Pytorch兼容性 (v1.4)
更新时间：
复制 MD 格式
PyTorch 2.4 unit tests
Test configuration
Ubuntu 22.04
PyTorch 2.4.0
CUDA 12.3 (PPU SDK 1.4 internal build)
Python 3.10
Test results
Total test cases | 305,622 |
Total test cases run | 305,622 |
Total passes | 228,293 |
Total fails | 5,688 |
Total skips | 71,641 |
Pass rate | 97.6% |
Note
Pass rate = Total passes / (Total test cases run - Total skips)
PyTorch 2.4 operator tests
Test configuration
Ubuntu 22.04
PyTorch 2.4.0
CUDA 12.3 (PPU SDK 1.4 internal build)
Python 3.10
Operator support
Module | Total operators | PPU-supported operators |
torch | 476 | 476 |
torch.nn | 218 | 218 |
torch.nn.functional | 115 | 115 |
torch.cuda | 82 | 82 |
torch.Tensor | 526 | 526 |
torch.distributed | 48 | 48 |
torch.linalg | 41 | 41 |
torch.optim | 33 | 33 |
Total | 1,539 | 1,539 |
Operator unit test results
Total test cases | 276,930 |
Total passes | 207,396 |
Total skips | 65,687 |
Total fails | 3,847 |
Pass rate | 98.2% |
Note
Pass rate = Total passes / (Total test cases - Total skips)
该文章对您有帮助吗？