Brief-details: A curated collection of popular CivitAI models optimized for Google Colab usage, featuring 15 high-quality models including Hassan blends, Dreamlike, and specialized artistic mixes.
Brief-details: Megatron GPT-2 345M: NVIDIA's transformer model with 380M params, trained on diverse text data. Strong perplexity scores on WikiText benchmarks.
Brief Details: A bilingual Chinese-English text-to-image model combining CLIP and diffusion technology, optimized for high-quality artistic image generation with modern style capabilities.
Brief Details: A Persian speech recognition model fine-tuned from Whisper-large-v2, achieving 26.37% WER on Common Voice 11.0, optimized for Persian language ASR tasks.
BRIEF DETAILS: RoBERTa-large fine-tuned model for corporate culture analysis, classifying text into four culture dimensions based on the Competing Values Framework.
Brief Details: A fine-tuned Whisper Large-v2 model specialized for Telugu ASR, achieving 9.65 WER on FLEURS test set. Trained on multiple Telugu speech corpora.
BRIEF DETAILS: A fine-tuned Whisper Large-v2 model optimized for Hausa language ASR, achieving 37.41% WER on Common Voice 11.0, trained with linear learning rate scheduling across 1000 steps.
Brief-details: A creative experimental AI model combining multiple Stable Diffusion checkpoints, specializing in traditional media rendering and image fixing capabilities, with three distinct mixing variations.
Brief-details: ESPnet-based ASR model trained on LibriSpeech-100 using E-Branchformer architecture, achieving 94.4% accuracy on test-clean with parallel MLP-attention design
Brief Details: A fine-tuned Whisper Large-v2 model for Bulgarian ASR, achieving 13.4% WER on Common Voice, trained with linear learning rate scheduling over 1000 steps.
BRIEF DETAILS: A Chinese speech recognition model based on Whisper-tiny architecture, fine-tuned on Common Voice 11.0, achieving 91.09% WER with Apache 2.0 license.
Brief Details: ruElectra-medium is a Russian language transformer model developed by ai-forever, optimized for embeddings with MIT license and research-backed architecture.
Brief Details: Portuguese-optimized Whisper ASR model achieving 6.59% WER on Common Voice 11.0, outperforming original Whisper Medium for Portuguese transcription.
Brief Details: German ASR model fine-tuned from Whisper-large-v2, achieving 5.76 WER on Common Voice 11.0, optimized for 16kHz audio with 1.55B parameters.
Brief-details: Small-scale ELECTRA model for Russian language embeddings, MIT-licensed, optimized for mean token embeddings with PyTorch/TensorFlow support
Brief Details: A specialized Stable Diffusion model for generating D&D-style magic item images with custom prompting system and specific item types/styles support
Brief-details: A specialized biomedical NER model based on PubMedBERT, fine-tuned on multiple disease-related datasets for accurate disease entity recognition in medical texts.
BRIEF-DETAILS: TVLT is a textless vision-language transformer that extends MAE for audio-visual pre-training, designed for multimodal learning tasks.
Brief Details: A specialized Arabic text summarization model built on mBART architecture, capable of generating concise summaries and news headlines from Arabic text
Brief-details: A Swedish speech recognition model based on Whisper-small, fine-tuned on Common Voice 11.0 dataset with PyTorch, offering automated speech recognition capabilities.
Brief Details: ERNIE-Layout-Pytorch is a PyTorch implementation of the ERNIE-Layout model for document understanding, supporting visual question answering with integrated layout analysis.
