Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Support vec_dot_q and dequantize_row_q for GGML_TYPE_Q8_1 #200

Closed
LoganDark opened this issue May 27, 2023 · 1 comment
Closed

Support vec_dot_q and dequantize_row_q for GGML_TYPE_Q8_1 #200

LoganDark opened this issue May 27, 2023 · 1 comment

Comments

@LoganDark
Copy link
Contributor

I had a crack at implementing a quantizer for this (as part of RWKV/rwkv.cpp#74) in order to try to run rwkv.cpp with it, but it looks like vec_dot_q at the very least is not implemented, resulting in a NULL pointer dereference at runtime. AKA inference isn't fully supported with this type.

Since there are a couple TODO comments implying that these might be implementable in the future, and I couldn't find any other tracking issues mentioning this, I'm opening this issue to show that at least someone cares.

I'm not well-versed in SIMD intrinsics or the inner workings of this library so I can't implement these operators atm.
@LoganDark LoganDark mentioned this issue May 27, 2023
Merged
@LoganDark
Copy link
Contributor Author

closing because this type does not seem intended / useful for inference (RWKV/rwkv.cpp#74 (comment))

@LoganDark LoganDark closed this as not planned Won't fix, can't repro, duplicate, stale May 27, 2023
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

No branches or pull requests
1 participant
@LoganDark