Qwen3-4B-Countdown-RLVR / checkpoint_selection.json
NotoriousH2's picture
Update Countdown RLVR 3072-token training artifacts
1a3685c verified
Raw
History Blame Contribute Delete
25.7 kB
{
"schema_version": 5,
"plan": {
"system_prompt": "제시된 숫자는 각 항목을 정확히 한 번씩 사용하여 목표값을 만드세요. 허용 연산은 +, -, *, /이며 괄호를 사용할 수 있습니다. 연산 기호는 두 수 사이에만 사용하세요. 숫자를 이어 붙여 새 수를 만들지 마세요. 풀이를 간결하게 검토한 뒤 <answer>와 </answer> 사이에는 등호와 계산 결과 없이 수식만 작성하세요.",
"training_lineage": {
"schema_version": 2,
"expected_checkpoint_steps": [
120,
240,
360,
480
],
"validation": "before_model_merge_and_evaluation",
"seal": {
"schema_version": 2,
"task": "countdown_rlvr",
"model": {
"repo_id": "Qwen/Qwen3-4B",
"revision": "1cfa9a7208912126459214e8b04321603b3df60c"
},
"training_data": {
"samples": 1024,
"sha256": "aa7abb6242d8ada72e55a6d8d0917e3618473ddf2f0f880b288814394b231131"
},
"system_prompt_sha256": "2f69064845b7e556c420d4617c4fd9a12786438eb534e4c0342dde9cb7402d44",
"reward_implementation_sha256": "d779da290b9c7376298160a4e9c722e17af6e60d9e45a7bfdb4952ccf19b07f3",
"phases": [
{
"name": "초기 구간",
"start_step": 1,
"end_step": 120,
"scheduler_end_step": 480,
"max_completion_length": 3072,
"vllm_gpu_memory_utilization": 0.25,
"resume_from_checkpoint": null
},
{
"name": "재개 구간",
"start_step": 121,
"end_step": 480,
"scheduler_end_step": 480,
"max_completion_length": 3072,
"vllm_gpu_memory_utilization": 0.25,
"resume_from_checkpoint": "checkpoint-120"
}
],
"metadata_sha256": {
"training_config.json": "525e7d779b93dc401e120382a2a7c3cb9876c79d5dd3093d759a14211d9df57f",
"training_provenance.json": "1d0e26bdf87cdbe82fc807baf7220fcad3275b653d7d4147060626b0aed7ea0b",
"history.json": "1de13f49b1cbb073d80d9428bca354b2cc71aadca881c7722e53f1fbf89aa3c5",
"phase_contract.json": "fd929179bb0d2447a9dd1422c60ed487a5c6cc533ab1e9ca1ca3b1f216b7d212"
},
"initial_resume_manifest_sha256": "264e28e46ca7712381320727c85faf1bbf9fe74f3591e7883e704912c4ea7c8c",
"checkpoints": [
{
"label": "checkpoint-120",
"step": 120,
"checkpoint_manifest_sha256": "ddcc84c06fe9643f6ab80ba183494d505a3568f28e4eeb81b04fb202b4e8a342",
"adapter_manifest_sha256": "5187f02b8831c48b250a515a9c02483fb72ddf7150bbe80046d234fe6a2512be",
"trainer_state_sha256": "a41022ab39cc63b52b47dfe5e00be141b14fb788058f0994ec3bea739ec9f4c8",
"reward_log_sha256": "a6844d02d3aee8b5b93157a82cb917f9611cbe94c6dd88cdee9618cc8f02d178"
},
{
"label": "checkpoint-240",
"step": 240,
"checkpoint_manifest_sha256": "4a5287d294cb5ee198ab6ce9193e5ded692457fdbb47913c6e36481854891797",
"adapter_manifest_sha256": "a8efb0730f3073efde6f77adedbf8a7bc7e4a91cb52ba7967dd428b224705074",
"trainer_state_sha256": "8447a22a7e83fc3f57aab69de1c64694c01d9d809a667129d251afc5f995e01e",
"reward_log_sha256": "dd4848fe682818ad7c7d253278ad716524bf15f61f8374d9b906810f3c812956"
},
{
"label": "checkpoint-360",
"step": 360,
"checkpoint_manifest_sha256": "d508cd52c23dfaa12d9c739ab96c9ee9545e47cbb08bb4a3713e688c268fb8aa",
