train pixels [108,768]; text [9,6]; embedding [B,16]; logits [9,9] test scores [45,9]; text-to-image scores [9,45] checks {"gradient_parameters": 32, "gradient_max_abs_error": 1.2918817127172133e-09, "duplicate_semantic_hit@1": 1.0, "duplicate_identity_hit@1": 0.3333333333333333, "duplicate_diagonal_ce": 1.0989754221652592, "duplicate_positive_mass_ce": 0.00036313349714963973, "duplicate_ce_gap": 1.0986122886681096} {"seed": 71, "baseline": {"i2t": {"hit@1": 0.022222222222222223, "relevant_recall@1": 0.022222222222222223, "hit@3": 0.17777777777777778, "relevant_recall@3": 0.17777777777777778, "hit@5": 0.3333333333333333, "relevant_recall@5": 0.3333333333333333, "mrr": 0.20538800705467372}, "t2i": {"hit@1": 0.0, "relevant_recall@1": 0.0, "hit@3": 0.0, "relevant_recall@3": 0.0, "hit@5": 0.2222222222222222, "relevant_recall@5": 0.044444444444444446, "mrr": 0.09494996380842674}}, "aligned": {"last_batch_loss": 0.0034472630868336647, "i2t": {"hit@1": 0.8222222222222222, "relevant_recall@1": 0.8222222222222222, "hit@3": 0.9777777777777777, "relevant_recall@3": 0.9777777777777777, "hit@5": 1.0, "relevant_recall@5": 1.0, "mrr": 0.8970370370370371}, "t2i": {"hit@1": 1.0, "relevant_recall@1": 0.2, "hit@3": 1.0, "relevant_recall@3": 0.5777777777777777, "hit@5": 1.0, "relevant_recall@5": 0.7999999999999999, "mrr": 1.0}}, "mismatched": {"last_batch_loss": 0.0034977708075967597, "i2t": {"hit@1": 0.0, "relevant_recall@1": 0.0, "hit@3": 0.3111111111111111, "relevant_recall@3": 0.3111111111111111, "hit@5": 0.6, "relevant_recall@5": 0.6, "mrr": 0.25797178130511467}, "t2i": {"hit@1": 0.0, "relevant_recall@1": 0.0, "hit@3": 0.0, "relevant_recall@3": 0.0, "hit@5": 0.0, "relevant_recall@5": 0.0, "mrr": 0.07785860744194079}}} {"seed": 72, "baseline": {"i2t": {"hit@1": 0.044444444444444446, "relevant_recall@1": 0.044444444444444446, "hit@3": 0.26666666666666666, "relevant_recall@3": 0.26666666666666666, "hit@5": 0.7111111111111111, "relevant_recall@5": 0.7111111111111111, "mrr": 0.2848059964726631}, "t2i": {"hit@1": 0.2222222222222222, "relevant_recall@1": 0.044444444444444446, "hit@3": 0.2222222222222222, "relevant_recall@3": 0.06666666666666668, "hit@5": 0.3333333333333333, "relevant_recall@5": 0.1111111111111111, "mrr": 0.30766036432703103}}, "aligned": {"last_batch_loss": 0.00374202494596602, "i2t": {"hit@1": 0.8888888888888888, "relevant_recall@1": 0.8888888888888888, "hit@3": 1.0, "relevant_recall@3": 1.0, "hit@5": 1.0, "relevant_recall@5": 1.0, "mrr": 0.937037037037037}, "t2i": {"hit@1": 1.0, "relevant_recall@1": 0.2, "hit@3": 1.0, "relevant_recall@3": 0.6, "hit@5": 1.0, "relevant_recall@5": 0.9555555555555555, "mrr": 1.0}}, "mismatched": {"last_batch_loss": 0.004547159430054881, "i2t": {"hit@1": 0.044444444444444446, "relevant_recall@1": 0.044444444444444446, "hit@3": 0.28888888888888886, "relevant_recall@3": 0.28888888888888886, "hit@5": 0.7111111111111111, "relevant_recall@5": 0.7111111111111111, "mrr": 0.3049206349206349}, "t2i": {"hit@1": 0.0, "relevant_recall@1": 0.0, "hit@3": 0.0, "relevant_recall@3": 0.0, "hit@5": 0.0, "relevant_recall@5": 0.0, "mrr": 0.0879135268024157}}} {"seed": 73, "baseline": {"i2t": {"hit@1": 0.022222222222222223, "relevant_recall@1": 0.022222222222222223, "hit@3": 0.4, "relevant_recall@3": 0.4, "hit@5": 0.6222222222222222, "relevant_recall@5": 0.6222222222222222, "mrr": 0.28247795414462085}, "t2i": {"hit@1": 0.1111111111111111, "relevant_recall@1": 0.022222222222222223, "hit@3": 0.1111111111111111, "relevant_recall@3": 0.044444444444444446, "hit@5": 0.2222222222222222, "relevant_recall@5": 0.06666666666666668, "mrr": 0.21079512051734273}}, "aligned": {"last_batch_loss": 0.003528784487823118, "i2t": {"hit@1": 0.9555555555555556, "relevant_recall@1": 0.9555555555555556, "hit@3": 1.0, "relevant_recall@3": 1.0, "hit@5": 1.0, "relevant_recall@5": 1.0, "mrr": 0.9777777777777777}, "t2i": {"hit@1": 1.0, "relevant_recall@1": 0.2, "hit@3": 1.0, "relevant_recall@3": 0.6, "hit@5": 1.0, "relevant_recall@5": 0.9555555555555555, "mrr": 1.0}}, "mismatched": {"last_batch_loss": 0.003683168753344475, "i2t": {"hit@1": 0.0, "relevant_recall@1": 0.0, "hit@3": 0.4888888888888889, "relevant_recall@3": 0.4888888888888889, "hit@5": 0.7111111111111111, "relevant_recall@5": 0.7111111111111111, "mrr": 0.29356261022927693}, "t2i": {"hit@1": 0.0, "relevant_recall@1": 0.0, "hit@3": 0.0, "relevant_recall@3": 0.0, "hit@5": 0.2222222222222222, "relevant_recall@5": 0.044444444444444446, "mrr": 0.13761022927689595}}} aggregate {"baseline": {"i2t": {"hit@1_mean": 0.02962962962962963, "hit@1_sample_sd": 0.012830005981991684}, "t2i": {"hit@1_mean": 0.1111111111111111, "hit@1_sample_sd": 0.1111111111111111}}, "aligned": {"i2t": {"hit@1_mean": 0.8888888888888888, "hit@1_sample_sd": 0.06666666666666671}, "t2i": {"hit@1_mean": 1.0, "hit@1_sample_sd": 0.0}}, "mismatched": {"i2t": {"hit@1_mean": 0.014814814814814815, "hit@1_sample_sd": 0.02566001196398337}, "t2i": {"hit@1_mean": 0.0, "hit@1_sample_sd": 0.0}}}