Open-Weight Models Under Creative Commons Attribution 4.0

Model

paper-embodied-ai

Bartosz G. Kowalczyk

This repository contains a working research note about Embodied AI. It organizes motivation, related work, a falsifiable hypothesis, and an evaluation plan. It is not presented as a completed paper or a release of trained models. - the scope of the research question and likely confounders - a proposed comparison with matched baselines - concrete evaluation context such as task-appropriate public benchmarks named in the main note - reproducibility checks, failure modes, and open questions - topic-relevant references Start with analysis.md for the full note. Sections labeled as plans or hypotheses should not be interpreted as experimental results. If results are added later, they should…

GGUF files of moondream/parakeet-redux for transcribe.cpp, keeping the model ternary: the encoder weights stay {−1, 0, +1} codes with one scale per 128 weights (1.75 bits/weight) and run on dedicated ternary kernels. Nothing is dequantized on disk. Parakeet Redux is Moondream's ternary sibling of parakeet-ultra, itself a post-training of NVIDIA's parakeet-tdt-0.6b-v3: FastConformer encoder + TDT transducer decoder, 25 European languages, 16 kHz mono input, punctuated and cased output with optional token/word timestamps. Not a streaming model; does not translate. - Format details, kernels, and how to reproduce: QUANTIZATION.md The ternary encoder weights (604 M, 1.75 bits/weight) are…

An INT8 ONNX build of nvidia/parakeet-tdtctc-0.6b-ja (TDT branch). It is the Japanese ASR model used by euhadra. Only the encoder's constant-weight MatMuls are quantised. On FLEURS-ja this keeps the CER of the FP32 export. A build that quantises every op loses 35 %. On the 321 utterances, this build's hypothesis is identical to FP32's on 257. Of the rest, 21 score worse and 23 better. Measurements and discussion: euhadra#173. The layout is the one parakeet-rs-featsize loads: it looks for encoder-model.onnx, then encoder-model.int8.onnx. Don't put FP32 encoder files in the same directory. 1. Original model: nvidia/parakeet-tdtctc-0.6b-ja, revision 44edb27eea9317daf89333e75eb830db4b1cc298…

GGUF quantizations of moondream/parakeet-ultra for transcribe.cpp. Parakeet Ultra is Moondream's full-precision post-training of NVIDIA's parakeet-tdt-0.6b-v3: the same FastConformer encoder (24 layers, dmodel 1024) and TDT transducer decoder, the same 25-language 8192-piece vocabulary, with better accuracy than v3 across English, European languages, noisy audio and long-form speech (upstream figures below). It takes 16 kHz mono audio and returns a punctuated, cased transcript with optional token/word timestamps. It is not a streaming model and does not translate. - How these files were made and how to reproduce them: QUANTIZATION.md Every file is at most the size of the parent checkpoint…

Model · Speech recognition

parakeet-ultra-glade

Josh Leverette

Glade is a native Swift library for on-device speech inference on Apple silicon. This repository supplies converted model assets and configuration for Glade's model-specific host runtime. Glade's runtime source is currently private. Glade conversion of Moondream Parakeet Ultra, a post-trained NVIDIA Parakeet TDT 0.6B V3 model. Source revision: 73175eb7aeb0d82f1e2a6b53b3aabc10a90bcd0b. The weights retain the upstream CC-BY-4.0 license. See LICENSE and NOTICE. This conversion is not endorsed by Moondream or NVIDIA. The repository contains one self-contained bundle. It preserves the model's native speech head to find pauses, selecting chunks of at most 30 seconds. All audio is retained…

Model · Speech recognition

parakeet-v2-glade

Josh Leverette

Glade is a native Swift library for on-device speech inference on Apple silicon. This repository supplies converted model assets and configuration for Glade's model-specific host runtime. Glade's runtime source is currently private. Glade conversion of NVIDIA Parakeet V2, pinned to ae9ad07059c7c739ffaf932226a8fe64ae2620b0. Converted weights retain the upstream Creative Commons Attribution 4.0 license. See LICENSE and NOTICE. This conversion is not endorsed by NVIDIA. - fast/: full finite Fourier relative attention, 192 encoder positions, fixed 15-second chunks, and an FP16 ANE decoder batching up to 128 independent chunks. - quality/: full sinusoidal relative attention, shared weights for…

