Open-Weight Models Under Apache License 2.0

Model

kev-queue-v4-4b

Fzkun

基于 KEV 框架微调的中文排队位置决策模型。输入现有队列和明确的插入规则，输出候选位置的概率分布及 positionN 选项。模型不生成自由文本，也不用于医疗诊断或判断患者病情紧急程度。 本仓库包含 LoRA 适配器和 KEV 决策指针头，不包含完整的基础模型权重。运行需要指定版本的基础模型、KEV 框架、分词器和推理代码；直接用通用文本生成管线加载适配器，无法复现本模型的决策输出。 公开网页是 确定性规则演示，不调用微调模型。本地版本通过服务端代理调用内网模型 API，并对照规则答案。网页后来新增的普通报到后移策略不属于本模型原有训练范围，不能用网页中的规则计算结果代表模型准确率。 原有患者的相对顺序保持不变。候选名称 position16 已表示 第 16 位，不要再加 1；模型置信度不是位置编号，也不是正确率保证。 call-system 的普通报到流程还会按排队号计算候选位置，跳过候选落点后连续的已调整患者，在第一个未调整患者之前停止。该流程与回诊、献血、家优的搜索算法不同，不能统一追加到所有策略中。 公开网页已增加这个规则的模拟场景，但 本模型尚未针对普通报到、调整状态、跨班次或加号规则重新训练。这些操作应由业务代码处理。网页保留源码 index < len(list)-1 的后移扫描边界，仅演示同班次普通号。 此次实际训练采用 多任务监督微调（SFT）和对称的候选排列一致性 KL 约束，没有进行 RL 或 RLCD。 建议在 Docker 中使用指定版本的 KEV 源码及兼容依赖。已验证部署采用 PyTorch 2.8.0、transformers…

Model

kfold

KAIST AI-Bio

\[Notice\] Preprint will be released soon! K-Fold predicts biomolecular complex structures and binding-induced conformational changes. Through an apo-to-holo diffusion bridge, K-Fold aims to capture conformational changes in systems such as G protein-coupled receptors (GPCRs). K-Fold supports proteins, DNA, RNA, small molecules, and chemical modifications without multiple sequence alignments (MSAs). This model repository provides K-Fold pretrained weights and configuration. Inference and training code, examples, and data preprocessing workflows are available on GitHub. K-Fold uses pretrained AtlasLM for protein sequence representations and TriProRep for protein structure representations.…

Model

kfold-assets

KAIST AI-Bio

This repository provides the chemical component dictionary and pretrained encoder weights used by K-Fold. These assets are downloaded automatically when needed. - assets/ccd.pkl: Precomputed chemical component dictionary (CCD). - weights/protstructbbtokenizer.pth: Protein backbone tokenizer weights copied from TriProRep-3B. - weights/protstructfatokenizer.pth: Protein full-atom tokenizer weights copied from TriProRep-3B. - weights/protstructencoder3b.pth: Protein structure encoder weights converted to bfloat16 from TriProRep-3B-finetune. - weights/rnaseq1b.pth: Pretrained RNA sequence encoder weights. Source code for the protein structure encoder and tokenizers is available in the TriProRep…

This model detects text lines and mathematical formula rows in scanned Cambodian textbooks for OCR line cropping and document layout analysis.

Model

Kiel-3-Cyber

Tech

Kiel-3-Cyber-7B is an elite autonomous cybersecurity expert model fine-tuned from Qwen2.5-Coder-7B-Instruct. Designed specifically for advanced vulnerability analysis, threat modeling, code security reviews, and CVE parsing, Kiel-3-Cyber incorporates explicit Deep Chain-of-Thought (CoT) reasoning inside tags before delivering precise mitigation strategies. saidutta69/CyberSec-Reasoning-Premium moro72842/cybersecurity-sft-dataset AlicanKiraz0/All-CVE-Records-Training-Dataset nvidia/OpenCodeReasoning Kiel-3-Cyber is trained on a standardized multi-turn system/user/assistant prompt layout with explicit step-by-step reasoning wrappers: You can load and run Kiel-3-Cyber-7B using the standard…

Experimental NAS/pruning/full LoRA recovery artifact. Created using Bonsai by Prism ML. Derived from original SFT, revision 011ee6c8d05d49144fa5747bedf23e8904883564. - q1/base-Q10.gguf + q1/adapter-f16.gguf: 169895008 bytes combined. Cap applies to this deployment pair, not this whole repository. - model/: matching pruned HF base and tokenizer, necessary for this adapter. Do not attach the adapter to the unpruned original base. - hfadapter/: final R08 LoRA weights; metadata/: pruning map, recipe, certificates and provenance. - contract/: original trained contract including recommendations. benchmark/: OOD comparison, recommendations ignored only in core-v1 scoring. One full epoch, 77,358…