"adapter_manifest_sha256": "26ca25a2e43806bf1e4f766993d5bb1b811e7e136d317846953f1bc93b6937c6",
"trainer_state_sha256": "080520a83dace5d342d1e7665207daf6af091fcf2959ef588762ba06392db0c8",
"reward_log_sha256": "5de32702a4a450e5e45544e6fdd000e8e25560664e8efe9a2003343be68871ed"
},
{
"label": "checkpoint-480",
"step": 480,
"checkpoint_manifest_sha256": "6f2ec5d48f5a9619494a37106d8d96b6924c38df5d3604b5613e3516d79ddea8",
"adapter_manifest_sha256": "04ba4a9cdbcf7a5c7c359149615dedf0ab171b18c71178dfd722ae332c99b016",
"trainer_state_sha256": "7cb0e9d1a7f2c1a6deeb676b1f0ea78fe0bf44f8fc4b8d444d813354e3feaff0",
"reward_log_sha256": "f07069b694188f4a6cfd7746820357713afcf55a5f850d696aabca3007fd82be"
}
],
"history": {
"records": 97,
"reward_records": 96,
"last_reward_step": 480,
"final_summary_scope": "continuation_process_steps_121_to_480"
},
"sha256": "a982d95e280ee9e2d985ff53a3da965f1ce0ad42b2a817ae2aab43747be37eab"
}
},
"baseline": {
"label": "Base",
"model_id": "Qwen/Qwen3-4B",
"revision": "1cfa9a7208912126459214e8b04321603b3df60c"
},
"checkpoints": [
{
"label": "checkpoint-120",
"step": 120
},
{
"label": "checkpoint-240",
"step": 240
},
{
"label": "checkpoint-360",
"step": 360
},
{
"label": "checkpoint-480",
"step": 480
}
],
"validation": {
"path": "data/countdown/validation.jsonl",
"subset": "all_rows",
"expected_samples": 128
},
"final_holdout": {
"path": "data/countdown/test.jsonl",
"expected_samples": 256,
"sealed_before_selection": true,
"used_for_selection": false
},
"decoding": {
"completion_count": 8,
"max_tokens": 3072,
"temperature": 0.6,
"top_p": 0.95,
"top_k": 20,
"seed": 42,
"request_seed_rule": "base_seed_plus_problem_index"
},
"eligibility": [
"complete_validation_and_completion_grid",
"valid_expression_rate_at_least_base",
"truncated_completion_rate_at_most_base_plus_one_standard_error"
],
"truncation_eligibility": {
"sample_unit": "completion",
"sample_count": 1024,
"base_rate": "base_truncated_completion_rate",
"base_standard_error_formula": "sqrt(base_rate * (1 - base_rate) / sample_count)",
"maximum_eligible_rate_formula": "base_rate + base_standard_error",
"comparison": "candidate_truncated_completion_rate <= maximum_eligible_rate"
},
"selection": {
"primary_metric": "pass_at_1",
"improvement_gate": "best_lower_one_se_above_base",
"tie_band": "within_one_se_of_best",
"tie_breakers": [
"pass_at_8",
"lower_truncated_completion_rate",
"earlier_checkpoint"
],
"fallback": "Base"
},
"server_reproducibility": {
"server_seed": 42,
"environment": {
"VLLM_BATCH_INVARIANT": "1"
}
},
"promotion": {
"adapter_dir": "adapter",
"merged_dir": ".",
"enabled": true
},
"workers": 16,
"selection_path": "checkpoint_selection.json"
},
"training_lineage": {
"schema_version": 2,
"task": "countdown_rlvr",
"model": {
"repo_id": "Qwen/Qwen3-4B",
"revision": "1cfa9a7208912126459214e8b04321603b3df60c"
},
"training_data": {
"samples": 1024,
"sha256": "aa7abb6242d8ada72e55a6d8d0917e3618473ddf2f0f880b288814394b231131"
},
