Affectra is a text-emotion foundation model: a single set of weights that solves 15 heterogeneous affective tasks — discrete-emotion classification, dimensional (VAD) regression, intensity regression, interpersonal judgments (irony, hate speech, stance), an aspect-level applied task, and multilingual emotion detection — in English, Japanese, and multilingual data. This is Affectra-1-8B, the first release. The series is versioned (Affectra-1, Affectra-2, …) and sized (-8B). (q,k,v,o,gate,up,down); 88.7 M trainable parameters (~334 MB). channel (label schema) is placed after the body, so under causal masking the body's hidden states are provably independent of the channel; the final [READ]…
Open weights
cc-by-nc-4.0
peft
This repository contains the fixed model files used by the AmySeg inference workflow. It is a model bundle, not a standalone executable. Runtime code and algorithm configuration are provided by the AmySeg Python package. The nnUNet dataset and plans JSON files are bundled with the Python package. The model weights in isonet2/model.pt and nnunet/checkpoint.pth are licensed under CC BY-NC 4.0. The full license text is included in LICENSE. Non-commercial research use is permitted under the license terms. Non-commercial sharing and adaptation are also permitted, subject to attribution and the other license conditions. Commercial use is not authorized by this license and requires separate…
Open weights
cc-by-nc-4.0
本仓库（Repository）所包含的所有人工神经网络（Artificial Neural Network）权重文件（.pth/.index）、训练日志及相关代码，均为计算声学（Computational Acoustics）与深度学习（Deep Learning）领域的技术研究实验产物。 这些文件本质上是高维张量（High-dimensional Tensors）的数值序列，通过随机梯度下降（SGD）与反向传播算法（Backpropagation）对公开可获取的音频数据进行统计建模（Statistical Modeling）得到。其技术形态与数字图像处理中的卷积核（Convolution Kernels）、自然语言处理中的词向量（Word Embeddings）并无本质差异。 本仓库不构成对任何第三方知识产权的故意侵犯，所有代码遵循 MIT License 开源协议，模型权重文件仅作为技术实现的副产品（By-products）存在。 本技术实验所使用的训练数据集（Training Dataset）包含以下角色的语音样本： - Mon3tr（モンスター）、Mon3tr 中文语音 等 - 声源版权归属：上海鹰角网络科技有限公司（Hypergryph Network Technology Co., Ltd.）及其关联公司 - 原始作品：《明日方舟》（Arknights） 上述角色的声音版权、肖像权、姓名权及相关知识产权均完全归属于鹰角网络及其合法授权方。本实验仅基于已公开发布的游戏内语音资源进行技术层面的信号处理（Signal Processing）与特征提取（Feature…
Open weights
cc-by-nc-4.0
Open weights
cc-by-nc-4.0
This is a collection that I have been trying to gather from LoRas that were banned or were from the very start way "too spicy" to ever upload into Civitai so hopefully users could try find some of them here for what they want to do. Im focusing mainly on LoRas for PonyXL, IllustriousXL and Anima. Because write what each LoRa respective trigger words is way too much, I uploaded a LoRa Viewer that you can use to load each of them to see what words trigger each of them. At the time of writing I can be contacted through my discord server perhaps if Im not quite busy with something (the usual) http://chat.nx34.xyz I also have another project over here: http://ubtl.nx34.xyz
Open weights
cc-by-nc-4.0
libraryname: transformers pipelinetag: text-generation - language-model - berna - dna-kernel-plexus - cells - research - non-commercial 1.2B parameter language model with self-growing cell architecture. 2. Saturation S = 0.8578 (just above threshold 0.85) Comparison: GPT-2 Small (124M) achieves HellaSwag 31.4%, ARC-Easy 44.0%, PIQA 64.1%. Berna R7 v3 (1.2B) matches this at ~15x less training data. code/ Training code config/ config.json cellsruntimecurrent/ Live cells + cognitive log cellsruntimefinal/ Final cell state checkpoints/step00010200/ Final model (with wrapper) checkpoints/step00001700/ Earlier checkpoint data/tokenized/ eng.npy, math.npy, code.npy data/tokenizer/ tokenizer.json…
Open weights
cc-by-nc-4.0
transformers
Each cell that saturates (S >= 0.68, K6 >= 0.03): 3. Gradient hooks - zero gradients for frozen neurons Result: Bounded Forgetting - old knowledge protected while learning new.
