brainevent.init_csr_config#
- brainevent.init_csr_config(*, save=True, device_kind=None, candidates=(HybridConfig(block_size=128, fixed_scatter_blocks=1024, tpr_threshold=128, task_nnz=2048), HybridConfig(block_size=128, fixed_scatter_blocks=2048, tpr_threshold=128, task_nnz=4096), HybridConfig(block_size=256, fixed_scatter_blocks=1024, tpr_threshold=128, task_nnz=4096), HybridConfig(block_size=256, fixed_scatter_blocks=2048, tpr_threshold=128, task_nnz=4096), HybridConfig(block_size=256, fixed_scatter_blocks=4096, tpr_threshold=128, task_nnz=4096), HybridConfig(block_size=512, fixed_scatter_blocks=2048, tpr_threshold=128, task_nnz=4096), HybridConfig(block_size=512, fixed_scatter_blocks=4096, tpr_threshold=128, task_nnz=4096), HybridConfig(block_size=256, fixed_scatter_blocks=2048, tpr_threshold=256, task_nnz=4096), HybridConfig(block_size=128, fixed_scatter_blocks=2048, tpr_threshold=512, task_nnz=2048), HybridConfig(block_size=256, fixed_scatter_blocks=1024, tpr_threshold=512, task_nnz=2048), HybridConfig(block_size=256, fixed_scatter_blocks=2048, tpr_threshold=512, task_nnz=2048), HybridConfig(block_size=512, fixed_scatter_blocks=2048, tpr_threshold=512, task_nnz=2048), HybridConfig(block_size=256, fixed_scatter_blocks=2048, tpr_threshold=512, task_nnz=4096), HybridConfig(block_size=256, fixed_scatter_blocks=2048, tpr_threshold=1024, task_nnz=4096), HybridConfig(block_size=256, fixed_scatter_blocks=4096, tpr_threshold=128, task_nnz=1024), HybridConfig(block_size=256, fixed_scatter_blocks=2048, tpr_threshold=128, task_nnz=8192)), **benchmark_kwargs)[source]#
Initialize the per-GPU CSR config: benchmark candidates and (by default) persist the winner.
Run this once manually before simulating; later processes read the saved config via
get_hybrid_config(). Returns the winningHybridConfig.- Return type: