サイト内検索
AI モデル 2119
- wavlm-base-plus-sd↗
Microsoft
The model was pretrained on 16kHz sampled speech audio with utterance and speaker contrastive loss. When using the model, make sure that your speech input is also sampled at 16kHz. - 60,000 hours of Libri-Light - 10,000 hours of GigaSpeech - 24,000 hours of VoxPopuli Authors: Sanyuan Chen, Chengyi Wang, Zhengyang Chen, Yu Wu, Shujie Liu, Zhuo Chen, Jinyu Li, Naoyuki Kanda, Takuya Yoshioka, Xiong Xiao, Jian Wu, Long Zhou, Shuo Ren, Yanmin Qian, Yao Qian, Jian Wu, Michael Zeng, Furu Wei
- Qwen-Image-2.1-Text-Encoder-Heretic-GGUF↗
pottokao
> [!IMPORTANT] > **Got `got input of size [1, 512, 12288]` with the GGUF? It's fixed.** > The GGUF files were fine all along — ComfyUI-GGUF just didn't load the Qwen3-VL vision tower. > Install the add-on node **ComfyUI-GGUF-Qwen3VL-TE**, > make sure the `mmproj` file sits next to the GGUF, restart ComfyUI. No need to re-download anything. > The same patch also fixes `Unknown model architecture!` for Qwen-Image-2.1 **DiT** GGUFs — so the whole > pipeline can run on GGUF. The sample / showcase images in this repo were generated with the technique behind **ComfyUI-QwenImage-PhotoStyles** — a small ComfyUI node (17 photographic styles; one short prompt becomes a full styled prompt via the PE-T2I rewriter). If you're interested in how the images are made, the node and its recipe are there. All three are in **this repo**. Every loader uses **type `qwen_image`** and feeds `TextEncodeQwenImage21`.
- Qwen3.8-27B-Uncensored-GGUF↗
shsunshining
Uncensored Qwen3.8-27B, published as GGUF quantizations with the multi token prediction (MTP) head retained and verified. Refusal behaviour has been **substantially reduced**, not eliminated. See Measured behaviour for the numbers. Capabilities, training data, and architecture are otherwise unchanged. The draft head ships at Q8_0 and Q4_0, the same weights at two precisions. Q8_0 is the default and is what the speculative decoding tables below were measured with. Q4_0 saves 1.5 GB for tight VRAM budgets, and its acceptance rate has not been measured here.
- wav2vec2-large-xlsr-53-telugu↗
anuragshas
Fine-tuned facebook/wav2vec2-large-xlsr-53 on Telugu using the OpenSLR SLR66 dataset. When using this model, make sure that your speech input is sampled at 16kHz. **Test Result**: 44.98% 70% of the OpenSLR Telugu dataset was used for training. Train Split of annotations is here
- NVIDIA-Nemotron-3.5-Bolt-74B-A9B-GGUF↗
AtomicChat
Atomic Dynamic (AD) quants, built from NVIDIA's original BF16 weights with our own importance matrix. At 21.9 GB, AD-IQ1_M_S stays closer to the original than NVIDIA's own 22.0 GB 2-bit file, measured side by side on one stand. The calibration corpora behind our builds are public. Every file here was measured against the BF16 original on the same stand as the other publishers' files. You can now run Nemotron 3.5 Bolt in Atomic Chat : 74B total parameters, 9B active per token. See our quantization analysis below for measurements and instructions. Lower is closer to the original. Each purple point is a file in this repo; the grey ones are other GGUF files of this model, downloaded and measured with exactly the same reference, text and llama.cpp build. NVIDIA's 2-bit file (the one Unsloth and LM Studio ship too) is a quantization-aware-trained checkpoint, so its distance to the BF16 weights also includes what that training changed; see below.
- Swift-1.5-Qwen3.8-27B-NInfer↗
CaptainArni
A single-file NInfer **v3 artifact** of ukisai/Swift-1.5-Qwen3.8-27b-NVFP4, UkisAI's second reasoning-efficient fine-tune of Qwen3.8-27B, with more post-training aimed at agentic and coding work, in NVIDIA ModelOpt's NVFP4/FP8 quantization. Converted with upstream's own converter and the recipe included here; weight bytes are imported, not requantized. The successor to CaptainArni/Swift-Qwen3.8-27B-NInfer (Swift 1.0). The two files are structural twins — same size to the byte, same layout, same VRAM — so one replaces the other with no other change. **What is inside:** all 64 MLP layers and nothing else at NVFP4 (block size 16); the 144 GDN and 64 full-attention projections at FP8 E4M3 with one BF16 scale per row; an FP8 output head and FP8 embeddings; Q4/Q5/Q6 Vision; Q8 MTP; z-lab's DFlash2 draft at W8G32. Norms, convolutions, the small GDN projections and the draft codebooks stay BF16.
