diff --git a/Jenkinsfile b/Jenkinsfile
index 474db51efdc8..c2357e280afb 100644
--- a/Jenkinsfile
+++ b/Jenkinsfile
@@ -126,72 +126,74 @@ pipeline {
         sh 'CUDA_VISIBLE_DEVICES="" NEMO_NUMBA_MINVER=0.53 pytest -m "not pleasefixme" --cpu --with_downloads --relax_numba_compat'
       }
     }
-//
-//     stage('L2: Multimodal Imagen Train') {
-//       when {
-//         anyOf {
-//           branch 'main'
-//           changeRequest target: 'main'
-//         }
-//       }
-//       failFast true
-//       steps {
-//         sh "rm -rf /home/TestData/multimodal/imagen_train"
-//         sh "pip install webdataset==0.2.48"
-//         sh "python examples/multimodal/text_to_image/imagen/imagen_training.py \
-//         trainer.precision=16 \
-//         trainer.num_nodes=1 \
-//         trainer.devices=1 \
-//         ++exp_manager.max_time_per_run=00:00:03:00 \
-//         trainer.max_steps=20 \
-//         model.micro_batch_size=1 \
-//         model.global_batch_size=1 \
-//         model.data.synthetic_data=True \
-//         exp_manager.exp_dir=/home/TestData/multimodal/imagen_train \
-//         model.inductor=False \
-//         model.unet.flash_attention=False \
-//         "
-//         sh "pip install 'webdataset>=0.1.48,<=0.1.62'"
-//         sh "rm -rf /home/TestData/multimodal/imagen_train"
-//       }
-//     }
-//
-//     stage('L2: Multimodal Stable Diffusion Train') {
-//       when {
-//         anyOf {
-//           branch 'main'
-//           changeRequest target: 'main'
-//         }
-//       }
-//       failFast true
-//       steps {
-//         sh "rm -rf /home/TestData/multimodal/stable_diffusion_train"
-//         sh "pip install webdataset==0.2.48"
-//         sh "python examples/multimodal/text_to_image/stable_diffusion/sd_train.py \
-//             trainer.precision=16 \
-//             trainer.num_nodes=1 \
-//             trainer.devices=1 \
-//             ++exp_manager.max_time_per_run=00:00:03:00 \
-//             trainer.max_steps=20 \
-//             model.micro_batch_size=1 \
-//             model.global_batch_size=1 \
-//             model.data.synthetic_data=True \
-//             exp_manager.exp_dir=/home/TestData/multimodal/stable_diffusion_train \
-//             model.inductor=False \
-//             model.cond_stage_config._target_=nemo.collections.multimodal.modules.stable_diffusion.encoders.modules.FrozenCLIPEmbedder \
-//             ++model.cond_stage_config.version=openai/clip-vit-large-patch14 \
-//             ++model.cond_stage_config.max_length=77 \
-//             ~model.cond_stage_config.restore_from_path \
-//             ~model.cond_stage_config.freeze \
-//             ~model.cond_stage_config.layer \
-//             model.unet_config.from_pretrained=null \
-//             model.first_stage_config.from_pretrained=null \
-//             model.unet_config.use_flash_attention=False \
-//             "
-//         sh "pip install 'webdataset>=0.1.48,<=0.1.62'"
-//         sh "rm -rf /home/TestData/multimodal/stable_diffusion_train"
-//       }
-//     }
+
+    stage('L2: Multimodal Imagen Train') {
+      when {
+        anyOf {
+          branch 'r1.23.0'
+          changeRequest target: 'r1.23.0'
+        }
+      }
+      failFast true
+      steps {
+        sh "rm -rf /home/TestData/multimodal/imagen_train"
+        sh "pip install webdataset==0.2.48"
+        sh "python examples/multimodal/text_to_image/imagen/imagen_training.py \
+        trainer.precision=16 \
+        trainer.num_nodes=1 \
+        trainer.devices=1 \
+        ++exp_manager.max_time_per_run=00:00:03:00 \
+        trainer.max_steps=20 \
+        model.conditioning.embed_dim=64 \
+        model.micro_batch_size=1 \
+        model.global_batch_size=1 \
+        model.data.synthetic_data=True \
+        exp_manager.exp_dir=/home/TestData/multimodal/imagen_train \
+        model.inductor=False \
+        model.unet.flash_attention=False \
+        "
+        sh "pip install 'webdataset>=0.1.48,<=0.1.62'"
+        sh "rm -rf /home/TestData/multimodal/imagen_train"
+      }
+    }
+    stage('L2: Multimodal Stable Diffusion Train') {
+      when {
+        anyOf {
+          branch 'r1.23.0'
+          changeRequest target: 'r1.23.0'
+        }
+      }
+      failFast true
+      steps {
+        sh "rm -rf /home/TestData/multimodal/stable_diffusion_train"
+        sh "pip install webdataset==0.2.48"
+        sh "python examples/multimodal/text_to_image/stable_diffusion/sd_train.py \
+            trainer.precision=16 \
+            trainer.num_nodes=1 \
+            trainer.devices=1 \
+            ++exp_manager.max_time_per_run=00:00:03:00 \
+            trainer.max_steps=20 \
+            model.micro_batch_size=1 \
+            model.global_batch_size=1 \
+            model.data.synthetic_data=True \
+            exp_manager.exp_dir=/home/TestData/multimodal/stable_diffusion_train \
+            model.inductor=False \
+            model.cond_stage_config._target_=nemo.collections.multimodal.modules.stable_diffusion.encoders.modules.FrozenCLIPEmbedder \
+            ++model.cond_stage_config.version=openai/clip-vit-large-patch14 \
+            ++model.cond_stage_config.max_length=77 \
+            ~model.cond_stage_config.restore_from_path \
+            ~model.cond_stage_config.freeze \
+            ~model.cond_stage_config.layer \
+            model.unet_config.from_pretrained=null \
+            model.first_stage_config.from_pretrained=null \
