tiny-think-sft

# Shekswess/tiny-think-sft ## Overview Merged supervised fine-tuning (SFT) dataset combining chat + instruct, math + STEM, and code subsets. Built from `allenai/Dolci-Think-SFT-7B` (plus GSM8K think-style SFT for math) using the `facebook/MobileLLM-R1-140M-base` tokenizer and chat template. ## Dataset Details - **Build date:** 2026-01-10 - **Rows:** 111,372 - **Tokens:** 179,998,434 - **Max sequence length:** 4096 tokens per example - **Columns:** `messages`, `dataset_source`, `token_count` ## Data Schema - `messages`: list of chat messages (role/content) - `dataset_source`: string identifier for the originating source dataset. - `token_count`: total tokens for the example under the tokenizer. ## Intended Use SFT for small models on single-GPU setups, providing a broad general-purpose mix.