trainer.precision=bf16

README.md

@@ -117,7 +117,7 @@ Alternatively, you can use NeMo Megatron training docker container with all depe
 git clone https://github.com/NVIDIA/NeMo.git 
 cd NeMo/examples/nlp/language_modeling
 git checkout v1.17.0
-python megatron_gpt_eval.py gpt_model_file=nemo_2b_bf16_tp1.nemo server=True tensor_model_parallel_size=1 trainer.devices=1
+python megatron_gpt_eval.py trainer.precision=bf16 gpt_model_file=nemo_2b_bf16_tp1.nemo server=True tensor_model_parallel_size=1 trainer.devices=1
 ```
 
 ### Step 3: Send prompts to your model!