rd_("AhT5 model implementation.AhYi model implementation.CfImplementation of the Descript Audio Codec (DAC) modelCeThe epsilon to be used in the group normalization \xe2\x80\xa6BmThe amount of noise to be added at each step.BnIntersection over union of two bounding boxes.CkModule implementing the MPT (Multi-Purpose Transformer) \xe2\x80\xa6CmCreates a new EulerAncestral Discrete scheduler given the \xe2\x80\xa6BbMicrosoft Phi model implementationBfVariational Auto-Encoder (VAE) Models.AlVGG-16 model implementation.BhVision Transformer (ViT) implementation.B`A bounding box around an object.BhBased on the BEIT vision-language model.CdBERT (Bidirectional Encoder Representations from \xe2\x80\xa6CaBased on the BLIP paper from Salesforce Research.BgContrastive Language-Image Pre-Training0BcDenoising Diffusion Implicit ModelsAoEVA-2 inference implementation.kFlux  ModelAoReturns the argument unchanged.00000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000AoGLM-4 inference implementation.BaCalls <code>U::from(self)</code>.00000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000jmimi modelBiOLMo (Open Language Model) implementationBdMicrosoft Phi-3 model implementationBjPerforms a backward step during inference.0EhTopk in the last dim. <code>values</code> retains a gradient but <code>indices</code>\xe2\x80\xa6BlBased from the Stanford Hazy Research group.AoGemma inference implementation.BmHiera inference implementation based on timm.AoLlama inference implementation.CfThe LLaVA (Large Language and Vision Assistant) model.AoMamba inference implementation.CgMix of Multi-scale Dilated and Traditional ConvolutionsCeQwen2 model implementation with quantization support.AkTrOCR model implementation.AkApply penalty and repeat_kvc\xe2\x80\xa6AeLinear interpolation.BcTakes as input some sampled values.CgImplementation of the DINOv2 models from Meta Research.BmReturns the distribution in the latent space.BnFalcon language model inference implementationCiGemma LLM architecture (Google) inference implementation.0CcThe number of groups to use in group normalization.B`Helium inference implementation.BaMarian Neural Machine TranslationCgCandle implementations for various deep learning modelsAoRepVGG inference implementationAeResNet ImplementationAfResNet Building BlocksAlSiglip model implementation.oUniPC SchedulerCkBigCode implementation in Rust based on the GPT-BigCode \xe2\x80\xa6CcImplementation of the ChatGLM2/3 models from THUDM.C`Colpali Model for text/image similarity scoring.CcEnCodec neural audio codec based on the Encodec \xe2\x80\xa6BnFastViT inference implementation based on timmCeGranite is a Long Context Transformer Language Model.C`Mixtral Model, based on the Mistral architectureClMixtral Model, a sparse mixture of expert model based on \xe2\x80\xa6BcPixtral Language-Image Pre-TrainingAmRWKV v5 model implementation.AmRWKV v6 model implementation.Ah2D UNet Denoising ModelsAlWhisper Model ImplementationCnAn enum variant representing the Embedding head dimensions \xe2\x80\xa6C`Timesteps will be separated by regular intervalsBiThe value of beta at the end of training.00AhConvNeXt implementation.CbNote that the returned tensor uses the CPU device.B`Llama2 inference implementation.BlOpen Contrastive Language-Image Pre-TrainingAjCreates a ResNet-18 model.AjCreates a ResNet-34 model.CgThis trait represents a scheduler for the diffusion \xe2\x80\xa6AoAttention Based Building BlocksBlImplementation of BLIP text encoder/decoder.AiConvMixer implementation.CgConfigure the UNIC corrector. By default it is disabledBaJinaBERT inference implementationAjMetaVoice Studio ML ModelsBjMixFormer (Microsoft\xe2\x80\x99s Phi Architecture)iMobileOneAnMoonDream Model vision-to-textChMultimodal multi-purpose model combining Gemma-based \xe2\x80\xa6oPersimmon ModelCkQwen2 model implementation with Mixture of Experts support.CmRepeats a key or value tensor for grouped query attention \xe2\x80\xa6ClSegformer model implementation for semantic segmentation \xe2\x80\xa6AnStableLM model implementation.c\xe2\x80\xa6CgTimesteps will be determined by interpolation of sigmasBoThe value of beta at the beginning of training.0C`The value of beta at the beginning of training.nChImplementation of the DINOv2 revision (4 regularization)CjImplementation of DistilBert, a distilled version of BERT.AmLogit Processing and SamplingCmMobile CLIP model, combining