BRIEF DETAILS: SPECTER2 base model for scientific paper embeddings with adapter support. Built on SciBERT, optimized for research tasks like classification and search.
Brief-details: A compact, efficient speech recognition model supporting 99 languages, built on CTranslate2 framework with MIT license and optimized for fast inference
Brief Details: BART-based dialogue summarization model fine-tuned on SAMSum dataset, achieving 41.32 ROUGE-1 score. Optimized for conversation summary tasks.
Brief-details: DeBERTa-v3-Large model fine-tuned for boolean question answering, achieving 88.35% accuracy on BoolQ dataset. 435M parameters, MIT license.
Brief-details: French Named Entity Recognition model using Flair embeddings & LSTM-CRF architecture. Achieves 90.61% F1-score on WikiNER for 4-class NER tagging.
Brief Details: Powerful T5-based time series forecasting model (710M params) that transforms numerical data into tokens for probabilistic predictions
BRIEF DETAILS: Multilingual speech recognition model built on wav2vec2-xlsr-53, fine-tuned for phoneme recognition across multiple languages using CommonVoice dataset.
Brief Details: Compact 39M-parameter speech recognition model supporting 99 languages, part of OpenAI's Whisper family. Ideal for lightweight ASR tasks.
Brief-details: XLS-R 1B-based speech recognition model for Portuguese, fine-tuned on Common Voice 8.0. Achieves 8.7% WER, drops to 6.04% with LM. 360K+ downloads.
Brief-details: RoBERTa-large model fine-tuned on MNLI dataset, optimized for zero-shot classification tasks with 356M parameters and state-of-the-art performance.
Brief Details: Multilingual NER model supporting 9 languages, fine-tuned on WikiNEuRal dataset with 177M parameters. Ideal for named entity recognition across languages.
Brief-details: E5-base-v2 is a 109M parameter text embedding model optimized for semantic similarity and retrieval tasks, featuring 12 layers and 768-dim embeddings
Brief Details: StarCoder2-15B GPTQ - Quantized 15B parameter code generation model supporting 600+ programming languages with 16K context window and Fill-in-the-Middle training
Brief Details: DeBERTa-v3 model fine-tuned for aspect-based sentiment analysis, trained on 180k examples across 8 datasets. 184M parameters, MIT licensed.
Brief-details: Qwen2-VL-7B-Instruct-AWQ is a quantized multimodal model offering advanced image/video understanding with 7B parameters, optimized for efficient deployment.
Brief-details: Advanced depth estimation AI model from Tencent that generates consistent depth sequences for videos, with 366K+ downloads and strong research backing.
Brief-details: BERT-based question encoder for open-domain Q&A, trained on Natural Questions dataset. Optimized for dense passage retrieval with 85.4% top-100 accuracy.
Brief Details: Multilingual punctuation restoration model supporting EN, DE, FR, IT. 559M params, high F1 scores for common punctuation marks. MIT licensed.
Brief Details: Llama-3.2-1B-Instruct-Q8_0-GGUF is a GGUF-converted 1.24B parameter LLM supporting 8 languages, optimized for efficient deployment using llama.cpp
Brief-details: CodeBERT model fine-tuned on Python code for masked language modeling, primarily used for code evaluation via CodeBERTScore framework. 372K+ downloads.
Brief-details: DPR context encoder model trained on Natural Questions dataset, specialized for open-domain QA with BERT architecture. High accuracy (78.4% Top-20) on NQ dataset.
