LogoLogo
Continuum WebsiteContinuum ApplicationsContinuum KnowledgeAxolotl Platform
Continuum Knowledge
Continuum Knowledge
  • Continuum
  • Data
    • Datasets
      • Pre Training Data
      • Types of Fine Tuning
      • Self Instruct Paper
      • Self-Alignment with Instruction Backtranslation
      • Systematic Evaluation of Instruction-Tuned Large Language Models on Open Datasets
      • Instruction Tuning
      • Instruction Fine Tuning - Alpagasus
      • Less is More For Alignment
      • Enhanced Supervised Fine Tuning
      • Visualising Data using t-SNE
      • UMAP: Uniform Manifold Approximation and Projection for Dimension Reduction
      • Training and Evaluation Datasets
      • What is perplexity?
  • MODELS
    • Foundation Models
      • The leaderboard
      • Foundation Models
      • LLama 2 - Analysis
      • Analysis of Llama 3
      • Llama 3.1 series
      • Google Gemini 1.5
      • Platypus: Quick, Cheap, and Powerful Refinement of LLMs
      • Mixtral of Experts
      • Mixture-of-Agents (MoA)
      • Phi 1.5
        • Refining the Art of AI Training: A Deep Dive into Phi 1.5's Innovative Approach
      • Phi 2.0
      • Phi-3 Technical Report
  • Training
    • The Fine Tuning Process
      • Why fine tune?
        • Does Fine-Tuning LLMs on New Knowledge Encourage Hallucinations?
        • Explanations in Fine Tuning
      • Tokenization
        • Tokenization Is More Than Compression
        • Tokenization - SentencePiece
        • Tokenization explore
        • Tokenizer Choice For LLM Training: Negligible or Crucial?
        • Getting the most out of your tokenizer for pre-training and domain adaptation
        • TokenMonster
      • Parameter Efficient Fine Tuning
        • P-Tuning
          • The Power of Scale for Parameter-Efficient Prompt Tuning
        • Prefix-Tuning: Optimizing Continuous Prompts for Generation
        • Harnessing the Power of PEFT: A Smarter Approach to Fine-tuning Pre-trained Models
        • What is Low-Rank Adaptation (LoRA) - explained by the inventor
        • Low Rank Adaptation (Lora)
        • Practical Tips for Fine-tuning LMs Using LoRA (Low-Rank Adaptation)
        • QLORA: Efficient Finetuning of Quantized LLMs
        • Bits and Bytes
        • The Magic behind Qlora
        • Practical Guide to LoRA: Tips and Tricks for Effective Model Adaptation
        • The quantization constant
        • QLORA: Efficient Finetuning of Quantized Language Models
        • QLORA and Fine-Tuning of Quantized Language Models (LMs)
        • ReLoRA: High-Rank Training Through Low-Rank Updates
        • SLoRA: Federated Parameter Efficient Fine-Tuning of Language Models
        • GaLora: Memory-Efficient LLM Training by Gradient Low-Rank Projection
      • Hyperparameters
        • Batch Size
        • Padding Tokens
        • Mixed precision training
        • FP8 Formats for Deep Learning
        • Floating Point Numbers
        • Batch Size and Model loss
        • Batch Normalisation
        • Rethinking Learning Rate Tuning in the Era of Language Models
        • Sample Packing
        • Gradient accumulation
        • A process for choosing the learning rate
        • Learning Rate Scheduler
        • Checkpoints
        • A Survey on Efficient Training of Transformers
        • Sequence Length Warmup
        • Understanding Training vs. Evaluation Data Splits
        • Cross-entropy loss
        • Weight Decay
        • Optimiser
        • Caching
      • Training Processes
        • Extending the context window
        • PyTorch Fully Sharded Data Parallel (FSDP)
        • Train Short, Test Long: Attention with Linear Biases Enables Input Length Extrapolation
        • YaRN: Efficient Context Window Extension of Large Language Models
        • Sliding Window Attention
        • LongRoPE
        • Reinforcement Learning
        • An introduction to reinforcement learning
        • Reinforcement Learning from Human Feedback (RLHF)
        • Direct Preference Optimization: Your Language Model is Secretly a Reward Model
  • INFERENCE
    • Why is inference important?
      • Grouped Query Attention
      • Key Value Cache
      • Flash Attention
      • Flash Attention 2
      • StreamingLLM
      • Paged Attention and vLLM
      • TensorRT-LLM
      • Torchscript
      • NVIDIA L40S GPU
      • Triton Inference Server - Introduction
      • Triton Inference Server
