feat: switch framework to vllm_fix_tokenizer

main.py

@@ -247,10 +247,10 @@ def check_queue_available() -> int:
 
 
 def build_config_params() -> str:
-    """构建 YAML 配置 - vllm"""
+    """构建 YAML 配置 - vllm_fix_tokenizer"""
     params = {
-        'framework': 'vllm',
+        'framework': 'vllm_fix_tokenizer',
-        'nv_framework': 'vllm',
+        'nv_framework': 'vllm_fix_tokenizer',
         'api': 'completion',
         'max_tokens': 1024,
         'temperature': 0.7,
@@ -295,7 +295,7 @@ def submit_model(model_url: str) -> tuple:
         'modelAddress': normalize_model_url(model_url),
         'taskType': 'text-generation',
         'targetGpu': TARGET_GPU,
-        'framework': 'vllm',
+        'framework': 'vllm_fix_tokenizer',
         'strategyId': STRATEGY_ID,
         'configParams': build_config_params(),
     }
@@ -568,7 +568,7 @@ class AgentHandler(BaseHTTPRequestHandler):
                     'modelAddress': 'https://www.modelscope.cn/models/Qwen/Qwen3-8B',
                     'taskType': 'text-generation',
                     'targetGpu': TARGET_GPU,
-                    'framework': 'vllm',
+                    'framework': 'vllm_fix_tokenizer',
                     'strategyId': STRATEGY_ID,
                     'configParams': build_config_params(),
                 },
@@ -686,7 +686,7 @@ def main():
                 json={
                     'modelAddress': 'https://www.modelscope.cn/models/Qwen/Qwen3-8B',
                     'taskType': 'text-generation', 'targetGpu': TARGET_GPU,
-                    'framework': 'vllm', 'strategyId': STRATEGY_ID,
+                    'framework': 'vllm_fix_tokenizer', 'strategyId': STRATEGY_ID,
                     'configParams': build_config_params(),
                 }, timeout=10
             )