Brief-details: EfficientNet B1 variant trained on ImageNet-1k + JFT-300M using Noisy Student method. 7.8M params, 240x240 input, optimized for efficient image classification.
BRIEF DETAILS: EXAONE-3.0-7.8B-Instruct is a bilingual (English/Korean) 7.8B parameter LLM, pre-trained on 8T tokens with competitive benchmark performance against similar models.
Brief-details: A state-of-the-art sentence segmentation model using 12 Transformer layers, designed for wtpsplit with advanced text segmentation capabilities
Brief-details: A minimalist Qwen2-based sequence classification model designed specifically for TRL library testing purposes, featuring a streamlined architecture.
Brief Details: Thai-English language model with 8B parameters, based on Llama3. Excels in Thai exam benchmarks and supports both languages with strong instruct capabilities.
🤖 Ultravox v0.5: A multimodal Speech LLM combining Llama 3.2-1B and Whisper for speech/text processing. MIT-licensed, supports audio input with text generation.
Brief Details: A compact variant of InternVL2 model created by katuni4ka, designed for experimental purposes with randomized weights for vision-language tasks.
BRIEF-DETAILS: Ancillary fluency model for the Parrot paraphraser framework, designed to enhance NLU model training through paraphrase-based utterance augmentation.
Brief Details: Line detection model focused on surya system implementation. Created by vikp, hosted on HuggingFace for specialized linear pattern recognition.
BRIEF-DETAILS: Large-scale outdoor depth estimation model with 335.3M params, using DPT architecture & DINOv2 backbone, trained on 600K synthetic + 62M real images
Brief-details: A randomly initialized tiny T5 model variant created by patrickvonplaten, useful for testing and baseline comparisons in NLP tasks
Brief Details: HuBERT base model trained on 200k hours of general audio data, designed for robust audio understanding and representation learning
BRIEF DETAILS: Neural MT model for English-Turkish translation, part of OPUS-MT project. Uses transformer-big architecture with SentencePiece tokenization and achieves 42.3 BLEU on Tatoeba test set.
Brief-details: Perfect Gap Blend v1 is a specialized AI model designed for enhanced image generation, focusing on improved detail consistency and gap handling in generated images.
Brief Details: Fooocus Inpaint - A specialized inpainting model by lllyasviel, designed for precise image manipulation and restoration tasks on Hugging Face.
Brief Details: Kokoro-82M-ONNX is a compact yet powerful text-to-speech model with 82M parameters, supporting multiple English voices and efficient quantization options for deployment.
Brief-details: OminiControl is a minimal and universal control framework for Diffusion Transformer, enabling enhanced control capabilities for image generation and manipulation.
Brief-details: OpenChat 3.5 7B GGUF model, achieving ChatGPT-comparable performance with 7.81 MT-bench score. Features multiple quantization options and GPU acceleration support.
BRIEF-DETAILS: A creative fusion model combining Synthwave and InkPunk artistic styles, offering dual-style generation with adjustable weightings via tokens 'snthwve style' and 'nvinkpunk'.
Brief Details: A specialized Chinese-focused adaptation of Whisper-tiny using federated learning techniques, optimized for Chinese speech recognition tasks and distributed training environments.
Brief Details: A Cantonese-optimized Whisper tiny model leveraging federated learning techniques for improved speech recognition and transcription tasks.
