LLM-Drop

The official implementation of the paper "What Matters in Transformers? Not All Attention is Needed".

Stars: 165

Visit

LLM-Drop is an official implementation of the paper 'What Matters in Transformers? Not All Attention is Needed'. The tool investigates redundancy in transformer-based Large Language Models (LLMs) by analyzing the architecture of Blocks, Attention layers, and MLP layers. It reveals that dropping certain Attention layers can enhance computational and memory efficiency without compromising performance. The tool provides a pipeline for Block Drop and Layer Drop based on LLaMA-Factory, and implements quantization using AutoAWQ and AutoGPTQ.

README:

What Matters in Transformers? Not All Attention is Needed

Shwai He*, Guoheng Sun*, Zheyu Shen, Ang Li

This is the official implementation of the paper What Matters in Transformers? Not All Attention is Needed . We conduct extensive experiments and analysis to reveal the architecture redundancy within transformer-based Large Language Models (LLMs). Pipeline for Block Drop and Layer Drop is based on the LLaMA-Factory. The quantization is implemented based on the AutoAWQ and AutoGPTQ.

Introduction

Transformer-based large language models (LLMs) often contain architectural redundancies. In this work, we systematically investigate redundancy across different types of modules, including Blocks, Attention layers, and MLP layers. Surprisingly, we found that Attention layers, the core component of transformers, are particularly redundant. For example, in the Llama-3-70B model, half of the Attention layers can be dropped while maintaining performance. Our observations indicate that this redundancy in Attention layers persists throughout the training process, necessitating Attention Drop. Additionally, dropping Attention layers significantly enhances computational and memory efficiency. Our findings are informative for the ML community and provide insights for future architecture design.

News

Nov 2024: Updated the code to support additional language models, including Gemma2, Baichuan, DeepSeek, Yi, Solar. Yi and Solar refer to Llama, and we checked their availability.
Sep 2024: Released checkpoints for dropped models using Block Drop and Layer Drop.
Jun 2024: Published preprint on arXiv along with the related codebase.

Quick Start

Installation

conda create -n llm-drop python=3.10
conda activate llm-drop

git clone https://github.com/CASE-Lab-UMD/LLM-Drop

#For Dropping:
cd ./LLM-Drop
pip install -e .

#For Quantization:
cd ./src/llmtuner/compression/quantization/AutoAWQ
pip install -e .

cd ./src/llmtuner/compression/quantization/AutoAWQ/AutoAWQ_kernels
pip install -e .

cd ./src/llmtuner/compression/quantization/AutoGPTQ
pip install -vvv --no-build-isolation -e .

Prepare Models

Download the models (e.g., Mistral-7B, Llama-2 and Llama-3) from HuggingFace. We create new config and modeling files to represent the models by layers or blocks. The key auto_map needs to be added in the config.json to utilize the new files. Take Mistral-7B as an example:

"auto_map": {
    "AutoConfig": "configuration_dropped_mistral.MistralConfig",
    "AutoModelForCausalLM": "modeling_dropped_mistral.MistralForCausalLM"
  },

Additionally, the key drop_attn_list and drop_mlp_list respectively mark which Attention layers and MLPs should be dropped based on their layer index. For instance,

Drop 4 Attention layers:

 "drop_mlp_list": [],
 "drop_attn_list": [25, 26, 24, 22],

Drop 4 MLPs layers:

 "drop_mlp_list": [26, 27, 25, 24],
 "drop_attn_list": [],

Drop 4 Blocks:

 "drop_mlp_list": [26, 25, 24, 27],
 "drop_attn_list": [26, 25, 24, 27],

Run Dropping

Block Drop

bash scripts/dropping/block_drop.sh

Layer Drop

bash scripts/dropping/layer_drop.sh

Joint Layer Drop

bash scripts/dropping/layer_drop_joint.sh

These bash scripts will generate the importance scores for blocks/layers, determine which blocks/layers to retain, and create new model configuration files indicating the dropped modules.

