queen-v6 / config.json
Crawlnet's picture
Queen v6: 896.5M params, trained from scratch on the Crawlnet crawl
eb40ba6 verified
Raw History Blame Contribute Delete
7.19 kB
{
"model_config": {
"sequence_len": 2048,
"vocab_size": 32768,
"n_layer": 20,
"n_head": 10,
"n_kv_head": 10,
"n_embd": 1280,
"window_pattern": "SSSL"
},
"queen": {
"version": 6,
"name": "queen-v6",
"label": null,
"toy": false,
"source": "sft",
"chat_format": "chat",
"sft_method": "synthetic+openbook",
"step": 734,
"base_step": 6312,
"val_bpb": 0.7836986502315648,
"sft_val_bpb": 0.69914275963766,
"params": 896533746,
"params_scaling": 435160240,
"dtype": "bfloat16",
"book_chars": 3200,
"openbook": {
"recipe": "raft2",
"book_chars": 6000,
"max_passages": 6
},
"base_tag": null,
"exported_at": "2026-10-14T18:24:28+00:00",
"train_summary": {
"stages": {
"base": {
"wall_seconds": 2793.2409036159515,
"invocations": 1
},
"sft": {
"wall_seconds": 533.4349920749664,
"invocations": 1
}
},
"gpu": {
"count": 8
},
"wall_hours": 0.924,
"gpu_hours": 7.393,
"base": {
"last_step": 6312,
"last_val_bpb": 0.7836986502315648,
"best_step": 5985,
"best_val_bpb": 0.7663782417430147,
"sft_from_step": 5985
},
"sft": {
"method": "synthetic+openbook",
"files": [
"sft/checked/mix.train.jsonl"
],
"conversations": 3699330,
"epochs": 1,
"base_step": 6312,
"init_lr_frac": 0.5,
"mix": {
"weights": {
"synthetic": 0.55,
"openbook": 0.25,
"headings": 0.2
},
"anchor": "synthetic",
"seed": 0,
"sources": {
"synthetic": {
"available": 2563837,
"used": 2563837
},
"openbook": {
"available": 212432,
"used": 212432
},
"headings": {
"available": 1213799,
"used": 932304
}
},
"shares": {
"synthetic": 0.6913,
"openbook": 0.0573,
"headings": 0.2514
},
"conversations": 3708573
},
"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": 9243,
"synthetic.val.jsonl": 150,
"openbook.val.jsonl": 11,
"headings.val.jsonl": 0
},
"checked": {
"mix.train.jsonl": 3708573,
"synthetic.val.jsonl": 52332,
"openbook.val.jsonl": 3610,
"headings.val.jsonl": 34321
},
"by_criterion": {
"mix.train.jsonl": {
"ngram8_70": 395,
"question": 9209,
"ngram13": 149,
"question_answer": 108
},
"synthetic.val.jsonl": {
"question": 150,
"ngram8_70": 10,
"ngram13": 5
},
"openbook.val.jsonl": {
"question": 11
}
},
"by_bench": {
"mix.train.jsonl": {
"qa.json": 9062,
"cryptomc.json": 568,
"cryptoterms.json": 15,
"projectid_xl.json": 6
},
"synthetic.val.jsonl": {
"qa.json": 150,
"cryptomc.json": 10
},
"openbook.val.jsonl": {
"qa.json": 11,
"cryptomc.json": 1
}
},
"bench_fingerprint": "656b4d7aa50c7d85e43d3915aae7dce568d54be7842219a97db529bbc349c125"
},
"openbook": {
"recipe": "raft2",
"book_chars": 6000,
"max_passages": 6
},
"synth": {
"coverage": {
"eligible": 580990,
"answered": 216813,
"with_pairs": 213805,
"answered_share": 0.3732,
"pairs_share": 0.368
},
"gate": 0.0,
"teacher": "a general-purpose LLM"
}
}
}
}
}