"system_prompt_sha256": "2f69064845b7e556c420d4617c4fd9a12786438eb534e4c0342dde9cb7402d44",
"reward_implementation_sha256": "d779da290b9c7376298160a4e9c722e17af6e60d9e45a7bfdb4952ccf19b07f3",
"phases": [
{
"name": "초기 구간",
"start_step": 1,
"end_step": 120,
"scheduler_end_step": 480,
"max_completion_length": 3072,
"vllm_gpu_memory_utilization": 0.25,
"resume_from_checkpoint": null
},
{
"name": "재개 구간",
"start_step": 121,
"end_step": 480,
"scheduler_end_step": 480,
"max_completion_length": 3072,
"vllm_gpu_memory_utilization": 0.25,
"resume_from_checkpoint": "checkpoint-120"
}
],
"metadata_sha256": {
"training_config.json": "525e7d779b93dc401e120382a2a7c3cb9876c79d5dd3093d759a14211d9df57f",
"training_provenance.json": "1d0e26bdf87cdbe82fc807baf7220fcad3275b653d7d4147060626b0aed7ea0b",
"history.json": "1de13f49b1cbb073d80d9428bca354b2cc71aadca881c7722e53f1fbf89aa3c5",
"phase_contract.json": "fd929179bb0d2447a9dd1422c60ed487a5c6cc533ab1e9ca1ca3b1f216b7d212"
},
"initial_resume_manifest_sha256": "264e28e46ca7712381320727c85faf1bbf9fe74f3591e7883e704912c4ea7c8c",
"checkpoints": [
{
"label": "checkpoint-120",
"step": 120,
"checkpoint_manifest_sha256": "ddcc84c06fe9643f6ab80ba183494d505a3568f28e4eeb81b04fb202b4e8a342",
"adapter_manifest_sha256": "5187f02b8831c48b250a515a9c02483fb72ddf7150bbe80046d234fe6a2512be",
"trainer_state_sha256": "a41022ab39cc63b52b47dfe5e00be141b14fb788058f0994ec3bea739ec9f4c8",
"reward_log_sha256": "a6844d02d3aee8b5b93157a82cb917f9611cbe94c6dd88cdee9618cc8f02d178"
},
{
"label": "checkpoint-240",
"step": 240,
"checkpoint_manifest_sha256": "4a5287d294cb5ee198ab6ce9193e5ded692457fdbb47913c6e36481854891797",
"adapter_manifest_sha256": "a8efb0730f3073efde6f77adedbf8a7bc7e4a91cb52ba7967dd428b224705074",
"trainer_state_sha256": "8447a22a7e83fc3f57aab69de1c64694c01d9d809a667129d251afc5f995e01e",
"reward_log_sha256": "dd4848fe682818ad7c7d253278ad716524bf15f61f8374d9b906810f3c812956"
},
{
"label": "checkpoint-360",
"step": 360,
"checkpoint_manifest_sha256": "d508cd52c23dfaa12d9c739ab96c9ee9545e47cbb08bb4a3713e688c268fb8aa",
"adapter_manifest_sha256": "26ca25a2e43806bf1e4f766993d5bb1b811e7e136d317846953f1bc93b6937c6",
"trainer_state_sha256": "080520a83dace5d342d1e7665207daf6af091fcf2959ef588762ba06392db0c8",
"reward_log_sha256": "5de32702a4a450e5e45544e6fdd000e8e25560664e8efe9a2003343be68871ed"
},
{
"label": "checkpoint-480",
"step": 480,
"checkpoint_manifest_sha256": "6f2ec5d48f5a9619494a37106d8d96b6924c38df5d3604b5613e3516d79ddea8",
"adapter_manifest_sha256": "04ba4a9cdbcf7a5c7c359149615dedf0ab171b18c71178dfd722ae332c99b016",
"trainer_state_sha256": "7cb0e9d1a7f2c1a6deeb676b1f0ea78fe0bf44f8fc4b8d444d813354e3feaff0",
"reward_log_sha256": "f07069b694188f4a6cfd7746820357713afcf55a5f850d696aabca3007fd82be"
}
],
"history": {
"records": 97,
"reward_records": 96,
"last_reward_step": 480,
"final_summary_scope": "continuation_process_steps_121_to_480"
},
"sha256": "a982d95e280ee9e2d985ff53a3da965f1ce0ad42b2a817ae2aab43747be37eab"
