatron-LM, SGLang, GRPO, Post-Training, GLM]
dependencies: [sglang-router>=0.2.3, ray, torch>=2.0.0, transformers>=4.40.0]
---

# slime: LLM Post-Training Framework for RL Scaling

slime is an LLM post-training framework from Tsinghua's THUDM team, powering GLM-4.5, GLM-4.6, and GLM-4.7. It connects Megatron-LM for training with SGLang for high-throughput rollout generation.

## When to Use slime

**Choose slime when you need:**
- Megatron-LM native training with SGLang inference
- Custom data generation workflows with flexible data buffers
- Training GLM, Qwen3, DeepSeek V3, or Llama 3 models
- Research-grade framework with production backing (Z.ai)

**Consider alternatives when:**
- You need enterprise-grade stability features → use **miles**
- You want flexible backend swapping → use **verl**
- You need PyTorch-native abstractions → use **torchforge**

## Key Features

- **Training**: Megatron-LM with full parallelism support (TP, PP, DP, SP)
- **Rollout**: SGLang-based high-throughput generation with router
- **Data Buffer**: Flexible prompt management and sample storage
- **Models**: GLM-4.x, Qwen3, DeepSeek V3/R1, Llama 3

## Architecture Overview

```
┌─────────────────────────────────────────────────────────┐
│                    Data Buffer                          │
│ - Prompt initialization and management                  │
│ - Custom data generation and filtering                  │
│ - Rollout sample storage                                │
└─────────────┬───────────────────────────┬───────────────┘
              │                           │
┌─────────────▼───────────┐ ┌─────────────▼───────────────┐
│ Training (Megatron-LM)  │ │ Rollout (SGLang + Router)   │
│ - Actor model training  │ │ - Response generation       │
│ - Critic (optional)     │ │ - Reward/verifier output    │
│ - Weight sync to rollout│ │ - Multi-turn support        │
└─────────────────────────┘ └─────────────────────────────┘
```

## Installation

```bash
# Recommended: Docker
docker pull slimerl/slime:latest
docker run --rm --gpus all --ipc=host --shm-size=16g \
  -it slimerl/slime:latest /bin/bash

# Inside container
cd /root/slime && pip install -e . --no-deps
```

### From Source

```bash
git clone https://github.com/THUDM/slime.git
cd slime
pip install -r requirements.txt
pip install -e .
```

## Quick Start: GRPO Training

```bash
# Source model configuration
source scripts/models/qwen3-4B.sh

# Launch training
python train.py \
    --actor-num-nodes 1 \
    --actor-num-gpus-per-node 4 \
    --rollout-num-gpus 4 \
    --advantage-estimator grpo \
    --use-kl-loss --kl-loss-coef 0.001 \
    --rollout-batch-size 32 \
    --n-samples-per-prompt 8 \
    --global-batch-size 256 \
    --num-rollout 3000 \
    --prompt-data /path/to/data.jsonl \
    ${MODEL_ARGS[@]} ${CKPT_ARGS[@]}
```

---

## Workflow 1: Standard GRPO Training

Use this workflow for training reasoning models with group-relative advantages.

### Prerequisites Checklist
- [ ] Docker environment or Megatron-LM + SGLang installed
- [ ] Model checkpoint (HuggingFace or Megatron format)
- [ ] Training data in JSONL format

### Step 1: Prepare Data

```python
# data.jsonl format
{"prompt": "What is 2 + 2?", "label": "4"}
{"prompt": "Solve: 3x = 12", "label": "x = 4"}
```

Or with chat format:
```python
{
    "prompt": [
        {"role": "system", "content": "You are a math tutor."},
        {"role": "user", "content": "What is 15 + 27?"}
    ],
    "label": "42"
}
```

### Step 2: Configure Model

Choose a pre-configured model script:

```bash
# List available models
ls scripts/models/
# glm4-9B.sh, qwen3-4B.sh, qwen3-30B-A3B.sh, deepseek-v3.sh, llama3-8B.sh, ...

# Source your model
source scripts/models/qwen3-4B.sh
```

### Step 3: Launch Training

```bash
python train.py \
    --actor-num-nodes 1 \
    --actor-num-gpus-per-node 8 \
    --rollout-num-gpus 8 \
    --advantage-estimator grpo \
    --use-kl-loss \
    --kl-loss-coef 0.001 \
    --prompt-data /path/to/train.jsonl \
    --input-key prompt \
    --label-key label \
    --apply-chat-template \
    --rollout-batch-size 32 \
    --n-samples-per-prompt 8 \
    --global-batch-size 256 \
    --num-rollout 3000 \
    --save-interval 100 \
    --eval-interval 50 \
    ${MODEL_ARGS[@]}
```

### Step 4: Monitor Training
- [ ] Check TensorBoard: `tensorboard --logdir outputs/`
- [ ] Verify reward curves are increasing
- [ ] Monitor GPU utilization across nodes

---

## Workflow 2: Asynchronous Training

Use async mode for higher throughput by overlapping rollout and training.

