00:43:02	Jasmine C.:	Is all of this in the context of backpropagation — circuits can also form under other optimization schemes (??) ? 

Trying to grasp the qualities of a “circuit” (human meaningful concept?) and what is stable and changeable.
00:45:35	Jasmine C.:	Replying to "Is all of this in th..."

is it that the computation associated with backdrop is good for error attribution (due to how it’s computed)
00:52:48	Jasmine C.:	I guess there can be many roads to Rome 🙂
00:53:40	Jesseba Fernando:	Replying to "I guess there can be..."

don’t ALL roads lead to Rome? 😆
00:54:13	Jasmine C.:	Replying to "I guess there can be..."

will report back once I’ve finished testing them 😀
00:54:47	Jesseba Fernando:	Reacted to "will report back onc..." with ❤️
00:59:52	Jasmine C.:	https://github.com/huggingface/transformers/blob/main/src/transformers/models/gpt_oss/modeling_gpt_oss.py
01:01:32	Jasmine C.:	^the forward() is defined here if anyone needs this, it’s a useful reference too for me :)
01:08:09	Jasmine C.:	Oh, like from before we talked about what is sufficient v. complete?
01:27:24	Jasmine C.:	Oh is the test like “there isn’t underlying structure you can find” so if you do that is a problem
01:42:29	Jasmine C.:	ahhh
01:43:07	Jasmine C.:	Do you have any more advice for avoiding measurement artifacts, like that what you see in your microscope is real and not just dirt on your slide