is the training running in a subprocess from the one that starts the ClearML task? I've noticed a bug that subprocesses ignore the parent's auto_connect_frameworks settings and always run with all frameworks
Answered
Hi! I Am Running A Clearml Task. It Is Training A Transformer Model Using Huggingface. I Have Disabled All Callbacks From The Trainer, And In The Task.Init Have Set All The Auto_Connect_Frameworks To False. However, The Model Checkpoints, The Optimizer, S
Hi! I am running a clearml task. It is training a transformer model using huggingface. I have disabled all callbacks from the trainer, and in the task.init have set all the auto_connect_frameworks to False. However, the model checkpoints, the optimizer, scheduler, scaler and random state are still logged in the artifacts? How do I disable the artifacts logging?
362 Views
1
Answer
17 days ago
17 days ago
Tags