Automatically updated from 모델 배포(742:TinyLlama-1.1B) by 그룹사용자(groupuser)

config.pbtxt

@@ -1,66 +1,110 @@
-# Triton Backend for vLLM.
+# Triton Backend for TransformerLLM.
-backend: "vllm"
+backend: "python"
-platform: "vllm"
 max_batch_size: 0
 
+# Triton should expect as input a single string
+# input of variable length named 'text_input'
 input [
+
   {
     name: "text_input"
     data_type: TYPE_STRING
     dims: [ 1 ]
+
+
   },
   {
     name: "max_length"
     data_type: TYPE_INT32
     dims: [ 1 ]
+
+
   },
   {
     name: "max_new_tokens"
     data_type: TYPE_INT32
     dims: [ 1 ]
+
+
   },
   {
     name: "do_sample"
     data_type: TYPE_BOOL
     dims: [ 1 ]
+
+
   },
   {
     name: "top_k"
     data_type: TYPE_INT32
     dims: [ 1 ]
+
+
   },
   {
     name: "top_p"
     data_type: TYPE_FP32
     dims: [ 1 ]
+
+
   },
   {
     name: "temperature"
     data_type: TYPE_FP32
     dims: [ 1 ]
+
+
   },
   {
     name: "repetition_penalty"
     data_type: TYPE_FP32
     dims: [ 1 ]
+
+
   },
   {
     name: "stream"
     data_type: TYPE_BOOL
     dims: [ 1 ]
+
+
   }
+
 ]
 
+
+# Triton should expect to respond with a single string
+# output of variable length named 'text_output'
 output [
+
   {
     name: "text_output"
     data_type: TYPE_STRING
     dims: [ 1 ]
+
   }
+
 ]
 
+parameters: [
+  {
+    key: "base_model_path",
+    value: {string_value: "/cheetah/input/model/groupuser/TinyLlama-1.1B-Chat-v1.0"}
+  },
+  {
+    key: "is_adapter_model",
+    value: {string_value: "false"}
+  },
+  {
+    key: "adapter_model_path",
+    value: {string_value: ""}
+  },
   
-
+  {
+    key: "quantization",
+    value: {string_value: "none"}
+  }
+]
 
 instance_group [
   {
@@ -68,3 +112,4 @@ instance_group [
     count: 1
   }
 ]
+