全站搜索
AI 模型 2119
- LFM2.5-VL-3B-Uncensored-GGUF↗
SC117
LFM2.5-VL-3B-Uncensored-GGUF English | 📖 中文文档 Uncensored vision-language edge model · abliterix Trial 65 merged into official LFM2.5-VL-3B · imatrix-calibrated GGUFs + mmproj LFM2.5-VL-3B is Liquid AI's official 3.1B vision-language edge model : the LFM2.5-2.6B hybrid language backbone (30 layers: 22 double-gated short-convolution blocks + 8 GQA, 128K context, 128K vocab) paired with a SigLIP2 NaFlex vision encoder (256×256, patch 16) and a language-aligned projector. This release merges the abliterix Trial 65 LoRA (from our LFM2.5-2.6B-Uncensored , rank-1, alpha = r = 1) into the language-model layers of the official VL checkpoint ( q/k/v/out_proj , feed_forward.w2 , conv.out_proj — 84 tensors across 30 layers). The vision encoder and projector are untouched , so multimodal capability is fully preserved. Build steps: LoRA merge into the VL checkpoint: W += (B @ A) * (alpha / r) , applied to model.language_model.layers.* . BF16 GGUF conversion of the merged model (llama.cpp lfm2 architecture) + mmproj export of the vision tower via convert_hf_to_gguf.py --mmproj --outtype bf16 . imatrix calibration — reused the 401-chunk (≈1.6M tokens) importance matrix from the 2.6B release (same architecture, weights closely related). Quantization with llama-quantize --imatrix into five tiers.
- dfine-large-coco↗
ustc-community
The D-FINE model was proposed in D-FINE: Redefine Regression Task in DETRs as Fine-grained Distribution Refinement by Yansong Peng, Hebei Li, Peixi Wu, Yueyi Zhang, Xiaoyan Sun, Feng Wu D-FINE, a powerful real-time object detector that achieves outstanding localization precision by redefining the bounding box regression task in DETR models. D-FINE comprises two key components: Fine-grained Distribution Refinement (FDR) and Global Optimal Localization Self-Distillation (GO-LSD). D-FINE is trained on COCO (Lin et al. [2014]) train2017 and validated on COCO val2017 dataset. We report the standard AP metrics (averaged over uniformly sampled IoU thresholds ranging from 0.50 − 0.95 with a step size of 0.05), and APval5000 commonly used in real scenarios.
- Qwen3.8-Flash-Next-ngram-table-NInfer-v3↗
WaveCut
The n-gram table of Qwen3.8-Flash-Next's per-layer embedding (PLE), as a NInfer **v3** artifact for **NInfer-all**, the `master` branch of iamwavecut/ninfer-all. Every NInfer conversion of Qwen3.8-Flash-Next published without its table reads its rows from this file: Stock NInfer builds cannot read it: the model family and the `gguf_*` formats exist only in that line; build `master` from commit `82f120de` on. The table holds 320,001,536 rows of 160 values (51.2B parameters), 16 hash heads over the 2- and 3-token n-grams of the context, each head's rows at its own offset. Every token reads the 16 rows its n-grams address. The rows are ISTA-DASLab's IQ4_NL blocks (`per_layer_token_embd` of their GGUF releases, 90 bytes a row), copied byte for byte: the second shard of every GSQ-RCO release, the Coder build's included, is the same file.
- K2-Horizon-MoVA-36B-A4B-ROCmFP4-GGUF↗
kingjones777
Four ROCmFP4/ROCmFPX tiers of `IFM/K2-Horizon-MoVA-36B-A4B`, quantised from the official BF16 GGUF and measured on a Ryzen AI Max+ 395 (Radeon 8060S, gfx1151, 128 GB unified, ROCm 7.2.4). This is the **dual-expert** `k2_horizon` MoVA architecture — 36B total parameters, **~4B active** — with two independent expert systems: a 64-way MoVA value-expert bank (top-4) alongside a 100-way MoE feed-forward (top-8), 48 layers, hidden 2560, and it is a **reasoning** model. **Every number on this card was measured on these exact files.** Nothing is estimated. As an A4B MoE, it decodes far faster than a dense 32B on the same hardware. **Start with `FAST`** — fastest on both axes here, all four answer correctly. Take `AGENT` only if you specifically want 8-bit weights.
- gemma-3-27b-it-AWQ-INT4↗
pytorch
Calibrated with 30 samples of `mmlu_philosophy`, got eval accuracy of 80.06, while gemma-3-27b-it-INT4 is 77.17, and bfloat16 baseline is 79.42 Note: please use `VLLM_DISABLE_COMPILE_CACHE=1` to disable compile cache when running this code, e.g. `VLLM_DISABLE_COMPILE_CACHE=1 python example.py`, since there are some issues with the composability of compile in vLLM and torchao, this is expected be resolved in pytorch 2.8. We rely on lm-evaluation-harness to evaluate the quality of the quantized model. Here we only run on mmlu for sanity check.
- xlm-roberta-base-finetuned-nace↗
erst
This model is xlm-roberta-base fine-tuned to classify descriptions of activities into NACE Rev. 2 codes. The data used to fine-tune the model consist of 2.5 million descriptions of activities from Norwegian and Danish businesses. To improve the model's multilingual performance, random samples of the Norwegian and Danish descriptions were machine translated into the following languages: - English - German - Spanish - French - Finnish - Polish
- gpt2-124m-tinystories↗
sampatil1010
- Samantha-1.1-70B-GGUF↗
TheBloke
- MiniMax-H3-Fun-Controlnet-Union↗
alibaba-pai
> **This is a legacy version. A newer, better version is now available: MiniMax-H3-Fun-Controlnet-Union-2.0.** > The 2.0 model produces **improved results** and is the recommended checkpoint for new usage. Please prefer it over this one. MiniMax-H3-Fun-Controlnet-Union is a ControlNet-Union for MiniMax-H3, trained with the VideoX-Fun pipeline. A single checkpoint conditions the MiniMax-H3 video generator on Canny, Depth, HED, MLSD or Pose control videos, and also runs video inpainting. All samples below are generated with `num_inference_steps = 40`, `guidance_scale = 1.0`, `control_context_scale = 1.00`, seed 43.
- Pixtral-12B-2409↗
Mistral AI
The performance of Pixtral-12B-2409 compared to multimodal models. All models were re-evaluated and benchmarked through the same evaluation pipeline. We recommend using Pixtral with the vLLM library to implement production-ready inference pipelines with Pixtral. After installing `mistral_inference`, a `mistral-chat` CLI command should be available in your environment. You can pass text and images or image urls to the model in *instruction-following* mode as follows: