ggml : fix GGML_MAX_N_THREADS + improve formatting (ggml/969)
443678a5
vulkan : argsort barriers must be under uniform control flow (ggml/951)
8fc92397
vulkan : fix build for GGML_VULKAN_RUN_TESTS, add TFLOPS to log (ggml…
45c860f0
vulkan : multithread pipeline creation (ggml/963)
ec8e9191
CUDA: remove bad assert (ggml/972)
44e2d393
cann: fix crash when llama-bench is running on multiple cann devices …
d2eac9f1
ggml : remove assert for AArch64 GEMV and GEMM Q4 kernels (llama/9217)
b0978f08
mtgpu: enable VMM (llama/9597)
04485511
Enable use to the rebar feature to upload buffers to the device. (lla…
fd5cb2bf
ggml : add run-time detection of neon, i8mm and sve (llama/9331)
a98b5fa0
ggml : define missing HWCAP flags (llama/9684)
9d176caa
ggml: fix gradient allocation logic (ggml/966)
034ed81a
ggml : fix ggml_cast (ggml/973)
ee8e29c6
vulkan : mul_mat: fix UB with small warps (ggml/952)
28a3391e
test: fix OPT_STEP_ADAMW for test-backend-ops (ggml/974)
2eda43aa
sync : ggml
fce227e5
metal : reduce command encoding overhead (llama/9698)
f0839085
talk-llama : sync llama.cpp
b4c96313
Reviewers
No reviews
Assignees
No one assigned
Labels
None yet
Milestone
No milestone
Login to write a write a comment.
Login via GitHub