Cannot set model_version

[gcaillaut]

Is there a way to specify the model_version field? I cannot find how to setup this field at the model level. It seems that it can be set on the prediction level by specifying model_version next to to result fied in the JSON output of the model, but it is ignored by Label-Studio in favour of the randomly generated version number.

[KonstantinKorotaev]

Hi @GaaH You can modify it through import\export:

1. Export your tasks
2. Modify model_version in json file
3. Import your tasks into Label Studio

[Goschjann]

running into the same problem as well - is there any other way then the heuristic described by @KonstantinKorotaev (thanks)?

[makseq]

@Goschjann do you use ML Backend and it generates predictions? As I remember you have to specify model_version in the dict that will be returned after model training in fit().