- Qwen3.8-27B-GGUF↗
DhruvalLabs
> [!IMPORTANT] > **This VLM requires TWO files** — a text backbone GGUF and the `mmproj` vision encoder GGUF. > Download one text backbone (e.g. Q4_K_M) **and** the mmproj file. Both must be in the same folder. 1. Search for `DhruvalLabs/Qwen3.8-27B` in LM Studio 2. Download the Q4_K_M text file **and** the mmproj file 3. Load the model — LM Studio automatically uses both files > **Why is mmproj always F16?** > The vision encoder maps image pixels to token embeddings. Quantizing it causes > visible visual artifacts and degraded image understanding. It stays at F16 (half precision) > which is already very efficient at ~1-2GB for most models.
- mms-tts-mon↗
facebook
This model is part of Facebook's Massively Multilingual Speech project, aiming to provide speech technology across a diverse range of languages. You can find more details about the supported languages and their ISO 639-3 codes in the MMS Language Coverage Overview, and see all MMS-TTS checkpoints on the Hugging Face Hub: facebook/mms-tts. VITS (**V**ariational **I**nference with adversarial learning for end-to-end **T**ext-to-**S**peech) is an end-to-end speech synthesis model that predicts a speech waveform conditional on an input text sequence. It is a conditional variational autoencoder (VAE) comprised of a posterior encoder, decoder, and conditional prior. A set of spectrogram-based acoustic features are predicted by the flow-based module, which is formed of a Transformer-based text encoder and multiple coupling layers. The spectrogram is decoded using a stack of transposed convolutional layers, much in the same style as the HiFi-GAN vocoder. Motivated by the one-to-many nature of the TTS problem, where the same text input can be spoken in multiple ways, the model also includes a stochastic duration predictor, which allows the model to synthesise speech with different rhythms from the same input text.
- FLUX.1-dev-ControlNet-Union-Pro-2.0↗
Shakker-Labs
This repository contains an unified ControlNet for FLUX.1-dev model released by Shakker Labs. We provide an online demo. A FP8 quantized version provided by community can be found in ABDALLALSWAITI/FLUX.1-dev-ControlNet-Union-Pro-2.0-fp8. In comparison with Shakker-Labs/FLUX.1-dev-ControlNet-Union-Pro, - Remove mode embedding, has smaller model size. - Improve on canny and pose, better control and aesthetics. - Add support for soft edge. Remove support for tile. - This ControlNet consists of 6 double blocks and 0 single block. Mode embedding is removed. - We train the model from scratch for 300k steps using a dataset of 20M high-quality general and human images. We train at 512x512 resolution in BFloat16, batch size = 128, learning rate = 2e-5, the guidance is uniformly sampled from [1, 7]. We set the text drop ratio to 0.20. - This model supports multiple control modes, including canny, soft edge, depth, pose, gray. You can use it just as a normal ControlNet. - This model can be jointly used with other ControlNets.
- hpc-assistant-qwen3-4b↗
VaibhavKKM
**ClusterPilot** answers supercomputer questions (Slurm job scripts, job arrays and dependencies, GPUs, MPI, modules, storage, EuroHPC systems) and, in agent mode, works on your real cluster: it checks, submits, debugs and resubmits your jobs. It is a 4B model distilled from 80B and 235B teachers, ships as a **3.3 GB GGUF for Ollama and llama.cpp**, and runs **fully offline**. The model is still version 1.3. The ClusterPilot app (the `clusterpilot` command) got 30+ fixes from testing real use on Leonardo (CINECA) and on a B300 cluster, on macOS, Linux and Windows. Highlights: `main` now holds **ClusterPilot-4B 1.3**: version 1.2 further distilled from Qwen3-235B-A22B-Instruct-2507 (FP8) with full-parameter training on 8 NVIDIA B300 GPUs (teacher answers with top-20 log-probabilities for prompts with the app's documentation excerpts). Same 4B size, same chat format and GGUF files. Version 1.2 stays on the `v1.2` branch, version 1.0 on `v1.0`.