diff --git a/dbgpt/configs/model_config.py b/dbgpt/configs/model_config.py
index f45f6503b..d27d741f4 100644
--- a/dbgpt/configs/model_config.py
+++ b/dbgpt/configs/model_config.py
@@ -184,6 +184,7 @@ LLM_MODEL_CONFIG = {
     # https://huggingface.co/google/gemma-2b-it
     "gemma-2b-it": os.path.join(MODEL_PATH, "gemma-2b-it"),
     "starling-lm-7b-beta": os.path.join(MODEL_PATH, "Starling-LM-7B-beta"),
+    "deepseek-v2-lite-chat": os.path.join(MODEL_PATH, "DeepSeek-V2-Lite-Chat"),
 }
 
 EMBEDDING_MODEL_CONFIG = {
diff --git a/dbgpt/model/adapter/hf_adapter.py b/dbgpt/model/adapter/hf_adapter.py
index 8a1f5d764..e3248d8f6 100644
--- a/dbgpt/model/adapter/hf_adapter.py
+++ b/dbgpt/model/adapter/hf_adapter.py
@@ -370,6 +370,32 @@ class Llama3Adapter(NewHFChatModelAdapter):
         return str_prompt
 
 
+class DeepseekV2Adapter(NewHFChatModelAdapter):
+    support_4bit: bool = False
+    support_8bit: bool = False
+
+    def do_match(self, lower_model_name_or_path: Optional[str] = None):
+        return (
+            lower_model_name_or_path
+            and "deepseek" in lower_model_name_or_path
+            and "v2" in lower_model_name_or_path
+            and "chat" in lower_model_name_or_path
+        )
+
+    def load(self, model_path: str, from_pretrained_kwargs: dict):
+        if not from_pretrained_kwargs:
+            from_pretrained_kwargs = {}
+        if "trust_remote_code" not in from_pretrained_kwargs:
+            from_pretrained_kwargs["trust_remote_code"] = True
+        model, tokenizer = super().load(model_path, from_pretrained_kwargs)
+
+        from transformers import GenerationConfig
+
+        model.generation_config = GenerationConfig.from_pretrained(model_path)
+        model.generation_config.pad_token_id = model.generation_config.eos_token_id
+        return model, tokenizer
+
+
 # The following code is used to register the model adapter
 # The last registered model adapter is matched first
 register_model_adapter(YiAdapter)
@@ -381,3 +407,4 @@ register_model_adapter(StarlingLMAdapter)
 register_model_adapter(QwenAdapter)
 register_model_adapter(QwenMoeAdapter)
 register_model_adapter(Llama3Adapter)
+register_model_adapter(DeepseekV2Adapter)
diff --git a/scripts/setup_autodl_env.sh b/scripts/setup_autodl_env.sh
index e4af3068e..5dc1f19eb 100644
--- a/scripts/setup_autodl_env.sh
+++ b/scripts/setup_autodl_env.sh
@@ -41,7 +41,7 @@ clone_repositories() {
 }
 
 install_dbgpt_packages() {
-    conda activate dbgpt && cd /root/DB-GPT && pip install -e ".[default]"
+    conda activate dbgpt && cd /root/DB-GPT && pip install -e ".[default]" && pip install transformers_stream_generator einops
     cp .env.template .env && sed -i 's/LLM_MODEL=vicuna-13b-v1.5/LLM_MODEL=qwen-1.8b-chat/' .env
 }