I’m trying to log test metrics in wandb logger but it’s not logging that. Only train and eval metrocs are being logged successfully in wandb. How can I do that?
if training_args.do_predict:
logger.info(“*** Test ***”)
metrics = trainer.predict(vectorized_datasets[“test”])
metrics = metrics._asdict()[“metrics”]
metrics[“test_samples”] = len(vectorized_datasets[“test”])
trainer.log_metrics(“test”, metrics)
trainer.save_metrics(“test”, metrics)