BRIEF-DETAILS: Llama-2-7B optimized with bfloat16 precision and model sharding, enabling efficient deployment and reduced memory footprint while maintaining performance.
Brief Details: Llama-2-13B-Chat-fp16 is a compressed 16-bit floating-point variant of Meta's 13B parameter chat model, optimized for efficient deployment while maintaining performance.
Brief-details: A 13B parameter uncensored Vicuna model with SuperHOT RoPE extensions enabling 8K context length, available in various GGML quantizations (2-6 bit) for CPU inference.
Brief-details: German BERT-large model fine-tuned for zero-shot classification using NLI, achieving 85.6% accuracy on XNLI test set. Trained on 847K translated pairs.
Brief Details: Wav2Vec2-based model fine-tuned for speaker identification, achieving 75.18% accuracy on VoxCeleb1. Built for 16kHz audio processing.
Brief Details: Wav2vec2-based model for keyword spotting tasks, achieving 96.4% accuracy on Speech Commands v1.0. Optimized for 16kHz speech processing.
Brief Details: LUKE-based NER model achieving SOTA 94.3 F1 on CoNLL-2003. Specializes in entity recognition with knowledge-enhanced transformers.
Brief Details: A fine-tuned version of CodeT5-small specifically optimized for Python code summarization tasks, based on Salesforce's CodeT5 architecture.
Brief-details: German ELECTRA discriminator model trained on 844GB of clean Common Crawl data (GC4). Focused on bias research and NLP tasks for German language.
Brief-details: Stanza NLP model for Simplified Chinese language processing, offering state-of-the-art linguistic analysis capabilities including syntactic parsing and entity recognition.
Brief-details: Stanza NLP model for Ukrainian language processing - provides state-of-the-art linguistic analysis capabilities including syntactic parsing and entity recognition
Brief-details: CoreNLP - Stanford's comprehensive Java-based NLP toolkit offering linguistic annotations, parsing, NER, sentiment analysis, and more.
Brief-details: A distilled BART model for text summarization, offering 1.68x speedup over baseline with 306M parameters while achieving superior ROUGE scores on XSUM dataset.
Brief-details: SepFormer speech separation model achieving 19.8dB SI-SNRi on WSJ0-3Mix dataset. Separates mixed audio into distinct speaker sources at 8kHz sampling rate.
BRIEF-DETAILS: SepFormer audio source separation model trained on WHAMR! dataset, achieving 13.5dB SI-SNRi performance. Handles 16kHz audio with environmental noise and reverberation.
Brief-details: SepFormer model for audio source separation, trained on WHAMR! dataset. Achieves 13.7dB SI-SNRi performance. Handles environmental noise and reverberation in speech separation tasks.
Brief-details: Small Russian language model for spaCy with NER, POS tagging, and dependency parsing. High accuracy (98.77% POS, 94.98% NER F-score).
Brief Details: A tiny random initialization of the Longformer architecture, useful for testing and development purposes. Created by patrickvonplaten for experimental workflows.
Brief-details: Qwen2.5-Coder-32B-Instruct is a 4-bit quantized code-specific LLM with 32B parameters, optimized for code generation and reasoning with 32K context window.
BRIEF-DETAILS: Evaluation metric for autonomous driving video QA, closely aligns with human judgment on LingoQA suite. Developed by WayveAI.
BRIEF-DETAILS: NoobAI-XL 1.1 is an advanced language model hosted on HuggingFace, developed by Laxhar, focusing on natural language processing capabilities.
