[https://nvbugs/5374016][fix] improve error message (#6893)

Signed-off-by: junq <[email protected]>

Author: QiJune

tensorrt_llm/_torch/attention_backend/trtllm.py

@@ -763,8 +763,14 @@ def prepare(self) -> None:
             self.kv_cache_block_offsets[:, :self.num_seqs].copy_(
                 self.host_kv_cache_block_offsets[:, :self.num_seqs],
                 non_blocking=True)
+
+            error_message = (
+                f"The max KV cache length of input sequences ({self.kv_lens[:self.num_seqs].max()}) "
+                f"exceeds the KV cache manager's maximum supported length "
+                f"({self.kv_cache_manager.max_seq_len}).")
+
             assert self.kv_lens[:self.num_seqs].max(
-            ) <= self.kv_cache_manager.max_seq_len, f"Please set max_seq_len to at least {self.kv_lens[:self.num_seqs].max()} for kv cache manager."
+            ) <= self.kv_cache_manager.max_seq_len, error_message
 
         self.kv_lens_cuda_runtime = self.kv_lens_cuda[:self.num_seqs]
         self.kv_lens_runtime = self.kv_lens[:self.num_seqs]