chore: apply patches

cpp/coreml/whisper-encoder.mm

@@ -24,9 +24,9 @@
 
     // select which device to run the Core ML model on
     MLModelConfiguration *config = [[MLModelConfiguration alloc] init];
-    config.computeUnits = MLComputeUnitsCPUAndGPU;
+    //config.computeUnits = MLComputeUnitsCPUAndGPU;
     //config.computeUnits = MLComputeUnitsCPUAndNeuralEngine;
-    //config.computeUnits = MLComputeUnitsAll;
+    config.computeUnits = MLComputeUnitsAll;
 
     const void * data = CFBridgingRetain([[whisper_encoder_impl alloc] initWithContentsOfURL:url_model configuration:config error:nil]);
 

cpp/whisper.cpp

@@ -2855,7 +2855,9 @@ struct whisper_state * whisper_init_state(whisper_context * ctx) {
         log("%s: kv cross size = %7.2f MB\n", __func__, memory_size / 1024.0 / 1024.0);
     }
 
+
 #ifdef WHISPER_USE_COREML
+    if (ctx->params.use_coreml) {
     const auto path_coreml = whisper_get_coreml_path_encoder(ctx->path_model);
 
     log("%s: loading Core ML model from '%s'\n", __func__, path_coreml.c_str());
@@ -2871,6 +2873,7 @@ struct whisper_state * whisper_init_state(whisper_context * ctx) {
     } else {
         log("%s: Core ML model loaded\n", __func__);
     }
+    }
 #endif
 
     state->logits.reserve(ctx->vocab.n_vocab * ctx->model.hparams.n_text_ctx);
@@ -3048,6 +3051,7 @@ int whisper_ctx_init_openvino_encoder(
 struct whisper_context_params whisper_context_default_params() {
     struct whisper_context_params result = {
         /*.use_gpu    =*/ true,
+        /*.use_coreml =*/ false,
     };
     return result;
 }

cpp/whisper.h

@@ -81,6 +81,7 @@ extern "C" {
 
     struct whisper_context_params {
         bool  use_gpu;
+        bool  use_coreml;
     };
 
     typedef struct whisper_token_data {