diff --git a/applications/DeepSpeed-Chat/dschat/utils/model/model_utils.py b/applications/DeepSpeed-Chat/dschat/utils/model/model_utils.py
index 97d3bff15..050819a22 100644
--- a/applications/DeepSpeed-Chat/dschat/utils/model/model_utils.py
+++ b/applications/DeepSpeed-Chat/dschat/utils/model/model_utils.py
@@ -10,7 +10,7 @@
     AutoModel,
 )
 from huggingface_hub import snapshot_download
-from transformers.deepspeed import HfDeepSpeedConfig
+from transformers.integrations.deepspeed import HfDeepSpeedConfig
 
 from dschat.utils.model.reward_model import RewardModel
 from dschat.utils.utils import load_state_dict_into_model, print_rank_0
diff --git a/applications/DeepSpeed-VisualChat/utils/model/modeling_dsvl.py b/applications/DeepSpeed-VisualChat/utils/model/modeling_dsvl.py
index eb9db9428..1407c1dfc 100755
--- a/applications/DeepSpeed-VisualChat/utils/model/modeling_dsvl.py
+++ b/applications/DeepSpeed-VisualChat/utils/model/modeling_dsvl.py
@@ -15,7 +15,7 @@
     os.path.abspath(os.path.join(os.path.dirname(__file__), os.path.pardir)))
 import data.DST as DST # default special tokens
 from torch.utils.data import DataLoader
-from transformers.deepspeed import HfDeepSpeedConfig
+from transformers.integrations.deepspeed import HfDeepSpeedConfig
 import numpy as np
 from .vis_proj import VisProjection_vit, VisProjection_perceiver
 
diff --git a/inference/huggingface/zero_inference/README.md b/inference/huggingface/zero_inference/README.md
index f6dd4850e..acca9404e 100644
--- a/inference/huggingface/zero_inference/README.md
+++ b/inference/huggingface/zero_inference/README.md
@@ -90,7 +90,7 @@ deepspeed --num_gpus 1 run_model.py --model bigscience/bloom-7b1 --batch-size 8
 Here is an example of running `meta-llama/Llama-2-7b-hf` with Zero-Inference using 4-bit model weights and offloading kv cache to CPU:
 
 ```sh
-deepspeed --num_gpus 1 run_model.py --model meta-llama/Llama-2-7b-hf` --batch-size 8 --prompt-len 512 --gen-len 32 --cpu-offload --quant-bits 4 --kv-offload
+deepspeed --num_gpus 1 run_model.py --model meta-llama/Llama-2-7b-hf --batch-size 8 --prompt-len 512 --gen-len 32 --cpu-offload --quant-bits 4 --kv-offload
 ```
 
 ## Performance Tuning Tips
diff --git a/inference/huggingface/zero_inference/run_model.py b/inference/huggingface/zero_inference/run_model.py
index 230d601cb..d0e16eca3 100644
--- a/inference/huggingface/zero_inference/run_model.py
+++ b/inference/huggingface/zero_inference/run_model.py
@@ -19,7 +19,7 @@
 from transformers import (AutoConfig, AutoTokenizer, AutoModelForCausalLM, 
                           BloomForCausalLM, OPTForCausalLM, LlamaForCausalLM,
                         )
-from transformers.deepspeed import HfDeepSpeedConfig
+from transformers.integrations.deepspeed import HfDeepSpeedConfig
 from utils import (GB, add_model_hooks, cache_bytes,
                    get_filename, get_quant_config, hidden_bytes, meta_to_cpu,
                    model_bytes, write_benchmark_log)