update logic of individual translation

Current logic would skip the unmatched inner method of set_translation in SRT class


Former-commit-id: 84535d2adcec9c824849f9606b628f663af8c2a4

pipeline.py

@@ -24,7 +24,7 @@ def parse_args():
     parser.add_argument("--download", help="download path", default='./downloads', type=str, required=False)
     parser.add_argument("--output_dir", help="translate result path", default='./results', type=str, required=False)
     parser.add_argument("--video_name", help="video name, if use video link as input, the name will auto-filled by youtube video name", default='placeholder', type=str, required=False)
-    parser.add_argument("--model_name", help="model name only support gpt-4 and gpt-3.5-turbo", type=str, required=False, default="gpt-3.5-turbo") # default change to gpt-4
+    parser.add_argument("--model_name", help="model name only support gpt-4 and gpt-3.5-turbo", type=str, required=False, default="gpt-4") # default change to gpt-4
     parser.add_argument("--log_dir", help="log path", default='./logs', type=str, required=False)
     parser.add_argument("-only_srt", help="set script output to only .srt file", action='store_true')
     parser.add_argument("-v", help="auto encode script with video", action='store_true')
@@ -192,10 +192,10 @@ def check_translation(sentence, translation):
     """
     sentence_count = sentence.count('\n\n') + 1
     translation_count = translation.count('\n\n') + 1
-    print("sentence length: ", len(sentence), sentence_count)
-    print("translation length: ",  len(translation), translation_count)
 
     if sentence_count != translation_count:
+        # print("sentence length: ", len(sentence), sentence_count)
+        # print("translation length: ",  len(translation), translation_count)
         return False
     else:
         return True
@@ -257,12 +257,16 @@ def translate(srt, script_arr, range_arr, model_name, video_name, video_link):
                 # if failure still happen, split into smaller tokens
                 if attempt_left == 0:
                     single_sentences = sentence.split("\n\n")
-                    print("merge sentence issue found: ", len(single_sentences), len(translate), single_sentences, translate)
+                    print("merge sentence issue found for range", range)
                     translate = ""
-                    for single_sentence in single_sentences:
-                        translate += get_response(model_name, single_sentence) + "\n\n"
-                    print("after correction: ", "chinese length: ", len(translate), translate)
-                # print(type(translate))
+                    for i, single_sentence in enumerate(single_sentences):
+                        if i == len(single_sentences) - 1:
+                            translate += get_response(model_name, single_sentence)
+                        else:
+                            translate += get_response(model_name, single_sentence) + "\n\n"
+                            # print(single_sentence, translate.split("\n\n")[-2])
+                    print("solved by individually translation!")    
+
             except Exception as e:
                 logging.debug("An error has occurred during translation:",e)
                 print("An error has occurred during translation:",e)