Model · Speech recognition

parakeet-v3-glade

Josh Leverette

Glade is a native Swift library for on-device speech inference on Apple silicon. This repository supplies converted model assets and configuration for Glade's model-specific host runtime. Glade's runtime source is currently private. Glade conversion of NVIDIA Parakeet V3, pinned to 541d1f99c6b0c3cd0b11a95167540bb8edefd82b. Converted weights retain the upstream Creative Commons Attribution 4.0 license. See LICENSE and NOTICE. This conversion is not endorsed by NVIDIA. - fast/: full finite Fourier relative attention, 192 encoder positions, fixed 15-second chunks, and an FP16 ANE decoder batching up to 128 independent chunks. - quality/: full sinusoidal relative attention, shared weights for…

Model · Speech recognition

phonon-2-glade-experimental

Josh Leverette

Glade is a native Swift library for on-device speech inference on Apple silicon. This repository supplies converted model assets and configuration for Glade's model-specific host runtime. Glade's runtime source is currently private. Three experimental conversions of Fermion Research's Phonon-2 for Apple silicon and Core AI APIs on macOS/iOS 27. LUT6 is recommended for speed. The other variants expose the tradeoff between download size, working memory, and inference time. Phonon-2 is an encoder-heavy FastConformer/TDT model derived from NVIDIA Parakeet V3. The conversions preserve its trained five-value encoder weights and original INT6 decoder values. They use FP16 activations and the…

Clone of Pocket TTS 6l model only. Use of our model must comply with all applicable laws and regulations and must not result in, involve, or facilitate any illegal, harmful, deceptive, fraudulent, or unauthorized activity. Prohibited uses include, without limitation, voice impersonation or cloning without explicit and lawful consent; misinformation, disinformation, or deception (including fake news, fraudulent calls, or presenting generated content as genuine recordings of real people or events); and the generation of unlawful, harmful, libelous, abusive, harassing, discriminatory, hateful, or privacy-invasive content. We disclaim all liability for any non-compliant use. - Kyutai for the…

Model · Text to speech

pocket-tts-glade

Josh Leverette

Specialized conversion of Kyutai Pocket TTS. This profile supports English text with the predefined alba voice and returns mono 24 kHz Float PCM, including incremental audio. 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 approximately 117.2 MB bundle uses W8A16 language…

Model · Text to speech

pocket-tts-onnx

Nitesh Sharma

Optimized ONNX exports of Pocket TTS for high-speed, memory-efficient text-to-speech and zero-shot voice cloning on CPU, mobile, and edge runtimes. Mirrored and validated from KevinAHM/pocket-tts-onnx at commit 58a6d00cf13d239b6748cb0769f35c580a8f606c. This repository provides comprehensive support for both compact 6-layer models and high-capacity 24-layer models: - 6-Layer Compact Models: english2026-04, german, italian, portuguese, spanish - Ideal for lightweight CPU inference, mobile devices, and low-latency streaming. - ~302 MB FP32 / ~76 MB INT8 Flow LM. - 24-Layer Capacity Models: french24l, german24l, italian24l, portuguese24l, spanish24l - High-fidelity modeling for complex prosody…

Model

prompt-engineering89

Nathan DAVIS

A structured set of research notes on Prompt Engineering, with concrete evaluation references and open questions. Plans and hypotheses are kept separate from completed results. - the scope of the research question and likely confounders - a proposed comparison with matched baselines - concrete evaluation context such as task-appropriate public benchmarks named in the main note - reproducibility checks, failure modes, and open questions - topic-relevant references Start with review.md for the full note. Sections labeled as plans or hypotheses should not be interpreted as experimental results. If results are added later, they should include dataset versions, commands, seeds, hardware, and raw…

Model

Proteus-Structures

Inserloft Research

Computationally generated and predicted molecular structures from Project Proteus, the biological discovery research program of Inserloft Research. This repository contains structural representations associated with computational protein-design experiments conducted within Project Proteus. Proteus Structures is a structural repository for molecular candidates generated or investigated through the Proteus research program. The structures in this repository are intended to support: Unless explicitly stated otherwise, structures contained here are computational models and should not be interpreted as experimentally determined structures. Additional structures may be added as Proteus research…

Model

Qwen3.6-27B-A3B-Coder-ROCMFP4_FAST

Alston Albarado

This model was dynamically quantized and generated using the ROCmFPX My Repo Space To run this model, you must use the specialized ROCmFPX fork of llama.cpp by charlie12345. Standard llama.cpp releases do not currently support ROCmFPX quantization formats. This model is distributed under the Creative Commons Attribution 4.0 International (CC BY 4.0) license. Please adhere to the original license constraints of the base model and provide appropriate attribution when sharing.

