⚡ WebGPU Benchmark Results (20.95x speedup)
#99
by
fantos
- opened
| Batch Size | WASM (int8) | WASM (fp16) | WASM (fp32) | WebGPU (fp16) | WebGPU (fp32) |
| 1 | 417.80 | 501.00 | 552.20 | 45.70 | 65.90 |
| 2 | 864.90 | 1331.00 | 1244.30 | 123.40 | 117.50 |
| 4 | 1766.80 | 2569.30 | 2098.90 | 157.10 | 232.30 |
| 8 | 3411.10 | 4781.60 | 4399.30 | 327.10 | 409.90 |
| 16 | 6772.20 | 11739.10 | 10643.90 | 613.70 | 842.50 |
| 32 | 15859.40 | 23870.10 | 18189.10 | 1139.20 | 1247.60 |
- Model: Xenova/all-MiniLM-L6-v2
- Tests run: WASM (int8), WASM (fp16), WASM (fp32), WebGPU (fp16), WebGPU (fp32)
- Sequence length: 512
- Browser: Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/129.0.0.0 Safari/537.36
- GPU: vendor=intel, architecture=gen-12lp, device=, description=