Open weights
cc-by-nc-4.0
transformers
Junhao Zhou 4, Zitian Chen 1, Linchao Zhu 2,† CaptchaAgent is a single 9B computer-use policy for all 20 CAPTCHA types in CaptchaArena, fine-tuned from Qwen3.5-9B. One set of weights, one tool schema and one observation format are shared across every type, with no per-task adapter or task-conditioned switching. Both subfolders hold full bf16 weights (9 safetensors shards each) with the tokenizer, processor and chat template, so each loads on its own. SFT/ is the final SFT checkpoint with the LoRA merged in (Table 5a of the paper); RL/ is the best GRPO checkpoint (Table 5b). CaptchaArena scores are on the 4,000-puzzle test split (20 types × 200), with 5 rollouts per puzzle and the unbiased…
Access requested at publisher
cc-by-nc-4.0
transformers
Modelos de tesis sobre ChimpACT Release v1 (23 comportamientos, multi-etiqueta, split oficial por video). Cada carpeta /models/ / contiene best.pt, spec.json (thresholds por clase), metrics.json (métricas de test e historial por época) y trainingcurve.png. Dashboard y API: https://huggingface.co/spaces/Rengatitos/chimpact-thesis-dashboard
Open weights
cc-by-nc-4.0
Open weights
cc-by-nc-4.0
Open weights
cc-by-nc-4.0
1.4B parameters
Trained weights accompanying a manuscript in preparation. Access is granted on request. Layout: /fold / — one directory per experiment and cross-validation fold. Diffusion experiments store the selected checkpoint plus its config.yaml; GAN baselines store the generator used at inference. Companion outputs and metadata are held separately and shared on request. Contact the repository owner for details.
Access requested at publisher
cc-by-nc-4.0
ONNX export of tardellirs/crivo-150m, retaining the trained ModernBERT encoder, question-type embedding, and option scorer. This is a format conversion and quantization of the original model, with no additional training. Credit for the original architecture, weights, and training belongs to the original authors. The graph reads one sequence per question, gathers hidden states at each [OPT] marker, adds the question-type embedding, and returns one logit per option. Apply softmax across those logits. The tokenizer is included in this repository. The TypeScript loader chooses INT8 on CPU/WASM and FP32 on GPU. Files download from the Hub and run locally; no hosted inference service is required.…
Open weights
cc-by-nc-4.0
7,999 tokens
transformers.js
S
Model · Text classification
Sxy
训练时对文本做了规范化（还原 HTML 转义、链接替换为 http），推理时建议做同样的处理， 见项目中的 triage/preprocess.py。 - 数据：Kaggle NLP with Disaster Tweets 的 train.csv，其中 id ≤ 8696 的 6090 条用于训练，其余 1523 条作为验证集。 - 超参数：BERT 学习率 2e-5，ModernBERT 学习率 5e-5；2 个 epoch；batch size 32；maxseqlength 128； weight decay 0.01；warmup 0.1；bf16；随机种子 42。 同一配置 3 个随机种子的 Macro-F1：BERT 81.63 ± 0.11，ModernBERT 81.70 ± 1.11，两者的差异在种子波动范围内。 - 仅支持英文；比喻性用法（如 "this mixtape is fire"）仍可能误判。 - 两个模型都只在推文（≤ 82 token）上微调，ModernBERT 在长文本上的效果没有经过标注数据验证。 - 模型权重：CC BY-NC 4.0。基于 Kaggle 比赛数据训练，仅限非商业用途；使用时还需遵守该比赛规则中对数据的限制。 - 基础模型：bert-base-uncased、ModernBERT-base（均为 Apache-2.0）。
Open weights
cc-by-nc-4.0
transformers
Experimental ONNX conversion of DreamLite-mobile, for the independent DreamLite Lab Android test application. Original implementation: ByteVisionLab/DreamLite. Download the single shared/ package (~1.57 GB total). dreamliteweights.bin holds one exact shared copy of the large weights. Four small ONNX execution graphs select UNet/decoder and 512/1024 tensor shapes. Keep all five files in the same directory without renaming. Sizes and SHA-256 hashes are listed in shared/manifest.json. APK 0.3 defaults to 512px / 1 step and allows 512/1024 and 1/2/4-step selection in Settings. Download once; changing resolution or step count does not download another weight set. 1024px inference is native…
Open weights
cc-by-nc-4.0
onnxruntime
