Update TRAIN.md

TRAIN.md

@@ -26,7 +26,7 @@ cd experiments/siamrpn_r50_l234_dwxcorr_8gpu
 
 ### Multi-processing Distributed Data Parallel Training
 
-Please refer to [Pytorch distributed training](https://pytorch.org/docs/stable/distributed.html) for detailed understanding
+Refer to [Pytorch distributed training](https://pytorch.org/docs/stable/distributed.html) for detailed description.
 
 #### Single node, multiple GPUs:
 ```bash