+            model.unet_config.use_flash_attention=False \
+            model.unet_config.attention_resolutions=[1] \
+            model.unet_config.channel_mult=[1] \
+            "
+        sh "pip install 'webdataset>=0.1.48,<=0.1.62'"
+        sh "rm -rf /home/TestData/multimodal/stable_diffusion_train"
+      }
+    }
 //     stage('L2: Multimodal ControlNet Train') {
 //       when {
 //         anyOf {
@@ -260,122 +262,122 @@ pipeline {
 //         sh "rm -rf /home/TestData/multimodal/dreambooth_train"
 //       }
 //     }
-//     stage('L2: Vision ViT Pretrain TP=1') {
-//       when {
-//         anyOf {
-//           branch 'main'
-//           changeRequest target: 'main'
-//         }
-//       }
-//       failFast true
-//       steps {
-//         sh "rm -rf /home/TestData/vision/vit_pretrain_tp1"
-//         sh "pip install webdataset==0.2.48"
-//         sh "python examples/vision/vision_transformer/megatron_vit_classification_pretrain.py \
-//             trainer.precision=16 \
-//             model.megatron_amp_O2=False \
-//             trainer.num_nodes=1 \
-//             trainer.devices=1 \
-//             trainer.val_check_interval=5 \
-//             ++exp_manager.max_time_per_run=00:00:03:00 \
-//             trainer.max_steps=20 \
-//             model.micro_batch_size=2 \
-//             model.global_batch_size=4 \
-//             model.tensor_model_parallel_size=1 \
-//             model.pipeline_model_parallel_size=1 \
-//             model.data.num_workers=0 \
-//             exp_manager.create_checkpoint_callback=False \
-//             model.data.data_path=[/home/TestData/multimodal/tiny-imagenet/train,/home/TestData/multimodal/tiny-imagenet/val] \
-//             exp_manager.exp_dir=/home/TestData/vision/vit_pretrain_tp1 "
-//         sh "pip install 'webdataset>=0.1.48,<=0.1.62'"
-//         sh "rm -rf /home/TestData/vision/vit_pretrain_tp1"
-//       }
-//     }
-//
-//     stage('L2: Multimodal CLIP Pretrain TP=1') {
-//       when {
-//         anyOf {
-//           branch 'main'
-//           changeRequest target: 'main'
-//         }
-//       }
-//       failFast true
-//       steps {
-//         sh "rm -rf /home/TestData/multimodal/clip_pretrain_tp1"
-//         sh "pip install webdataset==0.2.48"
-//         sh "python examples/multimodal/vision_language_foundation/clip/megatron_clip_pretrain.py  \
-//             trainer.precision=16 \
-//             model.megatron_amp_O2=False \
-//             trainer.num_nodes=1 \
-//             trainer.devices=1 \
-//             trainer.val_check_interval=10 \
-//             ++exp_manager.max_time_per_run=00:00:03:00 \
-//             trainer.max_steps=20 \
-//             model.micro_batch_size=1 \
-//             model.global_batch_size=1 \
-//             model.tensor_model_parallel_size=1 \
-//             model.pipeline_model_parallel_size=1 \
-//             exp_manager.create_checkpoint_callback=False \
-//             model.data.num_workers=0 \
-//             model.vision.num_layers=2 \
-//             model.text.num_layers=2 \
-//             model.vision.patch_dim=32 \
-//             model.vision.encoder_seq_length=49 \
-//             model.vision.class_token_length=7 \
-//             model.data.train.dataset_path=[/home/TestData/multimodal/tiny-clip/00000.tar] \
-//             model.data.validation.dataset_path=[/home/TestData/multimodal/tiny-clip/00000.tar] \
-//             model.data.webdataset.local_root_path=/ \
-//             exp_manager.exp_dir=/home/TestData/multimodal/clip_pretrain_tp1 "
-//         sh "pip install 'webdataset>=0.1.48,<=0.1.62'"
-//         sh "rm -rf /home/TestData/multimodal/clip_pretrain_tp1"
-//       }
-//     }
-//
-//     stage('L2: Multimodal NeVA Pretrain TP=1') {
-//       when {
-//         anyOf {
-//           branch 'main'
-//           changeRequest target: 'main'
-//         }
-//       }
-//       failFast true
-//       steps {
-//         sh "rm -rf /home/TestData/multimodal/neva_pretrain_tp1"
-//         sh "pip install webdataset==0.2.48"
-//         sh "python examples/multimodal/multimodal_llm/neva/neva_pretrain.py \
-//             trainer.precision=bf16 \
-//             model.megatron_amp_O2=False \
-//             trainer.num_nodes=1 \
-//             trainer.devices=1 \
-//             trainer.val_check_interval=10 \
-//             trainer.limit_val_batches=5 \
-//             trainer.log_every_n_steps=1 \
-//             ++exp_manager.max_time_per_run=00:00:03:00 \
-//             trainer.max_steps=20 \
-//             model.micro_batch_size=2 \
-//             model.global_batch_size=4 \
-//             model.tensor_model_parallel_size=1 \
-//             model.pipeline_model_parallel_size=1 \
-//             exp_manager.create_checkpoint_callback=False \
-//             model.data.data_path=/home/TestData/multimodal/tiny-neva/dummy.json \
-//             model.data.image_folder=/home/TestData/multimodal/tiny-neva/images \
-//             model.tokenizer.library=sentencepiece \
-//             model.tokenizer.model=/home/TestData/multimodal/tiny-neva/tokenizer_add_special.model \
-//             model.num_layers=2 \
-//             model.hidden_size=5120 \
-//             model.ffn_hidden_size=13824 \
-//             model.num_attention_heads=40 \
-//             model.normalization=rmsnorm \
-//             model.data.num_workers=0 \
-//             model.data.conv_template=llama_2 \