DuDE is a full-duplex conversation TTS (text-to-speech) model that generates two speakers together from an XML dialogue. It supports turn-taking, backchannels, and overlapping speech, with a separate voice for each speaker. The output is a 24 kHz stereo WAV: speaker A on the left, speaker B on the right. DuDE first self-distills Qwen3-TTS on independent utterances to adapt a shared backbone to two interleaved channels. It is then fine-tuned on approximately 1,922 hours of synchronized Seamless Interaction conversations to learn turn-taking, backchannels, and overlapping speech. The XML frontend resolves turn dependencies into interleaved text inputs. Reference-speaker embeddings condition…
Open weights
cc-by-nc-4.0
1.9B parameters
Paper: arXiv:2610.02638 · Code: github.com/adventists-ai/duplexjev · Research note: teaching a speech-decision LLM to hear stress DuplexJev-32B-Turn that also hears stress and pauses. "I didn't say she took the money" and "I didn't say she took the money" are the same words. This model can tell them apart: it answers which word is stressed, is any word stressed at all, and where the speaker pauses, in Mandarin and English. Spoken QA, gender, emotion and turn-taking stay at the 32B-Turn level. Same size, same serving stack, same prompt format as the other DuplexJev models. DuplexJev reads typed, closed-set decisions from speech as a single-token distribution of the LLM: zero decode steps…
Open weights
cc-by-nc-4.0
vllm
Paper: arXiv:2610.02638 · Code: github.com/adventists-ai/duplexjev · Plugins: pipecat-duplexjev, livekit-plugins-duplexjev The 4B DuplexJev tuned for live end-of-turn: is the user done, right after they go quiet? A turn-taking variant of DuplexJev-4B-Para v1.1, which stays the general 4B. Same recipe as DuplexJev-32B-Para v1.2: the 4B-Para v1.1 mix plus accent data and pause-level turn-taking data, one rank-16 LoRA, 4,000 steps, merged. Qwen3-ASR-0.6B encoder + connector + Qwen3-4B, 4.2 B parameters, bf16, ~10 GB GPU memory. Which 4B to pick. - This one for a voice agent's end-of-turn on a small GPU (LiveKit / Pipecat plugins). - 4B-Para for everything else (stress, non-verbal sounds…
Open weights
cc-by-nc-4.0
4.2B parameters
40,960 tokens
vllm
Paper: arXiv:2610.02638 · Code: github.com/adventists-ai/duplexjev · Research note: teaching a speech-decision LLM to hear stress DuplexJev-4B-Turn that also hears stress and pauses, for smaller GPUs. "I didn't say she took the money" and "I didn't say she took the money" are the same words; this model tells them apart. It answers which word is stressed, is any word stressed at all, where the speaker pauses, and what the stress implies, in Mandarin and English. Trade-off — read before choosing. At 4B, adding stress costs some spoken-QA accuracy: VoiceBench OBQA / MMSU is 42.0 / 38.2 against 46.4 / 41.7 for 4B-Turn, and qa100 / ZJU-ML are 1 point lower. We tried lower stress weights…
Open weights
cc-by-nc-4.0
4.2B parameters
40,960 tokens
vllm
This repository hosts released EasyWAM-Hidden checkpoints built on Wan2.2-TI2V-5B. EasyWAM-Hidden uses a separate Action DiT conditioned one-way on intermediate Video DiT features. Inference retains future-video prediction to provide the hidden visual conditioning. The checkpoints are trained with the EasyWAM codebase. The LoRA checkpoint uses rank 128 (alpha 128) on the backbone Video DiT target modules. Other trainable EasyWAM modules are stored in the same checkpoint, so it must be loaded with the matching lora task recipe. Task success rate (%) under the EasyWAM LIBERO evaluation protocol. All results use stateposition: sequence; higher is better. Models using the same Wan2.2-TI2V-5B…
Open weights
cc-by-nc-4.0
easywam