Model

reading-self-supervised-2023

Lucia Fernandez

This repository contains a working research note about Self Supervised. It organizes motivation, related work, a falsifiable hypothesis, and an evaluation plan. It is not presented as a completed paper or a release of trained models. - the scope of the research question and likely confounders - a proposed comparison with matched baselines - concrete evaluation context such as task-appropriate public benchmarks named in the main note - reproducibility checks, failure modes, and open questions - topic-relevant references Start with reading.md for the full note. Sections labeled as plans or hypotheses should not be interpreted as experimental results. If results are added later, they should…

Binary classifier (NORMAL vs PNEUMONIA): torchvision ResNet50 (ImageNet IMAGENET1KV2 weights) fully fine-tuned on Purino/chest-xray-pneumonia-256-fullframe-gray. Research/educational use only, not a diagnostic tool. Confusion matrix (rows = true, columns = predicted): Validation (15% stratified hold-out of train, 783 images) at the selected epoch 8: accuracy 0.9885, recall 0.9880, macro-F1 0.9851, AUC 0.9983. - The dataset has no patient-level split or deduplication, so validation (and possibly test) images may come from patients seen in training; scores can be optimistic. - Trained on one pediatric dataset (Kermany et al.); it will likely not generalize to other scanners, populations or…

Pretrained checkpoints for REVA-PO: Stabilizing Reinforcement Learning for Chest X-ray Report Generation, accepted to ECCV 2026. REVA-PO is a reinforcement learning framework for chest X-ray report generation. It combines: - Response-Weighted Regularization, which adjusts response-level KL regularization using the response advantage and reference-policy entropy. - Validation-Anchored Policy Reset, which periodically synchronizes the current and reference policies with the best validation checkpoint. - A three-stage training pipeline consisting of warm-up training, classifier-guided supervised fine-tuning, and reinforcement learning. This repository contains the PyTorch and DeepSpeed…

Model

robotics-vision-language-tutorial-2023

A. Ward

Reading notes and an experiment sketch for Robotics Vision Language. The repository emphasizes what still needs to be tested instead of manufacturing scores or release claims. - the scope of the research question and likely confounders - a proposed comparison with matched baselines - concrete evaluation context such as task-appropriate public benchmarks named in the main note - reproducibility checks, failure modes, and open questions - topic-relevant references Start with notes.md for the full note. Sections labeled as plans or hypotheses should not be interpreted as experimental results. If results are added later, they should include dataset versions, commands, seeds, hardware, and raw…

Model

self-supervised

Joaquín A. González

A structured set of research notes on Self Supervised, with concrete evaluation references and open questions. Plans and hypotheses are kept separate from completed results. - the scope of the research question and likely confounders - a proposed comparison with matched baselines - concrete evaluation context such as task-appropriate public benchmarks named in the main note - reproducibility checks, failure modes, and open questions - topic-relevant references Start with summary.md for the full note. Sections labeled as plans or hypotheses should not be interpreted as experimental results. If results are added later, they should include dataset versions, commands, seeds, hardware, and raw…

Model

semg-intent-restoration-pipeline

Jorge Ortiz Ceballos

Three chained PyTorch models (plus one exploratory variant) implementing the Each 200 ms / 400-sample sEMG window (2000 Hz, single channel, MyoWare-compatible) is routed through: intent detection → quality assessment → (optional) restoration. Operating thresholds (τintent = 0.40, τquality = 0.50) were selected via grid search on the validation split only, and evaluated once on a held-out test split. If you want to reproduce the pipeline exactly as described in Section 4.7 and Figure 1 of the paper, use only the first three files. The task-aware variant is a separate experiment reported transparently as a limitation, not a replacement for restorerbest.pt — see the "Downstream Motor-Intent…

Model · Text generation

semsimula-ladder-live-owt-d384-l2-attention-potential-rfqk-norc

Dimitar Gueorguiev

Settled is the mean of the last three evals (54.20, 54.39, 54.22); best 52.50 at step 31,000, final 54.22. Consecutive evals near the end differ by about ±1 PPL. - The field alone is a real but weaker route. It lowers the conservative-only model's perplexity by 6.0%; the register path alone lowers it by 8.0%. - Beside the registers it adds almost nothing: G3′ is 0.4% better than G2. The two routes largely substitute for each other. This model sees exactly 532,480,000 training tokens — 32,500 steps × 32 × 512 = 532,480,000 tokens — the same budget, corpus, tokenizer and validation batches as every other model in the table. No model is trained longer than another. - The base is F3.1: the…

Model · Text generation

semsimula-ladder-owt-d384-l2-attention

Dimitar Gueorguiev

One arm of the SemSimula mechanism ladder — OpenWebText, d=384, L=2: five models that differ by exactly one mechanism each, so that every gap in the ordering prices one component by construction rather than by attribution. Italic entries are pre-registered predictions, recorded before the run. The ordering is itself the prediction; any inversion is a result. Every arm above sees exactly 532,480,000 training tokens — 32,500 steps × 32 × 512 = 532,480,000 tokens — on the same corpus, tokenizer and validation batches. No arm is trained longer than another, so a gap in the table is a gap between mechanisms and never a difference in budget. The models below integrate a damped mechanical system…

Model · Text generation

semsimula-ladder-owt-d384-l2-attention-potential

Dimitar Gueorguiev

One arm of the SemSimula mechanism ladder — OpenWebText, d=384, L=2: five models that differ by exactly one mechanism each, so that every gap in the ordering prices one component by construction rather than by attribution. Italic entries are pre-registered predictions, recorded before the run. The ordering is itself the prediction; any inversion is a result. Every arm above sees exactly 532,480,000 training tokens — 32,500 steps × 32 × 512 = 532,480,000 tokens — on the same corpus, tokenizer and validation batches. No arm is trained longer than another, so a gap in the table is a gap between mechanisms and never a difference in budget. The models below integrate a damped mechanical system…

Model · Text generation

semsimula-ladder-owt-d384-l2-gpt2-matched

Dimitar Gueorguiev

One arm of the SemSimula mechanism ladder — OpenWebText, d=384, L=2: five models that differ by exactly one mechanism each, so that every gap in the ordering prices one component by construction rather than by attribution. Italic entries are pre-registered predictions, recorded before the run. The ordering is itself the prediction; any inversion is a result. Every arm above sees exactly 532,480,000 training tokens — 32,500 steps × 32 × 512 = 532,480,000 tokens — on the same corpus, tokenizer and validation batches. No arm is trained longer than another, so a gap in the table is a gap between mechanisms and never a difference in budget. The models below integrate a damped mechanical system…

Model · Text generation

semsimula-ladder-owt-d384-l2-none

Dimitar Gueorguiev

One arm of the SemSimula mechanism ladder — OpenWebText, d=384, L=2: five models that differ by exactly one mechanism each, so that every gap in the ordering prices one component by construction rather than by attribution. Italic entries are pre-registered predictions, recorded before the run. The ordering is itself the prediction; any inversion is a result. Every arm above sees exactly 532,480,000 training tokens — 32,500 steps × 32 × 512 = 532,480,000 tokens — on the same corpus, tokenizer and validation batches. No arm is trained longer than another, so a gap in the table is a gap between mechanisms and never a difference in budget. The models below integrate a damped mechanical system…

Model · Text generation

semsimula-ladder-owt-d384-l2-none-norc

Dimitar Gueorguiev

One arm of the SemSimula mechanism ladder — OpenWebText, d=384, L=2: five models that differ by exactly one mechanism each, so that every gap in the ordering prices one component by construction rather than by attribution. Italic entries are pre-registered predictions, recorded before the run. The ordering is itself the prediction; any inversion is a result. Every arm above sees exactly 532,480,000 training tokens — 32,500 steps × 32 × 512 = 532,480,000 tokens — on the same corpus, tokenizer and validation batches. No arm is trained longer than another, so a gap in the table is a gap between mechanisms and never a difference in budget. The models below integrate a damped mechanical system…