Experimental NAS/pruning/full LoRA recovery artifact. Created using Bonsai by Prism ML. Derived from original SFT, revision 011ee6c8d05d49144fa5747bedf23e8904883564. - q1/base-Q10.gguf + q1/adapter-f16.gguf: 197170592 bytes combined. Cap applies to this deployment pair, not this whole repository. - model/: matching pruned HF base and tokenizer, necessary for this adapter. Do not attach the adapter to the unpruned original base. - hfadapter/: final R08 LoRA weights; metadata/: pruning map, recipe, certificates and provenance. - contract/: original trained contract including recommendations. benchmark/: OOD comparison, recommendations ignored only in core-v1 scoring. One full epoch, 77,358…

A QLoRA adapter for OpenAI's gpt-oss-20b that turns it into KinShield's phone-call evidence detector. KinShield is an entry in the AWS "Zero to Shipped" hackathon. The model reads a call transcript and returns strict JSON with: - a risk level (LOW / MEDIUM / HIGH) and a score - every warning sign, with the exact quoted line, a timestamp, one of 7 fixed signals and a weight It answers directly on the harmony final channel, without writing out reasoning first. Status: this is a research result. The live KinShield app still uses the stock gpt-oss-20b on Amazon Bedrock. This adapter is not hosted anywhere public. The adapter matches gpt-oss-120b, a model about 6× larger, and edges out the stock…

A 22M-parameter phone-call scam classifier (MiniLM-L6, int8 ONNX, 22.9 MB) built for KinShield, an entry in the AWS "Zero to Shipped" hackathon. It reads a call transcript and returns a score from 0 to 1 for each of KinShield's 7 warning signs, plus a HIGH-risk probability. It runs on CPU in about 40–80 ms. What's new in v3: v3 was distilled from Solomonwilsonr/kinshield-gpt-oss-20b-lora, our fine-tuned gpt-oss-20b. v2 was distilled from the stock gpt-oss-20b. Both runs use int8 ONNX. v3 still misses scam016 (a Medicare card scam) and scam019 (a Hinglish UPI "beta in trouble" call), which it rates LOW, and rates scam014 MEDIUM. impersonation, emergency, secrecy, financialrequest…

Model · Text generation

knivesysl-typed-2b

SRSWTI Inc.

local, typed decisions from qwen3.5-2b. one shared state is prefetched once, each question is isolated, every allowed answer is scored as a complete token sequence, and python returns validated choice, score, and noul results. this is an inference system, not rlcd training and not a clone of typesafe jev. it never calls typesafe. the published qwen checkpoint is unchanged; fp8 changes execution precision only. probabilities are normalized support over the candidates you provide, not calibrated correctness probabilities. unlike ordinary autoregressive json generation, the model does not write a response token by token. it scores only the values supplied by the caller. complete-sequence…

Model

kodiak-large-v2-preview

Cortex Agent LLC

Kodiak is an encoder-only "System One" decision model by Cortex Agent LLC: a state (text, list of texts, or JSON) plus typed questions in, calibrated answers out, in one forward pass. Choice answers are always one of your labels, scores stay inside your range, and every question can come back as "not answerable from this state" with its own probability. The large model beats every open zero-shot classifier we tested on never-seen tasks (release criterion 1; decision D33). Details: - Run b-base-s1-v2-s1 (seed 1 of three, chosen by validation loss, not the eval set), step 6000. ModernBERT-large backbone (Apache-2.0). Calibration temperatures baked in; default abstain threshold from…

Model

kodiak-small-v2-preview

Cortex Agent LLC

Kodiak is an encoder-only "System One" decision model by Cortex Agent LLC: a state (text, list of texts, or JSON) plus typed questions in, calibrated answers out, in one forward pass. Choice answers are always one of your labels, scores stay inside your range, and every question can come back as "not answerable from this state" with its own probability. Trained on the public datasets plus Generator v2 synthetic data (9,137 training examples): a 971-document-type taxonomy, ~45% real web passages, questions labeled stated / inferred / unanswerable, and two AI critics. Compared with the same model trained on the v1 synthetic data at equal size, averaged over three training runs each: So v2…

This model was converted to MLX format from hexagrad/Kokoro-82M using mlx-audio version 0.0.1. Refer to the original model card for more details on the model.

Model · Text to speech

Kokoro-82M-CoreAI

Richard R.

