⚡ WebGPU Benchmark Results (17.82x speedup)
#68
by
hans00
- opened
| Batch Size | WASM (fp16) | WASM (fp32) | WebGPU (fp32) |
| 1 | 461.80 | 456.10 | 58.40 |
| 2 | 968.50 | 916.30 | 106.60 |
| 4 | 2025.00 | 1744.30 | 161.50 |
| 8 | 4272.70 | 3778.40 | 286.60 |
| 16 | 8781.40 | 8083.00 | 525.30 |
| 32 | 17967.20 | 15692.50 | 1008.20 |
- Model: Xenova/all-MiniLM-L6-v2
- Tests run: WASM (fp16), WASM (fp32), WebGPU (fp32)
- Sequence length: 512
- Browser: Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/122.0.0.0 Safari/537.36
- GPU: vendor=amd, architecture=gcn-5, device=, description=