datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
sherpa-onnx-tts-modelsPP-OCRv5_ONNXONNX-converted models from PaddlePaddle/PaddleOCR.
Models
Model
Purpose
PP-OCRv5_mobile_det.onnx
Text detection
PP-OCRv5_mobile_rec.onnx
Text recognition
PP-LCNet_x1_0_table_cls.onnx
Table classification
UVDoc.onnx
Document image correction
SLANeXt_wired.onnx
Table structure recognition (wired)
SLANeXt_wireless.onnx
Table structure recognition (wireless)
Source
Converted from PaddlePaddle format using Paddle2ONNX.
License… See the full description on the dataset page: https://huggingface.co/datasets/BMekiker/PP-OCRv5_ONNX.onnx-sdxl-webcpu-demostable-audio-open-small-onnx
Stable Audio Open Small - ONNX Runtime Assets (v1.0.0)
This repository contains optimized ONNX artifacts derived from the Stability AI Stable Audio Open Small model.
These files are specifically packaged for local, on-device inference in the Pi Day application.
🛠 Modifications
Format: Exported model components to ONNX format.
Optimization: Converted denoiser to FP16 for improved performance and reduced memory footprint.
Portability: Tailored for cross-platform runtime… See the full description on the dataset page: https://huggingface.co/datasets/cherub8128/stable-audio-open-small-onnx.pure-onnx
SPLADE Endpoint Golden Dataset (v1)
This dataset contains sparse vectors generated from a SPLADE inference service.
Source
Model: prithivida/Splade_PP_en_v1
Generated at (UTC): 2026-02-21T19:09:51.715028+00:00
Rows: 20
Digest (JSONL SHA-256): 050715eeb0a288daa486444e6bad006a98705f1c0a6e8f3780e0c46b1614105c
Notes
Inputs are short synthetic test prompts for regression parity.
Labels are included in this release for parity validation.
No private endpoint… See the full description on the dataset page: https://huggingface.co/datasets/tazarov/pure-onnx.onnx-inference
ONNX-Predict Dataset
This dataset is derived from ONNX-Predict, a dataset of ONNX model inference performance measurements collected from models hosted on the ONNX Model Zoo.
It contains inference data for 2,300+ ONNX models. Each model was optimized into four variants, and inference measurements were collected across 16 machines with different CPU core counts, memory sizes, cache configurations, and CPU providers.
Dataset Overview
Each row in the dataset combines… See the full description on the dataset page: https://huggingface.co/datasets/sidnarsipur/onnx-inference.onnx-shape-inference-poc-custom-function-recursion-2026-05-12
ONNX shape inference custom function recursion PoC
Private evidence package for authorized huntr / ONNX MFF research.
Target
Package: onnx==1.21.0
API: onnx.shape_inference.infer_shapes_path
Payload: payload.onnx
Result: process termination by SIGSEGV / shell return code 139
Reproduce
python3 -m venv /tmp/onnx-poc-venv
/tmp/onnx-poc-venv/bin/python -m pip install onnx==1.21.0
/tmp/onnx-poc-venv/bin/python -c 'import onnx; m = onnx.load("payload.onnx");… See the full description on the dataset page: https://huggingface.co/datasets/tonydav41/onnx-shape-inference-poc-custom-function-recursion-2026-05-12.cntk-onnx-double-raw-overread-2026-05-13
CNTK ONNX DOUBLE raw_data Over-Read PoC
Target: CNTK v2 / microsoft/CNTK
Bug: CNTK's ONNX importer decodes TensorProto.DOUBLE raw_data with a 4-byte stride while copying 8 bytes per decoded element. A malformed ONNX model loaded through cntk.load_model(path, format=C.ModelFormat.ONNX) corrupts decoded constants and Valgrind reports an invalid native read during public API import.
Main PoC
initializer_raw48.onnx
SHA-256:… See the full description on the dataset page: https://huggingface.co/datasets/tonydav41/cntk-onnx-double-raw-overread-2026-05-13.