      • FiDO: Fusion-in-Decoder optimised for stronger performance and faster inference
      • Is PUE a useful measure of data centre performance?
      • SLORA
  • KNOWLEDGE
    • Vector Databases
      • A Comprehensive Survey on Vector Databases
      • Vector database management systems: Fundamental concepts, use-cases, and current challenges
      • Using the Output Embedding to Improve Language Models
      • Decoding Sentence-BERT
      • ColBERT: Efficient and Effective Passage Search via Contextualized Late Interaction over BERT
      • SimCSE: Simple Contrastive Learning of Sentence Embeddings
      • Questions Are All You Need to Train a Dense Passage Retriever
      • Improving Text Embeddings with Large Language Models
      • Massive Text Embedding Benchmark
      • RocketQAv2: A Joint Training Method for Dense Passage Retrieval and Passage Re-ranking
      • LLM2Vec: Large Language Models Are Secretly Powerful Text Encoders
      • Embedding and Fine-Tuning in Neural Language Models
      • Embedding Model Construction
      • Demystifying Embedding Spaces using Large Language Models
      • Fine-Tuning Llama for Multi-Stage Text Retrieval
      • Large Language Model Based Text Augmentation Enhanced Personality Detection Model
      • One Embedder, Any Task: Instruction-Finetuned Text Embeddings
      • Vector Databases are not the only solution
      • Knowledge Graphs
        • Harnessing Knowledge Graphs to Elevate AI: A Technical Exploration
        • Unifying Large Language Models and Knowledge Graphs: A Roadmap
      • Approximate Nearest Neighbor (ANN)
      • High Dimensional Data
      • Principal Component Analysis (PCA)
      • Vector Similarity Search - HNSW
      • FAISS (Facebook AI Similarity Search)
      • Unsupervised Dense Retrievers
    • Retrieval Augmented Generation
      • Retrieval-Augmented Generation for Large Language Models: A Survey
      • Fine-Tuning or Retrieval?
      • Revolutionising Information Retrieval: The Power of RAG in Language Models
      • A Survey on Retrieval-Augmented Text Generation
      • REALM: Retrieval-Augmented Language Model Pre-Training
      • Retrieve Anything To Augment Large Language Models
      • Generate Rather Than Retrieve: Large Language Models Are Strong Context Generators
      • Active Retrieval Augmented Generation
      • DSPy: LM Assertions: Enhancing Language Model Pipelines with Computational Constraints
      • DSPy: Compiling Declarative Language Model Calls
      • DSPy: In-Context Learning for Extreme Multi-Label Classification
      • Optimizing Instructions and Demonstrations for Multi-Stage Language Model Programs
      • HYDE: Revolutionising Search with Hypothetical Document Embeddings
      • Enhancing Recommender Systems with Large Language Model Reasoning Graphs
      • Retrieval Augmented Generation (RAG) versus fine tuning
      • RAFT: Adapting Language Model to Domain Specific RAG
      • Summarisation Methods and RAG
      • Lessons Learned on LLM RAG Solutions
      • Stanford: Retrieval Augmented Language Models
      • Overview of RAG Approaches with Vector Databases
      • Mastering Chunking in Retrieval-Augmented Generation (RAG) Systems
    • Semantic Routing
    • Resource Description Framework (RDF)
  • AGENTS
    • What is agency?
      • Rephrase and Respond: Let Large Language Models Ask Better Questions for Themselves
      • Types of Agents
      • The risk of AI agency
      • Understanding Personality in Large Language Models: A New Frontier in AI Psychology
      • AI Agents - Reasoning, Planning, and Tool Calling
      • Personality and Brand
      • Agent Interaction via APIs
      • Bridging Minds and Machines: The Legacy of Newell, Shaw, and Simon
      • A Survey on Language Model based Autonomous Agents
      • Large Language Models as Agents
      • AI Reasoning: A Deep Dive into Chain-of-Thought Prompting
      • Enhancing AI Reasoning with Self-Taught Reasoner (STaR)
      • Exploring the Frontier of AI: The "Tree of Thoughts" Framework
      • Toolformer: Revolutionising Language Models with API Integration - An Analysis
      • TaskMatrix.AI: Bridging Foundational AI Models with Specialised Systems for Enhanced Task Completion
      • Unleashing the Power of LLMs in API Integration: The Rise of Gorilla
      • Andrew Ng's presentation on AI agents
      • Making AI accessible with Andrej Karpathy and Stephanie Zhan
