lint

Author: NotBioWaste905

dialogue2graph/pipelines/model_storage.py

@@ -104,7 +104,7 @@ def load(self, path: Path):
         try:
             with open(path, "r") as f:
                 loaded_storage = yaml.safe_load(f)
-                
+
                 for key, config in loaded_storage.items():
                     self.add(
                         key=key, config=config, model_type=config.pop("model_type")
@@ -169,7 +169,9 @@ def save(self, path: str):
                 for model_key in self.storage:
                     storage_dump[model_key] = {}
                     storage_dump[model_key]["config"] = self.storage[model_key].config
-                    storage_dump[model_key]["model_type"] = self.storage[model_key].model_type
+                    storage_dump[model_key]["model_type"] = self.storage[
+                        model_key
+                    ].model_type
                 yaml.dump(storage_dump, f)
             logger.info(f"Saved {len(self.storage)} models to {path}")
         except Exception as e:

docs/source/conf.py

@@ -11,9 +11,9 @@
 
 sys.path.insert(0, os.path.abspath("../../dialogue2graph"))
 
-project = 'Dialogue2Graph'
-copyright = '2024, Denis Kuznetsov, Anastasia Voznyuk, Andrey Chirkin'
-author = 'Denis Kuznetsov, Anastasia Voznyuk, Andrey Chirkin'
+project = "Dialogue2Graph"
+copyright = "2024, Denis Kuznetsov, Anastasia Voznyuk, Andrey Chirkin"
+author = "Denis Kuznetsov, Anastasia Voznyuk, Andrey Chirkin"
 
 # Get the deployment environment
 on_github = os.environ.get("GITHUB_ACTIONS") == "true"
@@ -71,7 +71,10 @@
 html_static_path = ["_static"]
 
 extlinks = {
-    'github_source_link': ("https://github.com/deeppavlov/dialogue2graph/tree/dev/%s", None),
+    "github_source_link": (
+        "https://github.com/deeppavlov/dialogue2graph/tree/dev/%s",
+        None,
+    ),
 }
 
 # Add these configurations

experiments/exp2025_03_12_rec_models_incrementation/exp2025_03_12_rec_models_incrementation/append_chain.py

@@ -17,9 +17,12 @@
 #     compare_graphs
 # )
 
+from dialogue2graph.metrics.no_llm_metrics import is_same_structure
+from dialogue2graph.metrics.llm_metrics import compare_graphs
 
 env_settings = EnvSettings()
 
+
 # @AlgorithmRegistry.register(input_type=list[Dialogue], path_to_result=env_settings.GENERATION_SAVE_PATH, output_type=BaseGraph)
 class AppendChain(GraphExtender):
     """
@@ -32,41 +35,48 @@ class AppendChain(GraphExtender):
     Returns:
         graph
     """
+
     prompt: str = ""
+
     def __init__(self):
         super().__init__()
         self.prompt = PromptTemplate.from_template(prompt_dialogs_and_graph)
 
-    def invoke(self, dialogues: list[Dialogue] = None, graph: Graph = None) -> BaseGraph:
-        print("model:  ",env_settings.GENERATION_MODEL_NAME)
-        base_model = ChatOpenAI(model=env_settings.GENERATION_MODEL_NAME, api_key=env_settings.OPENAI_API_KEY, base_url=env_settings.OPENAI_BASE_URL, temperature=0)
+    def invoke(
+        self, dialogues: list[Dialogue] = None, graph: Graph = None
+    ) -> BaseGraph:
+        print("model:  ", env_settings.GENERATION_MODEL_NAME)
+        base_model = ChatOpenAI(
+            model=env_settings.GENERATION_MODEL_NAME,
+            api_key=env_settings.OPENAI_API_KEY,
+            base_url=env_settings.OPENAI_BASE_URL,
+            temperature=0,
+        )
         model = base_model | PydanticOutputParser(pydantic_object=DialogueGraph)
 