},
"validation": {
"path": "data/countdown/validation.jsonl",
"sha256": "b06a1be3604d637aa19bd61af57aadf98fbbffcb8ef4db8d477fe8535c617497",
"samples": 128,
"problem_ids": [
"countdown-4-1-3-8-9-t70",
"countdown-4-1-6-7-7-t70",
"countdown-4-3-5-5-7-t20",
"countdown-4-2-3-5-7-t20",
"countdown-4-5-7-8-9-t37",
"countdown-4-3-4-5-6-t54",
"countdown-4-3-6-7-7-t11",
"countdown-4-2-3-4-9-t51",
"countdown-4-1-3-6-8-t28",
"countdown-4-2-6-7-8-t43",
"countdown-4-1-2-4-8-t34",
"countdown-4-3-7-7-9-t37",
"countdown-4-2-7-8-8-t62",
"countdown-4-2-3-6-8-t64",
"countdown-4-5-7-7-7-t91",
"countdown-4-1-4-5-7-t38",
"countdown-4-2-7-8-9-t35",
"countdown-4-1-1-4-7-t11",
"countdown-4-3-6-6-7-t40",
"countdown-4-3-6-7-9-t39",
"countdown-4-2-2-3-9-t90",
"countdown-4-1-1-3-9-t20",
"countdown-4-3-4-4-6-t48",
"countdown-4-2-2-3-8-t56",
"countdown-4-2-4-7-8-t12",
"countdown-4-2-2-7-8-t36",
"countdown-4-1-2-6-6-t71",
"countdown-4-1-2-4-9-t12",
"countdown-4-1-2-4-6-t27",
"countdown-4-2-3-4-5-t45",
"countdown-4-2-4-6-8-t100",
"countdown-4-4-5-7-9-t32",
"countdown-4-2-4-5-7-t84",
"countdown-4-1-5-7-9-t47",
"countdown-4-1-6-7-9-t24",
"countdown-4-2-5-7-9-t97",
"countdown-4-4-6-6-7-t35",
"countdown-4-1-4-6-7-t46",
"countdown-4-1-1-5-8-t26",
"countdown-4-1-2-6-8-t22",
"countdown-4-2-6-7-9-t73",
"countdown-4-2-3-4-7-t34",
"countdown-4-3-6-7-8-t95",
"countdown-4-2-4-9-9-t83",
"countdown-4-2-4-5-5-t20",
"countdown-4-2-3-8-9-t26",
"countdown-4-4-4-7-9-t18",
"countdown-4-2-2-3-8-t24",
"countdown-4-1-3-4-8-t44",
"countdown-4-1-8-8-9-t55",
"countdown-4-3-4-7-7-t74",
"countdown-4-1-6-7-9-t66",
"countdown-4-4-5-7-7-t34",
"countdown-4-1-2-4-4-t24",
"countdown-4-1-1-4-5-t18",
"countdown-4-3-4-5-8-t36",
"countdown-4-3-3-6-8-t11",
"countdown-4-1-6-8-8-t41",
"countdown-4-4-5-9-9-t31",
"countdown-4-6-8-8-9-t42",
"countdown-4-5-6-8-8-t11",
"countdown-4-1-3-6-8-t41",
"countdown-4-2-4-6-7-t34",
"countdown-4-3-5-6-9-t52",
"countdown-4-1-2-4-5-t16",
"countdown-4-1-5-7-8-t55",
"countdown-4-3-5-9-9-t21",
"countdown-4-3-4-5-7-t84",
"countdown-4-2-5-5-5-t32",
"countdown-4-3-4-6-7-t42",
"countdown-4-2-3-7-7-t27",
"countdown-4-1-2-6-8-t84",
"countdown-4-1-4-9-9-t28",
"countdown-4-3-4-8-9-t96",
"countdown-4-1-3-5-5-t72",
"countdown-4-2-2-4-9-t36",
"countdown-4-1-2-8-9-t66",
"countdown-4-3-4-8-9-t43",
"countdown-4-2-3-7-8-t91",
"countdown-4-1-3-7-7-t46",
"countdown-4-3-5-8-8-t11",
"countdown-4-3-6-9-9-t36",
"countdown-4-6-7-9-9-t18",
"countdown-4-3-5-5-7-t18",
"countdown-4-3-5-9-9-t99",
"countdown-4-2-6-6-7-t15",
"countdown-4-2-3-4-7-t63",
"countdown-4-1-2-2-5-t14",
"countdown-4-1-2-6-8-t32",
"countdown-4-2-6-7-8-t27",
"countdown-4-4-4-5-9-t72",
"countdown-4-4-4-5-8-t30",
"countdown-4-3-6-8-9-t34",
"countdown-4-2-3-8-9-t64",
"countdown-4-2-2-5-8-t39",
"countdown-4-5-5-6-7-t18",
"countdown-4-4-5-9-9-t61",
"countdown-4-1-1-6-7-t43",
"countdown-4-1-3-4-6-t36",
"countdown-4-1-7-9-9-t71",
"countdown-4-1-4-6-6-t47",
"countdown-4-2-2-5-9-t36",
"countdown-4-4-6-8-8-t48",
