问题如下：
ERROR: pip’s dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts.
nbconvert 7.16.4 requires beautifulsoup4, which is not installed.
nbconvert 7.16.4 requires bleach!=5.0.0, which is not installed.
nbconvert 7.16.4 requires defusedxml, which is not installed.
nbconvert 7.16.4 requires jinja2>=3.0, which is not installed.
nbconvert 7.16.4 requires jupyter-core>=4.7, which is not installed.
nbconvert 7.16.4 requires jupyterlab-pygments, which is not installed.
nbconvert 7.16.4 requires mistune<4,>=2.0.3, which is not installed.
nbconvert 7.16.4 requires nbclient>=0.5.0, which is not installed.
nbconvert 7.16.4 requires nbformat>=5.7, which is not installed.
nbconvert 7.16.4 requires pandocfilters>=1.4.1, which is not installed.
nbconvert 7.16.4 requires pygments>=2.4.1, which is not installed.
nbconvert 7.16.4 requires tinycss2, which is not installed.
nbconvert 7.16.4 requires traitlets>=5.1, which is not installed.

解决：

pip install -U nbconvert

更新最新版本的nbconvert即可，因为安装的Tensorflow-gpu版本很新，故相关nbconvert直接更新最新的，理论上可以适配

更新后一没有报error说conflict，同时仔细观察发现要求的依赖版本也都满足。
所以视为可行方案。
如果TensorFlow和nbconvert有版本冲突的话，可以查看版本对应关系。此处是取巧的方法