Best AI tools for< Quantize Weights >

FLUTE (Flexible Lookup Table Engine for LUT-quantized LLMs) is a tool designed for uniform quantization and lookup table quantization of weights in lower-precision intervals. It offers flexibility in mapping intervals to arbitrary values through a lookup table. FLUTE supports various quantization formats such as int4, int3, int2, fp4, fp3, fp2, nf4, nf3, nf2, and even custom tables. The tool also introduces new quantization algorithms like Learned Normal Float (NFL) for improved performance and calibration data learning. FLUTE provides benchmarks, model zoo, and integration with frameworks like vLLM and HuggingFace for easy deployment and usage.

qlora-pipe is a pipeline parallel training script designed for efficiently training large language models that cannot fit on one GPU. It supports QLoRA, LoRA, and full fine-tuning, with efficient model loading and the ability to load any dataset that Axolotl can handle. The script allows for raw text training, resuming training from a checkpoint, logging metrics to Tensorboard, specifying a separate evaluation dataset, training on multiple datasets simultaneously, and supports various models like Llama, Mistral, Mixtral, Qwen-1.5, and Cohere (Command R). It handles pipeline- and data-parallelism using Deepspeed, enabling users to set the number of GPUs, pipeline stages, and gradient accumulation steps for optimal utilization.

Bark.cpp is a C/C++ implementation of the Bark model, a real-time, multilingual text-to-speech generation model. It supports AVX, AVX2, and AVX512 for x86 architectures, and is compatible with both CPU and GPU backends. Bark.cpp also supports mixed F16/F32 precision and 4-bit, 5-bit, and 8-bit integer quantization. It can be used to generate realistic-sounding audio from text prompts.

The stable-diffusion.cpp repository provides an implementation for inferring stable diffusion in pure C/C++. It offers features such as support for different versions of stable diffusion, lightweight and dependency-free implementation, various quantization support, memory-efficient CPU inference, GPU acceleration, and more. Users can download the built executable program or build it manually. The repository also includes instructions for downloading weights, building from scratch, using different acceleration methods, running the tool, converting weights, and utilizing various features like Flash Attention, ESRGAN upscaling, PhotoMaker support, and more. Additionally, it mentions future TODOs and provides information on memory requirements, bindings, UIs, contributors, and references.

This repository contains the training code of LLM-QAT for large language models. The work investigates quantization-aware training for LLMs, including quantizing weights, activations, and the KV cache. Experiments were conducted on LLaMA models of sizes 7B, 13B, and 30B, at quantization levels down to 4-bits. Significant improvements were observed when quantizing weight, activations, and kv cache to 4-bit, 8-bit, and 4-bit, respectively.

The main goal of llama.cpp is to enable LLM inference with minimal setup and state-of-the-art performance on a wide range of hardware - locally and in the cloud. It provides a Plain C/C++ implementation without any dependencies, optimized for Apple silicon via ARM NEON, Accelerate and Metal frameworks, and supports various architectures like AVX, AVX2, AVX512, and AMX. It offers integer quantization for faster inference, custom CUDA kernels for NVIDIA GPUs, Vulkan and SYCL backend support, and CPU+GPU hybrid inference. llama.cpp is the main playground for developing new features for the ggml library, supporting various models and providing tools and infrastructure for LLM deployment.

llm-compressor is an easy-to-use library for optimizing models for deployment with vllm. It provides a comprehensive set of quantization algorithms, seamless integration with Hugging Face models and repositories, and supports mixed precision, activation quantization, and sparsity. Supported algorithms include PTQ, GPTQ, SmoothQuant, and SparseGPT. Installation can be done via git clone and local pip install. Compression can be easily applied by selecting an algorithm and calling the oneshot API. The library also offers end-to-end examples for model compression. Contributions to the code, examples, integrations, and documentation are appreciated.

Awesome-LLM-Quantization is a curated list of resources related to quantization techniques for Large Language Models (LLMs). Quantization is a crucial step in deploying LLMs on resource-constrained devices, such as mobile phones or edge devices, by reducing the model's size and computational requirements.

llama.cpp is a C++ implementation of LLaMA, a large language model from Meta. It provides a command-line interface for inference and can be used for a variety of tasks, including text generation, translation, and question answering. llama.cpp is highly optimized for performance and can be run on a variety of hardware, including CPUs, GPUs, and TPUs.

The worker-vLLM repository provides a serverless endpoint for deploying OpenAI-compatible vLLM models with blazing-fast performance. It supports deploying various model architectures, such as Aquila, Baichuan, BLOOM, ChatGLM, Command-R, DBRX, DeciLM, Falcon, Gemma, GPT-2, GPT BigCode, GPT-J, GPT-NeoX, InternLM, Jais, LLaMA, MiniCPM, Mistral, Mixtral, MPT, OLMo, OPT, Orion, Phi, Phi-3, Qwen, Qwen2, Qwen2MoE, StableLM, Starcoder2, Xverse, and Yi. Users can deploy models using pre-built Docker images or build custom images with specified arguments. The repository also supports OpenAI compatibility for chat completions, completions, and models, with customizable input parameters. Users can modify their OpenAI codebase to use the deployed vLLM worker and access a list of available models for deployment.