"countdown-4-1-3-4-8-t93",
"countdown-4-1-2-8-8-t14",
"countdown-4-5-5-5-7-t20",
"countdown-4-1-1-5-9-t54",
"countdown-4-1-4-5-6-t13",
"countdown-4-5-7-9-9-t69",
"countdown-4-1-3-4-8-t45",
"countdown-4-2-4-5-9-t47",
"countdown-4-1-2-5-5-t49",
"countdown-4-1-1-8-8-t14",
"countdown-4-4-6-8-9-t62",
"countdown-4-1-2-5-9-t50",
"countdown-4-2-3-5-8-t45",
"countdown-4-1-3-3-4-t24",
"countdown-4-1-3-8-9-t84",
"countdown-4-1-4-6-9-t17",
"countdown-4-1-2-4-6-t19",
"countdown-4-2-8-9-9-t19",
"countdown-4-2-5-7-7-t28",
"countdown-4-1-3-5-6-t85",
"countdown-4-1-2-3-7-t19",
"countdown-4-4-7-9-9-t78",
"countdown-4-4-6-7-8-t20",
"countdown-4-2-6-9-9-t56",
"countdown-4-4-7-8-8-t39"
]
},
"final_holdout": {
"path": "data/countdown/test.jsonl",
"sha256": "416c02076321875cccfeed19f742e56048269b4b9d24112f6a2bee82ba301815",
"samples": 256,
"evaluated": false
},
"candidates": [
{
"label": "Base",
"step": 0,
"model_path": "Qwen/Qwen3-4B",
"pass_at_1": 0.1015625,
"pass_at_8": 0.484375,
"truncated_completion_rate": 0.1630859375,
"valid_expression_rate": 0.119140625,
"mean_response_tokens": 1656.318359375,
"serving_manifest_sha256": "4021473f68a8794c8bba380119edf57d342c42c09d1ccd463b90ae8f31df6609"
},
{
"label": "checkpoint-120",
"step": 120,
"pass_at_1": 0.703125,
"pass_at_8": 0.9296875,
"truncated_completion_rate": 0.1796875,
"valid_expression_rate": 0.7041015625,
"mean_response_tokens": 1688.859375,
"serving_manifest_sha256": "c7f4df21e568dea8892061f29ea07e6cd7e55f9b4681479902e17e0c01b3c80f"
},
{
"label": "checkpoint-240",
"step": 240,
"pass_at_1": 0.703125,
"pass_at_8": 0.9453125,
"truncated_completion_rate": 0.1748046875,
"valid_expression_rate": 0.6923828125,
"mean_response_tokens": 1671.7431640625,
"serving_manifest_sha256": "2048ec6d6667549c65351d3bc4dbf56a1f616c0e0e59d1d8ece532957305c8c3"
},
{
"label": "checkpoint-360",
"step": 360,
"model_path": "NotoriousH2/Qwen3-4B-Countdown-RLVR",
"pass_at_1": 0.6484375,
"pass_at_8": 0.9453125,
"truncated_completion_rate": 0.171875,
"valid_expression_rate": 0.583984375,
"mean_response_tokens": 1701.4521484375,
"serving_manifest_sha256": "3af912b7de2cf6f1a745bf6a2a83c2a7c95e600eaae2aa01200ecb721b9c5379"
},
{
"label": "checkpoint-480",
"step": 480,
"pass_at_1": 0.5859375,
"pass_at_8": 0.9375,
"truncated_completion_rate": 0.185546875,
"valid_expression_rate": 0.59375,
"mean_response_tokens": 1718.89453125,
"serving_manifest_sha256": "61728c0221a70781324f4ed0852dda809c8a714892826c85740830d08de8298d"
}
],
"decisions": [
{
"label": "Base",
"step": 0,
"eligible": true,
"ineligibility_reasons": [],
"pass_at_1_standard_error": 0.026699654520492805,
"truncated_completion_rate": 0.1630859375,
"truncation_sample_count": 1024,
"base_truncated_completion_rate": 0.1630859375,
"base_truncated_completion_standard_error": 0.011545126918158703,
"maximum_eligible_truncated_completion_rate": 0.1746310644181587
},
{
"label": "checkpoint-120",
"step": 120,
"eligible": false,
