diff --git a/models/intel_optimized_models/resnet_50/solver.prototxt b/models/intel_optimized_models/resnet_50/solver.prototxt
index bf9922187..5693483fd 100644
--- a/models/intel_optimized_models/resnet_50/solver.prototxt
+++ b/models/intel_optimized_models/resnet_50/solver.prototxt
@@ -1,3 +1,4 @@
+#The prototxt files(solver.prototxt and train_val.prototxt) are copied from "models/intel_optimized_models/multinode/resnet_50_16_nodes_2k_batch", temporarily used for throughput benchmarking. It can't achieve expected accuracy for training. We'll provide the files for single node training in the future.
 net: "models/intel_optimized_models/resnet_50/train_val.prototxt"
 test_iter: 1000
 test_interval: 625
@@ -9,11 +10,9 @@ stepvalue:18750
 stepvalue:37500
 stepvalue:50000
 gamma: 0.1
-max_iter: 56300
+max_iter: 200
 warmup_iter: 3125 # 1281167 / 2048 * 5 epochs
 warmup_start_lr: 0.1
 momentum: 0.9
 weight_decay: 0.0001
-snapshot: 6250
-snapshot_prefix: "models/intel_optimized_models/resnet_50/resnet_50_"
 solver_mode: CPU