Brief Details: A lightweight EfficientNet variant with 4.7M params, optimized for mobile/edge devices. Trained on ImageNet-1k, balancing accuracy and efficiency.
BRIEF-DETAILS: A minimal GPT2 language model created for TRL library testing purposes. Designed for internal validation and unit testing workflows.
BRIEF DETAILS: Multi-language Scandinavian NER model supporting Danish, Norwegian, Swedish, Icelandic and Faroese. Achieves 89.08% average F1-score across languages. 676MB size.
Brief Details: A compact vision-language model derived from Phi-3, focusing on efficient visual processing with randomized parameters suitable for research and experimentation.
Brief-details: Jais-13b is a powerful bilingual LLM with 13B parameters, trained on 72B Arabic and 279B English/code tokens, featuring ALiBi positioning and SwiGLU activation.
Brief Details: A compact Llama 2-based model series trained on TinyStories, optimized for llama2.c implementation. Created by Karpathy for lightweight applications.
Brief Details: A 33B parameter language model merged by Tim Dettmers, based on the Guanaco architecture for enhanced instruction-following and dialogue capabilities.
Brief-details: Gemma-2-2b-jpn-it is Google's 2.2B parameter Japanese-focused instruction-tuned language model, requiring explicit license agreement for access via Hugging Face.
Brief Details: LiteLlama-460M-1T is a compact 460M parameter LLaMA variant trained on 1T tokens, achieving impressive performance despite its reduced size
Brief-details: Meta's Llama-3.2-90B-Vision is a large multimodal model with 90B parameters, combining advanced vision capabilities with language understanding.
Brief-details: ASR model with 110M parameters for English speech transcription. Features punctuation/capitalization support, 5300x real-time speed on A100, and 20-min audio processing capability.
Brief Details: Innovative protein sequence modeling using Bayesian Flow Networks - enables unconditional generation of novel protein sequences with structural motif preservation.
Brief Details: A fine-tuned Llama 2 model specialized for mental health support, offering therapeutic conversations with empathetic responses. Built for virtual counseling.
Brief Details: Qwen2.5-0.5B-Instruct-AWQ is a 4-bit quantized instruction-tuned LLM with 0.5B parameters, offering multilingual support and 32K context window
BRIEF-DETAILS: Qwen2.5's 72B parameter instruction-tuned model quantized to 8-bit, offering 128K context length, multilingual support, and enhanced capabilities in coding and mathematics.
BRIEF-DETAILS: Qwen2.5-14B-Instruct-GPTQ is a 4-bit quantized 14.7B parameter LLM supporting 29+ languages with 128K context length and specialized in coding/math tasks
Brief Details: A 72B parameter reward model designed to enhance Qwen2-Math training by providing detailed feedback on mathematical reasoning steps and quality.
Brief-details: EraX-VL-7B-V1.0 is a Vietnamese-focused multimodal LLM specializing in OCR and visual QA, fine-tuned from Qwen2-VL-7B for medical documents and forms
BRIEF-DETAILS: Vector illustration model specialized in flat shading with outline effects, created by mujibanget. Ideal for generating cute vector art with distinctive style.
Brief-details: Flow-Judge-v0.1 is a 3.8B parameter LLM evaluation model based on Phi-3.5-mini, offering customizable assessment capabilities across multiple scoring scales with structured feedback.
Brief-details: PaliGemma2-3B by Google - A 3 billion parameter model requiring license agreement for access through Hugging Face, representing an advanced AI development in the PaLI model family.