This repository hosts released EasyWAM-MoT checkpoints built on Wan2.2-TI2V-5B. EasyWAM-MoT uses separate Video DiT and Action DiT experts whose tokens interact through mixed self-attention. At inference time, it predicts actions without generating future video. The checkpoints are trained with the EasyWAM codebase. The LoRA checkpoint uses rank 128 (alpha 128) on the backbone Video DiT target modules. Other trainable EasyWAM modules are stored in the same checkpoint, so it must be loaded with the matching lora task recipe. Task success rate (%) under the EasyWAM LIBERO evaluation protocol. All results use stateposition: sequence; higher is better. Models using the same Wan2.2-TI2V-5B…
Open weights
cc-by-nc-4.0
easywam
This repository hosts released EasyWAM-Unified checkpoints built on Wan2.2-TI2V-5B. EasyWAM-Unified uses a single Video DiT to jointly represent and denoise video and action tokens, coupling both modalities bidirectionally through the same self-attention layers. The checkpoints are trained with the EasyWAM codebase. The LoRA checkpoint uses rank 128 (alpha 128) on the backbone Video DiT target modules. Other trainable EasyWAM modules are stored in the same checkpoint, so it must be loaded with the matching lora task recipe. Task success rate (%) under the EasyWAM LIBERO evaluation protocol. All results use stateposition: sequence; higher is better. Models using the same Wan2.2-TI2V-5B…
Open weights
cc-by-nc-4.0
easywam
Frozen V-JEPA 2.1 tokens on LED, tokenised as k frames (histogram) (16 time bins of 1.25 ms paired into 8 tubelet tokens per cell), scored per event: each event looks up the token at its (time bin / 2, patch) and is judged with its own E15 neighbourhood patch. Routes in this repo: Pixel-AC fuses the per-event scores into one decision per pixel and chunk (integrated/ /; result.json carries the Pixel-AC metrics with the AC-only metrics under ac). Run-name suffixes: nokv = reader bypassed, lb = scalar Local branch on, vitbase = ViT-B/384 tokens. All units: 6,000 training chunks, 4 epochs, official 713 chunks at a fixed threshold of 0.5, no selection on test. Reference on the same protocol: the…
Open weights
cc-by-nc-4.0
pytorch
Modèle ASR complet affiné pour transcrire le fongbe à partir de la parole. Base: facebook/mms-1b-all; tokenizer: Fon v0.4.1. Le bundle préparé contient les poids complets sélectionnés, leur configuration, l’espace de sortie, les artefacts du tokenizer et le runtime autonome. Arrêt définitif demandé par l’utilisateur après 19 246 updates au huitième epoch. Le meilleur checkpoint retenu sur le dev d’origine est le step 18 904. Le plafond prévu était de 54 016 updates et 20 epochs. Ancienne cohorte v07: 113 320 lignes d’entraînement, 243,8558 heures réellement décodées; dev de 1 381 audios. La provenance et les empreintes de cette cohorte restent celles du contrat d’origine. Les poids n’ont…
Open weights
cc-by-nc-4.0
Modèle ASR complet affiné pour transcrire le fongbe à partir de la parole. Base: facebook/mms-1b-all; tokenizer: reference. Le bundle préparé contient les poids complets sélectionnés, leur configuration, l’espace de sortie, les artefacts du tokenizer et le runtime autonome. Arrêt définitif demandé par l’utilisateur après 20 774 updates au huitième epoch. Le meilleur checkpoint retenu sur le dev d’origine est le step 20 000. Le plafond prévu était de 54 016 updates et 20 epochs. Ancienne cohorte v07: 113 320 lignes d’entraînement, 243,8558 heures réellement décodées; dev de 1 381 audios. La provenance et les empreintes de cette cohorte restent celles du contrat d’origine. Les poids n’ont pas…
Open weights
cc-by-nc-4.0
Open weights
cc-by-nc-4.0
ncnn
This is the supervised fine-tuning checkpoint of the Frontis-MA1 family, before OpenMLE reinforcement learning. It is intended for research on machine learning engineering and program transformation. The actual 24,350-row training input is distinct from the 26,259-example public SFT trace collection. Dataset cutoff configuration was 32,768 tokens; this does not establish validated long-context inference performance. No standalone SFT benchmark claim is made by this release. The separately released RL model is FrontisAI/Frontis-MA1-30B. Pin the published commit SHA for reproducibility. This model requires substantial memory; use the dependency versions recorded in the release validation…
