Brief Details: 8B parameter instruction-tuned model with multiple GGUF quantized versions - optimized for efficient deployment with sizes from 3.2GB to 16.4GB
Brief Details: PipaT1-500M is a 500M parameter language model fine-tuned from Qwen2-0.5b-instruct, developed by ksanjeeb under Apache-2.0 license.
BRIEF DETAILS: 12B parameter GGUF-quantized model with multiple compression variants (3.1GB-10.2GB), optimized for efficient deployment while maintaining performance.
BRIEF-DETAILS: 24B parameter Mistral model with multiple GGUF quantization options (7-25GB), optimized for different RAM/VRAM constraints and performance needs. Features Q2-Q8 quantization variants.
Brief Details: 4-bit quantized version of Qwen2-Audio-7B-Instruct for audio-text processing, offering reduced memory usage while maintaining core capabilities
Brief-details: Neural machine translation model for Polish to Norwegian translation. Transformer-align architecture with SentencePiece tokenization, achieving 27.5 BLEU score.
Brief-details: A specialized classifier model built on transformer architecture to distinguish between question and statement queries, primarily designed for neural search applications and Haystack integration.
BRIEF DETAILS: 1B parameter Llama2-based embedding model specialized for Malaysian text, with 8k training context length and 32k inference scaling capability.
BRIEF-DETAILS: Neural machine translation model for Polish to German translation, using transformer architecture with SentencePiece preprocessing. BLEU score: 47.8
Brief-details: German to Czech neural machine translation model trained on OPUS data, achieving BLEU scores of 20-42 across various test sets
BRIEF-DETAILS: TinyStories-1M: Language model trained on simple stories dataset, designed for generating coherent short narratives. Uses GPT-Neo tokenizer.
Brief-details: A tiny random model created by katuni4ka, with connections to other tiny-random variants. Primarily focused on experimental AI development with minimal architecture details available.
Brief-details: Chinese punctuation restoration model that supports 6 punctuation marks (， 、 。 ？ ！ ；). Uses transformer architecture to automatically insert correct punctuation in unpunctuated Chinese text.
Brief-details: Large-scale Norwegian ASR model (1B parameters) for Nynorsk dialect, achieving 11.32% WER. Built on XLS-R, trained on NPSC dataset.
Brief-details: A transformers-based model by conjuncts hosted on Hugging Face Hub. Limited documentation available but appears to be an experimental model implementation.
Brief-details: Personal model collection curated by lllyasviel, featuring ControlNet-related models and optimization tools. Focused on private use and development.
BRIEF-DETAILS: Alpaca-13B is a native implementation of the Alpaca model with 13B parameters, offering instruction-following capabilities without LoRA fine-tuning.
BRIEF DETAILS: RecurrentGemma-2B-IT: Google's instruction-tuned variant of RecurrentGemma-2B, requiring Hugging Face license acceptance for access. Built on Gemma architecture.
Brief-details: GGML-optimized version of LLaVA-1.5-7B for efficient local inference using llama.cpp, enabling end-to-end multimodal capabilities without extra dependencies
BRIEF DETAILS: Astra-v1-12B quantized model with multiple GGUF variants optimized for different size/performance tradeoffs. Features iMatrix quantization for improved efficiency.
Brief Details: Astra-v1-12B-GGUF is a quantized version of the Astra language model, offering multiple compression variants from 4.9GB to 13.1GB with varying quality-size tradeoffs.