Benchmarks

Performance

Evaluate the performance of the model with dropping some modules on specific tasks:

bash scripts/benchmark/benchmark_lm_eval.sh

The evaluation code is based on EleutherAI/lm-evaluation-harness. To fully reproduce our results, please use this version. It samples few-shot based on the index of the samples, avoiding the issue of result variation with the number of processes during data parallel inference. Remember to use the modeling files in src/llmtuner/model to load the Mistral and Llama models.

SpeedUp

Evaluate the speedup ratio of the model with dropping some modules:

bash scripts/benchmark/benchmark_speed.sh

Quantization

Please refer to AutoGPTQ and AutoAWQ. Ensure you carefully install the packages that correspond to your CUDA version. For quantization, use the following scripts:

bash scripts/quantization/awq.sh
bash scripts/quantization/gptq.sh

Citation

@misc{he2024matterstransformersattentionneeded,
      title={What Matters in Transformers? Not All Attention is Needed}, 
      author={Shwai He and Guoheng Sun and Zheyu Shen and Ang Li},
      year={2024},
      eprint={2406.15786},
      archivePrefix={arXiv},
      primaryClass={cs.LG},
      url={https://arxiv.org/abs/2406.15786}, 
}

Contact Us

If you have any questions, please contact:

Shwai He: [email protected]
Guoheng Sun: [email protected]

For Tasks:

Click tags to check more tools for each tasks

evaluate model performance speed up model quantize model drop attention layers analyze model architecture

For Jobs:

machine learning researcher data scientist ai engineer research scientist nlp engineer

Alternative AI tools for LLM-Drop

Similar Open Source Tools

LLM-Drop

github

: 165

SUPIR

SUPIR is an AI-based image processing and upscaling tool that leverages cutting-edge technology to enhance image quality and resolution. The tool provides users with the ability to upscale images with high generalization and quality, as well as specific settings for light degradation scenarios. It offers a range of models and checkpoints for different use cases, along with detailed instructions for installation and usage. SUPIR also includes features for color fixing, linear CFG adjustments, and various prompts for image enhancement. The tool is designed for non-commercial use only and comes with a contact email for inquiries and permission requests for commercial use.

github

: 4.0k

MMC

This repository, MMC, focuses on advancing multimodal chart understanding through large-scale instruction tuning. It introduces a dataset supporting various tasks and chart types, a benchmark for evaluating reasoning capabilities over charts, and an assistant achieving state-of-the-art performance on chart QA benchmarks. The repository provides data for chart-text alignment, benchmarking, and instruction tuning, along with existing datasets used in experiments. Additionally, it offers a Gradio demo for the MMCA model.

github

: 75

tensorrtllm_backend

The TensorRT-LLM Backend is a Triton backend designed to serve TensorRT-LLM models with Triton Inference Server. It supports features like inflight batching, paged attention, and more. Users can access the backend through pre-built Docker containers or build it using scripts provided in the repository. The backend can be used to create models for tasks like tokenizing, inferencing, de-tokenizing, ensemble modeling, and more. Users can interact with the backend using provided client scripts and query the server for metrics related to request handling, memory usage, KV cache blocks, and more. Testing for the backend can be done following the instructions in the 'ci/README.md' file.

github

: 814

embodied-agents

Embodied Agents is a toolkit for integrating large multi-modal models into existing robot stacks with just a few lines of code. It provides consistency, reliability, scalability, and is configurable to any observation and action space. The toolkit is designed to reduce complexities involved in setting up inference endpoints, converting between different model formats, and collecting/storing datasets. It aims to facilitate data collection and sharing among roboticists by providing Python-first abstractions that are modular, extensible, and applicable to a wide range of tasks. The toolkit supports asynchronous and remote thread-safe agent execution for maximal responsiveness and scalability, and is compatible with various APIs like HuggingFace Spaces, Datasets, Gymnasium Spaces, Ollama, and OpenAI. It also offers automatic dataset recording and optional uploads to the HuggingFace hub.

github

: 158

gemma

Gemma is a family of open-weights Large Language Model (LLM) by Google DeepMind, based on Gemini research and technology. This repository contains an inference implementation and examples, based on the Flax and JAX frameworks. Gemma can run on CPU, GPU, and TPU, with model checkpoints available for download. It provides tutorials, reference implementations, and Colab notebooks for tasks like sampling and fine-tuning. Users can contribute to Gemma through bug reports and pull requests. The code is licensed under the Apache License, Version 2.0.