Open weights
cc-by-nc-4.0
30.5B parameters
262,144 tokens
transformers
This is the supervised fine-tuning checkpoint of the Frontis-MA1 family, before OpenMLE reinforcement learning. It is intended for research on machine learning engineering and program transformation. The actual 24,350-row training input is distinct from the 26,259-example public SFT trace collection. Dataset cutoff configuration was 32,768 tokens; this does not establish validated long-context inference performance. No standalone SFT benchmark claim is made by this release. The separately released RL model is FrontisAI/Frontis-MA1-35B. Pin the published commit SHA for reproducibility. This model requires substantial memory; use the dependency versions recorded in the release validation…
Open weights
cc-by-nc-4.0
36B parameters
262,144 tokens
transformers
Voice conversion for Asante Twi (twi). An audio-to-audio model that converts speech into the Zephyr voice while preserving the language. This checkpoint is dedicated to Asante Twi: it was fine-tuned only on Asante Twi (twi) clips from starting from the zero-shot seed-uvit-whisper-small-wavenet base. Use it to convert audio in Asante Twi. For other languages — or as a cross-lingual fallback — use ghanaopenai/ghana-vc. The library auto-selects this model when the dataset/config name matches.
Open weights
cc-by-nc-4.0
Voice conversion for Avatime (avn). An audio-to-audio model that converts speech into the Zephyr voice while preserving the language. This checkpoint is dedicated to Avatime: it was fine-tuned only on Avatime (avn) clips from starting from the zero-shot seed-uvit-whisper-small-wavenet base. Use it to convert audio in Avatime. For other languages — or as a cross-lingual fallback — use ghanaopenai/ghana-vc. The library auto-selects this model when the dataset/config name matches.
Open weights
cc-by-nc-4.0
This repository contains the complete GRADE model, paper ablations, and baselines: 19 model checkpoint files plus one auxiliary TAESD weight file (20.safetensors files in total). It also includes the inference and metric code and the saved camera-ready quantitative results. The full GRADE model uses the radar, diffusion, and ControlNet checkpoints in checkpoints/grade/. - E1 — saved-result reproduction: regenerate the quantitative tables and figures from the supplied camera-ready CSVs and auxiliary evaluation data. This requires no inference or dataset download. - E2 — local reproduction: run inference on the released evaluation data, compute 2D/3D metrics, then regenerate the quantitative…
Open weights
cc-by-nc-4.0
pytorch
Open weights
cc-by-nc-4.0
26.9B parameters
transformers
weighted/imatrix quants of https://huggingface.co/brainnxdomain/Hemmingway-1-Heretic-MTP-V3-Final For a convenient overview and download list, visit our model page for this model. static quants are available at https://huggingface.co/mradermacher/Hemmingway-1-Heretic-MTP-V3-Final-GGUF If you are unsure how to use GGUF files, refer to one of TheBloke's READMEs for more details, including on how to concatenate multi-part files. (sorted by size, not necessarily quality. IQ-quants are often preferable over similar sized non-IQ quants) Here is a handy graph by ikawrakow comparing some lower-quality quant And here are Artefact2's thoughts on the matter…
Open weights
cc-by-nc-4.0
transformers
A LoRA adapter that teaches Altworld/Hemmingway-1 (a 27B Qwen3.5 GDN model) to reason in the Tree-of-Thought "reasoning-steps → final justified answer" style of the codex5.5ToT dataset. - Altworld/Hemmingway-1 — a 27B-parameter Qwen35ForCausalLM (Qwen3.5 / GDN architecture): 48 linear-attention layers + 16 full-attention layers, 64 layers total, plus an MTP (multi-token-prediction) head. BF16 ≈ 54.6 GB. - AmelieSchreiber/codex5.5ToT — 3,890 high-quality reasoning records across 10 domains (math, physics, CS/algorithms, chemistry, …). Each record is a [system, user, assistant] chat where the assistant response is a reasoning tree: it explores branches, compares alternatives, and ends with a…
