No matching distribution found for tfdlpack-gpu

[zhangxin9988]

hi,I want to use dgl with backend of tensorflow. However,I can't install tfdlpack in my computer,below is the error report:
(base) ### C:\Users\zx\PycharmProjects\Deep_learning>pip install tfdlpack-gpu
Collecting tfdlpack-gpu.
ERROR: Could not find a version that satisfies the requirement tfdlpack-gpu (from versions: none)
ERROR: No matching distribution found for tfdlpack-gpu
os: windows10
tensorflow_version: 2.0.0
cuda: 10.0.130
dgl_version: dgl-cu100

[awthomp]

I believe tfdlpack-gpu is only supported in Linux.

[zphy]

Similar issue encountered here trying to install on mac.

[VoVAllen]

@zphy it's already supported by the latest nightly tensorflow. (pip install tf-nightly) As the API is tf.experimental.dlpack.to_dlpack/from_dlpack

[Programmergg]

you need change some settings.1.update your tensorflow at least 2.1.0,mine is 2.3.0 2.from dgl/backend/tensorflow find tensor.py.then change import tfdlpack to import tensorflow.experimental.dlpack as tfdlpack. 3.you can import dgl dgl.backend.set_default_backend('tensorflow') or pytorch and mxnet.now youcan use it.