Kokoro-82M (StyleTTS2 + iSTFTNet) exported to Apple's Core AI.aimodel format for iOS 27 and macOS 26. Built for OpenReader iOS, an on-device document reader. Published because the export differs from the community recipe in one way that matters, and because a working Core AI TTS export was hard enough to arrive at that it is worth not repeating. Three fixed-shape bundles with two host steps between them. Kokoro's acoustic graph has exactly one data-dependent length — the duration-to-alignment expansion L = sum(preddur) — so the model is cut where that length appears rather than being made dynamic. Token bucket 128, frame bucket 512 — one pass renders 12.8 s of audio. Fixed buckets are not a…

Model · Text to speech

kokoro-82m-glade

Josh Leverette

Native-context conversion of Kokoro-82M v1.0. It accepts phonemes and generates mono 24 kHz Float PCM. The bundle includes afheart and amadam; measurements use afheart at speed 1. Glade is a native Swift library for speech inference on Apple silicon, using Core AI, Core ML, and model-specific host implementations. Its runtime source is currently private. This repository distributes the model assets and runtime metadata used by that library. Python is export/measurement tooling, not an application dependency. Mac qualification is included; iPhone and Watch qualification is not claimed. Download weights and configuration from one snapshot. The profile retains up to 510 phoneme symbols, with…

The fp32 onnx/model.onnx from onnx-community/Kokoro-82M-v1.0-ONNX at revision 1939ad2a, with the graph cut at decoder.generator.convpost. The weights are unchanged. Why: on onnxruntime-web 1.27 WebGPU, the full graph's iSTFT tail (ConvTranspose/ScatterND) returns zeros for every sample past about 65,528 (2.7 s at 24 kHz). Callers then have to feed short phrases, and that audibly flattens the prosody. Inputs: inputids int64 [1, n], style float32 [1, 256], speed float32 [1] (same as upstream). iSTFT to run in the caller (nfft 20, hop 5; istftbasis.json = the 22x20 inverse basis): This matches the full graph's waveform to within 6e-7 (max abs, CPU). Kokoro-82M is by hexgrad, released under…

GGUF quantizations of Aleph-Alpha/Kolibri-1 — trained from scratch, with tool calling, an explicit reasoning mode and a 262 K native context. Apache-2.0. Because only \~3.5 B parameters are active per token, it decodes like a small model while carrying the knowledge capacity of a large one: \~51 tokens/s on a DGX Spark (GB10) at 4.6 bpw. arch kolibri1 78.1 B params, vocab 128,000 layers 50 FFN MoE 384 experts, top-6, +1 shared (ungated); expert ffn 512 routing select top-6 on logits + per-expert bias, weight by sigmoid(logits), no renorm attention 48 heads / 4 KV, headdim 128, per-head q/k RMSNorm 4 sliding-window layers (window 513, RoPE): 1 full-attention layer (no RoPE) context 262,144…

Model · Text generation

Kolibri-1-GGUF

Insidious Fiddler

Unofficial GGUF quantization of Aleph Alpha's Kolibri-1. Requires the experimental Kolibri patch for llama.cpp. Get the patched runtime from Eliasfpv28/Kolibri-1-Q3KS-GGUF, and follow Eliasfpv28's build instructions to build the patched llama.cpp. The tested runtime uses base revision edd6e2bbdad5930899a93db8fa73c3b61c7b9bcc. Kolibri-1 is a text model for German and English, with reasoning and tool-calling modes. It has 78.10 billion parameters and activates about 3.46 billion per token. The following llama-bench runs use an RTX Pro 6000 Blackwell Server Edition on RunPod. The Pod has 96 GB VRAM, 188 GB host RAM and 32 allocated vCPUs. Measurements use three repetitions, 16 CPU threads…

Model · Text generation

Kolibri-1-GGUF

Bartowski

Using llama.cpp release b11716 for quantization. Don't know which to choose? Grab Q4KM (49.76GB) - usually a good mix of size and performance. Download instructions available here First, make sure you have the Hugging Face CLI installed: The files marked true in the Split column above are stored as multiple parts in a folder. To download all the parts to a local folder, run: You can either specify a new local-dir (Kolibri-1-Q80) or download them all in place (./) These quants run with llama.cpp - installable in one line via llama.app: llama-server includes a built-in chat web UI, served at http://localhost:8080 by default. These quants were made with llama.cpp release b11716 - if this…

A kraken 7 baseline-segmentation model fine-tuned on Cairo Genizah manuscript pages — the first Genizah-specific segmenter we know of that is publicly available. Recognition models for these hands exist (the MiDRASH project published its recognizers on Zenodo), but line segmentation is where a generic model loses most of the text on Genizah fragments: tight columns, damaged edges, script far larger than the segmenter was trained on, marginal glosses and rotated scans. This model closes a large part of that gap. 10.5281/zenodo.22879549, Apache-2.0), a general print-and-handwriting segmenter trained on Latin, Arabic and Greek script corpora. Cairo Genizah, drawn from the KTIV project of the…

