oneflow

oneflow copied to clipboard

## Summary When attempting to create or fill `IntTensor` with unsupported data types (e.g., strings or floats), OneFlow does not raise a Python-level error, but instead leads to a **segmentation...

deepliao

## Summary Can't import oneflow ## Code to reproduce bug  Just import oneflow in python code. ## System Information - What is your OneFlow installation (pip, source, dockerhub):python3 -m...

eayicati

Got Exception "libcudnn_cnn_train-fb9074d0.so.8.9.0: undefined symbol: _ZN5cudnn3cnn5infer22queryClusterPropertiesERPhS3_, version libcudnn_cnn_infer.so.8"

2

### Description When I install onediff that depend on oneflow,I get the above exception. Intall Command: pip install --pre oneflow -f https://oneflow-pro.oss-cn-beijing.aliyuncs.com/branch/community/cu121 --no-deps My pytorch version is 2.1.2+cu121

wangmouren609

### Description Major torch releases have cudnn 9.X backend, so I wonder is there a plan to support it? Without having to manually install older cudnn. ### Alternatives _No response_

dimentary

## Code to reproduce bug ``` import oneflow as flow lamb = flow.optim.LAMB(flow.nn.Parameter([float('nan')])) lamb.step() ``` output: ``` Stack trace (most recent call last): Object "/opt/conda/lib/python3.8/site-packages/oneflow/_oneflow_internal.cpython-38-x86_64-linux-gnu.so", at 0x7f35d91282cd, in Object "/opt/conda/lib/python3.8/site-packages/oneflow/_oneflow_internal.cpython-38-x86_64-linux-gnu.so",...

deepliao

### Description ## Summary The crash is triggered when the start value is set to Nan. ## Code to reproduce bug ``` import oneflow as flow output_tensor_nan_start = flow.logspace(start=float('nan'), end=1.0,...

deepliao

### Description ## Code to reproduce bug ``` import oneflow as flow result = flow.linalg.vector_norm(flow.zeros((0,)), ord=2, dim=None, keepdim=False) ``` output: ``` Stack trace (most recent call last) in thread 1756:...

deepliao

### Description A crash is triggered when value is set to a negative value. ## Code to reproduce bug ``` import oneflow as flow import numpy as np negative_number =...

deepliao

### Description ## A crash is triggered when K is set to a negative value. ## Code to reproduce bug ``` import oneflow as flow topk_result = flow.topk(input=flow.randn(2, 100), k=-100)...

deepliao

fix autocast to support global tensor

6

oneflow autocast 不支持 global tensor。下面代码直接报错 ```python import oneflow as flow placement = flow.placement("cuda", ranks=[0]) sbp = flow.sbp.broadcast a = flow.randn(2, 3).to_global(placement=placement, sbp=sbp) b = flow.randn(3, 4).to_global(placement=placement, sbp=sbp) with flow.autocast(device_type="cuda"): c...

fpzh2011