Open weights
cc-by-nc-4.0
transformers
weighted/imatrix quants of https://huggingface.co/Hcompany/Holo4-27B For a convenient overview and download list, visit our model page for this model. static quants are available at https://huggingface.co/mradermacher/Holo4-27B-GGUF This is a vision model - mmproj files (if any) will be in the static repository. If you are unsure how to use GGUF files, refer to one of TheBloke's READMEs for more details, including on how to concatenate multi-part files. (sorted by size, not necessarily quality. IQ-quants are often preferable over similar sized non-IQ quants) Here is a handy graph by ikawrakow comparing some lower-quality quant And here are Artefact2's thoughts on the matter…
Open weights
cc-by-nc-4.0
transformers
Open weights
cc-by-nc-4.0
Open weights
cc-by-nc-4.0
752M parameters
transformers
ONNX conversion of jinaai/jina-code-embeddings-0.5b (Qwen2.5-Coder-0.5B finetune, 494M params), laid out like weights under onnx/, tokenizer + config at root, transformers.js compatible. Dynamic int8 quantization was tried (ORT quantizedynamic, QInt8, per-channel + reduce-range, with and without quantpreprocess shape inference) and scored drift ≈ 0.145 — two orders of magnitude over the 1e-3 bar. This GQA-decoder + RMSNorm + last-token-pooling architecture is far more quant-sensitive than BERT-style encoders where int8 dynamic works. Per the sentence-transformers team's guidance, the recommended efficiency path for embeddings is embedding quantization (quantize output vectors to int8/binary…
Open weights
cc-by-nc-4.0
32,768 tokens
transformers.js
Open weights
cc-by-nc-4.0
Fine-tune of convaiinnovations/laya (subfolder="multilingual") for Thai typed decisions. The model does not write text. One forward pass stamps choice, score, and noul (yes/no) answers and returns a probability for each. The weights in this repo are epoch02. The config is the calibrated one: same weights, softer confidence. A single temperature does not change the chosen label. This release is published by the Hugging Face organization that owns the repo. No individual author is named. The model is a fine-tune, not a new base: Comparison numbers for iapp/OpenThai-SystemOne in the speed table were measured locally from those weights. iApp is not an author of this fine-tune. Route a Thai…
Open weights
cc-by-nc-4.0
322M parameters
Every weight in this model was produced by this project. There is no frompretrained, no base checkpoint, and no adapter: the tokenizer, the architecture code, the data pipeline and the training loop are all bespoke, and training started from a random initialisation. Trained on a single RTX 5070 Ti (16 GB). Written by hand — no transformers model classes were used to define it. 309,705,984 transformer blocks. Bits per byte is the headline metric. This model's vocabulary (32,768) differs from GPT-2's (50,257), and cross-entropy per token is not comparable across tokenizers. BPB divides that out and is directly comparable. Validation loss is measured on a held-out slice of the training corpus…
Open weights
cc-by-nc-4.0
transformers
A small multilingual decision model for classification, with a focus on Korean. Given a text, a question, and a set of criteria, Malkuth returns probabilities over the possible choices instead of generating text. Malkuth-2B is a LoRA adapter trained on Kev and built on a frozen empero-ai/Qwen3.8-2B-Distill. It has 17.9M trainable parameters and a 65 MB adapter. - choice: select one of up to 255 options - noul: yes or no - score: select an ordered level Multiple questions can be evaluated against the same input. Questions are evaluated independently, so they cannot see each other's answers. Probabilities are temperature-calibrated on a held-out validation split. Malkuth is served through Kev…
