]> git.djapps.eu Git - pkg/ggml/sources/ggml/commit
vulkan: clamp matmul and FA results to the max finite value (llama/15652)
authorJeff Bolz <redacted>
Sun, 31 Aug 2025 06:27:57 +0000 (01:27 -0500)
committerGeorgi Gerganov <redacted>
Fri, 5 Sep 2025 09:54:08 +0000 (12:54 +0300)
commitbb60a584a88aff7e42971b664a5d40b99c92284e
tree9ea70700990935c65387cc031940c040d2533b65
parent49135c8eca048a57a7555e02fed805b55062b11a
vulkan: clamp matmul and FA results to the max finite value (llama/15652)

* vulkan: clamp matmul and FA results to the max finite value

* only clamp for fp16
src/ggml-vulkan/vulkan-shaders/flash_attn.comp
src/ggml-vulkan/vulkan-shaders/flash_attn_cm1.comp
src/ggml-vulkan/vulkan-shaders/flash_attn_cm2.comp
src/ggml-vulkan/vulkan-shaders/flash_attn_split_k_reduce.comp
src/ggml-vulkan/vulkan-shaders/mul_mm.comp
src/ggml-vulkan/vulkan-shaders/mul_mm_cm2.comp
src/ggml-vulkan/vulkan-shaders/vulkan-shaders-gen.cpp