exp run: option to append unique id to duplicate exp names

For example, right now the behavior is:

$ dvc exp run -n alexnet
Reproduced experiment(s): alexnet
$ dvc exp run -n alexnet
ERROR: Reproduced experiment conflicts with existing experiment alexnet

Desired behavior would be:

$ dvc exp run -n alexnet
Reproduced experiment(s): alexnet
$ dvc exp run -n alexnet
WARNING: Reproduced experiment conflicts with existing experiment alexnet, so unique ID was appended to the experiment name.
Reproduced experiment(s): alexnet-fds23

There might need to be some option to either strictly fail or use a unique ID.

Extracted from https://github.com/iterative/dvc/issues/6071#issuecomment-924804238.

For anyone interested, please note the workaround mentioned by @iesahin:

dvc exp run -n myprefix-${RANDOM}

[dberenbaum]

Thanks @karajan1001, that makes it less problematic, although the original feature request from the user in #6071 was to alleviate the burden on users to be responsible for making unique names. It’s really just one implementation of #6071. I don’t see it as a likely feature to implement anytime soon, so we could set as low priority or move it to a discussion.

[dberenbaum]

Similarly, there are workarounds for #6071, right? Like:

lr=0.5
dvc exp run -n lr=$lr -S lr=$lr

There could be all kinds of shortcuts to make this easier. Maybe it’s better to focus on a doc or blog post showing off how these kinds of tricks could make for a more robust way to organize experiments?