         final_prompt = self.prompt.format(
-            orig_dial=dialogues[0],
-            orig_graph=graph.graph_dict,
-            new_dial=dialogues[1]
+            orig_dial=dialogues[0], orig_graph=graph.graph_dict, new_dial=dialogues[1]
         )
 
         result = call_llm_api(final_prompt, model, temp=0)
         if result is None:
             return Graph(graph_dict={})
 
         graph_dict = result.model_dump()
-        
-        if not all([e['target'] for e in graph_dict['edges']]):
+
+        if not all([e["target"] for e in graph_dict["edges"]]):
             return Graph(graph_dict={}), []
 
         result_graph = Graph(graph_dict=graph_dict)
         return result_graph
 
     async def ainvoke(self, *args, **kwargs):
         return self.invoke(*args, **kwargs)
-    
+
     async def evaluate(self, dialogues, graph, target_graph):
         result_graph = self.invoke(dialogues, graph)
-        # report = {
-        #     "is_same_structure": is_same_structure(result_graph, target_graph),
-        #     "graph_match": compare_graphs(result_graph, target_graph),
-        # }
+        report = {
+            "is_same_structure": is_same_structure(result_graph, target_graph),
+            "graph_match": compare_graphs(result_graph, target_graph),
+        }
         return report

experiments/exp2025_03_12_rec_models_incrementation/exp2025_03_12_rec_models_incrementation/embedder.py

@@ -7,8 +7,9 @@
 
 
 class EnvSettings(BaseSettings, case_sensitive=True):
-
-    model_config = SettingsConfigDict(env_file=os.environ["PATH_TO_ENV"], env_file_encoding="utf-8")
+    model_config = SettingsConfigDict(
+        env_file=os.environ["PATH_TO_ENV"], env_file_encoding="utf-8"
+    )
 
     OPENAI_API_KEY: Optional[str]
     OPENAI_BASE_URL: Optional[str]
@@ -21,24 +22,39 @@ class EnvSettings(BaseSettings, case_sensitive=True):
 preloaded_models = {}
 
 
-def compare_strings(first: str, second: str, embeddings: HuggingFaceEmbeddings, embedder_th: float = 0.001) -> bool:
+def compare_strings(
+    first: str,
+    second: str,
+    embeddings: HuggingFaceEmbeddings,
+    embedder_th: float = 0.001,
+) -> bool:
     """Calculate pairwise_embedding_distance between two strings based on embeddings
     and return True when threshold embedder_th not exceeded
     Return False othetwise"""
 
     evaluator_2 = load_evaluator("pairwise_embedding_distance", embeddings=embeddings)
-    score = evaluator_2.evaluate_string_pairs(prediction=first, prediction_b=second)["score"]
+    score = evaluator_2.evaluate_string_pairs(prediction=first, prediction_b=second)[
+        "score"
+    ]
     # print("SCORE: ", score)
     return score <= embedder_th
 
 
-def get_similarity(generated: list[str], golden: list[str], model_name: str = "BAAI/bge-m3"):
+def get_similarity(
+    generated: list[str], golden: list[str], model_name: str = "BAAI/bge-m3"
+):
     """ "Calculate similarity matrix between generated and golden using model model_name"""
 
     if model_name not in preloaded_models:
-        preloaded_models[model_name] = SentenceTransformer(model_name, device=env_settings.DEVICE)
-
-    golden_vectors = preloaded_models[model_name].encode(golden, normalize_embeddings=True)
-    generated_vectors = preloaded_models[model_name].encode(generated, normalize_embeddings=True)
+        preloaded_models[model_name] = SentenceTransformer(
+            model_name, device=env_settings.DEVICE
+        )
+
+    golden_vectors = preloaded_models[model_name].encode(
+        golden, normalize_embeddings=True
+    )
+    generated_vectors = preloaded_models[model_name].encode(
+        generated, normalize_embeddings=True
+    )
     similarities = generated_vectors @ golden_vectors.T
     return similarities