add logprob method to Model

serve/mlc_serve/model/paged_cache_model.py

@@ -660,6 +660,27 @@ def generate(
 
             return outputs
 
+    def logprob(
+        self,
+        context: str,
+        continuation: str,
+    ):
+        r"""Generate log probs for given context and continuation.
+        Return logprobs and is_greedy boolean
+        Parameters
+        ----------
+        context : str
+            The user input context string.
+        continuation : str
+            The user input continuation string.
+
+        Returns
+        -------
+        {"logprobs": float, "is_greedy": bool} : dict
+        """
+
+        return self.mod["loglikelihood"](context, continuation)
+
 
 def get_gpu_memory(gpu: int = 0) -> int:
     return torch.cuda.get_device_properties(gpu).total_memory