Missing gpu_batch_size in NVIDIA custom system script

mlcommons / inference_results_v2.0

This repository contains the results and code for the MLPerf™ Inference v2.0 benchmark.

https://mlcommons.org/en/inference-datacenter-20/

Apache License 2.0

9 stars 12 forks source link

Missing gpu_batch_size in NVIDIA custom system script #6

Closed mahmoodn closed 2 years ago

mahmoodn commented 2 years ago

Hi It seems that the NVIDIA script add_custom_system.py doesn't add gpu_batch_size in the generated config files. That is needed by benchmarks and therefore, has to be added manually.

nv-alicheng commented 2 years ago

Hi mahmoodn,

I believe this bug was fixed internally, as I cannot reproduce it. The fix will be released with the v2.1 submission.