Krea 2 model (covering both the Krea 2 Raw and Krea 2 Turbo variants) is released under a custom Krea 2 Community License Agreement for its weights, alongside an Apache-2.0 license for the inference code repository itself. For removal please contact me directly. All rights remain with their respective creators. By downloading or using any model from this repository, you acknowledge that: - You are solely responsible for how the model is used. - The repository maintainer assumes no responsibility or liability for any content generated by these models. - Krea AI, the original model authors (when applicable), and the repository maintainer cannot be held liable for misuse, illegal activities…

Model

kuza-qwen-3.5-4b

Kuza AI

Full training-run archive for Kuza (East Africa agricultural assistant), fine-tuned from unsloth/Qwen3.5-4B. Weights, logs, checkpoints, GGUFs, and provenance are stored with the same layout as $KUZAWORKDIR/kuza-qwen-3.5-4b/. This derivative is subject to the Apache 2.0 license of the Qwen base model. - 100% English train from kuzaai/kuzasftenglish - 35% Swahili from kuzaai/kuzasftswahili - 8% HuggingFaceH4/norobots - 5% adversarial from kuzaai/kuzasftadversarial - all multiturn from kuzaai/kuzasftmultiturn Sequence length 1024, 2 epochs, LR 0.0002. Thinking is off (--reasoning off when llama.cpp supports it). GGUFs are text-only (MTP/nextn kept; vision/audio dropped). ssmout is Q6K on…

Diffusion Policy treats visuomotor control as a generative diffusion process, producing smooth, multi-step action trajectories that excel at contact-rich manipulation. This policy has been trained and pushed to the Hub using LeRobot. See the full LeRobot documentation. The policy consumes these observation features and produces these action features. Inputs Outputs New to LeRobot? These guides cover the full workflow: - Install LeRobot — set up the lerobot package. - Hardware setup — assemble, wire, and calibrate your robot and cameras. - Record data & train a policy — the end-to-end imitation-learning walkthrough. - CLI cheat-sheet — quick reference for the lerobot- commands. The short…

World-action model for LAW + FastWAM, selected step-30,000 checkpoint. Use with the complete LAW-VLM-FastWAM Q1/Q2/Q3 VLM. Includes WAM weights, configuration and normalization statistics under wam/. The corresponding VLM is packaged separately. Original combined releases remain available in this repository's Git history. Code and download guide: hey-cjj/LAW. Historical RoboTwin overall success: 94.21%. These are selected historical results, not a new benchmark. Six bounded 32-GPU training save/load checks and both bounded real RoboTwin policy episodes passed before this packaging change; see release audit for scope and historical provenance limits. Tensor bytes are unchanged. LAW-authored…

World-action model for LAW + OpenWAM, selected step-120,815 checkpoint. Use with the complete LAW-VLM-OpenWAM Q1/Q2/Q3 VLM. Includes WAM weights, configuration and normalization statistics under wam/. The corresponding VLM is packaged separately. Original combined releases remain available in this repository's Git history. Code and download guide: hey-cjj/LAW. Historical RoboTwin overall success: 95.56%. These are selected historical results, not a new benchmark. Six bounded 32-GPU training save/load checks and both bounded real RoboTwin policy episodes passed before this packaging change; see release audit for scope and historical provenance limits. Tensor bytes are unchanged. LAW-authored…

Complete Qwen3-VL-2B VLM for LAW + FastWAM: Q1 planning, Q2 completion/failure assessment, and Q3 action-conditioned value estimation. Includes the frozen backbone, Q1/Q2 adapters, query embeddings and tokenizer at the repository root, plus the matching value checkpoint under q3/. Use this VLM with LAW-FastWAM. Matching training, inference, resource download and verification code: hey-cjj/LAW. Files are copied from the previously verified checkpoints without changing tensor bytes. Q2 and Q3 are the selected step-50,000 checkpoints. This packaging change does not retrain the model. The release previously passed six bounded 32-GPU two-optimizer-step training save/load checks and two bounded…

Complete Qwen3-VL-2B VLM for LAW + OpenWAM: Q1 planning, Q2 completion/failure assessment, and Q3 action-conditioned value estimation. Includes the frozen backbone, Q1/Q2 adapters, query embeddings and tokenizer at the repository root, plus the matching value checkpoint under q3/. Use this VLM with LAW-OpenWAM. Matching training, inference, resource download and verification code: hey-cjj/LAW. Files are copied from the previously verified checkpoints without changing tensor bytes. Q2 and Q3 are the selected step-50,000 checkpoints. This packaging change does not retrain the model. The release previously passed six bounded 32-GPU two-optimizer-step training save/load checks and two bounded…

Model · Text classification

laya

Convai Innovations

Multilingual, non-autoregressive System 1 decision model. Give it a state (text, email, ticket, or JSON) and typed questions; it returns typed answers with mathematically calibrated probabilities in a single forward pass (~33 ms) across 100+ languages. Trained with reinforcement learning against strictly proper scoring rules (RLCD), so reporting honest probabilities is the only way to maximise reward. It never generates text, so there is nothing to parse and nothing to hallucinate. This repo holds all three checkpoints and is the hub for the family. The English checkpoint is at the repo root; the other two are bundled subfolders, and only the one you request is downloaded: pip install -U…

The Laya System One decision model, exported to ONNX fp32 and packaged for Verdict -- a high-performance serving engine for System One decision models. Serve it with one command: Pick the image that matches the host. The -v mounts cache the model, so only the first run downloads it. Add -e VERDICTUSEMODELSCOPE=1 and use sgyaqing/laya to fetch from ModelScope instead of the hub; add -e HFENDPOINT=https://hf-mirror.com if the hub is slow or unreachable. Endpoints: the standard decision protocols POST /v1/systemone and POST /v1/decisions, native POST /v1/choice | /v1/score | /v1/noul, plus an OpenAI-compatible layer (/v1/chat/completions, /v1/models). Measured on one machine (RTX 4090, AMD…

A Laya (421M, ModernBERT-large encoder + typed decision heads) checkpoint fine-tuned to pick the next browser action — an operation (CLICK / TYPETEXT / SELECT) and a target element — from a page's DOM and a natural-language goal. Trained on Mind2Web (Deng et al., NeurIPS 2023 Datasets and Benchmarks Track, CC-BY-4.0). Built as the policy for laya-system-use, a fork of browser-use/jev-ultrafast that replaces the hosted TypeSafe Jev API with this open-weight model. This is a research/hobby-project checkpoint, not a production model. Read the Limitations section before using it. Given a page's interactive elements (label, role, current value) and a goal, it answers two typed questions in one…

Laya typed decisions on Apple Silicon, running on the Core AI runtime — the successor to Core ML. This is a.aimodel asset exported from via Apple's coreai-torch bridge. It outputs choice / score / noul probabilities (and RL action logits) with zero generated tokens and no PyTorch, Core ML, Transformers, or cloud API at inference time. macOS 27+ (Core AI runtime), Python 3.10+. Validated on M3 Max / macOS 27.2. Validate the download end-to-end (all three specializations, timing, contract checks): Snake demo with the model (terminal game, reuses the laya-coreml UI + safety shield; automatically uses the B3 asset when present for ~2x game throughput): ~3× faster per pass than the fastest Core…

Laya typed decisions on Apple Silicon, using CPU + GPU. This is a portable Core ML bundle for laya-coreml, converted from convaiinnovations/laya. It outputs choice, score, and noul probabilities with zero generated tokens. Inference needs no PyTorch, Transformers, MLX, remote code, or cloud API. Apple Silicon, macOS 15+, Python 3.11–3.13. Tested on M3 Max / macOS 27.2. To download explicitly and then run entirely offline: Use laya.load("aac6fef/laya-coreml", localfilesonly=True) for a cached snapshot or pass a local directory. Use revision=" " to pin a remote revision. This FP16 export retains the original model architecture and decision schema. The enumerated-length GPU export is the…

Two ONNX exports of Laya-multilingual (322M parameters, built on mmBERT) used by Laya-Heart, a texting copilot for couples that runs on your own computer. Laya reads the partner's message and answers short card questions with probabilities ("Does 没事 really mean it's fine?", a danger level from 1 to 10, a hidden safety question). Both run with onnxruntime and Laya's laya-ts runtime, exactly as the app runs them. Practice exam written and labelled by teacher B (Qwen3-32B-AWQ, a different model family from the teacher that wrote the training chats). One command in the repository recomputes every number from the raw answers, or re-runs both models: npm run reproduce / npm run reproduce…

Model · Text classification

laya-kvp10k-noul

Sothiara Em

A fine-tuned Laya model (Convai Innovations, 421M, ModernBERT-large backbone) that answers one typed noul question: does a value correctly match its key label? (e.g. firstname = John → true, firstname = 1992 → false). Fine-tuned on the IBM KVP-10K dataset via the pre-parsed community mirror (OCR pre-extracted; no OCR step needed). This is the v2 (remediated) release: document-level 80/10/10 data partitioning, mixed-class frozen test set, calib-split temperature fitting, and a frozen category-stratified release gate. It supersedes the v1 release pre-remediation, historical artifact). Laya is a non-autoregressive decision model: you give it a state (text/dict) plus typed questions (choice…

