{ "version": "v0.4", "git_tag": "model-v0.4", "built_from_commit": "a6b4e8af3290255ce861e63397a0ec399b9f49c8", "round": "r21-soft", "stage": "train", "training_commit": "d228b648f8a36628e46fd0e1e212f84af17522e2+dirty", "training_started": "2026-10-22 03:03:02", "stages": [ { "name": "train", "out": "runs/r21-soft/train", "init": "runs/r20-science/train" } ], "training_config": { "name": "r21-soft", "device": "cuda", "common": { "log_every": 100 }, "stages": [ { "name": "train", "kind": "train", "init": "runs/r20-science/train", "data": "data/n2/soft_mix.jsonl", "val": "data/v5/dev_mix.jsonl", "eval": { "sqa": "data/n2/dev_sqa.jsonl", "v7w": "data/n2/dev_v7w.jsonl", "spatial": "data/n2/dev_spatial.jsonl", "count": "data/n2/dev_count.jsonl", "iconqa": "data/n2/dev_iconqa.jsonl", "tallyqa": "data/n2/dev_tallyqa.jsonl", "sqa_synth": "data/n2/dev_sqa_synth.jsonl", "clevr": "data/n2/dev_clevr.jsonl" }, "params": { "epochs": 1, "lr_new": 2e-05, "lr_head": 1e-05, "lr_lora": 5e-05, "warmup": 300, "eval_every": 5000, "vision_unfreeze_top": 4, "lr_vision": 5e-06, "seed": 0 } } ], "evaluate": { "image_identity": "data/v5/image_identity.json", "controls": true, "sets": { "dev_mix": "data/v5/dev_mix.jsonl", "test_vsr": "data/v5/test_vsr.jsonl", "test_v7w": "data/v5/test_v7w.jsonl", "bench_pope": "data/v2/bench_pope.jsonl", "test_vqa_yesno": "data/v2/test_vqa_yesno.jsonl", "test_vqa_choice": "data/v2/test_vqa_choice.jsonl", "test_gqa": "data/v2/test_gqa.jsonl", "test_exist": "data/v2/test_exist.jsonl", "test_position": "data/v2/test_position.jsonl", "test_relation": "data/v2/test_relation.jsonl", "test_size": "data/v2/test_size.jsonl", "test_position_flip": "data/v6/test_position_flip.jsonl", "test_relation_flip": "data/v6/test_relation_flip.jsonl", "bench_lv_vqav2_yesno": "data/lv_bench/vqav2_yesno.jsonl", "bench_lv_aokvqa": "data/lv_bench/aokvqa.jsonl", "bench_lv_scienceqa": "data/lv_bench/scienceqa.jsonl", "bench_lv_pope": "data/lv_bench/pope.jsonl", "bench_whatsup": { "path": "data/r13/bench/whatsup.jsonl", "role": "heldout" }, "bench_countbench": { "path": "data/r13/bench/countbench.jsonl", "role": "heldout" }, "bench_cogent_b": { "path": "data/r13/bench/cogent_b.jsonl", "role": "heldout" }, "test_count_coco": { "path": "data/n2/test_count_coco.jsonl", "role": "heldout" }, "bench_mmbench": { "path": "data/n2/bench/mmbench.jsonl", "role": "heldout" }, "bench_cvbench": { "path": "data/n2/bench/cvbench.jsonl", "role": "heldout" }, "bench_mme": { "path": "data/n2/bench/mme.jsonl", "role": "heldout" }, "bench_naturalbench": { "path": "data/n2/bench/naturalbench.jsonl", "role": "heldout" }, "dev_sqa": "data/n2/dev_sqa.jsonl" } } }, "source_checkpoint_sha256": "fda0f969a79cf77e6bb9288ddd38e4bb377f3f677018e67275ef927bf76b0397" }