MMSoc_Memotion_corrected

# Memotion for MTEB This repository is a byte-preserving cleanup of [`mteb/MMSoc_Memotion`](https://huggingface.co/datasets/mteb/MMSoc_Memotion) at revision `f77e225ae55c1987b0b8cbf6badd1c10296f5f34` for [MTEB issue #5158](https://github.com/embeddings-benchmark/mteb/issues/5158). The pinned source and the [official Kaggle release](https://www.kaggle.com/datasets/williamscott701/memotion-dataset-7k) both contain the same truncated `train` image, `image_5119.png` (row 4578, SHA-256 `63175a3560ace1e74d4e7913206c94b6293473e09539df0e5be791df62e6a2a9`). The PNG is missing part of its IDAT payload. Permissive decoding fabricates 35 black rows, so this dataset excludes that one example rather than altering its visual content. All other source rows and embedded image bytes are preserved in their original order. All 6,991 remaining images pass both Pillow verification and a complete RGB decode. The validation and test splits are unchanged. Generated by `scripts/data/memotion/create_data.py` in MTEB.