# Large-scale pretraining

SigLIP 2 is an improved model based on SigLIP, integrating multiple technologies to enhance semantic understanding, localization, and dense feature extraction capabilities.

This model is a Russian automatic speech recognition (ASR) model fine-tuned on the common_voice_17_0 dataset based on facebook/wav2vec2-xls-r-300m, with a word error rate (WER) of 0.195.

This is a vision-language model based on the OpenCLIP framework, trained on the LAION-2B English subset, excelling in zero-shot image classification and cross-modal retrieval tasks.

A vision-language model trained on the LAION-2B English dataset based on the OpenCLIP framework, supporting zero-shot image classification and cross-modal retrieval

AIM-v2 is an image encoder model based on the timm library, with a scale of 1 billion parameters, suitable for image feature extraction tasks.

The EVA CLIP model is a vision-language model based on OpenCLIP and the timm framework, supporting zero-shot image classification tasks.

The EVA CLIP model is a vision-language model based on OpenCLIP and timm model weights, supporting tasks such as zero-shot image classification.

EVA-CLIP is a large-scale vision-language model based on the CLIP architecture, supporting tasks such as zero-shot image classification.

A vision transformer model based on the CLIP architecture, focused on image feature extraction, released by Apple.

Speech encoder module extracted from SeamlessM4Tv2-Large, excelling in cross-language and multilingual sequence-level audio classification tasks

A dual-framework compatible vision model trained on MetaCLIP-2.5B dataset, supporting both OpenCLIP and timm frameworks

Qwen2-Audio is the Tongyi Qianwen large audio language model series, supporting both voice chat and audio analysis interaction modes.

CLIP ViT-B/32 model trained on the LAION-2B dataset based on the OpenCLIP framework, supporting zero-shot image classification and cross-modal retrieval tasks

OWSM-CTC is an encoder-only speech foundation model based on hierarchical multi-task self-conditioned CTC, supporting multilingual speech recognition, speech translation, and language identification.

Chronos is a family of pretrained time series forecasting models based on language model architecture, which converts time series into token sequences through quantization and scaling for training, supporting probabilistic forecasting.

A speech recognition model fine-tuned on the Common Voice 16.0 dataset based on OpenAI Whisper Large V3

A speech encoder based on the Conformer architecture, pretrained on 4.5 million hours of unlabeled audio data, supporting over 143 languages

French sentence embedding model based on CamemBERT-large, providing powerful semantic search capabilities

CLIPA-v2 model, trained on the laion2B-en dataset, focusing on zero-shot image classification tasks

MetaCLIP is a large-scale vision-language model trained on 2.5 billion data points from CommonCrawl (CC), revealing CLIP's data filtering methodology

This is a CLIP ViT-B/32 model trained on the DataComp-1B dataset, designed for tasks like zero-shot image classification and image-text retrieval.

This is a BART large model pretrained from scratch with 400 million parameters, specifically designed for Romanian language.

Pile-T5 Large is an encoder-decoder model trained on The Pile dataset based on the T5x library, primarily used for English text-to-text generation tasks.

A vision Transformer model trained using the DINOv2 method for self-supervised image feature extraction

IDEFICS is an open-source multimodal model capable of processing both image and text inputs to generate text outputs, serving as an open-source reproduction of Deepmind's Flamingo model.

A vision-language pretrained model implemented based on OpenCLIP, supporting zero-shot image classification tasks

A vision-language model based on the CLIP architecture, supporting zero-shot image classification and cross-modal retrieval

A vision-language pretrained model based on the CLIP architecture, supporting zero-shot image classification and cross-modal retrieval tasks

A vision-language model based on the CLIP architecture, supporting zero-shot image classification tasks

Zero-shot image classification model based on CLIP architecture, supporting general pooling functionality

A vision-language model based on the CLIP architecture, supporting zero-shot image classification tasks

A vision-language model based on the CLIP architecture, supporting zero-shot image classification tasks

A vision-language model based on the CLIP architecture, supporting zero-shot image classification tasks

ARBERTv2 is an upgraded BERT model trained on Modern Standard Arabic (MSA) with a corpus of 243GB text, containing 27.8 billion tokens.

EVA02 is a large-scale vision-language model based on the CLIP architecture, supporting zero-shot image classification tasks.

An mT5 model fine-tuned on the XL-Sum dataset for 45 languages, designed for multilingual summarization tasks.

A multimodal vision-language model trained on the OpenCLIP framework, completed on the LAION-2B English dataset, supporting zero-shot image classification tasks

MaltBERTa is a large-scale pretrained language model based on Maltese text, using the RoBERTa architecture, developed by the MaCoCu project.

A language model pretrained on large-scale Bulgarian and Macedonian texts, part of the MaCoCu project

A Brazilian Portuguese speech recognition model fine-tuned on the Common Voice dataset based on Facebook's wav2vec2-large-xlsr-53-portuguese model