llama: give the qwen4exp full memory context its indexer cache

graph_reserve() walks a full memory context, and qwen4exp builds its
sparse attention only when the context exposes an indexer cache. the
full-context constructor left ctx_idx null, so the reserved worst case
was the dense fallback: a smaller graph than the one decode executes.
ggml-alloc then had to grow the compute buffer on the first decode,
past the size reported at load.

with -np 4 -c 32768 -fa on -ctk q8_0 -ctv q8_0 on an IQ1_S qwen4exp,
the reserved CUDA0 buffer was 217.00 MiB against 275.71 MiB actually
used, and CUDA_Host 42.31 MiB against 191.14 MiB. reserving the sparse
graph makes both match exactly, in unified and non-unified cache mode.

Co-authored-by: Pascal <admin@serveurperso.com>
Assisted-by: Claude

src/llama-memory-hybrid-idx.cpp

@@ -557,7 +557,14 @@ llama_memory_hybrid_idx_context::llama_memory_hybrid_idx_context(llama_memory_st
 
 llama_memory_hybrid_idx_context::llama_memory_hybrid_idx_context(llama_memory_hybrid_idx * mem) :
     llama_memory_hybrid_context(mem),
-    mem(mem) {}
+    mem(mem),
+    // graph reservation walks a full context, and qwen4exp builds the sparse attention only when
+    // this is set. without it the reserved worst case is the smaller dense graph, so ggml-alloc
+    // must grow the compute buffer on the first decode
+    ns_ubatch(mem->get_mem_idx() == nullptr ?
+        std::vector<uint32_t>() : std::vector<uint32_t>{ mem->get_mem_idx()->get_n_stream() }),
+    ctx_idx(mem->get_mem_idx() == nullptr ? nullptr :
+        new llama_kv_cache_context(mem->get_mem_idx())) {}
 
 llama_memory_hybrid_idx_context::llama_memory_hybrid_idx_context(
         llama_memory_hybrid_idx * mem,

src/llama-memory-hybrid-idx.h

@@ -157,7 +157,7 @@ public:
     // llama_memory_hybrid_idx_context specific API
     //
 
-    // nullptr with no indexer, and for the full and update contexts, which build no sparse graph
+    // nullptr with no indexer, and for the update context, which builds no sparse graph
     const llama_kv_cache_context * get_idx() const;
 
     // streams in the current slot info, the `ns` of get_k/get_v; 1 if unified
@@ -182,7 +182,7 @@ private:
     // declared first, so it is initialised while sinfos_idx is still intact
     const std::vector<uint32_t> ns_ubatch;
 
-    // null unless the model has an indexer and this is a batch context
+    // null unless the model has an indexer and this is a batch or full context
     const llama_memory_context_ptr ctx_idx;
 
     // mirrors the base class's ubatch cursor, which is private there