Open VertexC opened 3 years ago
Memory usage continue to using by monitored with
running torch2tvm round 0 duration: 0.24
name:torch2tvm model:resnet50 batch_size:1 params:{'backend': 'llvm'}: (0.23501062393188477,)
Used Memory: 1189.6640625 MB
running torch2tvm round 0 duration: 0.00
name:torch2tvm model:resnet50 batch_size:2 params:{'backend': 'cuda'}: (9.5367431640625e-07,)
Used Memory: 1376.0703125 MB
running torch2tvm round 0 duration: 0.00
name:torch2tvm model:resnet50 batch_size:2 params:{'backend': 'llvm'}: (9.5367431640625e-07,)
Used Memory: 1462.44140625 MB
running torch2tvm round 0 duration: 0.01
name:torch2tvm model:vgg16 batch_size:1 params:{'backend': 'cuda'}: (0.014052629470825195,)
Used Memory: 1731.8515625 MB
Seem to have memory leak https://discuss.tvm.apache.org/t/memory-leak-how-to-free-memory-manually-after-building-a-operator-for-long-running-building-case/8037 https://github.com/apache/tvm/issues/6590
Get error while run multiple tasks together by executor
but okay with python3 infer_perf/torch2tvm.py mobilenet --batch=2 --size=2