"ineligibility_reasons": [
"truncated_completion_rate_above_base_plus_one_standard_error"
],
"pass_at_1_standard_error": 0.04038295532838932,
"truncated_completion_rate": 0.1796875,
"truncation_sample_count": 1024,
"base_truncated_completion_rate": 0.1630859375,
"base_truncated_completion_standard_error": 0.011545126918158703,
"maximum_eligible_truncated_completion_rate": 0.1746310644181587
},
{
"label": "checkpoint-240",
"step": 240,
"eligible": false,
"ineligibility_reasons": [
"truncated_completion_rate_above_base_plus_one_standard_error"
],
"pass_at_1_standard_error": 0.04038295532838932,
"truncated_completion_rate": 0.1748046875,
"truncation_sample_count": 1024,
"base_truncated_completion_rate": 0.1630859375,
"base_truncated_completion_standard_error": 0.011545126918158703,
"maximum_eligible_truncated_completion_rate": 0.1746310644181587
},
{
"label": "checkpoint-360",
"step": 360,
"eligible": true,
"ineligibility_reasons": [],
"pass_at_1_standard_error": 0.042201739133460744,
"truncated_completion_rate": 0.171875,
"truncation_sample_count": 1024,
"base_truncated_completion_rate": 0.1630859375,
"base_truncated_completion_standard_error": 0.011545126918158703,
"maximum_eligible_truncated_completion_rate": 0.1746310644181587
},
{
"label": "checkpoint-480",
"step": 480,
"eligible": false,
"ineligibility_reasons": [
"truncated_completion_rate_above_base_plus_one_standard_error"
],
"pass_at_1_standard_error": 0.043536510010075705,
"truncated_completion_rate": 0.185546875,
"truncation_sample_count": 1024,
"base_truncated_completion_rate": 0.1630859375,
"base_truncated_completion_standard_error": 0.011545126918158703,
"maximum_eligible_truncated_completion_rate": 0.1746310644181587
}
],
"one_se_boundary": 0.6062357608665393,
"truncation_eligibility": {
"sample_unit": "completion",
"sample_count": 1024,
"base_rate": 0.1630859375,
"base_standard_error": 0.011545126918158703,
"maximum_eligible_rate": 0.1746310644181587
},
"selected": {
"label": "checkpoint-360",
"step": 360,
"reason": "checkpoint-360의 pass@1 0.6484에서 표준 오차 0.0422를 뺀 하한은 0.6062이고, Base pass@1은 0.1016입니다. 하한이 Base를 초과합니다. 동률 범위는 checkpoint-360이며, checkpoint-360의 동률 처리 값은 pass@8=0.9453, 잘림률=0.1719, step=360입니다.",
"source_model_path": "NotoriousH2/Qwen3-4B-Countdown-RLVR",
"promotion": {
"kind": "checkpoint",
"performed": true,
"adapter": "adapter",
"merged": ".",
"serving_manifest": {
"format_version": 1,
"sha256": "3af912b7de2cf6f1a745bf6a2a83c2a7c95e600eaae2aa01200ecb721b9c5379",
"files": [
{
"path": "chat_template.jinja",
"bytes": 4168,
"sha256": "a55ee1b1660128b7098723e0abcd92caa0788061051c62d51cbe87d9cf1974d8"
},
{
"path": "config.json",
"bytes": 1593,
"sha256": "855a4a562e47f29d00ff0c06e244cd27319c381abff813d8461d1184894e97f5"
},
{
"path": "generation_config.json",
"bytes": 214,
"sha256": "5292414bc5c987a1498cf50e6370ecf4939f451e023c74bb3facbdce8445a153"
},
{
"path": "model.safetensors",
"bytes": 8044982080,
