Saved toolsSubmit a tool
FOXAI SEARCH

Search FoxAI

AI Models 2119

  • ijepa_vith14_1k↗

    facebook

    **I-JEPA** is a method for self-supervised learning. At a high level, I-JEPA predicts the representations of part of an image from the representations of other parts of the same image: 1. without relying on pre-specified invariances to hand-crafted data transformations, which tend to be biased for particular downstream tasks, 2. and without having the model fill in pixel-level details, which tend to result in learning less semantically meaningful representations. As opposed to generative methods that have a pixel decoder, I-JEPA has a predictor that makes predictions in latent space. The predictor in I-JEPA can be seen as a primitive (and restricted) world-model that is able to model spatial uncertainty in a static image from a partially observable context. This world model is semantic in the sense that it predicts high level information about unseen regions in the image, rather than pixel-level details. We trained a stochastic decoder that maps the I-JEPA predicted representations back in pixel space as sketches. The model correctly captures positional uncertainty and produces high-level object parts with the correct pose (e.g., dog’s head, wolf’s front legs).

  • umt5-xxl↗

    Google

    UMT5 is pretrained on the an updated version of mC4 corpus, covering 107 languages: Afrikaans, Albanian, Amharic, Arabic, Armenian, Azerbaijani, Basque, Belarusian, Bengali, Bulgarian, Burmese, Catalan, Cebuano, Chichewa, Chinese, Corsican, Czech, Danish, Dutch, English, Esperanto, Estonian, Filipino, Finnish, French, Galician, Georgian, German, Greek, Gujarati, Haitian Creole, Hausa, Hawaiian, Hebrew, Hindi, Hmong, Hungarian, Icelandic, Igbo, Indonesian, Irish, Italian, Japanese, Javanese, Kannada, Kazakh, Khmer, Korean, Kurdish, Kyrgyz, Lao, Latin, Latvian, Lithuanian, Luxembourgish, Macedonian, Malagasy, Malay, Malayalam, Maltese, Maori, Marathi, Mongolian, Nepali, Norwegian, Pashto, Persian, Polish, Portuguese, Punjabi, Romanian, Russian, Samoan, Scottish Gaelic, Serbian, Shona, Sindhi, Sinhala, Slovak, Slovenian, Somali, Sotho, Spanish, Sundanese, Swahili, Swedish, Tajik, Tamil, Telugu, Thai, Turkish, Ukrainian, Urdu, Uzbek, Vietnamese, Welsh, West Frisian, Xhosa, Yiddish, Yoruba, Zulu. **Note**: UMT5 was only pre-trained on mC4 excluding any supervised training. Therefore, this model has to be fine-tuned before it is useable on a downstream task.

  • CyberRealisticPony↗

    cyberdelia

    **CyberRealistic Pony** is the awesome Pony Diffusion with some CyberRealistic elements. - **Photorealism**: Generates highly detailed and realistic pony images, capturing intricate textures and lighting. - **Ease of Use**: Achieves impressive results with straightforward prompts. - **Integrated VAE**: Comes with a baked-in Variational Autoencoder for enhanced image quality. - **Versatility**: Suitable for various applications, including character design, illustrations, and concept art. - May produce content that could be considered sensitive; use responsibly. - Some prompts involving abstract or non-pony content may not perform as well due to the model's specialized training. - Lighting and textures may occasionally be too clean or smooth depending on sampling choices.

  • Qwen3.5-4B-Heretic-GGUF↗

    darrellbest

    GGUF builds of darrellbest/Qwen3.5-4B-Heretic. Qwen/Qwen3.5-4B with its refusal behaviour removed by Heretic using full-weight Arbitrary-Rank Ablation: **6/100 refusals** (original: 99/100) at **KL divergence 0.0220**. See the main repository for how it was made and measured. Converted from the bf16 safetensors with llama.cpp's `convert_hf_to_gguf.py` and quantized with `llama-quantize` (no imatrix). The mmproj file carries the vision encoder; load it alongside any of the three for image input. Every file was loaded in `llama-server` with the mmproj and in Ollama. All three answered ordinary prompts correctly (BF16 and Q8_0 reproduce each other word for word), described a test image (a red circle and a blue square) correctly, and in Ollama answered 17 x 23 = 391 with thinking off. Thinking-mode reasoning, 4 arithmetic and word problems x 10 seeds with Qwen's recommended sampling: BF16 39/40, Q8_0 40/40, Q4_K_M 40/40 finished and correct (the original model in vLLM: 40/40).

  • parakeet-cpp-gguf↗

    mudler

    GGUF-format weights for parakeet.cpp, a C++/ggml port of NVIDIA NeMo Parakeet that matches the upstream PyTorch models on CPU. This single repo collects **every supported model × quantization** as a flat set of `.gguf` files — download just the one you need. **F16 is the recommended default** — same accuracy as F32, ~1.7× smaller, and typically the fastest on modern CPUs via ggml's F32×F16 matmul fast path. Source: nvidia/parakeet-tdt_ctc-110m · Hybrid TDT+CTC (FastConformer) · heads: TDT + CTC

  • Qwen3.5-0.8B-W4A16-AutoRound-AWQ↗

    Zerul

    This is a **W4A16 (4-bit weight, 16-bit activation) AWQ-format** quantized version of Qwen/Qwen3.5-0.8B, produced using AutoRound — Intel's sign gradient descent based quantization method designed for production-grade accuracy retention. - **AWQ format** — Exported in the standard AWQ format, optimized for efficient inference with activation-aware weight clipping. - **Ultra-high accuracy configuration** — 1000 iterations with 512 calibration samples ensures near-lossless quantization, especially critical at this model scale where parameter budget is tight. - **W4A16** — Weights are quantized to 4-bit integers; activations remain in FP16 for inference stability. - **Extremely lightweight** — The quantized 0.8B model is suitable for edge deployment, low-latency inference, and resource-constrained environments. - **~50% memory reduction** compared to the FP16 base model. This model is compatible with `transformers`, `AutoAWQ`, `vLLM`, and `SGLang` — any backend supporting AWQ-format weights works out of the box. For full model details, architecture, and capabilities, refer to the base model page.

  • SlimSAM-uniform-50↗

    Zigeng

  • vit-large-patch16-224↗

    Google

    Vision Transformer (ViT) model pre-trained on ImageNet-21k (14 million images, 21,843 classes) at resolution 224x224, and fine-tuned on ImageNet 2012 (1 million images, 1,000 classes) at resolution 224x224. It was introduced in the paper An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale by Dosovitskiy et al. and first released in this repository. However, the weights were converted from the timm repository by Ross Wightman, who already converted the weights from JAX to PyTorch. Credits go to him. Disclaimer: The team releasing ViT did not write a model card for this model so this model card has been written by the Hugging Face team. The Vision Transformer (ViT) is a transformer encoder model (BERT-like) pretrained on a large collection of images in a supervised fashion, namely ImageNet-21k, at a resolution of 224x224 pixels. Next, the model was fine-tuned on ImageNet (also referred to as ILSVRC2012), a dataset comprising 1 million images and 1,000 classes, at the same resolution, 224x224.

  • SambaLingo-Slovenian-Base-i1-GGUF↗

    mradermacher

  • coreml-depth-anything-v2-small↗

    apple

    Depth Anything V2 was introduced in the paper of the same name by Lihe Yang et al. It uses the same architecture as the original Depth Anything release, but uses synthetic data and a larger capacity teacher model to achieve much finer and robust depth predictions. The original Depth Anything model was introduced in the paper Depth Anything: Unleashing the Power of Large-Scale Unlabeled Data by Lihe Yang et al., and was first released in this repository. Depth Anything V2 leverages the DPT architecture with a DINOv2 backbone. The model is trained on ~600K synthetic labeled images and ~62 million real unlabeled images, obtaining state-of-the-art results for both relative and absolute depth estimation.