저장한 도구도구 추천
FOXAI SEARCH

전체 검색

AI 모델 2119

  • Infinity-Instruct-3M-0625-Llama3-8B↗

    BAAI

    Infinity-Instruct-3M-0625-Llama3-8B is an opensource supervised instruction tuning model without reinforcement learning from human feedback (RLHF). This model is just finetuned on Infinity-Instruct-3M and Infinity-Instruct-0625 and showing favorable results on AlpacaEval 2.0 and MT-Bench. Infinity-Instruct-3M-0625-Llama3-8B is tuned on Million-level instruction dataset Infinity-Instruct. First, we apply the foundational dataset Infinity-Instruct-3M to improve the foundational ability (math & code) of Llama3-8B, and get the foundational instruct model Infinity-Instruct-3M-Llama3-8B. Then we finetune the Infinity-Instruct-3M-Llama3-8B to get the stronger chat model Infinity-Instruct-3M-0625-Llama3-8B. Here is the training hyperparamers. Thanks to FlagScale, we could concatenate multiple training samples to remove padding token and apply diverse acceleration techniques to the traning procudure. It effectively reduces our training costs. We will release our code in the near future!

  • Qwen3.8-27B-nvfp4qat-NInfer↗

    cometkim

    The repository contains a QAT-sourced NVFP4 weight profile of Qwen/Qwen3.8-27B in the native NInfer `.ninfer` artifact format, with the z-lab/Qwen3.8-27B-DFlash2 block-diffusion speculative drafter embedded in the same image. It runs on NInfer **v3** engines — Text and MTP components through format/shape routes the upstream runtime already implements, the optional NVFP4-encoded DFlash2 companion through the fork's runtime support. It is not a Transformers checkpoint, Safetensors distribution, or GGUF file. This is a fourth weight profile for the existing `qwen3_8_27b` target — a peer of the official `groupwise-int` and `nvfp4` profiles and of the fork's fuller-requant `nvfp4full` profile. The Text weight stack is copied **word-for-word** from QUASAR-QAT/Qwen3.8-27B-QUASAR-NVFP4 — a quantization-aware-trained checkpoint (QUASAR loss-aware NVFP4 distillation against the frozen BF16 teacher, arXiv 2608.13966): - every one of the 496 text linear layers is NVFP4 (W4A4) — attention, gated-delta-net, and MLP alike, with no high-precision exceptions; - the QAT factory quantizes per site, so every constituent tensor of a fused parent shares one weight and input global scale; the converter enforces that sharing before copying a word, and the site input divisors ride the artifact directly; - the GDN control projections are decoded from their QAT NVFP4 words to the BF16 control parent the engine consumes.

  • Dr_Samantha-7b-i1-GGUF↗

    mradermacher

  • Dr.RP.Samantha-L3-8B-i1-GGUF↗

    mradermacher

  • Samrudh-3-Brahma-7B-GGUF↗

    mradermacher

  • Qwen3.5-9B-Cold-Fusion-GAIN-v1.0-Uncensored-Heretic-NEO-MAX-Imatrix-GGUF↗

    DavidAU

    IMPORTANT: The GAIN method of training maintains 99% of performance of BF16, at both 8 bit and 4 bit levels. COLD FUSION exceeds performance of both 9B and 27B Qwen 3.5 models. Cold Fusion exceeds Qwen 3.5 27B performance, at only 9B parameters (and at both 4 bit and 8 bit too) plus enhanced by "GAIN" method training. The "GAIN" method (programming) automatically (and dynamically) changes training on a per sample basis in real time during training AS THE MODEL LEARNS. This was designed to be used with all model types and datasets and plugs into the Unsloth training "stream".

  • krea2-turbo-sda↗

    F16

    A rank-32 LoRA for Krea 2 Turbo that **restores the sampling diversity the Turbo distillation removed**, without degrading image quality or prompt adherence. Trained with **SDA (Semantic Directional Alignment)** — a teacher-guided diversity alignment loss — wrapped in **Forward XM best-of-5** candidate exploration, on a single high-noise sigma node (σ = 0.9567). *"a cat sitting on a windowsill"* — the baseline regenerates the same tabby-and-white cat 16 times; the LoRA varies fur patterns, lighting, seasons and window settings: *"a still life with fruit on a table"* — one pineapple-centric template vs varied compositions, props and backgrounds:

  • mask2former-swin-tiny-ade-semantic↗

    facebook

    Mask2Former model trained on ADE20k semantic segmentation (tiny-sized version, Swin backbone). It was introduced in the paper Masked-attention Mask Transformer for Universal Image Segmentation and first released in this repository. Disclaimer: The team releasing Mask2Former did not write a model card for this model so this model card has been written by the Hugging Face team. Mask2Former addresses instance, semantic and panoptic segmentation with the same paradigm: by predicting a set of masks and corresponding labels. Hence, all 3 tasks are treated as if they were instance segmentation. Mask2Former outperforms the previous SOTA, MaskFormer both in terms of performance an efficiency by (i) replacing the pixel decoder with a more advanced multi-scale deformable attention Transformer, (ii) adopting a Transformer decoder with masked attention to boost performance without without introducing additional computation and (iii) improving training efficiency by calculating the loss on subsampled points instead of whole masks.

  • Ornith-1.5-35B-A3B-Abliterated-CyberTiel-Calibrated-MTPv2-ICE-GGUF↗

    gbuzhf

    Four ICE tiers of **huihui-ai/Huihui-Ornith-1.5-35B-A3B-abliterated**, quantized against a **trunk-matched** importance matrix, with the model's own trained **MTPv2** head and the **Qwen-Sharp v22.4.1** chat template embedded. The BF16 master they were all built from is published here too. > ⚠️ **Uncensored model.** This is an abliterated checkpoint: refusal behaviour has been > suppressed in the weights. It will comply with requests a stock model declines. You are > responsible for how you use it. Sandbox it at the OS level and control its network and > code-execution access — with no refusal backstop, a prompt injection from a hostile page > or third-party code has nothing to stop it. If you want the guardrailed equivalent, use > the non-abliterated TIEL-Calibrated ICE tiers. Rule of thumb: take the largest tier that leaves you the context you actually need at `q8_0` KV. If you already run a `UD` build of this model, pick the ICE tier at the same size and keep everything else the same.

  • MiniMax-H3-PTD-LoRAs↗

    taobao-content-ai

    We apply **PTD (Parametric Trajectory Distillation)** to MiniMax-H3, distilling the Ref2VA (reference-image → video + audio) trajectory into a **few-step** student. We release an **8-step** and a **4-step** accelerator for the Ref2VA task. > [!TIP] > **Want better quality or lower latency?** Check out `MiniMax-H3-PTD-RefAsymOPD-LoRAs` — the upgraded sibling trained with **PTD + RefAsymOPD** (Reference-Asymmetric On-Policy Distillation). > It ships an **8-step** and a new **6-step** variant, both delivering noticeably better **background plausibility, sharpness and realism** — surpassing the plain PTD checkpoints in this repo as well as the base model at 40 NFE. > Same checkpoint format and the same `apply_ptd_lora` call — a one-line filename swap. Unlike a single shared acceleration LoRA, PTD trains **one independent low-rank adapter per inference step** ("per-hop" LoRA) on the diffusion backbone, on top of the per-interval output heads introduced by Parallel Decoding Distillation (PDD). At inference, each denoising step selects its own hop's adapter and fuses that step's output head, so an 8-hop checkpoint runs in **8** transformer evaluations (NFE) and a 4-hop checkpoint in **4**.