  • Regulation and Ethics
    • Regulation and Ethics
      • Privacy
      • Detecting AI Generated content
      • Navigating the IP Maze in AI: The Convergence of Blockchain, Web 3.0, and LLMs
      • Adverse Reactions to generative AI
      • Navigating the Ethical Minefield: The Challenge of Security in Large Language Models
      • Navigating the Uncharted Waters: The Risks of Autonomous AI in Military Decision-Making
  • DISRUPTION
    • Data Architecture
      • What is a data pipeline?
      • What is Reverse ETL?
      • Unstructured Data and Generatve AI
      • Resource Description Framework (RDF)
      • Integrating generative AI with the Semantic Web
    • Search
      • BM25 - Search Engine Ranking Function
      • BERT as a reranking engine
      • BERT and Google
      • Generative Engine Optimisation (GEO)
      • Billion-scale similarity search with GPUs
      • FOLLOWIR: Evaluating and Teaching Information Retrieval Models to Follow Instructions
      • Neural Collaborative Filtering
      • Federated Neural Collaborative Filtering
      • Latent Space versus Embedding Space
      • Improving Text Embeddings with Large Language Models
    • Recommendation Engines
      • On Interpretation and Measurement of Soft Attributes for Recommendation
      • A Survey on Large Language Models for Recommendation
      • Model driven recommendation systems
      • Recommender AI Agent: Integrating Large Language Models for Interactive Recommendations
      • Foundation Models for Recommender Systems
      • Exploring the Impact of Large Language Models on Recommender Systems: An Extensive Review
      • AI driven recommendations - harming autonomy?
    • Logging
      • A Taxonomy of Anomalies in Log Data
      • Deeplog
      • LogBERT: Log Anomaly Detection via BERT
      • Experience Report: Deep Learning-based System Log Analysis for Anomaly Detection
      • Log-based Anomaly Detection with Deep Learning: How Far Are We?
      • Deep Learning for Anomaly Detection in Log Data: A Survey
      • LogGPT
      • Adaptive Semantic Gate Networks (ASGNet) for log-based anomaly diagnosis
  • Infrastructure
    • The modern data centre
      • Enhancing Data Centre Efficiency: Strategies to Improve PUE
      • TCO of NVIDIA GPUs and falling barriers to entry
      • Maximising GPU Utilisation with Kubernetes and NVIDIA GPU Operator
      • Data Centres
      • Liquid Cooling
    • Servers and Chips
      • The NVIDIA H100 GPU
      • NVIDIA H100 NVL
      • Lambda Hyperplane 8-H100
      • NVIDIA DGX Servers
      • NVIDIA DGX-2
      • NVIDIA DGX H-100 System
      • NVLink Switch
      • Tensor Cores
      • NVIDIA Grace Hopper Superchip
      • NVIDIA Grace CPU Superchip
      • NVIDIA GB200 NVL72
      • Hopper versus Blackwell
      • HGX: High-Performance GPU Platforms
      • ARM Chips
      • ARM versus x86
      • RISC versus CISC
      • Introduction to RISC-V
    • Networking and Connectivity
      • Infiniband versus Ethernet
      • NVIDIA Quantum InfiniBand
      • PCIe (Peripheral Component Interconnect Express)
      • NVIDIA ConnectX InfiniBand adapters
      • NVMe (Non-Volatile Memory Express)
      • NVMe over Fabrics (NVMe-oF)
      • NVIDIA Spectrum-X
      • NVIDIA GPUDirect
      • Evaluating Modern GPU Interconnect
      • Scalable Hierarchical Aggregation and Reduction Protocol (SHARP)
      • Next-generation networking in AI environments
      • NVIDIA Collective Communications Library (NCCL)
    • Data and Memory
      • NVIDIA BlueField Data Processing Units (DPUs)
      • Remote Direct Memory Access (RDMA)
      • High Bandwidth Memory (HBM3)
      • Flash Memory
      • Model Requirements
      • Calculating GPU memory for serving LLMs
      • Transformer training costs
      • GPU Performance Optimisation
    • Libraries and Complements
      • NVIDIA Base Command
      • NVIDIA AI Enterprise
      • CUDA - NVIDIA GTC 2024 presentation
      • RAPIDs
      • RAFT
    • Vast Data Platform
      • Vast Datastore
      • Vast Database
      • Vast Data Engine
      • DASE (Disaggregated and Shared Everything)
      • Dremio and VAST Data
    • Storage
      • WEKA: A High-Performance Storage Solution for AI Workloads
      • Introduction to NVIDIA GPUDirect Storage (GDS)
        • GDS cuFile API
      • NVIDIA Magnum IO GPUDirect Storage (GDS)
      • Vectors in Memory
Powered by GitBook
LogoLogo