-//             model.mm_cfg.vision_encoder.from_pretrained='openai/clip-vit-large-patch14' \
-//             model.mm_cfg.llm.from_pretrained=null \
-//             model.use_flash_attention=false \
-//             exp_manager.exp_dir=/home/TestData/multimodal/neva_pretrain_tp1 "
-//         sh "pip install 'webdataset>=0.1.48,<=0.1.62'"
-//         sh "rm -rf /home/TestData/multimodal/neva_pretrain_tp1"
-//       }
-//     }
+    stage('L2: Vision ViT Pretrain TP=1') {
+      when {
+        anyOf {
+          branch 'r1.23.0'
+          changeRequest target: 'r1.23.0'
+        }
+      }
+      failFast true
+      steps {
+        sh "rm -rf /home/TestData/vision/vit_pretrain_tp1"
+        sh "pip install webdataset==0.2.48"
+        sh "python examples/vision/vision_transformer/megatron_vit_classification_pretrain.py \
+            trainer.precision=16 \
+            model.megatron_amp_O2=False \
+            trainer.num_nodes=1 \
+            trainer.devices=1 \
+            trainer.val_check_interval=5 \
+            ++exp_manager.max_time_per_run=00:00:03:00 \
+            trainer.max_steps=20 \
+            model.micro_batch_size=2 \
+            model.global_batch_size=4 \
+            model.tensor_model_parallel_size=1 \
+            model.pipeline_model_parallel_size=1 \
+            model.data.num_workers=0 \
+            exp_manager.create_checkpoint_callback=False \
+            model.data.data_path=[/home/TestData/multimodal/tiny-imagenet/train,/home/TestData/multimodal/tiny-imagenet/val] \
+            exp_manager.exp_dir=/home/TestData/vision/vit_pretrain_tp1 "
+        sh "pip install 'webdataset>=0.1.48,<=0.1.62'"
+        sh "rm -rf /home/TestData/vision/vit_pretrain_tp1"
+      }
+    }
+
+    stage('L2: Multimodal CLIP Pretrain TP=1') {
+      when {
+        anyOf {
+          branch 'r1.23.0'
+          changeRequest target: 'r1.23.0'
+        }
+      }
+      failFast true
+      steps {
+        sh "rm -rf /home/TestData/multimodal/clip_pretrain_tp1"
+        sh "pip install webdataset==0.2.48"
+        sh "python examples/multimodal/vision_language_foundation/clip/megatron_clip_pretrain.py  \
+            trainer.precision=16 \
+            model.megatron_amp_O2=False \
+            trainer.num_nodes=1 \
+            trainer.devices=1 \
+            trainer.val_check_interval=10 \
+            ++exp_manager.max_time_per_run=00:00:03:00 \
+            trainer.max_steps=20 \
+            model.micro_batch_size=1 \
+            model.global_batch_size=1 \
+            model.tensor_model_parallel_size=1 \
+            model.pipeline_model_parallel_size=1 \
+            exp_manager.create_checkpoint_callback=False \
+            model.data.num_workers=0 \
+            model.vision.num_layers=2 \
+            model.text.num_layers=2 \
+            model.vision.patch_dim=32 \
+            model.vision.encoder_seq_length=49 \
+            model.vision.class_token_length=7 \
+            model.data.train.dataset_path=[/home/TestData/multimodal/tiny-clip/00000.tar] \
+            model.data.validation.dataset_path=[/home/TestData/multimodal/tiny-clip/00000.tar] \
+            model.data.webdataset.local_root_path=/ \
+            exp_manager.exp_dir=/home/TestData/multimodal/clip_pretrain_tp1 "
+        sh "pip install 'webdataset>=0.1.48,<=0.1.62'"
+        sh "rm -rf /home/TestData/multimodal/clip_pretrain_tp1"
+      }
+    }
+
+    stage('L2: Multimodal NeVA Pretrain TP=1') {
+      when {
+        anyOf {
+          branch 'r1.23.0'
+          changeRequest target: 'r1.23.0'
+        }
+      }
+      failFast true
+      steps {
+        sh "rm -rf /home/TestData/multimodal/neva_pretrain_tp1"
+        sh "pip install webdataset==0.2.48"
+        sh "python examples/multimodal/multimodal_llm/neva/neva_pretrain.py \
+            trainer.precision=16 \
+            model.megatron_amp_O2=False \
+            trainer.num_nodes=1 \
+            trainer.devices=1 \
+            trainer.val_check_interval=10 \
+            trainer.limit_val_batches=5 \
+            trainer.log_every_n_steps=1 \
+            ++exp_manager.max_time_per_run=00:00:03:00 \
+            trainer.max_steps=20 \
+            model.micro_batch_size=2 \
+            model.global_batch_size=4 \
+            model.tensor_model_parallel_size=1 \
+            model.pipeline_model_parallel_size=1 \
+            exp_manager.create_checkpoint_callback=False \
+            model.data.data_path=/home/TestData/multimodal/tiny-neva/dummy.json \
+            model.data.image_folder=/home/TestData/multimodal/tiny-neva/images \
+            model.tokenizer.library=sentencepiece \
+            model.tokenizer.model=/home/TestData/multimodal/tiny-neva/tokenizer_add_special.model \
+            model.num_layers=2 \
+            model.hidden_size=5120 \
+            model.ffn_hidden_size=13824 \
+            model.num_attention_heads=40 \
+            model.normalization=rmsnorm \
+            model.data.num_workers=0 \
+            model.data.conv_template=llama_2 \
+            model.mm_cfg.vision_encoder.from_pretrained='openai/clip-vit-large-patch14' \
+            model.mm_cfg.llm.from_pretrained=null \
+            model.use_flash_attention=false \
+            exp_manager.exp_dir=/home/TestData/multimodal/neva_pretrain_tp1 "
+        sh "pip install 'webdataset>=0.1.48,<=0.1.62'"
+        sh "rm -rf /home/TestData/multimodal/neva_pretrain_tp1"
+      }
+    }
 
     // TODO: this requires TE >= v0.11 which is not available in 23.06.
     //        please uncomment this test once mcore CI is ready.
diff --git a/examples/multimodal/multimodal_llm/neva/neva_evaluation.py b/examples/multimodal/multimodal_llm/neva/neva_evaluation.py
index 545a634ac7fb..bd3f975e4d54 100644
--- a/examples/multimodal/multimodal_llm/neva/neva_evaluation.py
+++ b/examples/multimodal/multimodal_llm/neva/neva_evaluation.py
@@ -71,15 +71,16 @@ def main(cfg) -> None:
     with open(cfg.prompt_file, 'r') as f:
         lines = f.readlines()
 
+    insert_image_token = cfg.inference.get("insert_image_token", None)
     final_prompts = []
     for line in lines:
         prompt_dict = json.loads(line)
         assert 'prompt' in prompt_dict or 'text' in prompt_dict
         if 'prompt' not in prompt_dict:
             prompt_dict['prompt'] = prompt_dict['text']
-        if cfg.inference.insert_image_token == 'left':
+        if insert_image_token == 'left':
             prompt_dict['prompt'] = '<image>' + prompt_dict['prompt']
-        elif cfg.inference.insert_image_token == 'right':
+        elif insert_image_token == 'right':
             prompt_dict['prompt'] = prompt_dict['prompt'] + '<image>'
         if 'image' in prompt_dict:
             prompt_dict['image_path'] = prompt_dict['image']
diff --git a/examples/multimodal/vision_language_foundation/clip/conf/megatron_clip_VIT-L-14.yaml b/examples/multimodal/vision_language_foundation/clip/conf/megatron_clip_VIT-L-14.yaml
index 8a21fccd0874..d8740bb98eb2 100644
--- a/examples/multimodal/vision_language_foundation/clip/conf/megatron_clip_VIT-L-14.yaml
+++ b/examples/multimodal/vision_language_foundation/clip/conf/megatron_clip_VIT-L-14.yaml
@@ -79,7 +79,7 @@ model:
     openai_gelu: False
     bias_activation_fusion: False
     megatron_legacy: True
-    activation: quick-gelu
+    activation: approx-gelu
 
