サイト内検索

AI モデル 2119

• Huihui-Qwen3.8-Flash-Next-abliterated-NVFP4-GPTQ-Strata↗

  Maximilian228

  > [!WARNING] > **Withdrawn from Strata NVFP4's installer (2026-10-09): this quant often loops in long thinking.** > We captured a real Claude Code request and replayed it. It had a 72K-token prompt, thinking on, and no temperature > sent, so the server decoded greedily, which is its default. > - **This quant looped in 9 of 10 runs**, on Strata NVFP4 v0.1.41-nvfp4.2 and on v0.1.40.3-nvfp4.1 alike. It > repeated the same 372-word passage of its reasoning until `max_tokens`, instead of closing the thinking and > answering. > - With Qwen's thinking-mode sampling (temperature 0.6, top_p 0.95, top_k 20) it still looped in 2 of 5 runs. > - Our OrcaRouter quant, > made with the same recipe (every expert NVFP4, GPTQ), **looped in 0 of 5 runs** on the same request. The installer > now offers that one instead (`--family orca-nvfp4`). > > We have not established whether the cause is huihui's abliteration itself or its quantization. The abliteration > edited the experts' `down_proj`, which is NVFP4 here. > > If you run this model anyway: > - send a temperature with your requests; > - turn on the server's loop guard, `"reasoning_loop_recovery": "recover"` in the server config. It brought the > reply to an answer in 2 of 2 looping runs. **Installer:** Strata NVFP4 v0.1.41-nvfp4.1 and .2 listed this repository as `--family huihui-nvfp4`. From v0.1.41-nvfp4.3 the installer no longer offers it (see the warning above) and installs `orca-nvfp4` by default. To run this model by hand, see Run it. - **Chain:** `Qwen/Qwen3.8-Flash-Next` → `huihui-ai/Huihui-Qwen3.8-Flash-Next-abliterated` (abliteration, BF16) → this quantization. - **4.5 bits a weight for the experts, 63.3 GiB:** the size and the decode cost of a ModelOpt NVFP4 checkpoint, about **34% of plain rounding's error** on held-out tokens (below). No layer is kept in 8 bits. - **Only for Strata NVFP4.** The files are Strata's own runtime formats (an expert pack, GGUFs the fork reads, the MTP draft head). They do not load in transformers, vLLM or llama.cpp.

• Qwen2.5-VL-7B-Instruct↗

  unsloth

  In the past five months since Qwen2-VL’s release, numerous developers have built new models on the Qwen2-VL vision-language models, providing us with valuable feedback. During this period, we focused on building more useful vision-language models. Today, we are excited to introduce the latest addition to the Qwen family: Qwen2.5-VL. We extend dynamic resolution to the temporal dimension by adopting dynamic FPS sampling, enabling the model to comprehend videos at various sampling rates. Accordingly, we update mRoPE in the time dimension with IDs and absolute time alignment, enabling the model to learn temporal sequence and speed, and ultimately acquire the ability to pinpoint specific moments. We enhance both training and inference speeds by strategically implementing window attention into the ViT. The ViT architecture is further optimized with SwiGLU and RMSNorm, aligning it with the structure of the Qwen2.5 LLM.

• sam-1-base-GGUF↗

  mradermacher

• SAM_3D_OBJECTS_GGUF↗

  Asher-1

  All GGUF weights belong in gguf/, never in benchmarks. Native raw reconstruction consumes the MoGe model plus five generative stages and exports a Gaussian PLY, a textured GLB and its base-color PNG. Vulkan inference hands its raw mesh and Gaussian outputs to a separate CUDA PBR process. The matching model repository is Asher-1/SAM_3D_OBJECTS_GGUF. Weights inherit the upstream SAM 3D Objects license. Use the helper from the checkout root; it refuses to overwrite an existing local model by default: The published MoGe ViT-L file is the F16 GGUF (`moge_vitl-f16.gguf`), also in the public model repository and part of the downloader's default f16 set. The helper still reconverts from the official `Ruicheng/moge-vitl` checkpoint (`huggingface_hub`) when another dtype or a fresh validation against upstream is needed. It preserves existing GGUF files unless explicitly given `--force`.

• LongCat-Video-Avatar-ComfyUI-GGUF↗

  vantagewithai

  We are excited to announce the release of LongCat-Video-Avatar, a unified model that delivers expressive and highly dynamic audio-driven character animation, supporting native tasks including Audio-Text-to-Video, Audio-Text-Image-to-Video, and Video Continuation with seamless compatibility for both single-stream and multi-stream audio inputs. --> The following videos showcase example generations from our model and have been compressed for easier viewing. Human evaluation on naturalness and realism of the synthesized videos. The benchmark EvalTalker [1] contains more than 400 testing samples with different difficulty levels for evaluating the single and multiple human video generation.

