MJPansa/qwen38-eagle4-native-features-37m-20261024-5277c167
Qwen3.8 Flash-Next native EAGLE feature pool: 37.7M tokens A byte-preserving research dataset of 37696946 physical training tokens across 21082 training conversations, with a separate fixed held-out split. The original corpus was expanded with local FP4-Local/MTP4 UltraChat continuations. This is native Flash-width EAGLE-style training data, not a DSpark bridge dataset, a trained drafter, model weights, or a full-vocabulary-logit cache. Upload status: Until ARCHIVE_COMPLETE.json… See the full description on the dataset page: https://huggingface.co/datasets/MJPansa/qwen38-eagle4-native-features
| Author / organization | MJPansa |
|---|---|
| License (as declared) | other |
| Downloads | 4.9K (as reported by Hugging Face) |
| Likes | 0 |
| Created | 23 October 2026 |
| Last updated at source | 24 October 2026 |
Popularity, last 3 days
4,885 now (+4,885 since 2026-10-23). Counts as reported by the source, snapshotted daily by Silytics.
More from MJPansa
Silytics shows public metadata from Hugging Face and links to the original. Information may be incomplete or out of date. Last refreshed 25 October 2026. Licenses and terms of the original apply to the work itself. Not financial, legal or professional advice.