Yaser Afshar

@sandeep-maddipatla would you please respond to regiss' comment? We would like to move forward

@skavulya - if the PR is ready for review, I can start on that. Please ping me here. - please run `GAUDI2_CI=1 RUN_SLOW=true python -m pytest tests/transformers/tests/models/ -s -v` before...

@skavulya please ignore my last comments if you do not like it. It is a personal style, I do not like the extra conditional check. The first comment for `.all`...

@skavulya thanks, everything sounds good to me. Would you please add a final test to make sure we do not miss anything. please run `GAUDI2_CI=1 RUN_SLOW=true python -m pytest tests/transformers/tests/models/...

@Gaurav7888 please provide the test cmd in the README, also how is perf in comparison with gpu?

@Gaurav7888 Please also run make style

> @yafshar Hi, I have made the changes as requested. Please review @Gaurav7888 thanks, please provide your test cmd?

> @yafshar > > python run_translation.py --model_name_or_path mt5-small --do_eval --source_lang en --target_lang ro --source_prefix "translate English to Romanian: " --dataset_name wmt16 --dataset_config_name ro-en --output_dir /tmp/tst-translation --per_device_eval_batch_size 4 --overwrite_output_dir --predict_with_generate --use_habana...

This is what I currently get ```python >> python run_translation.py --model_name_or_path google/mt5-small --do_eval --source_lang en --target_lang ro --source_prefix "translate English to Romanian: " --dataset_name wmt16 --dataset_config_name ro-en --output_dir /root/tst-translation --per_device_eval_batch_size...

@Gaurav7888 can you address the comments to finish this PR?