training-llms-megatron
关于
This Claude Skill trains large language models (2B-462B parameters) using NVIDIA's Megatron-Core framework with advanced parallelism strategies. Use it when training models over 1B parameters, needing maximum GPU efficiency (47% MFU on H100), or requiring tensor/pipeline/sequence parallelism. It's a production-ready framework proven on models like Nemotron, LLaMA, and DeepSeek.
技能文档
Megatron-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)
快速安装
/plugin add https://github.com/zechenzhangAGI/AI-research-SKILLs/tree/main/megatron-core在 Claude Code 中复制并粘贴此命令以安装该技能
GitHub 仓库
相关推荐技能
llamaguard
其他LlamaGuard是Meta推出的7-8B参数内容审核模型,专门用于过滤LLM的输入和输出内容。它能检测六大安全风险类别(暴力/仇恨、性内容、武器、违禁品、自残、犯罪计划),准确率达94-95%。开发者可通过HuggingFace、vLLM或Sagemaker快速部署,并能与NeMo Guardrails集成实现自动化安全防护。
sglang
元SGLang是一个专为LLM设计的高性能推理框架,特别适用于需要结构化输出的场景。它通过RadixAttention前缀缓存技术,在处理JSON、正则表达式、工具调用等具有重复前缀的复杂工作流时,能实现极速生成。如果你正在构建智能体或多轮对话系统,并追求远超vLLM的推理性能,SGLang是理想选择。
evaluating-llms-harness
测试该Skill通过60+个学术基准测试(如MMLU、GSM8K等)评估大语言模型质量,适用于模型对比、学术研究及训练进度追踪。它支持HuggingFace、vLLM和API接口,被EleutherAI等行业领先机构广泛采用。开发者可通过简单命令行快速对模型进行多任务批量评估。
langchain
元LangChain是一个用于构建LLM应用程序的框架,支持智能体、链和RAG应用开发。它提供多模型提供商支持、500+工具集成、记忆管理和向量检索等核心功能。开发者可用它快速构建聊天机器人、问答系统和自主代理,适用于从原型验证到生产部署的全流程。
