Update

file-structure-replica/.obsidian/plugins/Dual/skeleton/core.py

@@ -15,11 +15,11 @@ def __init__(self, root_dir):
         self.root_dir = root_dir
         self.cache_address = os.path.join(root_dir, '.obsidian/plugins/Dual/skeleton/cache.pickle')
         self.entry_regex = os.path.join(root_dir, '*md')
-        self.auxiliary_models_ready = False
-        self.main_model_ready = False
+        self.skeleton_ready = False
+        self.essence_ready = False
 
-        self.load_auxiliary_models()
-        self.load_main_model()
+        self.load_skeleton()
+        self.load_essence()
 
         if os.path.isfile(self.cache_address) is False:
             self.create_cache()
@@ -28,10 +28,10 @@ def __init__(self, root_dir):
             self.sync_cache()
 
     def fluid_search(self, query, considered_candidates=50, selected_candidates=5, second_pass=True):
-        self.load_main_model()
+        self.load_essence()
 
-        if self.main_model_ready == False:
-            return ['The aligned model is unavailable at the required location.']
+        if self.essence_ready == False:
+            return ['The essence is not present at the required location.']
 
         self.sync_cache()
         selected_candidates = min(selected_candidates, considered_candidates)
@@ -50,10 +50,10 @@ def fluid_search(self, query, considered_candidates=50, selected_candidates=5, s
             return [self.entry_filenames[hit['corpus_id']] for hit in hits[:selected_candidates]]
 
     def descriptive_search(self, claim, polarity=True, target='premise', considered_candidates=50, selected_candidates=5):
-        self.load_main_model()
+        self.load_essence()
 
-        if self.main_model_ready == False:
-            return ['The aligned model is unavailable at the required location.']
+        if self.essence_ready == False:
+            return ['The essence is not present at the required location.']
 
         selected_candidates = min(selected_candidates, considered_candidates)
         considered_candidates = min(considered_candidates, len(self.entry_filenames))     
@@ -79,10 +79,10 @@ def descriptive_search(self, claim, polarity=True, target='premise', considered_
         return results
 
     def open_dialogue(self, question, considered_candidates=3):
-        self.load_main_model()
+        self.load_essence()
 
-        if self.main_model_ready == False:
-            return ['The aligned model is unavailable at the required location.']
+        if self.essence_ready == False:
+            return ['The essence is not present at the required location.']
 
         candidate_entry_filenames = self.fluid_search(question, selected_candidates=considered_candidates)
         candidate_entry_contents = reversed([self.entries[e][0] for e in candidate_entry_filenames])
@@ -95,33 +95,33 @@ def open_dialogue(self, question, considered_candidates=3):
             max_length=len(input_ids[0]) + 100, 
             top_p=0.9, 
             top_k=40,
-            temperature=1
+            temperature=0.9
         )
 
         output_sample = self.gen_tokenizer.decode(generator_output[0], skip_special_tokens=True)[len(generator_prompt):]
         output_sample = re.sub(r'^[\W_]+|[\W_]+$', '', output_sample)
         output_sample = re.sub(r'[^a-zA-Z0-9\s]{3,}', '', output_sample)
-        output_sample = output_sample.split('Q:')[0].strip()
+        output_sample = output_sample.split('Q:')[0].split('\n\n')[0].strip()
         output_sample += '...'
 
         return [output_sample]
 
-    def load_auxiliary_models(self):
-        print('Loading auxiliary models...')
+    def load_skeleton(self):
+        print('Loading skeleton...')
         self.text_encoder = SentenceTransformer('msmarco-distilbert-base-v2')
         self.pair_encoder = CrossEncoder('cross-encoder/ms-marco-TinyBERT-L-4')
         self.nli = CrossEncoder('cross-encoder/nli-distilroberta-base')
-        self.auxiliary_models_ready = True
+        self.skeleton_ready = True
 
-    def load_main_model(self):
+    def load_essence(self):
         tentative_folder_path = os.path.join(self.root_dir, '.obsidian/plugins/Dual/essence')
         tentative_file_path = os.path.join(tentative_folder_path, 'pytorch_model.bin')
 
-        if self.main_model_ready == False and os.path.isfile(tentative_file_path):
-            print('Loading main model...')
+        if self.essence_ready == False and os.path.isfile(tentative_file_path):
+            print('Loading essence...')
             self.gen_tokenizer = GPT2Tokenizer.from_pretrained('gpt2-medium')
             self.gen_model = GPT2LMHeadModel.from_pretrained(pretrained_model_name_or_path=tentative_folder_path, pad_token_id=self.gen_tokenizer.eos_token_id)
-            self.main_model_ready = True
+            self.essence_ready = True
 
     def copy_snapshot(self):
         return {