Conversation
|
Anyone have any thoughts on what we should do with all of the existing calls to |
|
TODO: I'm curious if we'll run into anything interesting when it comes to Accelerate and logging from multiple processes. |
Also removes logging imports from files that don't do any logging and removes the example comment. Both of those changes are just minor non-functional changes to reduce the number of lines of non-functional code. The logging module is standard to python and there are plenty of examples in the official Python docs, so I don't think we gain much by including examples in each file.
|
@eihli Thanks for the update. I had meant to respond your comment on what to with the existing print statements, but got a little busy! The changes all seemed fine to me, but I am not an expert in logging. Regarding the interaction with Accelerate and logging from multiple processes, is this something we can test through? |
Trying to differentiate between logging and printing. Logging is for informational and debug text. Printing is like a GUI. Printing is how you interact with the program on every run. That's how I'm thinking about it in my head right now.
Fix issue with previous PR. I should have made the PR into your issue branch rather than your main branch.
|
Apologies @eihli, I was away on leave and missed your previous request. Is there anything else required at this stage? |
2 participants
Added debug logging as per issue 73 for more verbose error messages. I added to all files except for
trajectory_gpt2.pyas this already had logging imported fromtransformers.utils.I kept the
logger.debug(f'foobar')as an example of usage, but happy to update this to have more file specific message if required.