1D Convolution Performance on Conv1d Shapes 1.0 (random)
0.0126Latency (ms)KernelEvolve
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| KernelEvolvePrecision=FP16, Tensor Shape=32 × 64 × 64 × 512†2025.12 | 0.0126 | 2.19 | 1.41 | |
| KernelEvolvePrecision=FP32, Tensor Shape=32 × 64 × 64 × 512†2025.12 | 0.0157 | 1.74 | 1.26 | |
| KernelEvolvePrecision=FP16, Tensor Shape=32 × 256 × 256 × 1024†2025.12 | 0.0603 | 1.32 | 0.91 | |
| KernelEvolvePrecision=FP32, Tensor Shape=32 × 256 × 256 × 1024†2025.12 | 0.135 | 1 | 0.77 | |
| KernelEvolvePrecision=FP16, Tensor Shape=64 × 768 × 768 × 1024†2025.12 | 1.1278 | 0.63 | 0.49 | |
| KernelEvolvePrecision=FP32, Tensor Shape=64 × 768 × 768 × 1024†2025.12 | 2.645 | 0.48 | 0.39 |