[DLRMv2] Updating DLRMv2 dataset size

[keithachorn-intel]

The updated value represents the dataset size, not the 'performance_sample_count_override' value set in mlperf.conf. After adding the new submission checker test, this mismatch raises an error which was not previously present.

[github-actions]

MLCommons CLA bot All contributors have signed the MLCommons CLA ✍️ ✅

[keithachorn-intel]

If this could possibly be merged today, it would be very helpful in shipping our workload containers. Thanks! @arjunsuresh

[attafosu]

@arjunsuresh There seem to be a mismatch between the dataset size (204k) from mlcommons rules table, and what is generated (330k) from multihot_criteo.py's num_aggregated_samples
Note that it is also this count which is used as the total_samples_count by loadgen, and it makes its way into the mlperf_log_detail.txt as the qsl_reported_total_count.

[arjunsuresh]

Confirming that the value of qsl_reported_total_count is 330067 for Nvidia as well : https://raw.githubusercontent.com/mlcommons/inference_results_v5.0/refs/heads/main/closed/NVIDIA/results/GH200-GraceHopper-Superchip_GH200-144GB_aarch64x1_TRT/dlrm-v2-99/Offline/accuracy/mlperf_log_detail.txt

[arjunsuresh]

@pgmpablo157321 can you please make the relevant change in the rules?