Model

study-lightweight-multimodal

Macolombo

This is an exploratory note for Lightweight Multimodal. It records the intended comparison, likely confounders, and reproducibility requirements before any benchmark result is reported. - the scope of the research question and likely confounders - a proposed comparison with matched baselines - concrete evaluation context such as task-appropriate public benchmarks named in the main note - reproducibility checks, failure modes, and open questions - topic-relevant references Start with notes.md for the full note. Sections labeled as plans or hypotheses should not be interpreted as experimental results. If results are added later, they should include dataset versions, commands, seeds, hardware…

Model

study-lightweight-multimodal

M. Garcia

A structured set of research notes on Lightweight Multimodal, with concrete evaluation references and open questions. Plans and hypotheses are kept separate from completed results. - the scope of the research question and likely confounders - a proposed comparison with matched baselines - concrete evaluation context such as task-appropriate public benchmarks named in the main note - reproducibility checks, failure modes, and open questions - topic-relevant references Start with review.md for the full note. Sections labeled as plans or hypotheses should not be interpreted as experimental results. If results are added later, they should include dataset versions, commands, seeds, hardware, and…

Model

test-scvi

Scvi Tools

ScVI is a variational inference model for single-cell RNA-seq data that can learn an underlying latent space, integrate technical batches and impute dropouts. The learned low-dimensional latent representation of the data can be used for visualization and clustering. scVI takes as input a scRNA-seq gene expression matrix with cells and genes. We provide an extensive user guide. - See our original manuscript for further details of the model: - See our manuscript on scvi-hub how to leverage pre-trained models. This model can be used for fine tuning on new data using our Arches framework: scVI model trained on synthetic IID data and uploaded with the full training data. We provide here key…

Model

test-scvi-minified

Scvi Tools

ScVI is a variational inference model for single-cell RNA-seq data that can learn an underlying latent space, integrate technical batches and impute dropouts. The learned low-dimensional latent representation of the data can be used for visualization and clustering. scVI takes as input a scRNA-seq gene expression matrix with cells and genes. We provide an extensive user guide. - See our original manuscript for further details of the model: - See our manuscript on scvi-hub how to leverage pre-trained models. This model can be used for fine tuning on new data using our Arches framework: scVI model trained on synthetic IID data and uploaded with the minified data. We provide here key…

Model

test-scvi-no-anndata

Scvi Tools

ScVI is a variational inference model for single-cell RNA-seq data that can learn an underlying latent space, integrate technical batches and impute dropouts. The learned low-dimensional latent representation of the data can be used for visualization and clustering. scVI takes as input a scRNA-seq gene expression matrix with cells and genes. We provide an extensive user guide. - See our original manuscript for further details of the model: - See our manuscript on scvi-hub how to leverage pre-trained models. This model can be used for fine tuning on new data using our Arches framework: scVI model trained on synthetic IID data and uploaded with no data. We provide here key performance metrics…

Model

text-image-retrieval-reading-2023

Saud Alharbi

A structured set of research notes on Text Image Retrieval, with concrete evaluation references and open questions. Plans and hypotheses are kept separate from completed results. - the scope of the research question and likely confounders - a proposed comparison with matched baselines - concrete evaluation context such as Flickr30k and MS COCO Captions - reproducibility checks, failure modes, and open questions - topic-relevant references Start with papernotes.md for the full note. Sections labeled as plans or hypotheses should not be interpreted as experimental results. If results are added later, they should include dataset versions, commands, seeds, hardware, and raw logs. The note is…

Model

text-image-retrieval-small-2024

Freja Andersson

A structured set of research notes on Text Image Retrieval, with concrete evaluation references and open questions. Plans and hypotheses are kept separate from completed results. - the scope of the research question and likely confounders - a proposed comparison with matched baselines - concrete evaluation context such as Flickr30k and MS COCO Captions - reproducibility checks, failure modes, and open questions - topic-relevant references Start with papernotes.md for the full note. Sections labeled as plans or hypotheses should not be interpreted as experimental results. If results are added later, they should include dataset versions, commands, seeds, hardware, and raw logs. The note is…