Model

laya-moneytrail-injection

Chetan Choudhary

A fine-tune of convaiinnovations/laya that flags an evidence document carrying an instruction aimed at its reader. Built for MoneyTrail, a cross-organization payment investigation prototype. This is not the best available detector for the job, and it is not used in MoneyTrail's decision path. It is published so the experiment can be reproduced. - On attacks no detector could have trained on, it scores 0.688 AUC. deepset/deberta-v3-base-injection scores 0.894 and sheltron-ai/prompt-guard-68m 0.785 on the same data. If you want a screen, use one of those. - It reaches a perfect 1.000 on its own held-out synthetic test split. That number is an artefact of the generator that made the data: the…

Model · Text classification

laya-multilingual

Convai Innovations

Non-autoregressive System 1 decision model covering 100+ languages. Give it a state (text, email, ticket, or JSON) and typed questions; it returns typed answers with probabilities in a single forward pass. No text generation, so nothing to parse and nothing to hallucinate. Part of the Laya family — use this checkpoint for anything that is not English. Since laya 0.3.13 the default Router() keeps both english and this checkpoint resident, so a mixed workload no longer swaps checkpoints on every language change. For a server, load them up front so even the first request of each language is just a forward pass: router.attach("multilingual", agent) registers an Agent you already built, so a…

Non-autoregressive System 1 decision model covering 100+ languages. Give it a state (text, email, ticket, or JSON) and typed questions; it returns typed answers with probabilities in a single forward pass. No text generation, so nothing to parse and nothing to hallucinate. Part of the Laya family — use this checkpoint for anything that is not English. The default Router() keeps both english and this checkpoint resident, so a mixed workload no longer swaps checkpoints on every language change. For a server, load them up front so even the first request of each language is just a forward pass: router.attach("multilingual", agent) registers an Agent you already built, so a process that loaded…

The Laya Multilingual System One decision model (mmBERT-base, 100+ languages), exported to ONNX fp32 and packaged for Verdict -- a high-performance serving engine for System One decision models. Serve it with one command: Pick the image that matches the host. The -v mounts cache the model, so only the first run downloads it. Add -e VERDICTUSEMODELSCOPE=1 and use sgyaqing/laya-multilingual to fetch from ModelScope instead of the hub; add -e HFENDPOINT=https://hf-mirror.com if the hub is slow or unreachable. For English-only workloads, use verdict-models/laya. Endpoints: the standard decision protocols POST /v1/systemone and POST /v1/decisions, native POST /v1/choice | /v1/score | /v1/noul…

Laya typed decisions on Apple Silicon, using CPU + GPU. This is a portable Core ML bundle for laya-coreml, converted from convaiinnovations/laya-multilingual. It outputs choice, score, and noul probabilities with zero generated tokens. Inference needs no PyTorch, Transformers, MLX, remote code, or cloud API. Apple Silicon, macOS 15+, Python 3.11–3.13. Tested on M3 Max / macOS 27.2. To download explicitly and then run entirely offline: Use laya.load("aac6fef/laya-multilingual-coreml", localfilesonly=True) for a cached snapshot or pass a local directory. Use revision=" " to pin a remote revision. This FP16 export retains the original model architecture and decision schema. The…

Laya typed decisions on Apple Silicon, using CPU + Neural Engine. This is a portable Core ML bundle for laya-coreml, converted from convaiinnovations/laya-multilingual. It outputs choice, score, and noul probabilities with zero generated tokens. Inference needs no PyTorch, Transformers, MLX, remote code, or cloud API. Apple Silicon, macOS 15+, Python 3.11–3.13. Tested on M3 Max / macOS 27.2. To download explicitly and then run entirely offline: Use laya.load("aac6fef/laya-multilingual-coreml-ane", localfilesonly=True) for a cached snapshot or pass a local directory. Use revision=" " to pin a remote revision. This FP16 conversion retains the original trained parameters. The ANE graph's host…

Laya typed decisions on Apple Silicon, using CPU + Neural Engine. This is a portable Core ML bundle for laya-coreml, converted from convaiinnovations/laya-multilingual. It outputs choice, score, and noul probabilities with zero generated tokens. Inference needs no PyTorch, Transformers, MLX, remote code, or cloud API. Apple Silicon, macOS 15+, Python 3.11–3.13. Tested on M3 Max / macOS 27.2. To download explicitly and then run entirely offline: Use laya.load("aac6fef/laya-multilingual-coreml-ane-w8", localfilesonly=True) for a cached snapshot or pass a local directory. Use revision=" " to pin a remote revision. This is a separately identified 8-bit grouped K-means weight-palette variant.…

