mirrors/ollama-for-amd
1
0
Fork 0
mirror of https://github.com/likelovewant/ollama-for-amd.git synced 2025-12-22 23:03:55 +00:00
Code Issues Packages Projects Releases Wiki Activity

ggml update to b7108 (#12992)

Browse Source 
* Revert "vulkan: temporary cary of vulkan fixes (#12971)"

This reverts commit 3a9e8e9fd4.

* ggml update to b7087

* fix argsort on metal

* update to b7108

* fix bakllava regression

This model lacks the metadata for the projector type.

* update to b7209

* fix TopK perf

* only build arm code on arm
This commit is contained in:
Daniel Hiltgen
2025-12-03 19:43:29 -08:00
committed by GitHub
parent 854d40edc5
commit 0cf7794b16
303 changed files with 32711 additions and 23435 deletions
Download Patch File Download Diff File Expand all files Collapse all files

26
llama/patches/0015-ggml-Export-GPU-UUIDs.patch
View File

@@ -1,10 +1,8 @@
From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
From: Jesse Gross <jesse@ollama.com>
Date: Thu, 24 Apr 2025 14:48:51 -0700
From: Daniel Hiltgen <daniel@ollama.com>
Date: Sun, 30 Nov 2025 11:05:56 -0800
Subject: [PATCH] ggml: Export GPU UUIDs
This enables matching up devices and information reported by the backend
with tools (e.g. nvidia-smi) and system management libraries (e.g. nvml).
---
ggml/include/ggml-backend.h | 1 +
ggml/src/ggml-cuda/ggml-cuda.cu | 67 +++++++++++++++++++++++++++---
@@ -24,10 +22,10 @@ index c54ff98bf..229bf387b 100644
size_t memory_total;
// device type
diff --git a/ggml/src/ggml-cuda/ggml-cuda.cu b/ggml/src/ggml-cuda/ggml-cuda.cu
index aefc6935e..cc201afff 100644
index 8f3b1c173..e803f4af6 100644
--- a/ggml/src/ggml-cuda/ggml-cuda.cu
+++ b/ggml/src/ggml-cuda/ggml-cuda.cu
@@ -183,6 +183,51 @@ static int ggml_cuda_parse_id(char devName[]) {
@@ -185,6 +185,51 @@ static int ggml_cuda_parse_id(char devName[]) {
}
#endif // defined(GGML_USE_HIP)
@@ -79,7 +77,7 @@ index aefc6935e..cc201afff 100644
static ggml_cuda_device_info ggml_cuda_init() {
ggml_cuda_device_info info = {};
@@ -249,22 +294,24 @@ static ggml_cuda_device_info ggml_cuda_init() {
@@ -251,22 +296,24 @@ static ggml_cuda_device_info ggml_cuda_init() {
info.devices[id].cc += prop.minor * 0x10;
}
}
@@ -110,7 +108,7 @@ index aefc6935e..cc201afff 100644
std::string device_name(prop.name);
if (device_name == "NVIDIA GeForce MX450") {
turing_devices_without_mma.push_back({ id, device_name });
@@ -3268,6 +3315,7 @@ struct ggml_backend_cuda_device_context {
@@ -4048,6 +4095,7 @@ struct ggml_backend_cuda_device_context {
std::string name;
std::string description;
std::string pci_bus_id;
@@ -118,9 +116,9 @@ index aefc6935e..cc201afff 100644
};
static const char * ggml_backend_cuda_device_get_name(ggml_backend_dev_t dev) {
@@ -3280,6 +3328,11 @@ static const char * ggml_backend_cuda_device_get_description(ggml_backend_dev_t
return ctx->description.c_str();
@@ -4136,6 +4184,11 @@ static bool ggml_backend_cuda_get_available_uma_memory(long * available_memory_k
}
#endif // defined(__linux__)
+static const char * ggml_backend_cuda_device_get_id(ggml_backend_dev_t dev) {
+ ggml_backend_cuda_device_context * ctx = (ggml_backend_cuda_device_context *)dev->context;
@@ -130,7 +128,7 @@ index aefc6935e..cc201afff 100644
static void ggml_backend_cuda_device_get_memory(ggml_backend_dev_t dev, size_t * free, size_t * total) {
ggml_backend_cuda_device_context * ctx = (ggml_backend_cuda_device_context *)dev->context;
ggml_cuda_set_device(ctx->device);
@@ -3296,6 +3349,7 @@ static void ggml_backend_cuda_device_get_props(ggml_backend_dev_t dev, ggml_back
@@ -4176,6 +4229,7 @@ static void ggml_backend_cuda_device_get_props(ggml_backend_dev_t dev, ggml_back
props->name = ggml_backend_cuda_device_get_name(dev);
props->description = ggml_backend_cuda_device_get_description(dev);
@@ -138,7 +136,7 @@ index aefc6935e..cc201afff 100644
props->type = ggml_backend_cuda_device_get_type(dev);
props->device_id = ctx->pci_bus_id.empty() ? nullptr : ctx->pci_bus_id.c_str();
ggml_backend_cuda_device_get_memory(dev, &props->memory_free, &props->memory_total);
@@ -3869,6 +3923,7 @@ ggml_backend_reg_t ggml_backend_cuda_reg() {
@@ -4767,6 +4821,7 @@ ggml_backend_reg_t ggml_backend_cuda_reg() {
cudaDeviceProp prop;
CUDA_CHECK(cudaGetDeviceProperties(&prop, i));
dev_ctx->description = prop.name;
@@ -147,10 +145,10 @@ index aefc6935e..cc201afff 100644
char pci_bus_id[16] = {};
snprintf(pci_bus_id, sizeof(pci_bus_id), "%04x:%02x:%02x.0", prop.pciDomainID, prop.pciBusID, prop.pciDeviceID);
diff --git a/ggml/src/ggml-metal/ggml-metal.cpp b/ggml/src/ggml-metal/ggml-metal.cpp
index bf0962274..f2ff9f322 100644
index f2b7fe692..8fc1c2fb5 100644
--- a/ggml/src/ggml-metal/ggml-metal.cpp
+++ b/ggml/src/ggml-metal/ggml-metal.cpp
@@ -538,6 +538,7 @@ static enum ggml_backend_dev_type ggml_backend_metal_device_get_type(ggml_backen
@@ -547,6 +547,7 @@ static enum ggml_backend_dev_type ggml_backend_metal_device_get_type(ggml_backen
static void ggml_backend_metal_device_get_props(ggml_backend_dev_t dev, ggml_backend_dev_props * props) {
props->name = ggml_backend_metal_device_get_name(dev);
props->description = ggml_backend_metal_device_get_description(dev);