Model · Image segmentation

toxoplasma-plaque-segmentation-cpsam-r5

Einar Olafsson

Segments Toxoplasma gondii plaques in crystal-violet plaque assays. Round 5 is the first model trained across all four domains at once: whole-well microscope scans (two fixations), wells cropped from published figures, and phone-camera wells. Built for 488 curated fields, 27,582 plaques, deduplicated by pixel content. Split by FIGURE (literature) and by PLATE (microscope) so no figure or plate straddles; malnio's own held-out 20 are forced into test. Read the incumbent columns with care. r3 and r4 trained on every microscope well and on most of the literature, so their scores there measure memorisation. qc/summary.json also reports every model on the subset neither incumbent ever trained on…

Model

toy-multimodal-generation

Alvaro Martin

This is an exploratory note for Multimodal Generation. It records the intended comparison, likely confounders, and reproducibility requirements before any benchmark result is reported. - the scope of the research question and likely confounders - a proposed comparison with matched baselines - concrete evaluation context such as task-appropriate public benchmarks named in the main note - reproducibility checks, failure modes, and open questions - topic-relevant references Start with summary.md for the full note. Sections labeled as plans or hypotheses should not be interpreted as experimental results. If results are added later, they should include dataset versions, commands, seeds…

Model

vision-language-pretraining-study-2024

S. Kapoor

Reading notes and an experiment sketch for Vision Language Pretraining. The repository emphasizes what still needs to be tested instead of manufacturing scores or release claims. - the scope of the research question and likely confounders - a proposed comparison with matched baselines - concrete evaluation context such as task-appropriate public benchmarks named in the main note - reproducibility checks, failure modes, and open questions - topic-relevant references Start with summary.md for the full note. Sections labeled as plans or hypotheses should not be interpreted as experimental results. If results are added later, they should include dataset versions, commands, seeds, hardware, and…

Model

vision-language-pretraining-v1

Joaquín A. González

This is an exploratory note for Vision Language Pretraining. It records the intended comparison, likely confounders, and reproducibility requirements before any benchmark result is reported. - the scope of the research question and likely confounders - a proposed comparison with matched baselines - concrete evaluation context such as task-appropriate public benchmarks named in the main note - reproducibility checks, failure modes, and open questions - topic-relevant references Start with papernotes.md for the full note. Sections labeled as plans or hypotheses should not be interpreted as experimental results. If results are added later, they should include dataset versions, commands, seeds…

Model · Visual question answering

visual-question-answering-study

Samuel Moore

This is an exploratory note for Visual Question Answering. It records the intended comparison, likely confounders, and reproducibility requirements before any benchmark result is reported. - the scope of the research question and likely confounders - a proposed comparison with matched baselines - concrete evaluation context such as VQAv2, GQA, and OK-VQA - reproducibility checks, failure modes, and open questions - topic-relevant references Start with review.md for the full note. Sections labeled as plans or hypotheses should not be interpreted as experimental results. If results are added later, they should include dataset versions, commands, seeds, hardware, and raw logs. The note is…

Model · Visual question answering

visual-question-answering-warmup

Caleb JONES

A structured set of research notes on Visual Question Answering, with concrete evaluation references and open questions. Plans and hypotheses are kept separate from completed results. - the scope of the research question and likely confounders - a proposed comparison with matched baselines - concrete evaluation context such as VQAv2, GQA, and OK-VQA - reproducibility checks, failure modes, and open questions - topic-relevant references Start with summary.md for the full note. Sections labeled as plans or hypotheses should not be interpreted as experimental results. If results are added later, they should include dataset versions, commands, seeds, hardware, and raw logs. The note is…

Model

zero-shot-transfer-playground

Amitiyer1984

Reading notes and an experiment sketch for Zero Shot Transfer. The repository emphasizes what still needs to be tested instead of manufacturing scores or release claims. - the scope of the research question and likely confounders - a proposed comparison with matched baselines - concrete evaluation context such as task-appropriate public benchmarks named in the main note - reproducibility checks, failure modes, and open questions - topic-relevant references Start with summary.md for the full note. Sections labeled as plans or hypotheses should not be interpreted as experimental results. If results are added later, they should include dataset versions, commands, seeds, hardware, and raw logs.…