ver4a/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity

repo : update links to new url (#11886)

Browse source 
* repo : update links to new url

ggml-ci

* cont : more urls

ggml-ci
This commit is contained in:
Georgi Gerganov 2025-02-15 16:40:57 +02:00 • committed by GitHub
parent f355229692
commit 68ff663a04
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
66 changed files with 192 additions and 202 deletions
Download patch file Download diff file Expand all files Collapse all files

4
ggml/src/ggml-cpu/ggml-cpu.c
View file

@ -1816,7 +1816,7 @@ inline static float ggml_silu_f32(float x) {
#if __FINITE_MATH_ONLY__
#error "some routines in ggml.c require non-finite math arithmetics -- pass -fno-finite-math-only to the compiler to fix"
#error "ref: https://github.com/ggerganov/llama.cpp/pull/7154#issuecomment-2143844461"
#error "ref: https://github.com/ggml-org/llama.cpp/pull/7154#issuecomment-2143844461"
#endif
#if defined(__ARM_NEON) && defined(__aarch64__)
@ -7574,7 +7574,7 @@ UseGgmlGemm2:;
int64_t nchunk1 = (nr1 + chunk_size - 1) / chunk_size;
// If the chunking is poor for the number of threads on this setup, scrap the whole plan. Re-chunk it by thread.
// Also, chunking by thread was measured to have perform better on NUMA systems. See https://github.com/ggerganov/llama.cpp/pull/6915
// Also, chunking by thread was measured to have perform better on NUMA systems. See https://github.com/ggml-org/llama.cpp/pull/6915
// In theory, chunking should be just as useful on NUMA and non NUMA systems, but testing disagreed with that.
if (nchunk0 * nchunk1 < nth * 4 || ggml_is_numa()) {
// distribute the thread work across the inner or outer loop based on which one is larger

2
ggml/src/ggml-metal/ggml-metal.m
View file

@ -1983,7 +1983,7 @@ static void ggml_metal_encode_node(
const float m1 = powf(2.0f, -(max_bias / 2.0f) / n_head_log2);
// TODO: add ggml_metal_kargs struct
// TODO: optimize (see https://github.com/ggerganov/llama.cpp/pull/10238/commits/7941b6b9ec29a2866fec6fa6c51612515ca509f6)
// TODO: optimize (see https://github.com/ggml-org/llama.cpp/pull/10238/commits/7941b6b9ec29a2866fec6fa6c51612515ca509f6)
[encoder setComputePipelineState:pipeline];
[encoder setBuffer:id_src0 offset:offs_src0 atIndex:0];
if (id_src1) {

4
ggml/src/ggml-metal/ggml-metal.metal
View file

@ -1058,7 +1058,7 @@ kernel void kernel_soft_max(
}
// This barrier fixes a failing test
// ref: https://github.com/ggerganov/ggml/pull/621#discussion_r1425156335
// ref: https://github.com/ggml-org/ggml/pull/621#discussion_r1425156335
threadgroup_barrier(mem_flags::mem_none);
float sum = simd_sum(lsum);
@ -1163,7 +1163,7 @@ kernel void kernel_soft_max_4(
const float lsum = lsum4[0] + lsum4[1] + lsum4[2] + lsum4[3];
// This barrier fixes a failing test
// ref: https://github.com/ggerganov/ggml/pull/621#discussion_r1425156335
// ref: https://github.com/ggml-org/ggml/pull/621#discussion_r1425156335
threadgroup_barrier(mem_flags::mem_none);
float sum = simd_sum(lsum);