Update src/txagent/txagent.py

src/txagent/txagent.py

@@ -78,9 +78,10 @@ class TxAgent:
             model=self.model_name,
             dtype="float16",
             max_model_len=131072,
-            max_num_batched_tokens=32768,  # Increased for A100 80GB
-            gpu_memory_utilization=0.9,    # Higher utilization for better performance
-            trust_remote_code=True
+            max_num_batched_tokens=65536,  # Increased for A100 80GB
+            gpu_memory_utilization=0.95,    # Higher utilization for better performance
+            trust_remote_code=True,
+            engine_args={"max_num_seqs": 512},
         )
         self.chat_template = Template(self.model.get_tokenizer().chat_template)
         self.tokenizer = self.model.get_tokenizer()