fix: bug with AIM server availability check

[HarikrishnanBalagopal]

Description of the change

Fixes the bug in #131

Related issue number

#131

How to verify the PR

Run an end-to-end test using https://github.com/foundation-model-stack/fms-hf-tuning/blob/main/tuning/sft_trainer.py in a venv or conda environment which has the AIM package installed. Even without an AIM server running the training should succeed.

Was the PR tested

• I have added >=1 unit test(s) for every new method I have added.
• I have ensured all unit tests pass

[dushyantbehl]

@HarikrishnanBalagopal I believe we do not need so much changes to the code. The idea is if Aim is installed but AIMSTACK_DB and AIMSTACK_SERVER are not set then we should not use Aim.

Just change this one line

  if aim_server:
        aim_callback = AimCallback(
            repo="aim://" + aim_server + "/", experiment=aim_experiment
        )
    if aim_db:
        aim_callback = AimCallback(repo=aim_db, experiment=aim_experiment)
   else:
-       aim_callback = AimCallback(experiment=aim_experiment)
+       aim_callback = None

[HarikrishnanBalagopal]

@HarikrishnanBalagopal I believe we do not need so much changes to the code. The idea is if Aim is installed but AIMSTACK_DB and AIMSTACK_SERVER are not set then we should not use Aim.

Just change this one line

  if aim_server:
        aim_callback = AimCallback(
            repo="aim://" + aim_server + "/", experiment=aim_experiment
        )
    if aim_db:
        aim_callback = AimCallback(repo=aim_db, experiment=aim_experiment)
   else:
-       aim_callback = AimCallback(experiment=aim_experiment)
+       aim_callback = None

I think that will also require a change here to avoid adding a None into the list of callbacks

fms-hf-tuning/tuning/sft_trainer.py

Line 247 in c2f2f8c

callbacks.append(get_aimstack_callback())

[dushyantbehl]

Wait for #89 which fixes this.

[HarikrishnanBalagopal]

Wait for #89 which fixes this.

Sure, I have made the requested change in this PR as well.

[anhuong]

if this is in #89 perhaps we can close this PR

[ashokponkumar]

@HarikrishnanBalagopal If #89 fixes this issue, lets close this PR. Thanks @anhuong !