Update documentation for multi-gpu training

docs/source/faq.md

@@ -59,8 +59,7 @@ We tried to collect common issues and questions we receive about 🐸TTS. It is
 
 6. Train your model.
     - SingleGPU training: ```CUDA_VISIBLE_DEVICES="0" python train_tts.py --config_path config.json```
-    - MultiGPU training: ```CUDA_VISIBLE_DEVICES="0,1,2" python distribute.py --script train_tts.py --config_path config.json```
-        - This command uses all the GPUs given in ```CUDA_VISIBLE_DEVICES```. If you don't specify, it uses all the GPUs available.
+    - MultiGPU training: ```python3 -m trainer.distribute --gpus "0,1" --script TTS/bin/train_tts.py --config_path config.json```
 
 **Note:** You can also train your model using pure 🐍 python. Check ```{eval-rst} :ref: 'tutorial_for_nervous_beginners'```.