Dali logs model configuration even with --log-verbose "0" flag

[ioangatop]

Launching the tritonserver w/ or w/out log-verbose doesn't seems to affect dali backend. It would be nice to have the same behaviour as the others!

Here is the output with --log-verbose "0";

I0528 18:03:34.171215 1 dali_backend.cc:316] TRITONBACKEND_ModelInitialize: dali (version 1)
I0528 18:03:34.171221 1 dali_backend.cc:328] Repository location: /tmp/folderAQX4gi
I0528 18:03:34.171224 1 dali_backend.cc:339] backend state is 'backend state'
I0528 18:03:34.171954 1 dali_backend.cc:126] Loading DALI pipeline from file /tmp/folderAQX4gi/1/model.dali
I0528 18:03:34.171996 1 dali_backend.cc:79] model configuration:
{
    "name": "dali",
    "platform": "",
    "backend": "dali",
    "version_policy": {
        "latest": {
            "num_versions": 1
        }
    },
    "max_batch_size": 1,
    "input": [

While for other backends it looks like:

I0528 18:03:34.211270 1 model_repository_manager.cc:1043] loading: model:1
I0528 18:03:35.426644 1 libtorch.cc:981] TRITONBACKEND_ModelInitialize: model (version 1)
I0528 18:03:35.427339 1 libtorch.cc:1022] TRITONBACKEND_ModelInstanceInitialize: model_0 (device 0)
I0528 18:03:38.321389 1 model_repository_manager.cc:1210] successfully loaded 'model' version 1

[banasraf]

Hi @ioangatop!
Indeed, we don't check the verbosity level when we print logs in DALI backend, even though we should. I added a task to fix that to our backlog. Thanks for pointing it out!