fix bug

app.py

@@ -1,6 +1,5 @@
 import spaces
 import torch
-import tensorflow as tf
 
 import gradio as gr
 import os
@@ -20,14 +19,13 @@ class App:
         self.app = gr.Blocks(css=CSS, theme=self.args.theme)
         #self.whisper_inf = WhisperInference() if self.args.disable_faster_whisper else FasterWhisperInference()
         self.whisper_inf = WhisperInference()
+        # NOTE: Faster whisper is not able to use in HuggingFace space. see more info : https://huggingface.co/spaces/jhj0517/Whisper-WebUI/discussions/1
         print("Use Open AI Whisper implementation")
-        # print(f"Device \"{self.whisper_inf.device}\" is detected")
-        # cuda_version = torch.version.cuda
-        # print(f"CUDA version: {cuda_version}")
-        # cudnn_version = torch.backends.cudnn.version()
-        # print(f"cuDNN version: {cudnn_version}")
-        # tf_version = tf.__version__
-        # print(f"tensorflow version: {tf_version}")
+        print(f"Device \"{self.whisper_inf.device}\" is detected")
+        cuda_version = torch.version.cuda
+        print(f"CUDA version: {cuda_version}")
+        cudnn_version = torch.backends.cudnn.version()
+        print(f"cuDNN version: {cudnn_version}")
         self.nllb_inf = NLLBInference()
         self.deepl_api = DeepLAPI()
 

modules/faster_whisper_inference.py

@@ -26,7 +26,7 @@ class FasterWhisperInference(BaseInterface):
         self.available_models = whisper.available_models()
         self.available_langs = sorted(list(whisper.tokenizer.LANGUAGES.values()))
         self.translatable_models = ["large", "large-v1", "large-v2", "large-v3"]
-        self.device = "cuda" #if torch.cuda.is_available() else "cpu"
+        self.device = "cuda" if torch.cuda.is_available() else "cpu"
         self.available_compute_types = ["int8_float16", "float32", "bfloat16", "float16", "int8", "int8_float32", "int8_bfloat16"]
         self.current_compute_type = "float16"
         self.default_beam_size = 1

modules/nllb_inference.py

@@ -23,7 +23,7 @@ class NLLBInference(BaseInterface):
         self.available_models = NLLB_MODELS
         self.available_source_langs = list(NLLB_AVAILABLE_LANGS.keys())
         self.available_target_langs = list(NLLB_AVAILABLE_LANGS.keys())
-        self.device = 0 #if torch.cuda.is_available() else 1
+        self.device = 0 if torch.cuda.is_available() else 1
         self.pipeline = None
 
     def translate_text(self, text):

modules/whisper_Inference.py

@@ -23,7 +23,7 @@ class WhisperInference(BaseInterface):
         self.model = None
         self.available_models = whisper.available_models()
         self.available_langs = sorted(list(whisper.tokenizer.LANGUAGES.values()))
-        self.device = "cuda" #if torch.cuda.is_available() else "cpu"
+        self.device = "cuda" if torch.cuda.is_available() else "cpu"
         self.available_compute_types = ["float16", "float32"]
         self.current_compute_type = "float16" if self.device == "cuda" else "float32"
         self.default_beam_size = 1

requirements.txt

@@ -4,5 +4,4 @@ git+https://github.com/jhj0517/jhj0517-whisper.git
 faster-whisper
 transformers
 gradio==4.14.0
-pytube
-tensorflow
+pytube