feat: update vllm versions

Author: yanbasic

src/emd/models/engines.py

@@ -118,6 +118,13 @@ class KtransformersEngine(OpenAICompitableEngine):
             "default_cli_args": "--max_num_seq 256 --max_model_len 16000 --chat-template emd/models/chat_templates/deepseek_r1_distill.jinja"
 })
 
+vllm_deepseek_r1_distill_qwen_engine085 = VllmEngine(**{
+            **vllm_engine064.model_dump(),
+            "engine_dockerfile_config": {"VERSION":"v0.8.5"},
+            "default_cli_args": "--max_num_seq 256 --max_model_len 16000 --chat-template emd/models/chat_templates/deepseek_r1_distill.jinja"
+})
+
+
 vllm_deepseek_r1_distill_llama_engine071 = vllm_deepseek_r1_distill_qwen_engine071
 
 vllm_deepseek_r1_engine084 = VllmEngine(**{

src/emd/models/llms/deepseek.py

@@ -1,6 +1,7 @@
 from .. import Model
 from ..engines import (
     vllm_deepseek_r1_distill_qwen_engine071,
+    vllm_deepseek_r1_distill_qwen_engine085,
     vllm_deepseek_r1_distill_llama_engine071,
     ollama_deepseek_r1_qwen2d5_1d5b_engine057,
     llama_cpp_deepseek_r1_1d58_bit_engine_b9ab0a4,
@@ -307,7 +308,7 @@
 Model.register(
     dict(
         model_id = "DeepSeek-R1-0528-Qwen3-8B",
-        supported_engines=[vllm_deepseek_r1_distill_llama_engine071],
+        supported_engines=[vllm_deepseek_r1_distill_qwen_engine085],
         supported_instances=[
             g5dxlarge_instance,
             g5d2xlarge_instance,