logs: fix bogus "0 MiB free" log line (#12590)

On the llama runner, after the recent GGML bump a new log line reports
incorrect 0 MiB free after our patch to remove memory from the props.  This
adjusts the llama.cpp code to fetch the actual free memory of the active device.

llama/llama.cpp/src/llama.cpp

@@ -267,10 +267,12 @@ static struct llama_model * llama_model_load_from_file_impl(
     for (auto * dev : model->devices) {
         ggml_backend_dev_props props;
         ggml_backend_dev_get_props(dev, &props);
+        size_t memory_free, memory_total;
+        ggml_backend_dev_memory(dev, &memory_free, &memory_total);
         LLAMA_LOG_INFO("%s: using device %s (%s) (%s) - %zu MiB free\n", __func__,
                 ggml_backend_dev_name(dev), ggml_backend_dev_description(dev),
                 props.device_id ? props.device_id : "unknown id",
-                props.memory_free/1024/1024);
+                memory_free/1024/1024);
     }
 
     const int status = llama_model_load(path_model, splits, *model, params);

llama/patches/0024-ggml-Enable-resetting-backend-devices.patch

@@ -12,10 +12,11 @@ unused then it can be reset to free these data structures.
  ggml/src/ggml-backend.cpp        |  8 ++++++++
  ggml/src/ggml-cuda/ggml-cuda.cu  | 16 +++++++++++++++-
  ggml/src/ggml-cuda/vendors/hip.h |  1 +
- 5 files changed, 29 insertions(+), 1 deletion(-)
+ src/llama.cpp                    |  4 +++-
+ 6 files changed, 32 insertions(+), 2 deletions(-)
 
 diff --git a/ggml/include/ggml-backend.h b/ggml/include/ggml-backend.h
-index 1ff53ed0..ba181d09 100644
+index 1ff53ed03..ba181d09d 100644
 --- a/ggml/include/ggml-backend.h
 +++ b/ggml/include/ggml-backend.h
 @@ -178,6 +178,7 @@ extern "C" {
@@ -27,7 +28,7 @@ index 1ff53ed0..ba181d09 100644
      GGML_API ggml_backend_buffer_type_t    ggml_backend_dev_host_buffer_type(ggml_backend_dev_t device);
      GGML_API ggml_backend_buffer_t         ggml_backend_dev_buffer_from_host_ptr(ggml_backend_dev_t device, void * ptr, size_t size, size_t max_tensor_size);
 diff --git a/ggml/src/ggml-backend-impl.h b/ggml/src/ggml-backend-impl.h
-index 3c3f22fc..43c91d9f 100644
+index 3c3f22fc0..43c91d9f2 100644
 --- a/ggml/src/ggml-backend-impl.h
 +++ b/ggml/src/ggml-backend-impl.h
 @@ -195,6 +195,10 @@ extern "C" {
@@ -42,7 +43,7 @@ index 3c3f22fc..43c91d9f 100644
  
      struct ggml_backend_device {
 diff --git a/ggml/src/ggml-backend.cpp b/ggml/src/ggml-backend.cpp
-index 6ef5eeaf..0b757af5 100644
+index 6ef5eeafa..0b757af59 100644
 --- a/ggml/src/ggml-backend.cpp
 +++ b/ggml/src/ggml-backend.cpp
 @@ -526,6 +526,14 @@ ggml_backend_t ggml_backend_dev_init(ggml_backend_dev_t device, const char * par
@@ -61,7 +62,7 @@ index 6ef5eeaf..0b757af5 100644
      GGML_ASSERT(device);
      return device->iface.get_buffer_type(device);
 diff --git a/ggml/src/ggml-cuda/ggml-cuda.cu b/ggml/src/ggml-cuda/ggml-cuda.cu
-index 811462c7..87c6c34a 100644
+index 811462c79..87c6c34a4 100644
 --- a/ggml/src/ggml-cuda/ggml-cuda.cu
 +++ b/ggml/src/ggml-cuda/ggml-cuda.cu
 @@ -107,6 +107,11 @@ int ggml_cuda_get_device() {
@@ -109,7 +110,7 @@ index 811462c7..87c6c34a 100644
  
  // backend reg
 diff --git a/ggml/src/ggml-cuda/vendors/hip.h b/ggml/src/ggml-cuda/vendors/hip.h
-index 890c1036..1f06be80 100644
+index 890c10364..1f06be80e 100644
 --- a/ggml/src/ggml-cuda/vendors/hip.h
 +++ b/ggml/src/ggml-cuda/vendors/hip.h
 @@ -45,6 +45,7 @@
@@ -120,3 +121,21 @@ index 890c1036..1f06be80 100644
  #define cudaDeviceSynchronize hipDeviceSynchronize
  #define cudaError_t hipError_t
  #define cudaErrorPeerAccessAlreadyEnabled hipErrorPeerAccessAlreadyEnabled
+diff --git a/src/llama.cpp b/src/llama.cpp
+index fe5a7a835..d821a96a0 100644
+--- a/src/llama.cpp
++++ b/src/llama.cpp
+@@ -267,10 +267,12 @@ static struct llama_model * llama_model_load_from_file_impl(
+     for (auto * dev : model->devices) {
+         ggml_backend_dev_props props;
+         ggml_backend_dev_get_props(dev, &props);
++        size_t memory_free, memory_total;
++        ggml_backend_dev_memory(dev, &memory_free, &memory_total);
+         LLAMA_LOG_INFO("%s: using device %s (%s) (%s) - %zu MiB free\n", __func__,
+                 ggml_backend_dev_name(dev), ggml_backend_dev_description(dev),
+                 props.device_id ? props.device_id : "unknown id",
+-                props.memory_free/1024/1024);
++                memory_free/1024/1024);
+     }
+ 
+     const int status = llama_model_load(path_model, splits, *model, params);