Continuum - Accelerated Artificial Intelligence

  • Continuum Website
  • Axolotl Platform

Copyright Continuum Labs - 2023

On this page
  • Calculating GPU memory for serving LLMs
  • GPU memory required for serving Llama 70B
  • Using FP8 (8-bit floating-point)
  • General Process

Was this helpful?

  1. Infrastructure
  2. Data and Memory

Calculating GPU memory for serving LLMs

PreviousModel RequirementsNextTransformer training costs

Last updated 7 months ago

Was this helpful?

Calculating GPU memory for serving LLMs

How many GPUs do I need to be able to serve Llama3 70 billion parameter model?

In order to answer that, you need to know how much GPU memory will be required by the model. The formula is:

M=((P×4B)(32/Q))×1.2  M = \left(\frac{{(P \times 4B)}}{{(32/Q)}}\right) \times 1.2 \ M=((32/Q)(P×4B)​)×1.2 
Symbol
Description

GPU memory expressed in Gigabyte

The amount of parameters in the model

4 bytes, expressing the bytes used for each parameter

There are 32 bits in 4 bytes

The amount of bits that should be used for loading the model. - 16 bits, 8 bits or 4 bits.

Represents a 20% overhead of loading additional things in GPU memory.

GPU memory required for serving Llama 70B

Let's try it out for Llama 70 billion parameter model that we will load in 16 bit number format.

M=((70×109×4 bytes)(32/16))×1.2=(140×109 bytes)×1.2=168 GB M = \left(\frac{{(70 \times 10^9 \times 4 \text{ bytes})}}{{(32/16)}}\right) \times 1.2 = \left(140 \times 10^9 \text{ bytes}\right) \times 1.2 = 168 \text{ GB} M=((32/16)(70×109×4 bytes)​)×1.2=(140×109 bytes)×1.2=168 GB

To run this model, you would require two NVIDIA A-100 80GB memory models.

A single A100 80GB wouldn't be enough, although 2x A100 80GB should be enough to serve the Llama How to further reduce GPU memory required for Llama 2 70B?

Using FP8 (8-bit floating-point)

To calculate the GPU memory requirements for training a model like Llama3 with 70 billion parameters using different precision levels such as FP8 (8-bit floating-point), we need to adjust our formula to fit the new context.

Let's define a general formula first, and then apply it specifically for FP8.

For the FP8 precision:

  • Bytes per parameter (b) is 1 byte

  • Bit width used (Q) is 8 bits

  • P=70×109 P = 70 \times 10^9 P=70×109 - 70 billion parameters

  • b=1 b = 1 b=1 - byte per parameter (since 8 bits = 1 byte)

  • overhead=1.2\text{overhead} = 1.2overhead=1.2 - representing a 20% overhead

  • QQQ = 8 bits

Substitute the values into the formula:

M = \left(\frac{{(70 \times 10^9 \times 1 \text{ byte})}}{{(32/8)}}\right) \times 1.2 \

Next:

M = \left(\frac{{(70 \times 10^9 \text{ bytes})}}{{4}}\right) \times 1.2 \

Next:

M = (17.5 \times 10^9 \text{ bytes}) \times 1.2 \

Work out the memory requirements:

M = 21 \times 10^9 \text{ bytes} \
M = 21 \text{ GB} \

Therefore, the memory requirement for training the Llama3 model with 70 billion parameters using FP8 precision is a much lower 21 GB.

General Process

Determine the number of parameters in the model (P)

  • The model size is often expressed in billions (B) of parameters.

  • For example, a 7B model has 7 billion parameters.

Identify the data type used for the model parameters

  • Common data types include:

    • float (32-bit floating point): 4 bytes per parameter

    • half/BF16 (16-bit floating point): 2 bytes per parameter

    • int8 (8-bit integer): 1 byte per parameter

    • int4 (4-bit integer): 0.5 bytes per parameter

Calculate the storage size of the model (S)

  • Multiply the number of parameters (P) by the size of the data type.

  • For example, a 7B model using BF16 would have a storage size of: S = 7 billion * 2 bytes = 14 billion bytes ≈ 14 GB

Estimate the memory required for inference (M_inf)

  • The memory required for inference is approximately equal to the storage size (S).

  • M_inf ≈ S

Estimate the memory required for training (M_train)

  • Training typically requires 3 to 4 times the memory needed for inference.

  • A conservative estimate is to multiply the inference memory (M_inf) by a factor of 4.

  • M_train ≈ M_inf * 4

  • For example, training a 7B model using float parameters would require: M_train ≈ 7 billion * 4 bytes * 4 = 112 GB

Consider memory requirements for gradients and optimizer states

  • During training, additional memory is needed for gradients and optimiser states.

  • The memory required for gradients is equal to the number of parameters (P).

  • The memory required for optimizer states depends on the optimizer used:

    • AdamW optimizer: 2 * P

    • SGD optimizer: P

Adjust for additional memory overhead

  • Training may require additional memory for intermediate computations and data storage.

  • Add a safety margin of 10-20% to the estimated training memory (M_train).

Consider memory-efficient training techniques

  • Techniques like LoRA (Low-Rank Adaptation) and QLoRA can reduce memory requirements.

  • These techniques involve training a smaller model while running inference on the original model.

  • The total memory used is the sum of the memory required for inference on the original model and the memory needed for training the smaller model.

Here's an example calculation for training a 13B model using float parameters:

  • Number of parameters (P) = 13 billion

  • Data type: float (4 bytes per parameter)

  • Storage size (S) = 13 billion * 4 bytes ≈ 52 GB

  • Inference memory (M_inf) ≈ 52 GB

  • Training memory (M_train) ≈ 52 GB * 4 = 208 GB

  • Additional memory for gradients = 13 billion * 4 bytes ≈ 52 GB

  • Additional memory for AdamW optimizer states = 2 * 13 billion * 4 bytes ≈ 104 GB

  • Total estimated memory for training = 208 GB + 52 GB + 104 GB = 364 GB

  • With a 20% safety margin, the final estimate would be: 364 GB * 1.2 ≈ 437 GB

This framework provides a rough estimate of the memory requirements for training LLMs based on storage usage. However, actual memory usage may vary depending on the specific model architecture, implementation details, and hardware characteristics. It's always a good idea to have some extra memory available to accommodate any additional overhead or unexpected memory usage during training.

MMM
PPP
4B4B4B
323232
QQQ
1.21.21.2