{ "repo_folder": "wrap_inspired", "paper_setting": "WRAP-INSPIRED", "internal_run_name": "wrap", "selection_rule": "A uniform random sample of the remaining pool, rewritten with four style-specific format prompts rather than the two shared prompts.", "content": "rewritten", "composition": { "note": "This folder is the complete 10B-token mixture the models actually trained on: the shared 5B raw anchor plus this setting's 5B component, already merged and shuffled. The anchor is included in every setting folder.", "shuffle_seed": 42, "anchor_docs": 4120164, "anchor_tokens": 5000002332, "strategy_docs": 13022091, "strategy_tokens": 5000000087, "strategy_token_target": 5000000000, "style_distribution": { "distill": 2456200, "original": 4120164, "wrap_easy": 2646480, "wrap_hard": 2631853, "wrap_qa": 2643215, "wrap_wiki": 2644343 }, "distill_selection": "seeded_random_seed42_no_quality_sort" }, "rewriter_model": "Qwen2.5-7B-Instruct (greedy decoding)", "prompt_style": "four WRAP-style format prompts (easy, hard, wiki, qa), plus distillation to top up the budget", "doc_count": 17142255, "token_count_llama2": 10000002419, "shard_count": 35, "columns": [ "doc_id", "orig_doc_id", "text", "source_prompt", "train_tokens" ], "provenance": { "source_folder": "/scratch/bvandur1/zhuicon1/data_rewrite/pretrain/wrap/shuffled", "tokenized_folder": "/scratch/bvandur1/zhuicon1/nanotron_tokenized/wrap/tokenized", "source_evidence": "datatrove executor.json (reader data_folder), confirmed by decoding 50 documents out of the .ds and matching them into this folder (50/50)", "trained_from_config": "config_wrap_10B_1.5B.yaml", "config_dir": "/scratch/bvandur1/zhuicon1/projects/nanotron/examples", "seeds": [ 42, 43, 44 ], "doc_content_digest_sha256": "2d1c51ce1edd5cc6133d16e68da96a22badc41a64ef5d714acbaf412b9f3b382", "doc_id_digest_sha256": "65c5fc9c4cb290d85ca4b2c378031d65719f394e609b05f94a88f2f510b11047" } }