• blossom-v5.1-34b↗

  Azure99

  Blossom is a conversational large language model, fine-tuned on the Blossom Orca/Wizard/Chat/Math mixed dataset based on the Yi-1.5-34B pre-trained model. Blossom possesses robust general capabilities and context comprehension. Additionally, the high-quality Chinese and English datasets used for training have been made open source. Training was conducted in two stages. The first stage used 40K Wizard, 40K Orca, 10K Math single-turn instruction datasets, training for 1 epoch; the second stage used 10K Blossom chat multi-turn dialogue dataset, and 10% randomly sampled data from the first stage, training for 3 epochs.

• sam-vit-huge↗

  facebook

  0. TL;DR 1. Model Details 2. Usage 3. Citation The **Segment Anything Model (SAM)** produces high quality object masks from input prompts such as points or boxes, and it can be used to generate masks for all objects in an image. It has been trained on a dataset of 11 million images and 1.1 billion masks, and has strong zero-shot performance on a variety of segmentation tasks. The abstract of the paper states: > We introduce the Segment Anything (SA) project: a new task, model, and dataset for image segmentation. Using our efficient model in a data collection loop, we built the largest segmentation dataset to date (by far), with over 1 billion masks on 11M licensed and privacy respecting images. The model is designed and trained to be promptable, so it can transfer zero-shot to new image distributions and tasks. We evaluate its capabilities on numerous tasks and find that its zero-shot performance is impressive -- often competitive with or even superior to prior fully supervised results. We are releasing the Segment Anything Model (SAM) and corresponding dataset (SA-1B) of 1B masks and 11M images at https://segment-anything.com to foster research into foundation models for computer vision.

• CodeRankEmbed-flash-attn↗

  handwoven8588

  A **bf16 quantization of `nomic-ai/CodeRankEmbed`** with a custom `modeling_hf_nomic_bert.py`, shipped in this repo, that **picks one of three attention implementations for the hardware it runs on.** It is not a finetune — the weights are the original CodeRankEmbed weights cast to bf16 (no further training). Two of the three replace the original's padded `O(seq²)` attention with an `O(N)` unpadded path; the third keeps the original's algorithm as the correctness reference and universal fallback. `nomic-ai/CodeRankEmbed` loads through `trust_remote_code`, and its attention path is **eager only** — activation memory grows as `batch × heads × seq²`, which runs out of memory at large batches even though the model is only 137M params. This repo adds two attention paths that compute the same attention in `O(N)` memory by packing unpadded sequences, so the large batches that the eager path cannot fit run comfortably — with **parity embeddings** (no quality change): - `torch_varlen` — `torch.nn.attention.varlen.varlen_attn`, shipped in torch itself (no extra package), available from torch **2.10.0** onward. - `flash_attn` — the `flash_attn` package's varlen kernel. Kept as a fallback for older torch builds that don't yet have `torch.nn.attention.varlen` but do have `flash_attn` installed. This package is **optional**.

• StartLux-Decision-v1-0.8B-Q8_0-GGUF↗

  startlux-models

  StartLux-Decision-v1-0.8B in Q8_0 (8-bit) as a GGUF file for llama.cpp. The decision server in this repository turns it into the same typed decisions, with a probability for every option, as the original model. "Same decision" is the share of the 231 public JevBench items on which the file picks the same answer as the original weights run through the `startlux_decision` package, with the same prompts, readout and temperatures. Q8_0 and Q4_K_M are llama.cpp's standard quantizations of BF16, without an importance matrix. At this size Q4_K_M changes noticeably more answers (93.1% the same), so Q8_0 is the better choice. `-ngl 99` puts every layer on the GPU (CUDA or Metal); leave it out on a CPU-only machine. llama.cpp has to be recent enough for this model (build b10454 or newer).

• sam2-hiera-tiny↗

  facebook

  Repository for SAM 2: Segment Anything in Images and Videos, a foundation model towards solving promptable visual segmentation in images and videos from FAIR. See the SAM 2 paper for more information. SAM2 can be used for automatic mask generation to segment all objects in an image using the `mask-generation` pipeline: SAM2's key strength is its ability to track objects across video frames. Here's how to use it for video segmentation: