# Lightweight

This is a miniature model for debugging purposes, using random initialization, configured based on the meta-llama Llama-4-Maverick-17B-128E-Instruct model.

Kanana is a bilingual (Korean/English) language model series developed by Kakao. This 2.1B parameter version outperforms similar models in Korean while maintaining efficient computational costs.

This project is an open-source project following the MIT license and has certain open-source value.

OpenChat 3.5 7B is a 7B-parameter large language model based on the Mistral architecture, developed by the OpenChat team and released under the Apache 2.0 license.

This is a sentence encoder model based on sentence-transformers, capable of mapping sentences and paragraphs into a 256-dimensional dense vector space, suitable for tasks such as sentence similarity calculation, clustering, and semantic search.

An image classification model fine-tuned on a letter recognition dataset based on Google's ViT base model, achieving 98.81% accuracy

This is a plain old model with unspecified specific functions and purposes.

This model is released under the MIT License, with no specific details provided.

This is a speech recognition model fine-tuned on the superb dataset based on distilhubert, primarily used for Automatic Speech Recognition (ASR) tasks.

A vision Transformer (ViT)-based object detection model trained with DETR loss function, achieving excellent performance on the COCO dataset.

An audio classification model based on the wav2vec 2.0 architecture, designed to distinguish between adult and child voices

ConvNeXT is a pure convolutional model inspired by Vision Transformer designs, trained on the ImageNet-1k dataset, outperforming Transformers.