AirLLM is a tool that optimizes inference memory usage, enabling large language models to run on low-end GPUs without quantization, distillation, or pruning. It supports models like Llama3.1 on 8GB VRAM. The tool offers model compression for up to 3x inference speedup with minimal accuracy loss. Users can specify compression levels, profiling modes, and other configurations when initializing models. AirLLM also supports prefetching and disk space management. It provides examples and notebooks for easy implementation and usage.

Marlin is a highly optimized FP16xINT4 matmul kernel designed for large language model (LLM) inference, offering close to ideal speedups up to batchsizes of 16-32 tokens. It is suitable for larger-scale serving, speculative decoding, and advanced multi-inference schemes like CoT-Majority. Marlin achieves optimal performance by utilizing various techniques and optimizations to fully leverage GPU resources, ensuring efficient computation and memory management.

DashInfer is a C++ runtime tool designed to deliver production-level implementations highly optimized for various hardware architectures, including x86 and ARMv9. It supports Continuous Batching and NUMA-Aware capabilities for CPU, and can fully utilize modern server-grade CPUs to host large language models (LLMs) up to 14B in size. With lightweight architecture, high precision, support for mainstream open-source LLMs, post-training quantization, optimized computation kernels, NUMA-aware design, and multi-language API interfaces, DashInfer provides a versatile solution for efficient inference tasks. It supports x86 CPUs with AVX2 instruction set and ARMv9 CPUs with SVE instruction set, along with various data types like FP32, BF16, and InstantQuant. DashInfer also offers single-NUMA and multi-NUMA architectures for model inference, with detailed performance tests and inference accuracy evaluations available. The tool is supported on mainstream Linux server operating systems and provides documentation and examples for easy integration and usage.

AIMET is a library that provides advanced model quantization and compression techniques for trained neural network models. It provides features that have been proven to improve run-time performance of deep learning neural network models with lower compute and memory requirements and minimal impact to task accuracy. AIMET is designed to work with PyTorch, TensorFlow and ONNX models. We also host the AIMET Model Zoo - a collection of popular neural network models optimized for 8-bit inference. We also provide recipes for users to quantize floating point models using AIMET.

HF-Waitress is a powerful server application for deploying and interacting with HuggingFace Transformer models. It simplifies running open-source Large Language Models (LLMs) locally on-device, providing on-the-fly quantization via BitsAndBytes, HQQ, and Quanto. It requires no manual model downloads, offers concurrency, streaming responses, and supports various hardware and platforms. The server uses a `config.json` file for easy configuration management and provides detailed error handling and logging.

Torchtune is a PyTorch-native library for easily authoring, fine-tuning, and experimenting with LLMs. It provides native-PyTorch implementations of popular LLMs using composable and modular building blocks, easy-to-use and hackable training recipes for popular fine-tuning techniques, YAML configs for easily configuring training, evaluation, quantization, or inference recipes, and built-in support for many popular dataset formats and prompt templates to help you quickly get started with training.

GPTQModel is an easy-to-use LLM quantization and inference toolkit based on the GPTQ algorithm. It provides support for weight-only quantization and offers features such as dynamic per layer/module flexible quantization, sharding support, and auto-heal quantization errors. The toolkit aims to ensure inference compatibility with HF Transformers, vLLM, and SGLang. It offers various model supports, faster quant inference, better quality quants, and security features like hash check of model weights. GPTQModel also focuses on faster quantization, improved quant quality as measured by PPL, and backports bug fixes from AutoGPTQ.

AutoRound is an advanced weight-only quantization algorithm for low-bits LLM inference. It competes impressively against recent methods without introducing any additional inference overhead. The method adopts sign gradient descent to fine-tune rounding values and minmax values of weights in just 200 steps, often significantly outperforming SignRound with the cost of more tuning time for quantization. AutoRound is tailored for a wide range of models and consistently delivers noticeable improvements.

llmc is an off-the-shell tool designed for compressing LLM, leveraging state-of-the-art compression algorithms to enhance efficiency and reduce model size without compromising performance. It provides users with the ability to quantize LLMs, choose from various compression algorithms, export transformed models for further optimization, and directly infer compressed models with a shallow memory footprint. The tool supports a range of model types and quantization algorithms, with ongoing development to include pruning techniques. Users can design their configurations for quantization and evaluation, with documentation and examples planned for future updates. llmc is a valuable resource for researchers working on post-training quantization of large language models.

Llama3.java is a practical Llama 3 inference tool implemented in a single Java file. It serves as the successor of llama2.java and is designed for testing and tuning compiler optimizations and features on the JVM, especially for the Graal compiler. The tool features a GGUF format parser, Llama 3 tokenizer, Grouped-Query Attention inference, support for Q8_0 and Q4_0 quantizations, fast matrix-vector multiplication routines using Java's Vector API, and a simple CLI with 'chat' and 'instruct' modes. Users can download quantized .gguf files from huggingface.co for model usage and can also manually quantize to pure 'Q4_0'. The tool requires Java 21+ and supports running from source or building a JAR file for execution. Performance benchmarks show varying tokens/s rates for different models and implementations on different hardware setups.