queen-v5 / config.json
Crawlnet's picture
Queen v5 (matriarch d20): 645M params, pretrained from scratch on the Crawlnet crawl
1a4f94d verified
Raw History Blame Contribute Delete
7.19 kB
{
"model_config": {
"sequence_len": 2048,
"vocab_size": 16384,
"n_layer": 20,
"n_head": 10,
"n_kv_head": 10,
"n_embd": 1280,
"window_pattern": "SSSL"
},
"queen": {
"version": 5,
"name": "queen-v5",
"label": null,
"toy": false,
"source": "sft",
"chat_format": "chat",
"sft_method": "synthetic+openbook",
"step": 720,
"base_step": 4462,
"val_bpb": 0.8303629654584476,
"sft_val_bpb": 0.6567396414703545,
"params": 644875506,
"params_scaling": 414188720,
"dtype": "bfloat16",
"book_chars": 3200,
"openbook": {
"recipe": "raft2",
"book_chars": 6000,
"max_passages": 6
},
"base_tag": null,
"exported_at": "2026-10-11T03:35:32+00:00",
"train_summary": {
"stages": {
"base": {
"wall_seconds": 3712.7017335891724,
"invocations": 1
},
"sft": {
"wall_seconds": 867.9689128398895,
"invocations": 1
}
},
"gpu": {
"count": 4
},
"wall_hours": 1.272,
"gpu_hours": 5.09,
"base": {
"last_step": 4462,
"last_val_bpb": 0.8303629654584476,
"best_step": 4014,
"best_val_bpb": 0.7964583312161406,
"sft_from_step": 4014
},
"sft": {
"method": "synthetic+openbook",
"files": [
"sft/checked/mix.train.jsonl"
],
"conversations": 3231406,
"epochs": 1,
"base_step": 4462,
"init_lr_frac": 0.5,
"mix": {
"weights": {
"synthetic": 0.55,
"openbook": 0.25,
"headings": 0.2
},
"anchor": "synthetic",
"seed": 0,
"sources": {
"synthetic": {
"available": 2342880,
"used": 2342880
},
"openbook": {
"available": 211975,
"used": 211975
},
"headings": {
"available": 683478,
"used": 683478
}
},
"shares": {
"synthetic": 0.7235,
"openbook": 0.0655,
"headings": 0.2111
},
"conversations": 3238333
},
"contamination": {
"bench_files": {
"arc_easy_test.json": {
"sha256": "76e55faa15bce1b6ef26212c0c67a242a39952088d3b9e1a4d6071ea33c15190",
"items": 2376
},
"cryptomc.json": {
"sha256": "c81b3367d654f1e207d2dcbb48f490ff5feb496e2d08a049d39fb8aa8de35f87",
"items": 150
},
"cryptoterms.json": {
"sha256": "918b257c12deac8818cebf271187eedd88f026f875908b5a2e6987434b991f90",
"items": 150
},
"ctslang.json": {
"sha256": "d967e9196888027a4055592381892fca6185ee3a41b70be6a7351fa06d0bfc70",
"items": 150
},
"projectid.json": {
"sha256": "6feec798ea7269d2931fba111c48b33af2030a885e4c589235072ca6e488f51e",
"items": 200
},
"projectid_xl.json": {
"sha256": "859b5fc0d6a197bdf38e09e980d4577bd32f321e4fb3aec674725b6e4ed9a709",
"items": 2593
},
"qa.json": {
"sha256": "d5f699bdec9dca94e355332380adf317ef497784a56d3d9277ca6c0f993af9c1",
"items": 100
}
},
"skipped_files": [
"dmind_mcq.json",
"fresh_pages.json",
"fresh_pages_log.json",
"grade-0.json",
"grade-1.json",
"grade-2.json",
"grade-3.json",
"grade-4.json",
"grade-key.json",
"grade2-a.json",
"grade2-b.json",
"grade2-c.json",
"grade2-grades.json",
"grade2-key.json",
"grade3-a.json",
"grade3-b.json",
"grade3-c.json",
"grade3-grades.json",
"grade3-key.json",
"grade4-a.json",
"grade4-b.json",
"grade4-c.json",
"grade4-grader-a.json",
"grade4-grader-b.json",
"grade4-grader-c.json",
"grade4-grades.json",
"grade4-key.json",
"models.json",
"projectid_bench_protocols.json",
"projectid_defillama_raw.json",
"projectid_meta.json",
"projectid_results.json",
"projectid_results_detail.json",
"projectid_split.json",
"projectid_xl_meta.json",
"qa_answers.json",
"report_q31.json",
"report_q31_pilots.json",
"report_q4_card.json",
"results.json",
"results_ctslang.json",
"results_ctslang_detail.json",
"results_detail.json",
"results_domain.json",
"results_domain_detail.json",
"results_q31.json",
"results_q31_detail.json",
"smoke.json",
"web_snapshot.json"
],
"criteria": {
"ngram13": 13,
"ngram8": 8,
"ngram8_share": 0.7,
"question_min_words_in_text": 5,
"answer_within_words": 100
},
"dropped": {
"mix.train.jsonl": 6927,
"synthetic.val.jsonl": 143,
"openbook.val.jsonl": 8,
"headings.val.jsonl": 0
},
"checked": {
"mix.train.jsonl": 3238333,
"synthetic.val.jsonl": 47826,
"openbook.val.jsonl": 3525,
"headings.val.jsonl": 16890
},
"by_criterion": {
"mix.train.jsonl": {
"question": 6903,
"ngram13": 102,
"ngram8_70": 297,
"question_answer": 79
},
"synthetic.val.jsonl": {
"question": 142,
"ngram13": 1,
"ngram8_70": 7,
"question_answer": 1
},
"openbook.val.jsonl": {
"question": 8
}
},
"by_bench": {
"mix.train.jsonl": {
"qa.json": 6785,
"cryptomc.json": 405,
"cryptoterms.json": 9,
"projectid_xl.json": 6
},
"synthetic.val.jsonl": {
"qa.json": 141,
"cryptomc.json": 9
},
"openbook.val.jsonl": {
"qa.json": 8
}
},
"bench_fingerprint": "656b4d7aa50c7d85e43d3915aae7dce568d54be7842219a97db529bbc349c125"
},
"openbook": {
"recipe": "raft2",
"book_chars": 6000,
"max_passages": 6
},
"synth": {
"coverage": {
"eligible": 381992,
"answered": 203795,
"with_pairs": 198882,
"answered_share": 0.5335,
"pairs_share": 0.5206
},
"gate": 0.0,
"teacher": "a general-purpose LLM"
}
}
}
}
}