Laya typed decisions on Apple Silicon, using CPU + GPU. This is a portable Core ML bundle for laya-coreml, converted from convaiinnovations/laya-multilingual. It outputs choice, score, and noul probabilities with zero generated tokens. Inference needs no PyTorch, Transformers, MLX, remote code, or cloud API. Apple Silicon, macOS 15+, Python 3.11–3.13. Tested on M3 Max / macOS 27.2. To download explicitly and then run entirely offline: Use laya.load("aac6fef/laya-multilingual-coreml-snake", localfilesonly=True) for a cached snapshot or pass a local directory. Use revision=" " to pin a remote revision. This FP16 export retains the original model architecture and decision schema. The…

Model · Text classification

laya-onnx

weXare

An unofficial ONNX export of convaiinnovations/laya, the open-source "System One" decision model by Convai Innovations, made to run client-side with ONNX Runtime Web on WebGPU or WebAssembly. Laya takes a state and typed questions and scores every option in one forward pass. It returns calibrated probabilities, not generated text. This is the default model of @wexare/laya-web (source), which loads it automatically: sha256 of layafp16.onnx: b2125c7a1044e78807ee718ff4e65c9892b6b55aeeeee64f31fffffd76871b0a Batch, sequence and option count are all dynamic. Masked option slots return -1e4. Turning logits into answers (per-option-count temperature, softmax, confidence) follows the reference laya…

Model · Text classification

laya-pdpa-onnx-v4-32k

MY

ONNX export of Suarify/laya-pdpa-my-v4 — a Malaysian PDPA classifier for call transcripts. Ten typed questions per call: eight yes/no classes, one ordered severity, one choice for who disclosed. Exported with torch.onnx.export(dynamo=True), opset 17, external data. Accuracy is over all 960 held-out decisions (96 rows × 10 questions), by argmax, against a majority-class baseline of 72.8%. Parity of the fp32 export against the PyTorch checkpoint is 1.07e-05 max absolute logit difference across batches 1, 3, 8 and 16. fp32 and q4f16 both score 825/960. That identical total is a coincidence of to 10.15 absolute logit difference, and the flips happen to break 23 each way. The disagreements are…

Laya 拼音输入法（macOS）用的候选词排序模型，基于 convaiinnovations/laya（revision 7b928d82）微调。 prior 对于词来说是词频，对短语来说是词格路径得分。模型看不到用户输入的拼音。 训练样本由输入法自己的引擎（backend/，commit 9d87301）录制，每条样本都对应一次真实的模型调用：上文、按引擎 - 从基座 Laya 开始，除 embedding 以外的全部参数参与训练。 - 损失：对每次调用的候选做 listwise cross-entropy，得分为 logit + betakind prior，每种调用类型各学一个 beta；每个 epoch 打乱候选顺序。 - AdamW，LR 3e-5，weight decay 0.01，bf16 autocast，gradient clipping 1.0。 - 学习率：1,000 step warmup，之后保持不变，最后 20% 线性衰减到 0。 - 166,432 step，batch 100（1,664 万条，约 1.15 轮），单张 RTX 5090 约 6 小时。 bootstrap 区间。对照为未公开的上一个版本 w28。 - 依赖本引擎： 模型针对本引擎的候选列表和 prior 微调，不是通用语言模型，不能单独当输入法用。 权重为 Apache 2.0，是 convaiinnovations/laya（Apache 2.0）的衍生作品。inference/、scripts/ 和 eval/ 中的代码为 MIT，与输入法仓库相同。

Model · Text classification

laya-pt-es-typed

Telepatia

This checkpoint fine-tunes convaiinnovations/laya-multilingual for native choice, score, and noul decisions in Portuguese and Spanish. It keeps the original 322M-parameter mmBERT architecture. It adds no inference component and does not generate text. It returns typed answers and probabilities in one forward pass. This is a text model. Inference takes a textual state plus typed questions. The second training stage used text decisions derived from public speech corpora, but this checkpoint does not accept audio by itself. The separate audio projector is not included. The official Laya SDK defines these primitives as follows: - choice: selects one key from a runtime-defined criteria object.…

Classifies GitHub issues written in any language as bug, feature, question or docs. A fine-tune of Laya multilingual (mmBERT-base) used by the laya-triage GitHub Action for non-English issues, next to the English model laya-triage-en. The same 500 NLBSE'23 validation issues, machine-translated with NLLB-200 into 13 languages. Accuracy (±3 points per language): laya-triage and Jev are within noise of each other across languages; both are far ahead of the untuned base. Translations can flatter a model trained on translations, so we also checked real issues: on 367 non-English issues opened in 2026 (never seen, written by people, not translated) accuracy went from 47.1% to 65.7%. Use it…

