juliannunezb/mixed-pretrain-100b-gpt2
Mixed-Pretrain-100B-GPT2 A 100-billion-token pretrain mix tokenized with the GPT-2 BPE tokenizer (vocab=50257) and packed into flat np.memmap-ready uint16 binary files. Designed as a universal pretrain corpus for GPT-2-vocab models in the 500M to 5B parameter range (1× to 200× Chinchilla optimal). Mix Composition (actual, after topups) Inspired by SmolLM2 / Llama-3 / DCLM recipes. Carefully balanced for: educational density, factual breadth, scientific reasoning… See the full description on the dataset page: https://huggingface.co/datasets/juliannunezb/mixed-pretrain-100b-gpt2.
| Author / organization | juliannunezb |
|---|---|
| License (as declared) | apache-2.0 |
| Downloads | 172 (as reported by Hugging Face) |
| Likes | 0 |
| Created | 14 May 2026 |
| Last updated at source | 14 May 2026 |
Popularity, last 4 days
172 now (+1 since 2026-10-16). Counts as reported by the source, snapshotted daily by Silytics.
Recent changes
- 16 Oct 2026 Description: Mixed-Pretrain-100B-GPT2 A 100-billion-token pretrain mix tokenized with the GPT-2 BPE tokenizer (vocab=50257) and packed into flat np.memmap-ready uint16 binary files. Designed as a universal pretrain corpus for GPT-2-vocab models in the 5 → Mixed-Pretrain-100B-GPT2 A 100-billion-token pretrain mix tokenized with the GPT-2 BPE tokenizer (vocab=50257) and packed into flat np.memmap-ready uint16 binary files. Designed as a universal pretrain corpus for GPT-2-vocab models in the 5
Changes detected by Silytics between refreshes of the source.
More from juliannunezb
Silytics shows public metadata from Hugging Face and links to the original. Information may be incomplete or out of date. Last refreshed 19 October 2026. Licenses and terms of the original apply to the work itself. Not financial, legal or professional advice.