 
 
@@ -144,7 +144,7 @@ model:
     fp8_amax_history_len: 1 # Number of steps for which amax history is recorded per tensor
     fp8_amax_compute_algo: most_recent # 'most_recent' or 'max'. Algorithm for computing amax from history
     use_emha: False # Use fused multi-head attention for large sequence-length. Note this is not yet supported. Please set to False.
-    activation: quick-gelu
+    activation: approx-gelu
 
   # Megatron O2-style half-precision
   megatron_amp_O2: False # Enable O2-level automatic mixed precision using main parameters
diff --git a/examples/multimodal/vision_language_foundation/nsfw/conf/megatron_nsfw_config.yaml b/examples/multimodal/vision_language_foundation/nsfw/conf/megatron_nsfw_config.yaml
index 11dc65155cf5..be820e8d731d 100644
--- a/examples/multimodal/vision_language_foundation/nsfw/conf/megatron_nsfw_config.yaml
+++ b/examples/multimodal/vision_language_foundation/nsfw/conf/megatron_nsfw_config.yaml
@@ -117,7 +117,7 @@ model:
     openai_gelu: false
     bias_activation_fusion: false
     megatron_legacy: true
-    activation: quick-gelu
+    activation: approx-gelu
 
   text:
     precision: ${trainer.precision}
@@ -171,7 +171,7 @@ model:
     fp8_amax_history_len: 1
     fp8_amax_compute_algo: most_recent
     use_emha: false
-    activation: quick-gelu
+    activation: approx-gelu
 
   # Megatron O2-style half-precision
   megatron_amp_O2: False # Enable O2-level automatic mixed precision using main parameters
diff --git a/nemo/collections/multimodal/data/neva/neva_dataset.py b/nemo/collections/multimodal/data/neva/neva_dataset.py
index 4dd6b120c8c8..90f862869369 100644
--- a/nemo/collections/multimodal/data/neva/neva_dataset.py
+++ b/nemo/collections/multimodal/data/neva/neva_dataset.py
@@ -397,7 +397,7 @@ def preprocess_nvgpt(sources: dict, tokenizer, cfg,) -> Dict:
                 if 'label' not in turn:
                     turn[
                         'label'
-                    ] = "quality:6,toxicity:0,humor:0,creativity:0,violence:0,helpfulness:6,not_appropriate:0"
+                    ] = "quality:4,toxicity:0,humor:0,creativity:0,helpfulness:4,correctness:4,coherence:4,complexity:4,verbosity:4"
                 value = DEFAULT_LABELS_TOKEN + turn['label'] + '\n' + turn['value']
                 conv.append_message(turn['from'], value)
                 if not turn["value"]:
diff --git a/nemo/collections/multimodal/models/multimodal_llm/neva/neva_model.py b/nemo/collections/multimodal/models/multimodal_llm/neva/neva_model.py
index 5fd0fa830dd0..3f4156d0fa73 100644
--- a/nemo/collections/multimodal/models/multimodal_llm/neva/neva_model.py
+++ b/nemo/collections/multimodal/models/multimodal_llm/neva/neva_model.py
@@ -612,8 +612,8 @@ def forward(self, tokens, text_position_ids, attention_mask, labels, media=None)
         output_tensor = self.model(**forward_args)
         return output_tensor
 
