Merge pull request #36 from shijinpjlab/dev

feat: add llm demo, including local and remote.（deepseek support）

Author: e06084

dingo/model/llm/base_openai.py

@@ -41,10 +41,10 @@ def build_messages(cls, input_data: MetaData) -> List:
 
     @classmethod
     def send_messages(cls, messages: List):
-        if cls.dynamic_config.model is None:
-            model_name = cls.client.models.list().data[0].id
-        else:
+        if cls.dynamic_config.model:
             model_name = cls.dynamic_config.model
+        else:
+            model_name = cls.client.models.list().data[0].id
 
         params = cls.dynamic_config.parameters
         cls.validate_config(params)

examples/app_huggingface/app.py

@@ -5,7 +5,7 @@
 from dingo.io import InputArgs
 
 
-def dingo_demo(input_path, data_format, column_content, rule_list, prompt_list, key, api_url):
+def dingo_demo(input_path, data_format, column_content, rule_list, prompt_list, model, key, api_url):
     if not input_path:
         return 'ValueError: input_path can not be empty, please input.'
     if not data_format:
@@ -27,6 +27,7 @@ def dingo_demo(input_path, data_format, column_content, rule_list, prompt_list,
                     {
                         "detect_text_quality_detail":
                             {
+                                "model": model,
                                 "key": key,
                                 "api_url": api_url,
                             }
@@ -55,8 +56,9 @@ def dingo_demo(input_path, data_format, column_content, rule_list, prompt_list,
                 column_content = gr.Textbox(value="content", placeholder="please input column name of content in dataset", label="column_content")
                 rule_list = gr.CheckboxGroup(choices=rule_options, label="rule_list")
                 prompt_list = gr.CheckboxGroup(choices=prompt_options, label="prompt_list")
-                key = gr.Textbox(placeholder="If want to use llm, please input the key of it.", label="key")
-                api_url = gr.Textbox(placeholder="If want to use llm, please input the api_url of it.", label="api_url")
+                model = gr.Textbox(placeholder="If want to use llm, please input model, such as: deepseek-chat", label="model")
+                key = gr.Textbox(placeholder="If want to use llm, please input key, such as: 123456789012345678901234567890xx", label="key")
+                api_url = gr.Textbox(placeholder="If want to use llm, please input api_url, such as: https://api.deepseek.com/v1", label="api_url")
                 with gr.Row():
                     submit_single = gr.Button(value="Submit", interactive=True, variant="primary")
             with gr.Column():
@@ -65,7 +67,7 @@ def dingo_demo(input_path, data_format, column_content, rule_list, prompt_list,
 
         submit_single.click(
             fn=dingo_demo,
-            inputs=[input_path, data_format, column_content, rule_list, prompt_list, key, api_url],
+            inputs=[input_path, data_format, column_content, rule_list, prompt_list, model, key, api_url],
             outputs=output
         )
 

examples/llm/local_llm.py

@@ -0,0 +1,28 @@
+from dingo.exec import Executor
+from dingo.io import InputArgs
+
+input_data = {
+    "input_path": "../../test/data/test_local_jsonl.jsonl",  # local filesystem dataset
+    "save_data": True,
+    "save_correct": True,
+    "dataset": "local",
+    "data_format": "jsonl",
+    "column_content": "content",
+    "custom_config":
+        {
+            "prompt_list": ["PromptRepeat"],
+            "llm_config":
+                {
+                    "detect_text_quality":
+                        {
+                            "key": "enter your key, such as:EMPTY",
+                            "api_url": "enter your local llm api url, such as:http://127.0.0.1:8080/v1",
+                        }
+                }
+        },
+    "log_level": "INFO"
+}
+input_args = InputArgs(**input_data)
+executor = Executor.exec_map["local"](input_args)
+result = executor.execute()
+print(result)

examples/llm/remote_llm.py

@@ -0,0 +1,29 @@
+from dingo.exec import Executor
+from dingo.io import InputArgs
+
+input_data = {
+    "input_path": "../../test/data/test_local_jsonl.jsonl",  # local filesystem dataset
+    "save_data": True,
+    "save_correct": True,
+    "dataset": "local",
+    "data_format": "jsonl",
+    "column_content": "content",
+    "custom_config":
+        {
+            "prompt_list": ["PromptRepeat"],
+            "llm_config":
+                {
+                    "detect_text_quality":
+                        {
+                            "model": "enter your llm, such as:deepseek-chat",
+                            "key": "enter your key, such as:sk-123456789012345678901234567890xx",
+                            "api_url": "enter remote llm api url, such as:https://api.deepseek.com/v1",
+                        }
+                }
+        },
+    "log_level": "INFO"
+}
+input_args = InputArgs(**input_data)
+executor = Executor.exec_map["local"](input_args)
+result = executor.execute()
+print(result)