gguf-py : add support for sub_type (in arrays) in GGUFWriter add_key_value method #13561
Conversation
|
@compilade gentle ping |
|
@compilade Did NorthSec go well? :) |
|
failing checks |
It's because I'm requiring a package that doesn't exist yet. :) See CI on previous commits. |
|
@cebtenzzre If you have time a review would be appreciated, seems @compilade is on vacation after NorthSec. :) |
@CISC It was fun, but surprisingly exhausting1, so I did take a break afterwards.
Kind of :) Footnotes
|
There was a problem hiding this comment.
This seems good for arrays with a single sub-type (which in practice is the only kind of arrays).
If we ever need nested sub-types (e.g. for nested arrays), then it will probably be simple to extend the sub_type argument to be Sequence[GGUFValueType] | GGUFValueType | None (but that's not required for now).
I did not test running this (yet), but the changes seem good and I assume you tested by fixing #13557.
| if sub_type is not None: | ||
| ltype = sub_type |
There was a problem hiding this comment.
Adding some dynamic type-checking here could be useful (to check if all the elements of the array have the same type and are compatible with the given sub_type), but a traceback from this function isn't that useful because it's called when packing, not when adding the value to the writer.
Not sure what would be best here. I guess it's fine like this.
There was a problem hiding this comment.
Yeah, I don't think it's very useful, it will mostly likely be used for passing existing GGUF arrays anyway.
We can always revisit if it becomes an issue.
Well earned, glad you had fun at least. :)
No worries, welcome back! |
Sure, that will probably require a bit of work anyway. :)
Yes, and some other GGUFs for good measure. I will merge and tag later today. |
Without this it's not possible to pass on original metadata array sub-types, thus coercing them to whatever GGUFWriter decides based on the value type.
Fixes #13557