Skip to content

Conversation

@0cc4m
Copy link
Collaborator

@0cc4m 0cc4m commented Jan 30, 2024

Two small fixes for bugs I had missed during development. And I had missed putting Vulkan into the common.cpp dump alongside the other backends.

Fixes #5179
Fixes #5203
Fixes #5217

@stduhpf
Copy link
Contributor

stduhpf commented Jan 30, 2024

It doesn't seem to fix #5217. It still behaves pretty much the same.

@0cc4m
Copy link
Collaborator Author

0cc4m commented Jan 30, 2024

It doesn't seem to fix #5217. It still behaves pretty much the same.

Really? I tried interactive with your commands and it was fine.

@stduhpf
Copy link
Contributor

stduhpf commented Jan 30, 2024

It doesn't seem to fix #5217. It still behaves pretty much the same.

Really? I tried interactive with your commands and it was fine.

Ah that's very strange then. Maybe it's a GPU architecture-dependant thing, or something is broken with my hardware...

@stduhpf
Copy link
Contributor

stduhpf commented Jan 31, 2024

It's fixed now! Thank you.

@0cc4m 0cc4m merged commit f8e9140 into master Jan 31, 2024
@0cc4m 0cc4m deleted the 0cc4m/vulkan-fixes branch January 31, 2024 10:44
@teleprint-me
Copy link
Contributor

F16 is working now. Just needs some optimization. CPU is much faster. I'm just happy that it's not crashing anymore. Thank you @0cc4m.

jordankanter pushed a commit to jordankanter/llama.cpp that referenced this pull request Feb 3, 2024
* Fix Vulkan F16 models

* Fix Vulkan context shift crash

* Add Vulkan to common.cpp dump_non_result_info_yaml function

* Fix bug in Vulkan CPY op

* Fix small matrix multiplication errors in AMD GPUs on Windows or with amdvlk

Co-authored-by: Engininja2 <[email protected]>

---------

Co-authored-by: Engininja2 <[email protected]>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

6 participants