{ "gpu": "Tesla T4", "total_vram_gb": 14.56317138671875, "target_vram_gb": 13.5, "hard_ceiling_fraction": 0.965, "hard_ceiling_gb": 14.053460388183593, "selected_inference_batch": 527, "selected_peak_reserved_gb": 14.041015625, "selected_peak_allocated_gb": 12.066949844360352, "measurements": [ { "batch_size": 64, "success": true, "peak_allocated_gb": 2.224550247192383, "peak_reserved_gb": 2.7890625, "above_target": false, "below_ceiling": true }, { "batch_size": 128, "success": true, "peak_allocated_gb": 3.584657669067383, "peak_reserved_gb": 4.666015625, "above_target": false, "below_ceiling": true }, { "batch_size": 256, "success": true, "peak_allocated_gb": 6.304872512817383, "peak_reserved_gb": 8.41796875, "above_target": false, "below_ceiling": true }, { "batch_size": 512, "success": true, "peak_allocated_gb": 11.745302200317383, "peak_reserved_gb": 13.666015625, "above_target": true, "below_ceiling": true }, { "batch_size": 520, "success": true, "peak_allocated_gb": 11.916261672973633, "peak_reserved_gb": 13.86328125, "above_target": true, "below_ceiling": true }, { "batch_size": 524, "success": true, "peak_allocated_gb": 12.001253128051758, "peak_reserved_gb": 13.962890625, "above_target": true, "below_ceiling": true }, { "batch_size": 526, "success": true, "peak_allocated_gb": 12.04667854309082, "peak_reserved_gb": 14.013671875, "above_target": true, "below_ceiling": true }, { "batch_size": 527, "success": true, "peak_allocated_gb": 12.066949844360352, "peak_reserved_gb": 14.041015625, "above_target": true, "below_ceiling": true }, { "batch_size": 527, "success": true, "peak_allocated_gb": 12.066949844360352, "peak_reserved_gb": 14.041015625, "above_target": true, "below_ceiling": true }, { "batch_size": 528, "success": true, "peak_allocated_gb": 12.086244583129883, "peak_reserved_gb": 14.0625, "above_target": true, "below_ceiling": false }, { "batch_size": 544, "success": false, "above_target": false, "below_ceiling": false, "error": "CUDA OOM" }, { "batch_size": 576, "success": false, "above_target": false, "below_ceiling": false, "error": "CUDA OOM" }, { "batch_size": 640, "success": false, "above_target": false, "below_ceiling": false, "error": "CUDA OOM" }, { "batch_size": 768, "success": false, "above_target": false, "below_ceiling": false, "error": "CUDA OOM" }, { "batch_size": 1024, "success": false, "above_target": false, "below_ceiling": false, "error": "CUDA OOM" } ] }