Open weights
cc-by-nc-4.0
peft
A small multilingual decision model for classification, with a focus on Korean. Given a text, a question, and a set of criteria, Malkuth returns probabilities over the possible choices instead of generating text. Malkuth-4B is a LoRA adapter trained on Kev and built on a frozen Qwen/Qwen3.5-4B-Base. It has 33.8M trainable parameters and a 124 MB adapter. - choice: select one of up to 255 options - noul: yes or no - score: select an ordered level Multiple questions can be evaluated against the same input. Questions are evaluated independently, so they cannot see each other's answers. Probabilities are temperature-calibrated on a held-out validation split. Malkuth is served through Kev and…
Open weights
cc-by-nc-4.0
peft
Open weights
cc-by-nc-4.0
9B parameters
mlx
Open weights
cc-by-nc-4.0
onnx
基于明前奶绿歌声数据集 v2（1,057 段 / 50.3 分钟，全量去混响）训练的 RVC v2 声音转换模型：48 kHz、HiFi-GAN 声码器、Titan 底模微调 200 轮（26,800 步）、RMVPE 音高提取。 - mingqiannailvv2final.pth：推理用模型权重（RVC 格式，内含生成器与配置） - mingqiannailvv2.index：音色检索索引 1. 两个文件放入 logs/mingqiannailv/ 目录（没有就新建） 3. 参数建议：音高调整 0、索引率 0.3~0.5、protect 0.33
Open weights
cc-by-nc-4.0
MoSE3 is a feed-forward model that predicts world-space SE(3) motion for every pixel of a monocular RGB video, generalizing across rigid, articulated and deformable objects. The code repository has the full inference and visualization scripts. The weights are released under CC BY-NC 4.0 (non-commercial). They include the weights of π³, which are released under the same license.
Open weights
cc-by-nc-4.0
1.5B parameters
A small causal Qwen3 model that predicts 32 MOSS v2 audio-codebook indices from the 12 indices of the current frame and up to ten previously generated HQ frames. It is trained from scratch on German, English, Spanish and French speech. This is an experimental enhancement model. Validation cross entropy measures token prediction under teacher forcing. It does not establish an improvement in perceived audio quality. Use the generated-history listening evaluation to judge speech content, speaker identity, artifacts and long-utterance stability. Root inference weights: step 64,699, selected by held-out extra-20-codebook cross entropy 5.818123. - Step 12,940: 100 originals / 200 input cases…
Open weights
cc-by-nc-4.0
pytorch
Open weights
cc-by-nc-4.0
MLX port of Agisight/nllb-rus-tyv-v3-dict16.5k, a Russian ↔ Tuvan translation model fine-tuned from Meta's NLLB-200-distilled-600M with a vocabulary pruned to 16.5K tokens. Runs fully offline on Apple Silicon (Mac, iPhone, iPad) via MLX. mlx-lm does not support encoder-decoder models, so the M2M100 architecture is re-implemented in nllbmlx.py (included), with a KV cache for fast greedy decoding. The float32 port reproduces the PyTorch model exactly: 100/100 identical translations on a spot check and identical chrF++/BLEU on the full clean test set. Clean test set: held-out test split of Agisight/tyv-rus-200k used for NLLB v3 (shuffle(seed=42), pairs 2000–4000), with pairs whose Tuvan or…
Open weights
cc-by-nc-4.0
1,024 tokens
mlx
MLX port of Agisight/nllb-rus-tyv-v3-dict16.5k, a Russian ↔ Tuvan translation model fine-tuned from Meta's NLLB-200-distilled-600M with a vocabulary pruned to 16.5K tokens. Runs fully offline on Apple Silicon (Mac, iPhone, iPad) via MLX. mlx-lm does not support encoder-decoder models, so the M2M100 architecture is re-implemented in nllbmlx.py (included), with a KV cache for fast greedy decoding. The float32 port reproduces the PyTorch model exactly: 100/100 identical translations on a spot check and identical chrF++/BLEU on the full clean test set. Clean test set: held-out test split of Agisight/tyv-rus-200k used for NLLB v3 (shuffle(seed=42), pairs 2000–4000), with pairs whose Tuvan or…