a lightweight vision encoder \xe2\x80\xa6jModernBERTkNV-Embed-v2CmParler Model implementation for parler_tts text-to-speech \xe2\x80\xa6AnDiffusion pipelines and modelsCiStarCoder model implementation with quantization support.BoChinese contrastive Language-Image Pre-TrainingBgContrastive Language-Image Pre-TrainingCkText encoder as used in most OpenCLIP pretrained models \xe2\x80\xa6BdW\xc3\xbcrstchen Efficient Diffusion ModelCmOption to predicted sample between -1 and 1 for numerical \xe2\x80\xa6lMobileNet-v4CmThis represents how beta ranges from its minimum value to \xe2\x80\xa6C`Linear interpolation of the square root of beta.7CbCodeGeeX4 - A multi-language code generation modelClImplementation of EfficientBert, an efficient variant of \xe2\x80\xa6CkEfficientViT (MSRA) inference implementation based on timm.BnSame as forward pass but normalizes the outputEc<code>shift</code> is a triple <code>(image_seq_len, base_shift, max_shift)</code>.CjInitialize a new `stella_en_1.5B_v5`` model with given \xe2\x80\xa6BmInitialize new <code>stella_en_400M_v5</code>ClThe number of output channels, defaults to the number of \xe2\x80\xa6BgUtilities for quanitized network layersCbT5 model implementation with quantization support.DaThe solver order which can be <code>1</code> or higher. It is \xe2\x80\xa6AoStella v5 model implementation.CkAdjust the indexes of the inference schedule by this value.0CnWhether to use the \xe2\x80\x9cdynamic thresholding\xe2\x80\x9d method. This \xe2\x80\xa6BoChinese contrastive Language-Image Pre-TrainingBgContrastive Language-Image Pre-TrainingAcThe DDIM scheduler.BaHow beta evolved during training.00BcQuantized MPT model implementation.CdPhi2 model implementation with quantization support.EhTopk in the last dim. <code>values</code> retains a gradient but <code>indices</code>\xe2\x80\xa6CmOption to clip the variance used when adding noise to the \xe2\x80\xa6CdBLIP model implementation with quantization support.CdPhi3 model implementation with quantization support.C`Determines how sigma relates to a given timestepAg2D UNet Building BlocksE`When <code>None</code> no cross-attn is used, when <code>Some(d)</code> then \xe2\x80\xa6AeGlide cosine scheduleBlTime step spacing for the diffusion process.Djprediction type of the scheduler function, one of <code>epsilon</code> \xe2\x80\xa6Biprediction type of the scheduler function1BiPrediction type of the scheduler functionBeQuantized llama model implementation.CeQwen2 model implementation with quantization support.BdRecurrent Gemma model implementationCanumber of diffusion steps used to train the modelCbnumber of diffusion steps used to train the model.1CkWhether to use a 2D convolution in the skip connection. \xe2\x80\xa6B`Llama2 inference implementation.AoBounding Boxes and IntersectionBmThe threshold value for dynamic thresholding.AlSegment Anything Model (SAM)A`Stable DiffusionBktime step spacing for the diffusion process0CeImplementation of the Depth Anything model from FAIR.CfWhether to use lower-order solvers in the final steps.CgMistral model implementation with quantization support.CgRWKV v5 model implementation with quantization support.CgRWKV v6 model implementation with quantization support.CkEnsures interchangeability with schedulers that need to \xe2\x80\xa6000AeDetermines the pointsBfQuantized Llama2 model implementation.AoA CLIP transformer based model.0BiThe configuration for the DDIM scheduler.BaConfiguration for a ResNet block.CeThe final output is scaled by dividing by this value.BjQuantized BLIP text module implementation.BiQuantized MetaVoice model implementation.CkModule containing quantized MixFormer model implementation.CkImplementation of a quantized Moondream vision language \xe2\x80\xa6BmModule for quantized StableLM implementation.BeConfiguration for an attention block.E`Type of position embedding. Choose one of <code>&quot;absolute&quot;</code>, \xe2\x80\xa6lreferer: \xe2\x80\xa6greferer0BaVarbuilder for Loading gguf filesCiGenerate the mask and IOU predictions from some image \xe2\x80\xa6BlA text transformer as used in CLIP variants.BkAncestral sampling with Euler method steps.CjRecurrent Gemma model implementation with quantization \xe2\x80\xa6CmThe ratio for the dynamic thresholding method. Valid only \xe2\x80\xa6BfThe EulerAncestral Discrete scheduler.CeThe configuration for the EulerAncestral Discrete \xe2\x80\xa6")