]> git.djapps.eu Git - pkg/ggml/sources/llama.cpp/commit
overview / pkg / ggml / sources / llama.cpp / commit
summary | shortlog | log | commit | commitdiff | tree
(parent: e4e6f67)
Vulkan Embedding Fix (#7360)
author0cc4m <redacted>
Sun, 19 May 2024 15:19:53 +0000 (17:19 +0200)
committerGitHub <redacted>
Sun, 19 May 2024 15:19:53 +0000 (17:19 +0200)
commitf030ec1f7a72aa825b2104823946551b9ec5dfc1
tree705d08bb8df40049ba48bf57629accbd0b4ee953tree
parente4e6f67be6a8a697f5f89a28c98934e53c99c359commit | diff
Vulkan Embedding Fix (#7360)

* Fix empty Vulkan host buffers

Add fp32 fp16 matmul shader

Fix matmul shader alignment

* Remove deprecated tensor->backend uses

* Fix Vulkan validation errors on embedding models with no offloaded layers

* Fix Vulkan llava segfault when not offloading layers
ggml-vulkan-shaders.hpp diff | blob | history
ggml-vulkan.cpp diff | blob | history
ggml_vk_generate_shaders.py diff | blob | history
Packaging of ggml-org/llama.cpp