Llama-3.1-8B-onnx-ryzenai-npu is an optimized version developed by AMD based on Meta's Llama-3.1-8B model, specifically optimized for deployment on the AMD Ryzen AI NPU. Through Quark quantization, OGA model builder, and NPU-specific post-processing technology, this model significantly improves the inference efficiency on AMD hardware while maintaining high-quality text generation capabilities.

Natural Language ProcessingOnnxEnglish

This model is the 8-billion parameter version in the Llama 3.1 series released by Meta. It has been processed and post-processed by the AMD Quark quantization tool, optimizing the deployment and inference efficiency on specific hardware. It is a multilingual text generation model that supports multiple languages and performs excellently in benchmark tests such as MMLU.

Natural Language ProcessingOnnxMultiple Languages

This project is a text generation model obtained by quantizing the Qwen3-4B model using AMD's quantization tool. It can be used in scenarios such as chatting. The model is prepared using the AMD Quark quantization tool and undergoes necessary post-processing.

Natural Language ProcessingOnnxEnglish

PARD is a high-performance speculative decoding method that can convert autoregressive draft models into parallel draft models at low cost, significantly improving the inference speed of large language models and reducing deployment complexity and adaptation costs.

Natural Language ProcessingTransformers

This is a derivative model based on Meta AI's LLaMA 3.1 base model, specifically optimized for FastFlowLM on the AMD Ryzen™ AI NPU and only applicable to the XDNA2 architecture. The model retains the core architecture and weights released by Meta and may be fine-tuned, quantized, or adapted for specific applications.

This is an optimized variant of the LLaMA 3.2 1B Instruct model released by Meta AI, specifically optimized for FastFlowLM on the AMD Ryzen™ AI NPU (XDNA2 architecture). The model retains the original architecture and weights, and improves the operation efficiency on the NPU through technologies such as quantization and low-level tuning.

Motif 2.6B is a language model with 2.6 billion parameters, trained from scratch on AMD Instinct™ MI250 GPUs, aiming to build AI that aligns with human values, is useful, and reliable.

Natural Language ProcessingSafetensorsMultiple Languages

Nitro-T is a series of text-to-image diffusion models focused on efficient training, which are competitive in image generation benchmark tests and have extremely short training times.

MultimodalDiffusersOther

Instella-Long is an open-source language model with 3B parameters developed by AMD, supporting a context length of 128K and performing excellently in long-context benchmark tests.

Natural Language ProcessingTransformers

PARD is a high-performance speculative decoding method that can convert autoregressive draft models into parallel draft models at low cost, significantly accelerating the inference of large language models.

Natural Language ProcessingTransformers

Based on the OLMo-2-0325-32B-Instruct model, fine-tuned using the Open R1 math dataset on AMD MI300X GPUs, focusing on enhancing mathematical reasoning capabilities

Natural Language ProcessingTransformersEnglish

Instella is a completely open-source language model series with 3 billion parameters trained from scratch by AMD based on the Instinct™ MI300X GPU. Its performance surpasses that of open-source models of the same scale and competes with advanced open-weight models such as Llama-3.2-3B and Gemma-2-2B.

Natural Language ProcessingTransformers

AMD's fully open 3-billion-parameter language model family trained on Instinct MI300X GPUs, outperforming open models of similar scale

Natural Language ProcessingTransformers

Instella is a series of 3-billion-parameter open-source language models developed by AMD, trained on AMD Instinct™ MI300X GPUs, outperforming other fully open-source models of the same scale.

Natural Language ProcessingTransformers

AMD Ryzen™ AI-optimized SDXL-Turbo, the world's first text-to-image model using Block FP16 format, designed specifically for AMD XDNA™ 2 NPU

MultimodalOnnxEnglish

AMD Nitro Diffusion is a series of efficient text-to-image models, distilled from mainstream diffusion models on AMD Instinct™ GPUs. PixArt-Sigma Nitro is a high-resolution single-step inference model based on Transformer architecture.

Computer VisionDiffusers

An efficient text-to-image model series distilled from mainstream diffusion models on AMD Instinct™ GPUs

Computer VisionDiffusers

AMD-OLMo is a series of 1 billion parameter language models trained from scratch by AMD on the AMD Instinct™ MI250 GPU, including pre-trained models, supervised fine-tuned models, and alignment models based on direct preference optimization (DPO). This series of models is based on the OLMo architecture, pre-trained on the Dolma dataset, and fine-tuned on multiple instruction datasets.

Natural Language ProcessingTransformers

AMD-OLMo is a series of 1-billion-parameter language models trained from scratch by AMD on the AMD Instinct™ MI250 GPU.

Natural Language ProcessingSafetensors

Phi-3.5-mini-instruct is an efficient small language model developed by Microsoft, optimized using advanced quantization techniques and designed specifically for NPU deployment. This model performs excellently in text generation tasks and supports natural language processing and code-related scenarios.

Natural Language ProcessingTransformersOther