github

: 3.7k

inspectus

Inspectus is a versatile visualization tool for large language models. It provides multiple views, including Attention Matrix, Query Token Heatmap, Key Token Heatmap, and Dimension Heatmap, to offer insights into language model behaviors. Users can interact with the tool in Jupyter notebooks through an easy-to-use Python API. Inspectus allows users to visualize attention scores between tokens, analyze how tokens focus on each other during processing, and explore the relationships between query and key tokens. The tool supports the visualization of attention maps from Huggingface transformers and custom attention maps, making it a valuable resource for researchers and developers working with language models.

github

: 600

lorax

LoRAX is a framework that allows users to serve thousands of fine-tuned models on a single GPU, dramatically reducing the cost of serving without compromising on throughput or latency. It features dynamic adapter loading, heterogeneous continuous batching, adapter exchange scheduling, optimized inference, and is ready for production with prebuilt Docker images, Helm charts for Kubernetes, Prometheus metrics, and distributed tracing with Open Telemetry. LoRAX supports a number of Large Language Models as the base model including Llama, Mistral, and Qwen, and any of the linear layers in the model can be adapted via LoRA and loaded in LoRAX.

github

: 2.3k

Dispider

Dispider is an implementation enabling real-time interactions with streaming videos, providing continuous feedback in live scenarios. It separates perception, decision-making, and reaction into asynchronous modules, ensuring timely interactions. Dispider outperforms VideoLLM-online on benchmarks like StreamingBench and excels in temporal reasoning. The tool requires CUDA 11.8 and specific library versions for optimal performance.

github

: 92

raid

RAID is the largest and most comprehensive dataset for evaluating AI-generated text detectors. It contains over 10 million documents spanning 11 LLMs, 11 genres, 4 decoding strategies, and 12 adversarial attacks. RAID is designed to be the go-to location for trustworthy third-party evaluation of popular detectors. The dataset covers diverse models, domains, sampling strategies, and attacks, making it a valuable resource for training detectors, evaluating generalization, protecting against adversaries, and comparing to state-of-the-art models from academia and industry.

github

: 55

DemoGPT

DemoGPT is an all-in-one agent library that provides tools, prompts, frameworks, and LLM models for streamlined agent development. It leverages GPT-3.5-turbo to generate LangChain code, creating interactive Streamlit applications. The tool is designed for creating intelligent, interactive, and inclusive solutions in LLM-based application development. It offers model flexibility, iterative development, and a commitment to user engagement. Future enhancements include integrating Gorilla for autonomous API usage and adding a publicly available database for refining the generation process.

github

: 1.7k

GraphRAG-SDK

Build fast and accurate GenAI applications with GraphRAG SDK, a specialized toolkit for building Graph Retrieval-Augmented Generation (GraphRAG) systems. It integrates knowledge graphs, ontology management, and state-of-the-art LLMs to deliver accurate, efficient, and customizable RAG workflows. The SDK simplifies the development process by automating ontology creation, knowledge graph agent creation, and query handling, enabling users to interact and query their knowledge graphs effectively. It supports multi-agent systems and orchestrates agents specialized in different domains. The SDK is optimized for FalkorDB, ensuring high performance and scalability for large-scale applications. By leveraging knowledge graphs, it enables semantic relationships and ontology-driven queries that go beyond standard vector similarity, enhancing retrieval-augmented generation capabilities.

github

: 292

VMind

VMind is an open-source solution for intelligent visualization, providing an intelligent chart component based on LLM by VisActor. It allows users to create chart narrative works with natural language interaction, edit charts through dialogue, and export narratives as videos or GIFs. The tool is easy to use, scalable, supports various chart types, and offers one-click export functionality. Users can customize chart styles, specify themes, and aggregate data using LLM models. VMind aims to enhance efficiency in creating data visualization works through dialogue-based editing and natural language interaction.

