{ "C1": { "model_name": "C1", "device": "cuda", "epochs_completed": 5, "batch_size": 16, "learning_rate": 0.0001, "parameter_count": 8162816, "training_seconds": 951.2637423379999, "training_minutes": 15.854395705633332, "training_batches": 1250, "samples_per_second": 21.02466341337087, "units_per_second": 10176.239847224653, "peak_memory_gb": { "allocated_gb": 0.14497995376586914, "reserved_gb": 9.294921875 }, "best_validation_loss": 5.037060260772705, "checkpoint": "outputs/checkpoints/C1/C1_best.pt", "latest_checkpoint": "outputs/checkpoints/C1/C1_latest.pt", "test_metrics": { "n_samples": 32, "bit_level_accuracy": 61.09117657430434, "sequence_accuracy": 0.0, "levenshtein_distance": 1530.375, "bleu": 0.5574633917156633, "rouge1": 5.195147860876352, "rougeL": 4.803999355784968 } }, "C2": { "model_name": "C2", "device": "cuda", "epochs_completed": 5, "batch_size": 16, "learning_rate": 0.0001, "parameter_count": 8162816, "training_seconds": 999.1606595490002, "training_minutes": 16.652677659150005, "training_batches": 1250, "samples_per_second": 20.016800910703964, "units_per_second": 9722.496484584215, "peak_memory_gb": { "allocated_gb": 0.14460420608520508, "reserved_gb": 9.294921875 }, "best_validation_loss": 5.03157976269722, "checkpoint": "outputs/checkpoints/C2/C2_best.pt", "latest_checkpoint": "outputs/checkpoints/C2/C2_latest.pt", "test_metrics": { "n_samples": 32, "bit_level_accuracy": 61.25138467068273, "sequence_accuracy": 0.0, "levenshtein_distance": 1471.46875, "bleu": 0.5273644151049959, "rouge1": 4.806079282844072, "rougeL": 4.395284162362044 } }, "C3": { "model_name": "C3", "device": "cuda", "epochs_completed": 5, "batch_size": 16, "learning_rate": 0.0001, "parameter_count": 7373312, "training_seconds": 935.0997815259998, "training_minutes": 15.584996358766663, "training_batches": 1250, "samples_per_second": 21.388091832683113, "units_per_second": 10367.23587313816, "peak_memory_gb": { "allocated_gb": 0.13343381881713867, "reserved_gb": 9.294921875 }, "best_validation_loss": 5.283080533146858, "checkpoint": "outputs/checkpoints/C3/C3_best.pt", "latest_checkpoint": "outputs/checkpoints/C3/C3_latest.pt", "test_metrics": { "n_samples": 32, "bit_level_accuracy": 61.69053910328242, "sequence_accuracy": 0.0, "levenshtein_distance": 1314.0625, "bleu": 0.5761428553819107, "rouge1": 5.4076538870892685, "rougeL": 5.021402898929718 } }, "C4": { "model_name": "C4", "device": "cuda", "epochs_completed": 5, "batch_size": 16, "learning_rate": 0.0001, "parameter_count": 8157440, "training_seconds": 949.5453518940003, "training_minutes": 15.825755864900005, "training_batches": 1250, "samples_per_second": 21.062711707352594, "units_per_second": 10246.571141223525, "peak_memory_gb": { "allocated_gb": 0.14508819580078125, "reserved_gb": 9.294921875 }, "best_validation_loss": 5.037295088171959, "checkpoint": "outputs/checkpoints/C4/C4_best.pt", "latest_checkpoint": "outputs/checkpoints/C4/C4_latest.pt", "test_metrics": { "n_samples": 32, "bit_level_accuracy": 60.87581936178038, "sequence_accuracy": 0.0, "levenshtein_distance": 1325.3125, "bleu": 0.7907915102933201, "rouge1": 7.158535269850726, "rougeL": 6.582769941576963 } }, "C5": { "model_name": "C5", "device": "cuda", "epochs_completed": 5, "batch_size": 16, "learning_rate": 0.0001, "parameter_count": 3448641, "training_seconds": 354.66554691299916, "training_minutes": 5.911092448549986, "training_batches": 1250, "samples_per_second": 56.39115548177584, "units_per_second": 57019.93942185984, "peak_memory_gb": { "allocated_gb": 0.06936883926391602, "reserved_gb": 9.294921875 }, "best_validation_loss": 2.4401081427931786, "checkpoint": "outputs/checkpoints/C5/C5_best.pt", "latest_checkpoint": "outputs/checkpoints/C5/C5_latest.pt", "test_metrics": { "n_samples": 32, "bit_level_accuracy": 67.79718048595763, "sequence_accuracy": 0.0, "levenshtein_distance": 483.96875, "bleu": null, "rouge1": null, "rougeL": null } } }