Curators: KaraKaraWitch (under featherless.ai)
Language: English
License: Apache-2.0
Volume: Approx 10 B tokens (UTF-8, whitespace tokenized)
Recent releases frequently couple large corpora with conversational or instruction-tuning objectives, making it hard to isolate the contribution of the raw text itself.
MIKU returns to a… See the full description on the dataset page:
https://huggingface.co/datasets/KaraKaraWitch/MIKU-Alpha.