Merge 4372071fc5 into ed4409d2a9

README.md

@@ -207,14 +207,46 @@ python3 -m sglang.launch_server --model deepseek-ai/DeepSeek-R1-Distill-Qwen-32B
 3. For mathematical problems, it is advisable to include a directive in your prompt such as: "Please reason step by step, and put your final answer within \boxed{}."
 4. When evaluating model performance, it is recommended to conduct multiple tests and average the results.
 
-## 7. License
+## 7. Message Handling Capability
+
+DeepSeek-R1 now supports successive user or assistant messages without causing an `invalid_request_error`. The message handling logic has been modified to allow successive messages of the same role. Here are some examples of how to format messages with successive user or assistant messages:
+
+### Example 1: Interleaved Messages
+
+```json
+[
+  {"role": "user", "content": "Hello!"},
+  {"role": "assistant", "content": "Hi! How can I help?"},
+  {"role": "user", "content": "Tell me about DeepSeek R1."}
+]
+```
+
+### Example 2: Successive User Messages
+
+```json
+[
+  {"role": "user", "content": "Hello!"},
+  {"role": "user", "content": "Tell me about DeepSeek R1."}
+]
+```
+
+### Example 3: Successive Assistant Messages
+
+```json
+[
+  {"role": "assistant", "content": "Hello!"},
+  {"role": "assistant", "content": "How can I assist you today?"}
+]
+```
+
+## 8. License
 This code repository and the model weights are licensed under the [MIT License](https://github.com/deepseek-ai/DeepSeek-R1/blob/main/LICENSE).
 DeepSeek-R1 series support commercial use, allow for any modifications and derivative works, including, but not limited to, distillation for training other LLMs. Please note that:
 - DeepSeek-R1-Distill-Qwen-1.5B, DeepSeek-R1-Distill-Qwen-7B, DeepSeek-R1-Distill-Qwen-14B and DeepSeek-R1-Distill-Qwen-32B are derived from [Qwen-2.5 series](https://github.com/QwenLM/Qwen2.5), which are originally licensed under [Apache 2.0 License](https://huggingface.co/Qwen/Qwen2.5-1.5B/blob/main/LICENSE), and now finetuned with 800k samples curated with DeepSeek-R1.
 - DeepSeek-R1-Distill-Llama-8B is derived from Llama3.1-8B-Base and is originally licensed under [llama3.1 license](https://huggingface.co/meta-llama/Llama-3.1-8B/blob/main/LICENSE).
 - DeepSeek-R1-Distill-Llama-70B is derived from Llama3.3-70B-Instruct and is originally licensed under [llama3.3 license](https://huggingface.co/meta-llama/Llama-3.3-70B-Instruct/blob/main/LICENSE).
 
-## 8. Citation
+## 9. Citation
 ```
 @misc{deepseekai2025deepseekr1incentivizingreasoningcapability,
       title={DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning}, 
@@ -228,5 +260,5 @@ DeepSeek-R1 series support commercial use, allow for any modifications and deriv
 
 ```
 
-## 9. Contact
+## 10. Contact
 If you have any questions, please raise an issue or contact us at [service@deepseek.com](service@deepseek.com).

deepseek_reasoner.py

@@ -0,0 +1,46 @@
+import json
+
+class DeepSeekReasoner:
+    def __init__(self):
+        pass
+
+    def process_messages(self, messages):
+        """
+        Process the input messages to ensure they follow the required format.
+        """
+        if not messages:
+            raise ValueError("Messages cannot be empty")
+
+        # Merge successive messages of the same role
+        merged_messages = []
+        for message in messages:
+            if merged_messages and merged_messages[-1]['role'] == message['role']:
+                merged_messages[-1]['content'] += " " + message['content']
+            else:
+                merged_messages.append(message)
+
+        return merged_messages
+
+    def handle_request(self, request):
+        """
+        Handle the incoming request and process the messages.
+        """
+        try:
+            messages = request.get('messages', [])
+            processed_messages = self.process_messages(messages)
+            return {"status": "success", "processed_messages": processed_messages}
+        except Exception as e:
+            return {"status": "error", "message": str(e)}
+
+# Example usage
+if __name__ == "__main__":
+    reasoner = DeepSeekReasoner()
+    request = {
+        "messages": [
+            {"role": "user", "content": "Hello!"},
+            {"role": "user", "content": "Tell me about DeepSeek R1."},
+            {"role": "assistant", "content": "DeepSeek R1 is a reasoning model."}
+        ]
+    }
+    response = reasoner.handle_request(request)
+    print(json.dumps(response, indent=2))

test_deepseek_reasoner.py

@@ -0,0 +1,46 @@
+import unittest
+from deepseek_reasoner import DeepSeekReasoner
+
+class TestDeepSeekReasoner(unittest.TestCase):
+
+    def setUp(self):
+        self.reasoner = DeepSeekReasoner()
+
+    def test_interleaved_messages(self):
+        request = {
+            "messages": [
+                {"role": "user", "content": "Hello!"},
+                {"role": "assistant", "content": "Hi! How can I help?"},
+                {"role": "user", "content": "Tell me about DeepSeek R1."}
+            ]
+        }
+        response = self.reasoner.handle_request(request)
+        self.assertEqual(response["status"], "success")
+        self.assertEqual(len(response["processed_messages"]), 3)
+
+    def test_successive_user_messages(self):
+        request = {
+            "messages": [
+                {"role": "user", "content": "Hello!"},
+                {"role": "user", "content": "Tell me about DeepSeek R1."}
+            ]
+        }
+        response = self.reasoner.handle_request(request)
+        self.assertEqual(response["status"], "success")
+        self.assertEqual(len(response["processed_messages"]), 1)
+        self.assertEqual(response["processed_messages"][0]["content"], "Hello! Tell me about DeepSeek R1.")
+
+    def test_successive_assistant_messages(self):
+        request = {
+            "messages": [
+                {"role": "assistant", "content": "Hello!"},
+                {"role": "assistant", "content": "How can I assist you today?"}
+            ]
+        }
+        response = self.reasoner.handle_request(request)
+        self.assertEqual(response["status"], "success")
+        self.assertEqual(len(response["processed_messages"]), 1)
+        self.assertEqual(response["processed_messages"][0]["content"], "Hello! How can I assist you today?")
+
+if __name__ == "__main__":
+    unittest.main()