General activation Quantization-DeQuantization (QDQ) module. Supports both Symmetric and Asymmetric (Affine) quantization. Uses torch.qin

\brief Class interface for specifying NDRange values.

===----------------------------------------------------------------------===// RadixTree Node Key ===-------------------------------------------------

Fetch maximum data range of all tensors to be concatenated

Main MiniCPM-o Model

Fixed activation Quantization-DeQuantization (QDQ) module. Uses pre-determined scale and zero_point instead of dynamic observation. Suppo

* @brief Configuration for Llama 3.x models used in QNN AOT compilation. * * This configuration is designed to support Llama 3.2 3B Instruct and sim

Convenience pipeline resembling common torchvision usage: transforms = Compose([Resize, (optional) CenterCrop, ToTensor, (optional) Normalize]) - User

Hidden States Key & Value Cache

* @brief Iterator for streaming multimodal generation * * This iterator provides Python-like streaming generation interface: * ```cpp * auto strea

* @brief Main Vocos model * * Consists of: * 1. Feature extractor (MelSpectrogramFeatures) * 2. Backbone (VocosBackbone) * 3. Head (ISTFTHead) *

Audio pooling wrapper module

Audio Projection Layer for projecting audio features to text embedding space

* Exception class for build errors to carry build info */

* ConditionalChatTTS: A conditional text-to-speech model that can generate speech from text with speaker conditioning. * * This model extends ARGene

* @class ConfigFile * @brief A utility class for handling JSON-based configuration files. * * This class wraps the nlohmann::json library to provid

Actually is LlamaTokenizer

===----------------------------------------------------------------------===// Elementwise Add ===----------------------------------------------------

! \brief Exception class * * This may be thrown by API functions when CL_HPP_ENABLE_EXCEPTIONS is defined. */

Hidden States Key & Value Cache This KVCache is managed in Paged method [B, S, H, D] is the format that received.

* @brief ISTFTHead module for Vocos * * This module transforms features to STFT coefficients and applies inverse STFT */

Multi-headed attention from 'Attention Is All You Need' paper