Open weights
cc-by-nc-4.0
1,024 tokens
mlx
Noah-Decide is a 0.6B-parameter decision model from Beinex AI. You give it a situation (the state), a question and a set of options. It returns a calibrated probability for each option. In certified mode, the model returns the set of options it cannot rule out. When that set is not exactly one option, it escalates to a human. Licence: CC BY-NC 4.0. It is free for research and other non-commercial use; commercial use needs a licence from Beinex. The base model, Qwen3-0.6B, is Apache-2.0. - Base model. Qwen3-0.6B, fully fine-tuned and then blended back towards the base weights (60% fine-tuned, 40% base). The blend ratio was chosen on held-out decision types. - Option order does not matter.…
Open weights
cc-by-nc-4.0
597M parameters
Re-exported llmdecoder sub-model for OmniVoice Web, a fully in-browser (WebGPU) TypeScript runtime for k2-fsa/OmniVoice. The public onnx-community/OmniVoice-Onnx llmdecoder uses causal attention (GroupQueryAttention + 2-D causal mask). OmniVoice is a bidirectional masked-diffusion LM, so with causal attention the model ignores the text prompt and emits generic speech. This repo hosts the Qwen3-0.6B core re-exported with a 4-D all-zero (non-causal) attention mask, which restores exact text conditioning. All three reproduce the PyTorch reference with WER 0 (round-trip Whisper) on the evaluation sentence. The embeddings/heads/tokenizer/codec sub-models remain on tools/exportllm.py --dtype…
Open weights
cc-by-nc-4.0
N
Model · Text to speech
Non
Open weights
cc-by-nc-4.0
Bienvenido al repositorio de checkpoints de entrenamiento (.ckpt) del modelo de texto a voz Capybara. Este modelo ha sido entrenado utilizando Piper. Capybara es un modelo de texto a voz para español de México. Este repositorio contiene sus archivos de checkpoint (.ckpt) generados durante el entrenamiento con Piper. - Continuación del entrenamiento: Los checkpoints permiten reanudar el proceso de afinación con Piper-Neo. Esta licencia establece las condiciones bajo las cuales se permite el uso de este modelo de voz. Al utilizar este modelo de voz, aceptas cumplir con estas condiciones. - Crédito: Dar crédito al creador del modelo es opcional, pero se agradece. - Prohibido el uso en…
Access requested at publisher
cc-by-nc-4.0
piper
Creator, Sole Author & Lead Architect: Alan Canto Architecture: ModernBERT-large (System 1 Non-Autoregressive Decision Engine) Hardware Runtime: Trained, fine-tuned, and certified locally on NVIDIA GeForce RTX 3050 Attribution: Conceived, engineered, and published by Alan Canto. All rights reserved. ProKope 421M (named after the classical Stoic concept of disciplined, measured progress and operational mastery) is a high-speed, non-autoregressive System 1 decision model engineered by Alan Canto. Unlike autoregressive language models (which incur significant token latency, JSON parsing errors, and hallucination loops), ProKope 421M processes structured input states in a single forward pass…
Access requested at publisher
cc-by-nc-4.0
421M parameters
transformers
Access requested at publisher
cc-by-nc-4.0
peft
C
Model · Speech recognition
Cstr
GGUF conversion of the speech-to-text subset of KRAFTON/Raon-Speech-9B for CrispASR (--backend raon-speech, or -m raon-speech to auto-download Q4K). The talker, code predictor, Mimi codec, output adaptor and speaker encoder serve speech output only, so they are not converted. CrispASR reproduces RaonPipeline.stt's front end: - the input is resampled 16k → 24k, as the processor does; - the audio is cut into 8 s chunks, each resampled 24k → 16k (a port of torchaudio's sinc resampler); - each chunk gets its own log-mel (per-chunk max) and is encoded on its own; - the 13 Hz encoder output is truncated to the 12.5 Hz frame grid. The prompt is "Transcribe the audio into text"; --ask "..."…
Open weights
cc-by-nc-4.0
Open weights
cc-by-nc-4.0
Open weights
cc-by-nc-4.0