### When to Use Async
- Large models with long generation times
- High GPU idle time in synchronous mode
- Sufficient memory for buffering

### Launch Async Training

```bash
python train_async.py \
    --actor-num-nodes 1 \
    --actor-num-gpus-per-node 8 \
    --rollout-num-gpus 8 \
    --advantage-estimator grpo \
    --async-buffer-size 4 \
    --prompt-data /path/to/train.jsonl \
    ${MODEL_ARGS[@]}
```

### Async-Specific Parameters

```bash
--async-buffer-size 4        # Number of rollouts to buffer
--update-weights-interval 2  # Sync weights every N rollouts
```

---

## Workflow 3: Multi-Turn Agentic Training

Use this workflow for training agents with tool use or multi-step reasoning.

### Prerequisites
- [ ] Custom generate function for multi-turn logic
- [ ] Tool/environment interface

### Step 1: Define Custom Generate Function

```python
# custom_generate.py
async def custom_generate(args, samples, evaluation=False):
    """Multi-turn generation with tool calling."""
    for sample in samples:
        conversation = sample.prompt

        for turn in range(args.max_turns):
            # Generate response
            response = await generate_single(conversation)

            # Check for tool call
            tool_call = extract_tool_call(response)
            if tool_call:
                tool_result = execute_tool(tool_call)
                conversation.append({"role": "assistant", "content": response})
                conversation.append({"role": "tool", "content": tool_result})
            else:
                break

        sample.response = response
        sample.reward = compute_reward(sample)

    return samples
```

### Step 2: Launch with Custom Function

```bash
python train.py \
    --custom-generate-function-path custom_generate.py \
    --max-turns 5 \
    --prompt-data /path/to/agent_data.jsonl \
    ${MODEL_ARGS[@]}
```

See `examples/search-r1/` for a complete multi-turn search example.

---

## Configuration Reference

### Three Argument Categories

slime uses three types of arguments:

**1. Megatron Arguments** (passed directly):
```bash
--tensor-model-parallel-size 2
--pipeline-model-parallel-size 1
--num-layers 32
--hidden-size 4096
```

**2. SGLang Arguments** (prefixed with `--sglang-`):
```bash
--sglang-mem-fraction-static 0.8
--sglang-context-length 8192
--sglang-log-level INFO
```

**3. slime Arguments**:
```bash
# Resource allocation
--actor-num-nodes 1
--actor-num-gpus-per-node 8
--rollout-num-gpus 8
--colocate  # Share GPUs between training/inference

# Data
--prompt-data /path/to/data.jsonl
--input-key prompt
--label-key label

# Training loop
--num-rollout 3000
--rollout-batch-size 32
--n-samples-per-prompt 8
--global-batch-size 256

# Algorithm
--advantage-estimator grpo  # or: gspo, ppo, reinforce_plus_plus
--use-kl-loss
--kl-loss-coef 0.001
```

### Key Constraints

```
rollout_batch_size × n_samples_per_prompt = global_batch_size × num_steps_per_rollout
```

Example: 32 × 8 = 256 × 1

---

## Data Buffer System

slime's data buffer enables flexible data management:

### Basic Data Source

```python
class RolloutDataSource:
    def get_samples(self, num_samples):
        """Fetch prompts from dataset."""
        return self.dataset.sample(num_samples)

    def add_samples(self, samples):
        """Called after generation (no-op by default)."""
        pass
```

### Buffered Data Source (Off-Policy)

```python
class RolloutDataSourceWithBuffer(RolloutDataSource):
    def __init__(self):
        self.buffer = []

    def add_samples(self, samples):
        """Store generated samples for reuse."""
        self.buffer.extend(samples)

    def buffer_filter(self, args, buffer, num_samples):
        """Custom selection logic (prioritized, stratified, etc.)."""
        return select_best(buffer, num_samples)
```

---

## Common Issues and Solutions

### Issue: SGLang Engine Crash

**Symptoms**: Inference engine dies mid-training

**Solutions**:
```bash
# Enable fault tolerance
--use-fault-tolerance

# Increase memory allocation
--sglang-mem-fraction-static 0.85

# Reduce batch size
--rollout-batch-size 16
```

### Issue: Weight Sync Timeout

**Symptoms**: Training hangs after rollout

**Solutions**:
```bash
# Increase sync interval
--update-weights-interval 5

# Use colocated mode (no network transfer)
--colocate
```

### Issue: OOM During Training

**Symptoms**: CUDA OOM in backward pass

**Solutions**:
```bash
# Enable gradient checkpointing
--recompute-activations

# Reduce micro-batch size
--micro-batch-size 1

# Enable sequence parallelism
--sequence-parallel
```

