ResearchDatasetsBERT, GPT-2, and OPT Inference WorkloadFollowBenchmarksTask NameDataset NameSOTA MethodSortMost resultsRecently updatedMost papersApplyTask NameDataset NameSOTA ResultTrendResultsInference ThroughputBERT, GPT-2, and OPT Inference Workload BS=2, SL=25657,117.4Original Throughput (tokens/s)6
Inference ThroughputBERT, GPT-2, and OPT Inference Workload BS=2, SL=25657,117.4Original Throughput (tokens/s)6