- route decode benchmark through InferenceEngine generate path - add enable_cuda_graph toggle to engine, scheduler, and executor - make benchmark --cuda-graph/--no-cuda-graph control the toggle - hoist local time imports to module top
- route decode benchmark through InferenceEngine generate path - add enable_cuda_graph toggle to engine, scheduler, and executor - make benchmark --cuda-graph/--no-cuda-graph control the toggle - hoist local time imports to module top