可使用
./tests/benchmark/tts_benchmark.py
复现
测试平台
- GPU:
GeForce RTX 2080 Ti
- CPU:
3.4hz 24core
以下为 batch size 为 8 时的结果,完整扫描看 performance_results.csv
Batch size | Use decoder | Half precision | Compile model | Use CPU | GPU Memory | Duration | RTF |
---|---|---|---|---|---|---|---|
8 | ✅ | ❌ | ✅ | ❌ | 1.72 | 36.78 | 0.22 |
8 | ✅ | ✅ | ✅ | ❌ | 0.89 | 39.34 | 0.24 |
8 | ❌ | ❌ | ✅ | ❌ | 1.72 | 36.78 | 0.23 |
8 | ❌ | ✅ | ✅ | ❌ | 0.90 | 39.34 | 0.24 |
8 | ❌ | ❌ | ❌ | ❌ | 1.70 | 36.78 | 0.29 |
8 | ✅ | ❌ | ❌ | ❌ | 1.72 | 36.78 | 0.29 |
8 | ❌ | ✅ | ❌ | ❌ | 1.02 | 35.75 | 0.40 |
8 | ✅ | ✅ | ❌ | ❌ | 0.95 | 35.75 | 0.40 |
8 | ❌ | ❌ | ❌ | ✅ | N/A | 49.92 | 0.58 |
8 | ❌ | ❌ | ✅ | ✅ | N/A | 49.92 | 0.58 |
8 | ✅ | ❌ | ✅ | ✅ | N/A | 49.92 | 0.58 |
8 | ✅ | ❌ | ❌ | ✅ | N/A | 49.92 | 0.60 |
8 | ❌ | ✅ | ❌ | ✅ | N/A | N/A | N/A |
8 | ❌ | ✅ | ✅ | ✅ | N/A | N/A | N/A |
8 | ✅ | ✅ | ❌ | ✅ | N/A | N/A | N/A |
8 | ✅ | ✅ | ✅ | ✅ | N/A | N/A | N/A |