Model · Text classification

laya-typed-decisions

Convai Innovations

Non-autoregressive System 1 decision model, fine-tuned on the typed-decisions workflows: agent-trace observability, customer service, invoice processing and security incidents. Part of the Laya family. 400 test cases, 2,000 decisions, measured on the official test split. +3.9 points over Jev's published 0.727, above the 0.735 teacher ceiling, with 2.4x better Brier and 1.6x better score MAE. Jev figures are third-party published, not measured here — there is no TypeSafe API access in this project, and sample sizes and prompts differ. Treat the comparison as indicative. Router will not select this checkpoint automatically unless you construct it with autotaskdetection=True — it is…

Laya typed decisions on Apple Silicon, using CPU + GPU. This is a portable Core ML bundle for laya-coreml, converted from convaiinnovations/laya-typed-decisions. It outputs choice, score, and noul probabilities with zero generated tokens. Inference needs no PyTorch, Transformers, MLX, remote code, or cloud API. Apple Silicon, macOS 15+, Python 3.11–3.13. Tested on M3 Max / macOS 27.2. To download explicitly and then run entirely offline: Use laya.load("aac6fef/laya-typed-decisions-coreml", localfilesonly=True) for a cached snapshot or pass a local directory. Use revision=" " to pin a remote revision. This FP16 export retains the original model architecture and decision schema. The…

This qwen2 model was trained 2x faster with Unsloth and Huggingface's TRL library.

Model · Text generation

less-is-moe-gpt-oss-120b-gpqa-main-64-intdim-e-50

XINKAI ZOU

This checkpoint was structurally pruned with the released Less-is-MoE mean-absolute-gradient method. It removes exactly 50% of routed-expert FFN neurons using 64 calibration samples from the gpqamain configuration of Idavidrein/gpqa revision 633f5ee89ab8ad4522a9f850766b73f62147ffdd. The released loader settings are preserved: train, Question plus shuffled choices, Explanation, selectionseed=1234, BF16, and no optimizer step. The samples are full length: no tokenizer maxlength, truncation, or padding. The longest input for this tokenizer is 1,511 tokens. The source MXFP4 checkpoint was explicitly dequantized to BF16 before scoring and pruning. The source-row selection hash is…

Model · Text generation

less-is-moe-qwen3.5-122b-a10b-gpqa-main-64-intdim-e-50

XINKAI ZOU

This checkpoint was structurally pruned with the released Less-is-MoE mean-absolute-gradient method. It removes exactly 50% of routed-expert FFN neurons using 64 calibration samples from the gpqamain configuration of Idavidrein/gpqa revision 633f5ee89ab8ad4522a9f850766b73f62147ffdd. The released loader settings are preserved: train, Question plus shuffled choices, Explanation, selectionseed=1234, BF16, and no optimizer step. The samples are full length: no tokenizer maxlength, truncation, or padding. The longest input for this tokenizer is 1,632 tokens. The source checkpoint was loaded and pruned in BF16. The source-row selection hash is…

Model · Text generation

less-is-moe-qwen3.5-122b-a10b-gpqa-main-64-intdim-g-50

XINKAI ZOU

This checkpoint was structurally pruned with the released Less-is-MoE mean-absolute-gradient method. It removes exactly 50% of routed-expert FFN neurons using 64 calibration samples from the gpqamain configuration of Idavidrein/gpqa revision 633f5ee89ab8ad4522a9f850766b73f62147ffdd. The released loader settings are preserved: train, Question plus shuffled choices, Explanation, selectionseed=1234, BF16, and no optimizer step. The samples are full length: no tokenizer maxlength, truncation, or padding. The longest input for this tokenizer is 1,632 tokens. The source checkpoint was loaded and pruned in BF16. The source-row selection hash is…

Model · Text generation

less-is-moe-qwen3.5-122b-a10b-gpqa-main-64-intdim-l-50

XINKAI ZOU

This checkpoint was structurally pruned with the released Less-is-MoE mean-absolute-gradient method. It removes exactly 50% of routed-expert FFN neurons using 64 calibration samples from the gpqamain configuration of Idavidrein/gpqa revision 633f5ee89ab8ad4522a9f850766b73f62147ffdd. The released loader settings are preserved: train, Question plus shuffled choices, Explanation, selectionseed=1234, BF16, and no optimizer step. The samples are full length: no tokenizer maxlength, truncation, or padding. The longest input for this tokenizer is 1,632 tokens. The source checkpoint was loaded and pruned in BF16. The source-row selection hash is…