github

: 263

MInference

MInference is a tool designed to accelerate pre-filling for long-context Language Models (LLMs) by leveraging dynamic sparse attention. It achieves up to a 10x speedup for pre-filling on an A100 while maintaining accuracy. The tool supports various decoding LLMs, including LLaMA-style models and Phi models, and provides custom kernels for attention computation. MInference is useful for researchers and developers working with large-scale language models who aim to improve efficiency without compromising accuracy.

github

: 853

CodeFuse-ModelCache

Codefuse-ModelCache is a semantic cache for large language models (LLMs) that aims to optimize services by introducing a caching mechanism. It helps reduce the cost of inference deployment, improve model performance and efficiency, and provide scalable services for large models. The project caches pre-generated model results to reduce response time for similar requests and enhance user experience. It integrates various embedding frameworks and local storage options, offering functionalities like cache-writing, cache-querying, and cache-clearing through RESTful API. The tool supports multi-tenancy, system commands, and multi-turn dialogue, with features for data isolation, database management, and model loading schemes. Future developments include data isolation based on hyperparameters, enhanced system prompt partitioning storage, and more versatile embedding models and similarity evaluation algorithms.

github

: 626

Search-R1

Search-R1 is a tool that trains large language models (LLMs) to reason and call a search engine using reinforcement learning. It is a reproduction of DeepSeek-R1 methods for training reasoning and searching interleaved LLMs, built upon veRL. Through rule-based outcome reward, the base LLM develops reasoning and search engine calling abilities independently. Users can train LLMs on their own datasets and search engines, with preliminary results showing improved performance in search engine calling and reasoning tasks.

github

: 1.6k

For similar tasks

LLM-Drop

github

: 165

lighteval

LightEval is a lightweight LLM evaluation suite that Hugging Face has been using internally with the recently released LLM data processing library datatrove and LLM training library nanotron. We're releasing it with the community in the spirit of building in the open. Note that it is still very much early so don't expect 100% stability ^^' In case of problems or question, feel free to open an issue!

github

: 1.9k

Firefly

Firefly is an open-source large model training project that supports pre-training, fine-tuning, and DPO of mainstream large models. It includes models like Llama3, Gemma, Qwen1.5, MiniCPM, Llama, InternLM, Baichuan, ChatGLM, Yi, Deepseek, Qwen, Orion, Ziya, Xverse, Mistral, Mixtral-8x7B, Zephyr, Vicuna, Bloom, etc. The project supports full-parameter training, LoRA, QLoRA efficient training, and various tasks such as pre-training, SFT, and DPO. Suitable for users with limited training resources, QLoRA is recommended for fine-tuning instructions. The project has achieved good results on the Open LLM Leaderboard with QLoRA training process validation. The latest version has significant updates and adaptations for different chat model templates.

github

: 4.8k

Awesome-Text2SQL

Awesome Text2SQL is a curated repository containing tutorials and resources for Large Language Models, Text2SQL, Text2DSL, Text2API, Text2Vis, and more. It provides guidelines on converting natural language questions into structured SQL queries, with a focus on NL2SQL. The repository includes information on various models, datasets, evaluation metrics, fine-tuning methods, libraries, and practice projects related to Text2SQL. It serves as a comprehensive resource for individuals interested in working with Text2SQL and related technologies.

github

: 1.5k

create-million-parameter-llm-from-scratch

The 'create-million-parameter-llm-from-scratch' repository provides a detailed guide on creating a Large Language Model (LLM) with 2.3 million parameters from scratch. The blog replicates the LLaMA approach, incorporating concepts like RMSNorm for pre-normalization, SwiGLU activation function, and Rotary Embeddings. The model is trained on a basic dataset to demonstrate the ease of creating a million-parameter LLM without the need for a high-end GPU.

github

: 65

StableToolBench

StableToolBench is a new benchmark developed to address the instability of Tool Learning benchmarks. It aims to balance stability and reality by introducing features such as a Virtual API System with caching and API simulators, a new set of solvable queries determined by LLMs, and a Stable Evaluation System using GPT-4. The Virtual API Server can be set up either by building from source or using a prebuilt Docker image. Users can test the server using provided scripts and evaluate models with Solvable Pass Rate and Solvable Win Rate metrics. The tool also includes model experiments results comparing different models' performance.

github

: 59

BetaML.jl

The Beta Machine Learning Toolkit is a package containing various algorithms and utilities for implementing machine learning workflows in multiple languages, including Julia, Python, and R. It offers a range of supervised and unsupervised models, data transformers, and assessment tools. The models are implemented entirely in Julia and are not wrappers for third-party models. Users can easily contribute new models or request implementations. The focus is on user-friendliness rather than computational efficiency, making it suitable for educational and research purposes.

github

: 90

AI-TOD

AI-TOD is a dataset for tiny object detection in aerial images, containing 700,621 object instances across 28,036 images. Objects in AI-TOD are smaller with a mean size of 12.8 pixels compared to other aerial image datasets. To use AI-TOD, download xView training set and AI-TOD_wo_xview, then generate the complete dataset using the provided synthesis tool. The dataset is publicly available for academic and research purposes under CC BY-NC-SA 4.0 license.

github

: 173

For similar jobs

sweep

Sweep is an AI junior developer that turns bugs and feature requests into code changes. It automatically handles developer experience improvements like adding type hints and improving test coverage.

github

: 7.1k

teams-ai

The Teams AI Library is a software development kit (SDK) that helps developers create bots that can interact with Teams and Microsoft 365 applications. It is built on top of the Bot Framework SDK and simplifies the process of developing bots that interact with Teams' artificial intelligence capabilities. The SDK is available for JavaScript/TypeScript, .NET, and Python.

github

: 502

ai-guide

This guide is dedicated to Large Language Models (LLMs) that you can run on your home computer. It assumes your PC is a lower-end, non-gaming setup.

github

: 159

classifai

Supercharge WordPress Content Workflows and Engagement with Artificial Intelligence. Tap into leading cloud-based services like OpenAI, Microsoft Azure AI, Google Gemini and IBM Watson to augment your WordPress-powered websites. Publish content faster while improving SEO performance and increasing audience engagement. ClassifAI integrates Artificial Intelligence and Machine Learning technologies to lighten your workload and eliminate tedious tasks, giving you more time to create original content that matters.

github

: 669

chatbot-ui

Chatbot UI is an open-source AI chat app that allows users to create and deploy their own AI chatbots. It is easy to use and can be customized to fit any need. Chatbot UI is perfect for businesses, developers, and anyone who wants to create a chatbot.

github

: 27.7k

BricksLLM

BricksLLM is a cloud native AI gateway written in Go. Currently, it provides native support for OpenAI, Anthropic, Azure OpenAI and vLLM. BricksLLM aims to provide enterprise level infrastructure that can power any LLM production use cases. Here are some use cases for BricksLLM: * Set LLM usage limits for users on different pricing tiers * Track LLM usage on a per user and per organization basis * Block or redact requests containing PIIs * Improve LLM reliability with failovers, retries and caching * Distribute API keys with rate limits and cost limits for internal development/production use cases * Distribute API keys with rate limits and cost limits for students

github

: 953

uAgents

uAgents is a Python library developed by Fetch.ai that allows for the creation of autonomous AI agents. These agents can perform various tasks on a schedule or take action on various events. uAgents are easy to create and manage, and they are connected to a fast-growing network of other uAgents. They are also secure, with cryptographically secured messages and wallets.

github

: 1.3k

griptape

Griptape is a modular Python framework for building AI-powered applications that securely connect to your enterprise data and APIs. It offers developers the ability to maintain control and flexibility at every step. Griptape's core components include Structures (Agents, Pipelines, and Workflows), Tasks, Tools, Memory (Conversation Memory, Task Memory, and Meta Memory), Drivers (Prompt and Embedding Drivers, Vector Store Drivers, Image Generation Drivers, Image Query Drivers, SQL Drivers, Web Scraper Drivers, and Conversation Memory Drivers), Engines (Query Engines, Extraction Engines, Summary Engines, Image Generation Engines, and Image Query Engines), and additional components (Rulesets, Loaders, Artifacts, Chunkers, and Tokenizers). Griptape enables developers to create AI-powered applications with ease and efficiency.

github

: 2.2k