shuai22/mmduet2-training-data

MMDuet2 Training Data (Prepared Frames) Prepared training inputs used for local MMDuet2 SFT and GRPO runs. The media are pre-extracted frames: standalone JPEGs, JPEG sequences presented through the video interface, or JPEG frames packed in HDF5 containers. Layout metadata/sft/: the five SFT JSONL files. metadata/rl/: the 2-second-per-frame RL train/validation parquet files and user-turn JSON. metadata/rl-1fps/: the 1-second-per-frame RL train/validation parquet… See the full description on the dataset page: https://huggingface.co/datasets/shuai22/mmduet2-training-data.

<iframe src="https://silytics.si/embed/item/8686b9d7-5f77-43e2-bc27-22c9a933417c" width="420" height="190" style="border:0" loading="lazy" title="Silytics entry"></iframe>

Silytics shows public metadata from Hugging Face and links to the original. Information may be incomplete or out of date. Last refreshed 17 October 2026. Licenses and terms of the original apply to the work itself. Not financial, legal or professional advice.