diff --git a/dlib/dnn/trainer.h b/dlib/dnn/trainer.h
index 5d6afdcc4..934b74a54 100644
--- a/dlib/dnn/trainer.h
+++ b/dlib/dnn/trainer.h
@@ -535,7 +535,15 @@ namespace dlib
             std::vector<tensor*> reference_params;
             visit_layer_parameters(devices[0]->net, [&](size_t, tensor& t) { reference_params.push_back(&t); });
 
-            thread_pool tp(devices.size());
+            // We make separate thread pools with just one thread in them because we want
+            // to make sure each device is always executed on the same thread.  We care
+            // about this because there are thread_local context variables for some cuda
+            // components and they get regenerated when the current cuda device changes.
+            // Recreating them over and over is somewhat expensive so we want to avoid
+            // that.
+            std::vector<std::shared_ptr<thread_pool>> tp;
+            for (size_t i = 0; i < devices.size(); ++i)
+                tp.push_back(std::make_shared<thread_pool>(1));
 
 
             size_t iteration = 0;
@@ -546,7 +554,7 @@ namespace dlib
                 // right version for unsupervised or supervised training based on the type
                 // of label_type.
                 for (size_t i = 0; i < devices.size(); ++i)
-                    tp.add_task_by_value([&,i](double& loss){ loss = compute_parameter_gradients(i, next_job, pick_which_run_update); }, losses[i]);
+                    tp[i]->add_task_by_value([&,i](double& loss){ loss = compute_parameter_gradients(i, next_job, pick_which_run_update); }, losses[i]);
                 // aggregate loss values from all the network computations.
                 double theloss = 0;
                 for (auto&& loss : losses)
@@ -597,9 +605,10 @@ namespace dlib
 
                 // Now apply all the updates to each device.
                 for (size_t i = 0; i < devices.size(); ++i)
-                    tp.add_task_by_value([&,i](){ if (next_job.have_data[i]) update_parameters(i); });
+                    tp[i]->add_task_by_value([&,i](){ if (next_job.have_data[i]) update_parameters(i); });
                 // and wait for the updates to all happen.
-                tp.wait_for_all_tasks();
+                for (size_t i = 0; i < devices.size(); ++i)
+                    tp[i]->wait_for_all_tasks();
 
 
                 // Evey now and then force all the parameters to be the same just to make