This is the "RadarML" dataset a publicly available dataset of experimentally captured modulated radar pulses recorded over-the-air across multiple independent receive channels. The dataset was captured on the UCL ARESTOR platform which is a Radio-Frequency-System-on-a-chip (RFSoC). Each waveform was designed based on its central frequency, bandwidth, duration and modulation type. The overall dataset contains over 2 million waveforms across 7 different modulation types. The goal of this dataset is to enable comparative analysis of Radar Modulation Classification techniques on real data and stimulate research into multi-channel signal detection methods. Baseline detection and classification results are presented in the linked publication, showing how the performance varies as a function of Signal to Noise ratio.