Brief-details: Spanish Named Entity Recognition model using XLM-RoBERTa architecture. Achieves 90.54% F1-score on CoNLL-03 Spanish, identifying PER, LOC, ORG, and MISC entities.
Brief-details: Dutch Named Entity Recognition model using BERT and LSTM-CRF architecture. Achieves 92.58% F1-score on CoNLL-03, detects PER/LOC/ORG/MISC entities.
Brief-details: Large-scale Dutch Named Entity Recognition model using FLERT architecture. Achieves 95.25% F1-score on CoNLL-03 Dutch. Identifies PER, LOC, ORG, and MISC entities.
Brief-details: RoBERTa-based targeted sentiment classifier for news articles, specialized in analyzing sentiment for specific entities within text contexts
Brief Details: T5-small model fine-tuned on XSum dataset for abstractive text summarization, optimized for generating concise news summaries.
BRIEF DETAILS: Qwen2.5-3B optimized with Unsloth's Dynamic 4-bit quantization. Offers 2x faster performance, 60% less memory usage, and specialized quantization for improved accuracy.
Brief Details: Ai-Thalli is a fine-tuned LLaMA model optimized for multi-language text generation tasks, featuring easy integration with the Transformers library
Brief-details: A specialized sentence transformer model fine-tuned for course recommendation, based on BGE-base-en-v1.5, outputting 768-dimensional vectors for semantic search and similarity tasks
Brief Details: 72B parameter Qwen2.5-based model optimized for generalist and roleplay tasks, featuring FP8 dynamic quantization and ChatML format support.
BRIEF-DETAILS: Facebook's MMS-300M is a multilingual speech model pretrained on 500K hours across 1400+ languages, ideal for ASR tasks after fine-tuning.
Brief Details: Llama-3.2-1B by Meta, optimized with Unsloth's Dynamic 4-bit quantization. Offers multilingual capabilities with 70% reduced memory footprint.
Brief-details: A high-resolution (768x768) image generation model focused on realistic visuals, known for creative compositions and detailed human portraits. Final version before composition restrictions.
Brief-details: ClimateBERT model for classifying climate-related text specificity, fine-tuned on paragraph-level data for identifying specific vs non-specific climate discussions
Brief-details: OCR-free document understanding transformer model combining Swin Transformer vision encoder with BART decoder, fine-tuned on RVL-CDIP dataset for document classification tasks.
Brief Details: A compact question-answering model derived from DistilBERT, fine-tuned on SQuAD, offering efficient performance for Q&A tasks while maintaining reasonable accuracy.
Brief-details: Moirai-1.1-R-base is Salesforce's upgraded time series forecasting model with ~20% improvement for low-frequency data prediction, specifically optimized for yearly and quarterly forecasting tasks.
BRIEF-DETAILS: Meta's largest Llama-2 variant (70B parameters) fine-tuned for chat. Advanced language model with strong dialogue capabilities and Meta privacy compliance.
BRIEF-DETAILS: TinyLlama 1.1B model variant implementing compressed tensors and optimized KV cache scheme for improved memory efficiency
BRIEF DETAILS: Advanced multilingual TTS model trained on 150k hours of audio data covering English, Chinese & Japanese. Released under BY-CC-NC-SA-4.0 license.
Brief-details: Saiga Mistral 7B GGUF is a lightweight, Llama.cpp-compatible language model requiring only 10GB RAM, optimized for efficient deployment and inference
Brief-details: Nemotron-3-8B is NVIDIA's 3.8 billion parameter foundation model with 4K context window, requiring NVIDIA AI Foundation Models license for usage