### Issue: Slow Data Loading

**Symptoms**: GPU idle during data fetch

**Solutions**:
```bash
# Increase data workers
--num-data-workers 4

# Use streaming dataset
--streaming-data
```

---

## Supported Models

| Model Family | Configurations |
|--------------|----------------|
| GLM | GLM-4.5, GLM-4.6, GLM-4.7, GLM-Z1-9B |
| Qwen | Qwen3 (4B, 8B, 30B-A3B), Qwen3-MoE, Qwen2.5 |
| DeepSeek | V3, V3.1, R1 |
| Llama | Llama 3 (8B, 70B) |
| Others | Kimi K2, Moonlight-16B |

Each model has pre-configured scripts in `scripts/models/`.

---

## Advanced Topics

### Co-location Mode

Share GPUs between training and inference to reduce memory:

```bash
python train.py \
    --colocate \
    --actor-num-gpus-per-node 8 \
    --sglang-mem-fraction-static 0.4 \
    ${MODEL_ARGS[@]}
```

### Custom Reward Model

```python
# custom_rm.py
class CustomRewardModel:
    def __init__(self, model_path):
        self.model = load_model(model_path)

    def compute_reward(self, prompts, responses):
        inputs = self.tokenize(prompts, responses)
        scores = self.model(inputs)
        return scores.tolist()
```

```bash
--custom-rm-path custom_rm.py
```

### Evaluation Multi-Task

```bash
--eval-prompt-data aime /path/to/aime.jsonl \
--eval-prompt-data gsm8k /path/to/gsm8k.jsonl \
--n-samples-per-eval-prompt 16
```

---

## Resources

- **Documentation**: https://thudm.github.io/slime/
- **GitHub**: https://github.com/THUDM/slime
- **Blog**: https://lmsys.org/blog/2025-07-09-slime/
- **Examples**: See `examples/` directory for 14+ worked examples

Overview

This skill provides practical guidance for post-training large language models using slime, a Megatron+SGLang framework tailored for RL scaling. It explains core workflows, deployment notes, and configuration patterns to run GRPO, async, and multi-turn agentic training for GLM-family and similar models. Use it to integrate Megatron-LM training with high-throughput SGLang rollouts and custom data buffers.

How this skill works

The skill inspects and documents slime's architecture: a data buffer that feeds prompts to Megatron-LM training and SGLang rollout processes, with weight synchronization between training and inference. It outlines workflows for synchronous GRPO, asynchronous overlapping of rollouts and training, and agentic multi-turn generation with custom generate hooks. It also summarizes resource, argument, and debugging knobs to tune throughput and stability.

When to use it

  • You need native Megatron-LM training combined with SGLang inference routing for high-throughput rollouts.
  • You are training GLM-4.x, Qwen3, DeepSeek V3, Llama 3, or custom large models and need tight scaling control.
  • You require custom data generation, buffering, or off-policy replay in RL post-training.
  • You want research-grade RL algorithms (GRPO, GPPO variants) with production-oriented integrations.
  • You must overlap generation and optimization to maximize GPU utilization for large models.

Best practices

  • Match rollout_batch_size × n_samples_per_prompt to global_batch_size × num_steps_per_rollout to avoid imbalance.
  • Start with colocated mode for easier weight sync debugging, then scale to distributed actors once stable.
  • Use async-buffer-size and update-weights-interval to tune throughput vs staleness for async training.
  • Enable fault tolerance and increase sglang memory fraction if inference engine crashes under load.
  • Profile GPU utilization and TensorBoard reward curves regularly to catch divergence early.

Example use cases

  • GRPO training of a reasoning model (GLM or Qwen) with n-samples-per-prompt and KL regularization to preserve policy priors.
  • Asynchronous large-model training where rollouts are buffered to keep GPUs busy and reduce idle time.
  • Multi-turn agentic training with custom_generate implementing tool calls, tool results folded into conversation, and custom reward computation.
  • Off-policy experiments: use RolloutDataSourceWithBuffer to store and prioritize generated samples for replay.
  • Custom reward model integration: load a separate RM to score responses and pass scores into the training loop.

FAQ

What argument groups should I tune first?

Tune slime resource args (actor/rollout GPU counts), rollout_batch_size and n_samples_per_prompt first, then Megatron parallelism sizes and SGLang memory settings.

How do I fix SGLang crashes under heavy load?

Enable --use-fault-tolerance, increase --sglang-mem-fraction-static, reduce rollout batch sizes, and monitor logs for OOMs or worker failures.

When should I use async mode vs sync?

Use async for large models or long generations where synchronous rollout stalls training; choose sync when determinism and simpler debugging matter.