Apply suggestions from code review

Author: CISC

examples/speculative-simple/speculative-simple.cpp

@@ -59,13 +59,7 @@ int main(int argc, char ** argv) {
     }
 
     params.cpuparams_batch.n_threads = params.speculative.cpuparams_batch.n_threads;
-
-    // Apply tensor overrides for draft model
-    if (!params.speculative.tensor_buft_overrides.empty()) {
-        params.tensor_buft_overrides = params.speculative.tensor_buft_overrides;
-    } else {
-        params.tensor_buft_overrides.clear();
-    }
+    params.tensor_buft_overrides     = params.speculative.tensor_buft_overrides;
 
     common_init_result llama_init_dft = common_init_from_params(params);
 

examples/speculative/speculative.cpp

@@ -85,13 +85,7 @@ int main(int argc, char ** argv) {
     }
 
     params.cpuparams_batch.n_threads = params.speculative.cpuparams_batch.n_threads;
-
-    // Apply tensor overrides for draft model
-    if (!params.speculative.tensor_buft_overrides.empty()) {
-        params.tensor_buft_overrides = params.speculative.tensor_buft_overrides;
-    } else {
-        params.tensor_buft_overrides.clear();
-    }
+    params.tensor_buft_overrides     = params.speculative.tensor_buft_overrides;
 
     common_init_result llama_init_dft = common_init_from_params(params);
 

tools/server/server.cpp

@@ -2011,12 +2011,7 @@ struct server_context {
             params_dft.cache_type_k = params_base.speculative.cache_type_k;
             params_dft.cache_type_v = params_base.speculative.cache_type_v;
 
-            // Apply tensor overrides for draft model
-            if (!params_base.speculative.tensor_buft_overrides.empty()) {
-                params_dft.tensor_buft_overrides = params_base.speculative.tensor_buft_overrides;
-            } else {
-                params_dft.tensor_buft_overrides.clear(); // ensure no main overrides leak in
-            }
+            params_dft.tensor_buft_overrides = params_base.speculative.tensor_buft_overrides;
 
             llama_init_dft = common_init_from_params(params_dft);