diff --git a/examples/nlp/language_modeling/megatron_ckpt_to_nemo.py b/examples/nlp/language_modeling/megatron_ckpt_to_nemo.py
index 1161614e0a534..c4c0394e3892b 100644
--- a/examples/nlp/language_modeling/megatron_ckpt_to_nemo.py
+++ b/examples/nlp/language_modeling/megatron_ckpt_to_nemo.py
@@ -31,7 +31,7 @@
 import torch
 from genericpath import isdir
 from megatron.core import parallel_state
-from omegaconf import open_dict
+from omegaconf import OmegaConf, open_dict
 from pytorch_lightning.plugins.environments import TorchElasticEnvironment
 from pytorch_lightning.trainer.trainer import Trainer
 
@@ -42,7 +42,12 @@
 from nemo.collections.nlp.models.language_modeling.megatron_retrieval_model import MegatronRetrievalModel
 from nemo.collections.nlp.models.language_modeling.megatron_t5_model import MegatronT5Model
 from nemo.collections.nlp.models.machine_translation.megatron_nmt_model import MegatronNMTModel
-from nemo.collections.nlp.parts.nlp_overrides import NLPDDPStrategy, NLPSaveRestoreConnector
+from nemo.collections.nlp.parts.nlp_overrides import (
+    GradScaler,
+    NLPDDPStrategy,
+    NLPSaveRestoreConnector,
+    PipelineMixedPrecisionPlugin,
+)
 from nemo.utils import AppState, logging
 from nemo.utils.distributed import initialize_distributed
 from nemo.utils.model_utils import inject_model_parallel_rank
@@ -92,6 +97,14 @@ def get_args():
     )
     parser.add_argument("--local_rank", type=int, required=False, default=os.getenv('LOCAL_RANK', -1))
     parser.add_argument("--bcp", action="store_true", help="Whether on BCP platform")
+    parser.add_argument(
+        "--precision",
+        type=str,
+        required=False,
+        default='16-mixed',
+        choices=['32-true', '16-mixed', 'bf16-mixed'],
+        help="Precision value for the trainer that matches with precision of the ckpt",
+    )
 
     args = parser.parse_args()
     return args
@@ -109,9 +122,27 @@ def convert(local_rank, rank, world_size, args):
     if args.model_type == 'gpt':
         strategy = NLPDDPStrategy()
 
-    trainer = Trainer(
-        devices=args.gpus_per_node, num_nodes=num_nodes, accelerator='gpu', plugins=plugins, strategy=strategy
-    )
+    cfg = {
+        'trainer': {
+            'devices': args.gpus_per_node,
+            'num_nodes': num_nodes,
+            'accelerator': 'gpu',
+            'precision': args.precision,
+        },
+        'model': {'native_amp_init_scale': 2 ** 32, 'native_amp_growth_interval': 1000, 'hysteresis': 2},
+    }
+    cfg = OmegaConf.create(cfg)
+
+    scaler = None
+    # If FP16 create a GradScaler as the build_model_parallel_config of MegatronBaseModel expects it
+    if cfg.trainer.precision == '16-mixed':
+        scaler = GradScaler(
+            init_scale=cfg.model.get('native_amp_init_scale', 2 ** 32),
+            growth_interval=cfg.model.get('native_amp_growth_interval', 1000),
+            hysteresis=cfg.model.get('hysteresis', 2),
+        )
+    plugins.append(PipelineMixedPrecisionPlugin(precision=cfg.trainer.precision, device='cuda', scaler=scaler))
+    trainer = Trainer(plugins=plugins, strategy=strategy, **cfg.trainer)
 
     app_state.pipeline_model_parallel_size = args.pipeline_model_parallel_size
     app_state.tensor_model_parallel_size = args.tensor_model_parallel_size
diff --git a/examples/nlp/token_classification/punctuation_capitalization_train_evaluate.py b/examples/nlp/token_classification/punctuation_capitalization_train_evaluate.py
index 137ee3a30622a..e983540a68b21 100644
--- a/examples/nlp/token_classification/punctuation_capitalization_train_evaluate.py
+++ b/examples/nlp/token_classification/punctuation_capitalization_train_evaluate.py
@@ -103,6 +103,10 @@
 
 @hydra_runner(config_path="conf", config_name="punctuation_capitalization_config")
 def main(cfg: DictConfig) -> None:
+    # PTL 2.0 has find_unused_parameters as False by default, so its required to set it to True
+    # when there are unused parameters like here
+    if cfg.trainer.strategy == 'ddp':
+        cfg.trainer.strategy = "ddp_find_unused_parameters_true"
     torch.manual_seed(42)
     cfg = OmegaConf.merge(OmegaConf.structured(PunctuationCapitalizationConfig()), cfg)
     trainer = pl.Trainer(**cfg.trainer)