server: can_speculate() requires a task instance

Author: srogmann

tools/server/server-context.cpp

@@ -1177,6 +1177,8 @@ struct server_context_impl {
             slot.smpl.reset();
         }
 
+        slot.task = std::make_unique<const server_task>(std::move(task));
+
         // initialize draft batch
         // TODO: rework speculative decoding [TAG_SERVER_SPEC_REWORK]
         if (slot.can_speculate()) {
@@ -1185,8 +1187,6 @@ struct server_context_impl {
             slot.batch_spec = llama_batch_init(task.params.speculative.n_max + 1, 0, 1);
         }
 
-        slot.task = std::make_unique<const server_task>(std::move(task));
-
         slot.state = slot.task->is_child()
             ? SLOT_STATE_WAIT_OTHER // wait for the parent to process prompt
             : SLOT_STATE_STARTED;