WandBLogger and TensorboardLogger have different APIs for logging audio #2791
Description
🚀 Feature
The following code doesn't work:
logger = WandBLogger()
logger.writer
This is how you would typically add audio with a tensorboard logger:
logger.writer.add_audio('mixture', x.t(), engine.state.epoch)
The workaround (similar to discussed in Lightning-AI/pytorch-lightning#7028) would be to use the underlying _wandb object:
logger._wandb.log({"validation": [wandb.Audio(x.t(), caption="mixture", sample_rate=44100)]})
logger._wandb.log({"validation": [wandb.Audio(x.t(), caption="vocals", sample_rate=44100)]})
Is there a proposal for a standardized media logging API?