Option Type/Default Description
autotuning::allow_lower_precision_training bool/false Enables to use FP16 in FP32 convolutions.
autotuning::max_runs int/100 Max number of runs per library and layer.
autotuning::not_improved int/5 Number of runs without improvement when autotuning will be stopped.
compiler::debug bool/false Generates code with debug symbols, prints execution times per cluster and inserts assertions to check correctness of memory allocations/frees.
compiler::debug_graph bool/false Generates SVGs in .sol/debug with the network structure.
compiler::debug_memory_consumption bool/false Generates memory consumption estimations in .sol/debug.
compiler::debug_text bool/false Textural network output in .sol/debug.
compiler::name string/Unknown Name used for the debug files.
compiler::performance_warning bool/false Shows warning, if chose hyper-parameters have negative impact on performance.
compiler::profile bool/false Shows performance of fused layers.
compiler::remove_unused_params bool/false Removes unused model parameters.
conv::sampling bool/true
dfp::debug bool/false Generates SVGs for the DFP computation graph.
dfp::debug_details bool/false Shows more detailed information in the DFP computation graphs.
heuristic::[backend]::[layer]::[inf, fwd, bwd, fil] integer Defines the heuristic value used by [backend] for [layer]
jit::debug bool/false Enables debugging in the generated code.
onnx::debug bool/false Generates SVGs for the ONNX computation graph.
reporting bool/false Enables GCC/NCC reporting on the generated code.
tvmc::tune bool/false Tune the computation graph platform specifically (compilation will take significantly longer).