fix(server): align layer split load plan with target load plan

Author: weicj

server/src/common/layer_split_utils.h

@@ -26,10 +26,11 @@ struct LayerSplitShardMeta {
     ggml_backend_t backend = nullptr;
 };
 
-inline TargetLoadPlan make_layer_split_load_plan(
+template <typename LoadPlan>
+inline LoadPlan make_layer_split_load_plan(
         const LayerSplitShardMeta & shard,
         bool is_last_shard) {
-    TargetLoadPlan plan;
+    LoadPlan plan;
     plan.layer_begin = shard.layer_begin;
     plan.layer_end = shard.layer_end;
     plan.load_output = is_last_shard;

server/src/internal.h

@@ -23,7 +23,6 @@
 #include "ggml-backend.h"
 #include "gguf.h"
 
-#include "common/layer_split_utils.h"
 #include "dflash27b.h"
 
 namespace dflash::common {

server/src/qwen35/layer_split_daemon_loop.cpp

@@ -49,7 +49,7 @@ int run_layer_split_daemon(const LayerSplitDaemonConfig & cfg) {
     // Load partial target weights + caches.
     for (auto & shard : shards) {
         const TargetLoadPlan plan =
-            make_layer_split_load_plan(shard, &shard == &shards.back());
+            make_layer_split_load_plan<TargetLoadPlan>(shard, &shard == &shards.back());
         if (!load_target_gguf_partial(cfg.target_path, shard.backend, plan, shard.weights) ||
             !create_target_cache_partial(shard.weights, cfg.max_ctx, cfg.max_verify_tokens,
                                          shard.backend, shard.cache,

server/src/qwen35/qwen35_layer_split_adapter.cpp

@@ -62,7 +62,7 @@ bool Qwen35LayerSplitAdapter::init() {
 
     for (auto & shard : shards_) {
         const TargetLoadPlan plan =
-            make_layer_split_load_plan(shard, &shard == &shards_.back());
+            make_layer_split_load_plan<TargetLoadPlan>(shard, &shard == &shards_.back());
         if (!load_target_gguf_partial(cfg_.target_path, shard.backend, plan,
                                       shard.weights) ||
             !create_target_cache_partial(shard.weights, cfg_.device.max_ctx,

server/test/test_dflash.cpp

@@ -344,7 +344,7 @@ static int run_target_layer_split_harness(
     (void)enable_layer_split_peer_access(target_gpus, peer_access);
     for (auto & shard : shards) {
         const TargetLoadPlan plan =
-            make_layer_split_load_plan(shard, &shard == &shards.back());
+            make_layer_split_load_plan<TargetLoadPlan>(shard, &shard == &shards.back());
         if (!load_target_gguf_partial(target_path, shard.backend, plan, shard.weights)) {
             std::fprintf(stderr, "target-split load gpu=%d: %s\n",
                          shard.gpu, dflash27b_last_error());