"sha256": "6101934ce09c25c36b6637ade43b80497499ab1b39cdb28b4e91212a64590796"
},
{
"path": "tokenizer.json",
"bytes": 11422650,
"sha256": "be75606093db2094d7cd20f3c2f385c212750648bd6ea4fb2bf507a6a4c55506"
},
{
"path": "tokenizer_config.json",
"bytes": 692,
"sha256": "1cc816812993bff176eb4f7495433b736f06fba9b6e7b05cac7b4a1780650c95"
}
]
}
}
},
"final_models": [
{
"label": "Base",
"kind": "base_model",
"model_path": "Qwen/Qwen3-4B",
"model_id": "Qwen/Qwen3-4B",
"serving_manifest": {
"format_version": 1,
"sha256": "4021473f68a8794c8bba380119edf57d342c42c09d1ccd463b90ae8f31df6609",
"files": [
{
"path": "config.json",
"bytes": 726,
"sha256": "8ba006f74fecfaaeb392872a60f4a480e7ec9860153d2e1b769ec81f9a147f8a"
},
{
"path": "generation_config.json",
"bytes": 239,
"sha256": "2325da0f15bb848e018c5ae071b7943332e9f871d6b60e2ed22ca97d4cb993d2"
},
{
"path": "merges.txt",
"bytes": 1671853,
"sha256": "8831e4f1a044471340f7c0a83d7bd71306a5b867e95fd870f74d0c5308a904d5"
},
{
"path": "model-00001-of-00003.safetensors",
"bytes": 3957900840,
"sha256": "328a91d3122359d5547f9d79521205bc0a46e1f79a792dfe650e99fc2d651223"
},
{
"path": "model-00002-of-00003.safetensors",
"bytes": 3987450520,
"sha256": "6cd087b316306a68c562436b5492edbcf6e16c6dba3a1308279caa5a58e21ca5"
},
{
"path": "model-00003-of-00003.safetensors",
"bytes": 99630640,
"sha256": "e4bf436957184f4eeb86a80e9db394503f1f56446b2e6b7edeac5b81470f4ca1"
},
{
"path": "model.safetensors.index.json",
"bytes": 32819,
"sha256": "6dc0981b8829fead746441f68f38f24c5ca4a3a66351f652c26c6df0efc43ab2"
},
{
"path": "tokenizer.json",
"bytes": 11422654,
"sha256": "aeb13307a71acd8fe81861d94ad54ab689df773318809eed3cbe794b4492dae4"
},
{
"path": "tokenizer_config.json",
"bytes": 9732,
"sha256": "d5d09f07b48c3086c508b30d1c9114bd1189145b74e982a265350c923acd8101"
},
{
"path": "vocab.json",
"bytes": 2776833,
"sha256": "ca10d7e9fb3ed18575dd1e277a2579c16d108e32f27439684afa0e10b1440910"
}
]
}
},
{
"label": "RLVR",
"kind": "selected_checkpoint",
"checkpoint_label": "checkpoint-360",
"model_path": "NotoriousH2/Qwen3-4B-Countdown-RLVR",
"serving_manifest": {
"format_version": 1,
"sha256": "3af912b7de2cf6f1a745bf6a2a83c2a7c95e600eaae2aa01200ecb721b9c5379",
"files": [
{
"path": "chat_template.jinja",
"bytes": 4168,
"sha256": "a55ee1b1660128b7098723e0abcd92caa0788061051c62d51cbe87d9cf1974d8"
},
{
"path": "config.json",
"bytes": 1593,
"sha256": "855a4a562e47f29d00ff0c06e244cd27319c381abff813d8461d1184894e97f5"
},
{
"path": "generation_config.json",
"bytes": 214,
"sha256": "5292414bc5c987a1498cf50e6370ecf4939f451e023c74bb3facbdce8445a153"
},
{
"path": "model.safetensors",
"bytes": 8044982080,
"sha256": "6101934ce09c25c36b6637ade43b80497499ab1b39cdb28b4e91212a64590796"
},
{
"path": "tokenizer.json",
"bytes": 11422650,
"sha256": "be75606093db2094d7cd20f3c2f385c212750648bd6ea4fb2bf507a6a4c55506"
},
{
"path": "tokenizer_config.json",
"bytes": 692,
"sha256": "1cc816812993bff176eb4f7495433b736f06fba9b6e7b05cac7b4a1780650c95"
}
]
}
}
]
}