-    def fwd_bwd_step(self, dataloader_iter, batch_idx, forward_only):
-        return MegatronGPTModel.fwd_bwd_step(self, dataloader_iter, batch_idx, forward_only)
+    def fwd_bwd_step(self, dataloader_iter, batch_idx, forward_only, first_val_step=None):
+        return MegatronGPTModel.fwd_bwd_step(self, dataloader_iter, batch_idx, forward_only, first_val_step)
 
     def training_step(self, dataloader_iter, batch_idx):
         """
@@ -623,7 +623,7 @@ def training_step(self, dataloader_iter, batch_idx):
         """
         return MegatronGPTModel.training_step(self, dataloader_iter, batch_idx)
 
-    def get_forward_output_and_loss_func(self, validation_step=False):
+    def get_forward_output_and_loss_func(self, validation_step=False, tuning=False):
         def loss_func(output_tensor, loss_mask):
             loss_for_ub = self.loss_func(loss_mask, output_tensor)
             if validation_step and not self.cfg.data.get('validation_drop_last', True):
@@ -921,7 +921,7 @@ def list_available_models(cls) -> Optional[PretrainedModelInfo]:
         Returns:
             List of available pre-trained models.
         """
-        return []
+        return None
 
     def setup_test_data(self, cfg):
         pass
diff --git a/nemo/collections/multimodal/models/text_to_image/instruct_pix2pix/ldm/ddpm_edit.py b/nemo/collections/multimodal/models/text_to_image/instruct_pix2pix/ldm/ddpm_edit.py
index 901745f09421..9bb490fb8fc8 100644
--- a/nemo/collections/multimodal/models/text_to_image/instruct_pix2pix/ldm/ddpm_edit.py
+++ b/nemo/collections/multimodal/models/text_to_image/instruct_pix2pix/ldm/ddpm_edit.py
@@ -40,7 +40,9 @@
 
 
 class LatentDiffusionEdit(LatentDiffusion):
-    def init_from_ckpt(self, path, ignore_keys=list(), only_model=False):
+    def init_from_ckpt(
+        self, path, ignore_keys=list(), only_model=False, load_vae=True, load_unet=True, load_encoder=True,
+    ):
         pl_sd = torch.load(path, map_location="cpu")
         if "state_dict" in list(pl_sd.keys()):
             pl_sd = pl_sd["state_dict"]
diff --git a/nemo/collections/multimodal/models/text_to_image/stable_diffusion/ldm/ddpm.py b/nemo/collections/multimodal/models/text_to_image/stable_diffusion/ldm/ddpm.py
index 31b56443846f..36dfb74fbfaf 100644
--- a/nemo/collections/multimodal/models/text_to_image/stable_diffusion/ldm/ddpm.py
+++ b/nemo/collections/multimodal/models/text_to_image/stable_diffusion/ldm/ddpm.py
@@ -557,9 +557,9 @@ def __init__(self, cfg, model_parallel_config):
 
         self.restarted_from_ckpt = False
         if ckpt_path is not None:
-            load_vae = True if cfg.load_vae is None else cfg.load_vae
-            load_unet = True if cfg.load_unet is None else cfg.load_unet
-            load_encoder = True if cfg.load_encoder is None else cfg.load_encoder
+            load_vae = True if cfg.get("load_vae", None) is None else cfg.load_vae
+            load_unet = True if cfg.get("load_unet", None) is None else cfg.load_unet
+            load_encoder = True if cfg.get("load_encoder", None) is None else cfg.load_encoder
 
             self.init_from_ckpt(
                 ckpt_path, ignore_keys, load_vae=load_vae, load_unet=load_unet, load_encoder=load_encoder,