Extend CI for i8mm kernels as well #16234
Conversation
|
Hi @ggerganov , could you trigger the CI again? |
|
@ggerganov thanks for running the CI! Let me know if I can make any more changes |
|
|
||
| float result = INFINITY; | ||
| qfns_cpu->vec_dot(test_size, &result, 0, tmp_q1.data(), 0, tmp_q2.data(), 0, 1); | ||
| qfns_cpu->vec_dot(test_size, &result, 0, tmp_q1.data(), 0, tmp_q2.data(), 0, nrc); |
There was a problem hiding this comment.
I think we need to expand the input data depending on nrc?
There was a problem hiding this comment.
Sure. Can you tell me an appropriate test_size for nrc=2 ? Would it just be double?
|
I went through the logs of some of the failed tests and it looks like they may be unrelated to the modified CI in this PR. @ggerganov or the team, can you help me with this/ tell me if I'm missing something? Thank you! |
|
@Rohanjames1997 , how about you rebase the PR to the main? that will trigger CI and you can check if the current failures have been fixed on the mainline already. |
|
Some CI failed again due to the Cloudflare/Github outage yesterday. |
3 participants
The i8mm kernels require
nrc == 2to actually get triggered.But currently, the CI's test-quantize-fns.cpp tests only the scenario where
nrc=1.