Open-Weight Models

Model

Occamy-1.0-oQ6e-mtp

Qinghua

本仓库是什么：把官方 occamy-1.0（35B-A3B MoE + 视觉编码器）与官方 occamy-1.0-MTP 草稿头合并进同一份权重（mtpnumhiddenlayers: 1），再用 oMLX 的 oQ6e 量化导出，专供 Apple Silicon 上的 oMLX 推理。非官方发布，未经 Accio-Lab 验证。 - 实测（本机真实使用日志，113 次请求，Mac Studio M4 Max 64GB）：草稿接受率 77.95%（48,501/62,218），分深度 d1 79.4% / d2 76.3% / d3 77.3%；中位 2.83 tokens/cycle；中位解码 110.5 tok/s（p10 78.0 / p90 146.9）。工具调用类结构化输出接受率 85.7%，自由文本 75.8%。 - 量化配方：oQ6e = oMLX 通用动态量化，6-bit 基准（专家层 6-bit/group 64；语言部分 ≈6.6 bpw，整文件含视觉塔 ≈6.7 bpw）；嵌入、注意力/GDN 投影、共享专家提升到 8-bit；全模型 norms、MoE 路由、MTP 融合层与整个视觉塔保持未量化；imatrix 校准（oqecodemultilingual，128×512，专家覆盖 100%）。 - 体积：30.28 GB（28.20 GiB），6 个分片；64GB 统一内存可运行（实测占用约 29 GB）。 Upstream ships the MTP draft head as a separate artifact that must be…

static quants of https://huggingface.co/OceanLabs/Ocean-1-4B

This model is a fine-tuned version of allenai/Olmo-3-7B-Instruct. It has been trained using TRL. This model was trained with SFT.

This model is a fine-tuned version of allenai/Olmo-3-7B-Instruct. It has been trained using TRL. This model was trained with SFT.

Standalone merged BF16 model from epoch 3.970178926441352, step 1500. Dataset: CompassioninMachineLearning/urban12738cleaned at ef7c0e742df63ea319e35d02d9f6ba63d6e7c68d. Training: 10,072 distinct documents plus 2,000 repeat exposures per epoch; 200 disjoint validation documents. Merged with Unsloth's native savepretrainedmerged(savemethod="merged16bit"). Weights are validated BF16 and packaged losslessly into eight safetensors shards. No adapter is required to load this model. See runmanifest.json for base revision, document selection hashes, training parameters and export validation. Training does not establish an improvement in compassion; evaluate that separately.

This model is a fine-tuned version of Qwen/Qwen3.8-27B on the on the Omni-Edu-70K dataset. The following hyperparameters were used during training: - learningrate: 5e-06 - trainbatchsize: 1 - evalbatchsize: 8 - distributedtype: multi-GPU - numdevices: 16 - gradientaccumulationsteps: 8 - totaltrainbatchsize: 128 - totalevalbatchsize: 128 - lrschedulertype: cosine - lrschedulerwarmupsteps: 0.1 - numepochs: 3.0 - Transformers 5.2.0 - Pytorch 2.10.0 - Datasets 4.0.0 - Tokenizers 0.22.2

Open Foundation Models for Learning and Teaching OmniEdu-27B-FP8 is the FP8-quantized release of OmniEdu-27B. The architecture, tokenizer, chat template and instruction-tuning data are unchanged: this checkpoint is the same model with its linear layers stored in FP8, which shrinks the weights from 50.96 GiB to 28.30 GiB and is intended for deployment on smaller GPUs and for higher-throughput serving. This checkpoint accompanies OmniEdu: Open Foundation Models for Learning and University · University of the Chinese Academy of Sciences · Zhongguancun Academy Quantized with LLM Compressor into the compressed-tensors FP8BLOCK scheme (library version 0.19.0), without calibration data: The vision…

Open Foundation Models for Learning and Teaching OmniEdu-4B-FP8 is the FP8-quantized release of OmniEdu-4B. The architecture, tokenizer, chat template and instruction-tuning data are unchanged: this checkpoint is the same model with its linear layers stored in FP8, which shrinks the weights from 8.46 GiB to 5.13 GiB and is intended for deployment on smaller GPUs and for higher-throughput serving. This checkpoint accompanies OmniEdu: Open Foundation Models for Learning and University · University of the Chinese Academy of Sciences · Zhongguancun Academy Quantized with LLM Compressor into the compressed-tensors FP8BLOCK scheme (library version 0.19.0), without calibration data: The vision…

Open Foundation Models for Learning and Teaching OmniEdu-9B-FP8 is the FP8-quantized release of OmniEdu-9B. The architecture, tokenizer, chat template and instruction-tuning data are unchanged: this checkpoint is the same model with its linear layers stored in FP8, which shrinks the weights from 17.53 GiB to 11.09 GiB and is intended for deployment on smaller GPUs and for higher-throughput serving. This checkpoint accompanies OmniEdu: Open Foundation Models for Learning and University · University of the Chinese Academy of Sciences · Zhongguancun Academy Quantized with LLM Compressor into the compressed-tensors FP8BLOCK scheme (library version 0.19.0), without calibration data: The vision…

Model · Text to speech

OmniVoice

Aero

Single-file checkpoints of k2-fsa/OmniVoice (Qwen3 LLM + HiggsAudioV2 codec), repacked for ComfyUI-OmniVoice-Native. Config, codec config, and BPE tokenizer are embedded as safetensors metadata — no sidecar files needed. - omnivoice.safetensors (3.3GB, full quality) — most robust, recommended for tags + voice-design prompts. - omnivoice-int8.safetensors (1.1GB, INT8-ConvRot) — same quality on plain prompts, smaller download. ~86% top-1 agreement with full. 1. Install the pack: ComfyUI Manager → search OmniVoice, or git clone https://github.com/Aero-Ex/ComfyUI-OmniVoice into ComfyUI/customnodes/. No pip install needed. 2. Put a file from this repo into ComfyUI/models/omnivoice/. 3. Load…

Self-contained ONNX export of k2-fsa/OmniVoice for CPU inference, including voice cloning: a dynamic-INT8 Qwen3-0.6B diffusion backbone, the Higgs audio-v2 encoder and vocoder in FP32, and the tokenizer/config files. Nothing is downloaded from any other model repo. Backbone: inputids[B,8,S] int64, audiomask[B,S] bool, attentionmask[B,1,S,S] bool → logits[B,8,S,1025]. Encoder: wav[1,1,N] float32 (24 kHz, N a multiple of 960) → audiocodes[1,8,N/960] int64. Vocoder: audiocodes[B,8,T] int64 → audiovalues[B,1,960T] (24 kHz). B, S, N and T are dynamic. --threads should be the physical core count. The omnivoice package supplies only the diffusion loop and text front-end code; every weight comes…