training-llms-megatron
Scannednpx machina-cli add skill Orchestra-Research/AI-Research-SKILLs/megatron-core --openclawMegatron-Core - Large-Scale LLM Training
Quick start
Megatron-Core trains LLMs from 2B to 462B parameters with up to 47% Model FLOP Utilization on H100 GPUs through advanced parallelism strategies.
Installation:
# Docker (recommended)
docker run --gpus all -it --rm nvcr.io/nvidia/pytorch:25.04-py3
# Or pip
pip install megatron-core
Simple distributed training:
# Train with 2 GPUs using data parallelism
torchrun --nproc_per_node=2 examples/run_simple_mcore_train_loop.py
# Or LLaMA-3 8B training
./examples/llama/train_llama3_8b_fp8.sh
Common workflows
Workflow 1: Train LLaMA-style model with 3D parallelism
Copy this checklist:
LLaMA Training Setup:
- [ ] Step 1: Choose parallelism configuration
- [ ] Step 2: Configure training hyperparameters
- [ ] Step 3: Launch distributed training
- [ ] Step 4: Monitor performance metrics
Step 1: Choose parallelism configuration
Model size determines parallelism strategy:
| Model Size | GPUs | Tensor Parallel | Pipeline Parallel | Data Parallel | Context Parallel |
|---|---|---|---|---|---|
| 7B | 8 | 1 | 1 | 8 | 1 |
| 13B | 8 | 2 | 1 | 4 | 1 |
| 70B | 64 | 4 | 4 | 4 | 1 |
| 405B | 128 | 8 | 8 | 2 | 2 |
Step 2: Configure training hyperparameters
#!/bin/bash
# train_llama_70b.sh
GPUS_PER_NODE=8
NNODES=8 # 64 GPUs total
TP=4 # Tensor parallel
PP=4 # Pipeline parallel
CP=1 # Context parallel
# LLaMA 70B configuration
MODEL_SIZE=70 # Billion parameters
HIDDEN_SIZE=8192
NUM_LAYERS=80
NUM_HEADS=64
SEQ_LENGTH=4096
# Training hyperparameters
MICRO_BATCH=1
GLOBAL_BATCH=1024
LR=3e-4
torchrun \
--nproc_per_node=$GPUS_PER_NODE \
--nnodes=$NNODES \
pretrain_gpt.py \
--tensor-model-parallel-size $TP \
--pipeline-model-parallel-size $PP \
--context-parallel-size $CP \
--sequence-parallel \
--num-layers $NUM_LAYERS \
--hidden-size $HIDDEN_SIZE \
--num-attention-heads $NUM_HEADS \
--seq-length $SEQ_LENGTH \
--max-position-embeddings $SEQ_LENGTH \
--micro-batch-size $MICRO_BATCH \
--global-batch-size $GLOBAL_BATCH \
--lr $LR \
--train-iters 100000 \
--lr-decay-style cosine \
--lr-warmup-iters 2000 \
--weight-decay 0.1 \
--clip-grad 1.0 \
--bf16 \
--use-mcore-models \
--transformer-impl transformer_engine \
--data-path /path/to/data \
--vocab-file /path/to/vocab.json \
--merge-file /path/to/merges.txt
Step 3: Launch distributed training
# Single node (8 GPUs)
bash train_llama_70b.sh
# Multi-node with SLURM
sbatch --nodes=8 --gpus-per-node=8 train_llama_70b.sh
Step 4: Monitor performance metrics
Key metrics to track:
Model FLOP Utilization (MFU): Target >40% on H100
Throughput: Tokens/sec/GPU
Memory usage: <80GB per GPU for 70B model
Loss: Should decrease steadily
Workflow 2: Configure Mixture of Experts (MoE) training
For sparse MoE models like Mixtral.
MoE Training:
- [ ] Step 1: Configure expert parallelism
- [ ] Step 2: Set MoE hyperparameters
- [ ] Step 3: Launch training with EP
Step 1: Configure expert parallelism
# Mixtral 8x7B example
TENSOR_PARALLEL=2
PIPELINE_PARALLEL=1
EXPERT_PARALLEL=4 # Split 8 experts across 4 GPUs
DATA_PARALLEL=4
TOTAL_GPUS=$((TENSOR_PARALLEL * PIPELINE_PARALLEL * EXPERT_PARALLEL * DATA_PARALLEL))
# = 2 * 1 * 4 * 4 = 32 GPUs
Step 2: Set MoE hyperparameters
torchrun \
--nproc_per_node=8 \
pretrain_gpt.py \
--tensor-model-parallel-size 2 \
--pipeline-model-parallel-size 1 \
--expert-model-parallel-size 4 \
--num-experts 8 \
--moe-router-topk 2 \
--moe-router-load-balancing-type aux_loss \
--moe-aux-loss-coeff 0.01 \
--hidden-size 4096 \
--num-layers 32 \
--num-attention-heads 32 \
--seq-length 4096 \
--max-position-embeddings 4096 \
--bf16 \
--use-mcore-models \
--transformer-impl transformer_engine \
--data-path /path/to/data \
--vocab-file /path/to/vocab.json \
--merge-file /path/to/merges.txt
Step 3: Launch training with EP
Expert parallelism distributes different experts across GPUs, reducing memory while maintaining capacity.
Memory without EP: 8 experts × 7B = 56GB per GPU
Memory with EP=4: 2 experts × 7B = 14GB per GPU
Savings: 75% memory reduction
Workflow 3: Optimize for maximum throughput
Achieve 47% MFU on H100.
Performance Optimization:
- [ ] Step 1: Enable Flash Attention
- [ ] Step 2: Use FP8 precision (H100)
- [ ] Step 3: Optimize micro-batch size
- [ ] Step 4: Tune parallelism degrees
Step 1: Enable optimizations
--use-mcore-models # Use Megatron Core models
--transformer-impl transformer_engine # Use Transformer Engine
--sequence-parallel # Reduce activation memory (use with TP)
Step 2: Use FP8 precision (H100 only)
--fp8-hybrid # FP8 mixed precision training
# Transformer Engine handles FP8 automatically
Result: 1.5-2x speedup on H100 vs BF16.
Step 3: Optimize micro-batch size
Find largest micro-batch that fits in memory:
# Start with 1, increase until OOM
for MBS in 1 2 4 8; do
echo "Testing micro-batch-size=$MBS"
torchrun ... --micro-batch-size $MBS
done
Typical values:
- 7B model: 4-8
- 70B model: 1-2
- 405B model: 1
Step 4: Tune parallelism degrees
Rules of thumb:
Tensor Parallel: Use ≤8 (limited by NVLink within node)
Pipeline Parallel: Use for >70B models
Context Parallel: Use for sequences >8K tokens
Data Parallel: Fill remaining GPUs
Example 405B on 128 H100s:
TP=8 (1 node)
PP=8 (across nodes)
CP=2 (long sequences)
DP=1
Total = 8 × 8 × 2 × 1 = 128 GPUs
When to use vs alternatives
Use Megatron-Core when:
- Training models >10B parameters
- Need maximum efficiency (target >40% MFU)
- Using NVIDIA GPUs (A100, H100)
- Production training at scale
- Want fine-grained parallelism control
Use alternatives instead:
- PyTorch FSDP: Models <70B, simpler API, PyTorch native
- DeepSpeed: Easier setup, good for <100B models
- HuggingFace Accelerate: Prototyping, simpler workflows
- LitGPT: Educational, single-file implementations
Common issues
Issue: Low GPU utilization (<30% MFU)
Causes:
- Micro-batch too small
- Too much parallelism overhead
- Not using Flash Attention
Fixes:
# Increase micro-batch
--micro-batch-size 4 # Was 1
# Enable optimizations
--use-flash-attn
--sequence-parallel
# Reduce TP if >8
--tensor-model-parallel-size 4 # Was 16
Issue: Out of memory
Reduce memory with:
--tensor-model-parallel-size 2 # Split model across GPUs
--recompute-granularity full # Gradient checkpointing
--recompute-method block # Checkpoint transformer blocks
--recompute-num-layers 1 # Checkpoint every layer
Or use CPU/NVMe offloading:
--cpu-optimizer # Offload optimizer to CPU
--cpu-optimizer-type ADAM # CPU Adam variant
Issue: Training slower than expected
Check:
- Network bottleneck: Ensure InfiniBand/NVLink enabled
- Pipeline bubbles: Use interleaved pipeline schedule
--num-layers-per-virtual-pipeline-stage 2 - Data loading: Use fast data loader
--dataloader-type cyclic
Issue: Diverging loss
Stabilize training:
--lr-warmup-iters 2000 # Longer warmup
--clip-grad 1.0 # Gradient clipping
--init-method-std 0.006 # Smaller init
--attention-dropout 0.0 # No dropout in attention
--hidden-dropout 0.0 # No dropout in FFN
Advanced topics
Parallelism strategies: See references/parallelism-guide.md for detailed comparison of TP/PP/DP/CP/EP with performance analysis and when to use each.
Performance benchmarks: See references/benchmarks.md for MFU numbers across different model sizes and GPU configurations.
Production configurations: See references/production-examples.md for real-world setups from LLaMA 3 405B, Nemotron-4 340B, and DeepSeek-V3 671B.
Training recipes: See references/training-recipes.md for complete hyperparameter configurations for GPT/LLaMA/Mixtral architectures.
Hardware requirements
- GPU: NVIDIA Ampere+ (A100, H100, B200)
- Turing works but slower
- FP8 requires Hopper/Ada/Blackwell
- Network: InfiniBand or 400Gb+ Ethernet for multi-node
- Memory per GPU:
- 7B model: 40GB+
- 70B model: 80GB (with TP=4)
- 405B model: 80GB (with TP=8, PP=8)
- Storage: Fast NVMe for checkpoints (1TB+ for 70B+ models)
Resources
- Docs: https://docs.nvidia.com/megatron-core/
- GitHub: https://github.com/NVIDIA/Megatron-LM
- Papers:
- "Megatron-LM: Training Multi-Billion Parameter Language Models Using Model Parallelism" (2019)
- "Efficient Large-Scale Language Model Training on GPU Clusters Using Megatron-LM" (2021)
- NeMo Framework: https://docs.nvidia.com/nemo-framework/ (built on Megatron-Core)
Source
git clone https://github.com/Orchestra-Research/AI-Research-SKILLs/blob/main/08-distributed-training/megatron-core/SKILL.mdView on GitHub Overview
Megatron-Core trains LLMs from 2B to 462B parameters using advanced parallelism strategies. It targets maximum GPU efficiency (up to 47% MFU on H100) and supports tensor, pipeline, sequence, context, and expert parallelism, making it suitable for production-grade models like Nemotron, LLaMA, and DeepSeek.
How This Skill Works
The framework partitions model weights across GPUs with multiple parallelism dimensions (tensor, pipeline, sequence/context/expert) and coordinates data flow for large-scale training. It supports 3D parallel configurations and MoE workflows to scale across nodes while optimizing GPU utilization.
When to Use It
- Training large language models from 2B up to 462B parameters and needing high GPU utilization on H100.
- When you require advanced parallelism (tensor, pipeline, sequence, context, or expert parallelism) to fit models across many GPUs.
- If you need a production-ready framework used publicly for Nemotron, LLaMA, and DeepSeek.
- When scaling training across multiple GPUs/nodes with structured parallelism (as in LLaMA-style 3D parallelism workflows).
- To explore sparse Mixture-of-Experts (MoE) training like Mixtral.
Quick Start
- Step 1: Install Megatron-Core via Docker (recommended) or pip install megatron-core.
- Step 2: Run a simple distributed training example (e.g., torchrun --nproc_per_node=2 examples/run_simple_mcore_train_loop.py) or use a small LLaMA script like train_llama_70b.sh.
- Step 3: Launch distributed training with your chosen parallelism configuration and monitor MFU, throughput, and memory.
Best Practices
- Start with a small data-parallel run (e.g., 2 GPUs) to validate environment before scaling.
- Choose the parallelism configuration (TP/PP/CP/sequence-context) that matches the model size using the published guidelines.
- Aim for high GPU utilization (target MFU > ~40-47% on H100) and monitor MFU, throughput, and memory.
- Use the docker container or a pinned CPU/GPU environment for reproducibility and install dependencies (megatron-core, torch, apex, transformer-engine).
- For MoE, correctly configure expert, tensor, and data parallelism to balance sparsity and load.
Example Use Cases
- Nemotron model training using Megatron-Core for large-scale deployments.
- LLaMA-family training with 3D parallelism to scale to tens of billions of parameters.
- DeepSeek model training leveraging tensor/pipeline/context/expert parallelism.
- Mixture-of-Experts (MoE) training workflows like Mixtral.
- A 405B model trained across a multi-GPU setup (e.g., 128 GPUs) as an example in the workflow.
Frequently Asked Questions
Related Skills
deepspeed
Orchestra-Research/AI-Research-SKILLs
Expert guidance for distributed training with DeepSpeed - ZeRO optimization stages, pipeline parallelism, FP16/BF16/FP8, 1-bit Adam, sparse attention
langchain
Orchestra-Research/AI-Research-SKILLs
Framework for building LLM-powered applications with agents, chains, and RAG. Supports multiple providers (OpenAI, Anthropic, Google), 500+ integrations, ReAct agents, tool calling, memory management, and vector store retrieval. Use for building chatbots, question-answering systems, autonomous agents, or RAG applications. Best for rapid prototyping and production deployments.
langsmith-observability
Orchestra-Research/AI-Research-SKILLs
LLM observability platform for tracing, evaluation, and monitoring. Use when debugging LLM applications, evaluating model outputs against datasets, monitoring production systems, or building systematic testing pipelines for AI applications.
nemo-guardrails
Orchestra-Research/AI-Research-SKILLs
NVIDIA's runtime safety framework for LLM applications. Features jailbreak detection, input/output validation, fact-checking, hallucination detection, PII filtering, toxicity detection. Uses Colang 2.0 DSL for programmable rails. Production-ready, runs on T4 GPU.
tensorrt-llm
Orchestra-Research/AI-Research-SKILLs
Optimizes LLM inference with NVIDIA TensorRT for maximum throughput and lowest latency. Use for production deployment on NVIDIA GPUs (A100/H100), when you need 10-100x faster inference than PyTorch, or for serving models with quantization (FP8/INT4), in-flight batching, and multi-GPU scaling.
serving-llms-vllm
Orchestra-Research/AI-Research-SKILLs
Serves LLMs with high throughput using vLLM's PagedAttention and continuous batching. Use when deploying production LLM APIs, optimizing inference latency/throughput, or serving models with limited GPU memory. Supports OpenAI-compatible endpoints, quantization (GPTQ/AWQ/FP8), and tensor parallelism.