⚡ WebGPU Benchmark Results (31.32x speedup)
#75
by
robottxd
- opened
| Batch Size | WASM (fp32) | WebGPU (fp32) |
| 1 | 515.90 | 18.40 |
| 2 | 1115.60 | 68.30 |
| 4 | 2145.90 | 73.80 |
| 8 | 4032.30 | 175.70 |
| 16 | 7978.00 | 288.70 |
| 32 | 15871.50 | 506.80 |
- Model: Xenova/all-MiniLM-L6-v2
- Tests run: WASM (fp32), WebGPU (fp32)
- Sequence length: 512
- Browser: Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/123.0.0.0 Safari/537.36
- GPU: vendor=nvidia, architecture=turing, device=, description=