whisper : remove whisper_load_backends function (ggml-org#3196)

* whisper : remove whisper_load_backends function

This commit removes the `whisper_load_backends` function, which was used
to load all GGML backends.

The motivation for this change push the responsibility of loading
backends to user applications to give them more control over which
backends to load and when. See the references below for more context.

Resolves: ggml-org#3182
Refs: ggml-org#3042 (comment)
Refs: ggml-org#3042 (comment)

* ruby : add check for rwc is NULL

This commit adds a check to ensure that the `rwc` pointer is not NULL
before attempting to mark its members in the garbage collector.

The motivation for this is an attempt to see if this fixed the CI build
as I'm not able to reproduce the issue locally.

Refs: https://github.com/ggml-org/whisper.cpp/actions/runs/15299612277/job/43036694928?pr=3196

Author: danbev

bindings/ruby/ext/ruby_whisper_params.c

@@ -77,6 +77,8 @@ static ID id_vad_params;
 static void
 rb_whisper_callbcack_container_mark(ruby_whisper_callback_container *rwc)
 {
+  if (rwc == NULL) return;
+
   rb_gc_mark(rwc->user_data);
   rb_gc_mark(rwc->callback);
   rb_gc_mark(rwc->callbacks);

examples/bench/bench.cpp

@@ -156,6 +156,8 @@ static int whisper_bench_full(const whisper_params & params) {
 }
 
 int main(int argc, char ** argv) {
+    ggml_backend_load_all();
+
     whisper_params params;
 
     if (whisper_params_parse(argc, argv, params) == false) {

examples/cli/cli.cpp

@@ -909,6 +909,8 @@ static void output_lrc(struct whisper_context * ctx, std::ofstream & fout, const
 static void cb_log_disable(enum ggml_log_level , const char * , void * ) { }
 
 int main(int argc, char ** argv) {
+    ggml_backend_load_all();
+
 #if defined(_WIN32)
     // Set the console output code page to UTF-8, while command line arguments
     // are still encoded in the system's code page. In this way, we can print
@@ -988,7 +990,6 @@ int main(int argc, char ** argv) {
     }
 
     // whisper init
-
     struct whisper_context_params cparams = whisper_context_default_params();
 
     cparams.use_gpu    = params.use_gpu;

examples/command/command.cpp

@@ -678,6 +678,8 @@ static int process_general_transcription(struct whisper_context * ctx, audio_asy
 }
 
 int main(int argc, char ** argv) {
+    ggml_backend_load_all();
+
     whisper_params params;
 
     if (whisper_params_parse(argc, argv, params) == false) {

examples/lsp/lsp.cpp

@@ -424,6 +424,8 @@ static void process_loop(struct whisper_context * ctx, audio_async &audio, const
 }
 
 int main(int argc, char ** argv) {
+    ggml_backend_load_all();
+
     whisper_params params;
     if (whisper_params_parse(argc, argv, params) == false) {
         return 1;

examples/quantize/quantize.cpp

@@ -1,4 +1,5 @@
 #include "ggml.h"
+#include "ggml-backend.h"
 
 #include "common.h"
 #include "common-ggml.h"
@@ -176,6 +177,8 @@ static bool whisper_model_quantize(const std::string & fname_inp, const std::str
 }
 
 int main(int argc, char ** argv) {
+    ggml_backend_load_all();
+
     if (argc != 4) {
         fprintf(stderr, "usage: %s model-f32.bin model-quant.bin type\n", argv[0]);
         ggml_print_ftypes(stderr);

examples/server/server.cpp

@@ -516,6 +516,8 @@ void get_req_parameters(const Request & req, whisper_params & params)
 }  // namespace
 
 int main(int argc, char ** argv) {
+    ggml_backend_load_all();
+
     whisper_params params;
     server_params sparams;
 

examples/stream/stream.cpp

@@ -116,6 +116,8 @@ void whisper_print_usage(int /*argc*/, char ** argv, const whisper_params & para
 }
 
 int main(int argc, char ** argv) {
+    ggml_backend_load_all();
+
     whisper_params params;
 
     if (whisper_params_parse(argc, argv, params) == false) {

examples/talk-llama/talk-llama.cpp

@@ -291,6 +291,8 @@ The transcript only includes text, it does not include markup like HTML and Mark
 {0}{4})";
 
 int main(int argc, char ** argv) {
+    ggml_backend_load_all();
+
     whisper_params params;
 
     if (whisper_params_parse(argc, argv, params) == false) {

examples/vad-speech-segments/speech.cpp

@@ -83,6 +83,8 @@ static bool vad_params_parse(int argc, char ** argv, cli_params & params) {
 static void cb_log_disable(enum ggml_log_level , const char * , void * ) { }
 
 int main(int argc, char ** argv) {
+    ggml_backend_load_all();
+
     cli_params cli_params;
 
     if (!vad_params_parse(argc, argv, cli_params)) {