Merge pull request #190 from jhj0517/fix/translation-long-input

app.py

@@ -20,7 +20,7 @@ class App:
         print(f"Device \"{self.whisper_inf.device}\" is detected")
         self.nllb_inf = NLLBInference(
             model_dir=self.args.nllb_model_dir,
-            output_dir=self.args.output_dir
+            output_dir=os.path.join(self.args.output_dir, "translations")
         )
         self.deepl_api = DeepLAPI(
             output_dir=self.args.output_dir
@@ -375,6 +375,8 @@ class App:
                                                              choices=self.nllb_inf.available_source_langs)
                             dd_nllb_targetlang = gr.Dropdown(label="Target Language",
                                                              choices=self.nllb_inf.available_target_langs)
+                        with gr.Row():
+                            nb_max_length = gr.Number(label="Max Length Per Line", value=200, precision=0)
                         with gr.Row():
                             cb_timestamp = gr.Checkbox(value=True, label="Add a timestamp to the end of the filename",
                                                        interactive=True)
@@ -388,7 +390,7 @@ class App:
                             md_vram_table = gr.HTML(NLLB_VRAM_TABLE, elem_id="md_nllb_vram_table")
 
                     btn_run.click(fn=self.nllb_inf.translate_file,
-                                  inputs=[file_subs, dd_nllb_model, dd_nllb_sourcelang, dd_nllb_targetlang, cb_timestamp],
+                                  inputs=[file_subs, dd_nllb_model, dd_nllb_sourcelang, dd_nllb_targetlang, nb_max_length, cb_timestamp],
                                   outputs=[tb_indicator, files_subtitles])
 
                     btn_openfolder.click(fn=lambda: self.open_folder(os.path.join("outputs", "translations")),

modules/translation/nllb_inference.py

@@ -21,9 +21,13 @@ class NLLBInference(TranslationBase):
         self.pipeline = None
 
     def translate(self,
-                  text: str
+                  text: str,
+                  max_length: int
                   ):
-        result = self.pipeline(text)
+        result = self.pipeline(
+            text,
+            max_length=max_length
+        )
         return result[0]['translation_text']
 
     def update_model(self,

modules/translation/translation_base.py

@@ -24,7 +24,8 @@ class TranslationBase(ABC):
 
     @abstractmethod
     def translate(self,
-                  text: str
+                  text: str,
+                  max_length: int
                   ):
         pass
 
@@ -42,6 +43,7 @@ class TranslationBase(ABC):
                        model_size: str,
                        src_lang: str,
                        tgt_lang: str,
+                       max_length: int,
                        add_timestamp: bool,
                        progress=gr.Progress()) -> list:
         """
@@ -57,6 +59,8 @@ class TranslationBase(ABC):
             Source language of the file to translate from gr.Dropdown()
         tgt_lang: str
             Target language of the file to translate from gr.Dropdown()
+        max_length: int
+            Max length per line to translate
         add_timestamp: bool
             Boolean value from gr.Checkbox() that determines whether to add a timestamp at the end of the filename.
         progress: gr.Progress
@@ -84,7 +88,7 @@ class TranslationBase(ABC):
                     total_progress = len(parsed_dicts)
                     for index, dic in enumerate(parsed_dicts):
                         progress(index / total_progress, desc="Translating..")
-                        translated_text = self.translate(dic["sentence"])
+                        translated_text = self.translate(dic["sentence"], max_length=max_length)
                         dic["sentence"] = translated_text
                     subtitle = get_serialized_srt(parsed_dicts)
 
@@ -99,7 +103,7 @@ class TranslationBase(ABC):
                     total_progress = len(parsed_dicts)
                     for index, dic in enumerate(parsed_dicts):
                         progress(index / total_progress, desc="Translating..")
-                        translated_text = self.translate(dic["sentence"])
+                        translated_text = self.translate(dic["sentence"], max_length=max_length)
                         dic["sentence"] = translated_text
                     subtitle = get_serialized_vtt(parsed_dicts)
 
@@ -124,7 +128,6 @@ class TranslationBase(ABC):
             print(f"Error: {str(e)}")
         finally:
             self.release_cuda_memory()
-            self.remove_input_files([fileobj.name for fileobj in fileobjs])
 
     @staticmethod
     def get_device():