update packages version and fix bug

- add torch and torchaudio version in requirements, remove torchvision - faster whisper now use the v0.10.0 - updated stable-ts, whisper, and webrtcvad version - add pause to downloading faster whisper model - now dialog interaction is disabled disable when downloading model from dialog - fix for #54, the problem was that i forgot that fg was set in the parameter - now enforce ffmpeg instead of asking continue or not when failed to install
Dadangdut33 · Nov 25, 2023 · 165d25a · 165d25a
1 parent 82e7041
commit 165d25a
Show file tree

Hide file tree

Showing 10 changed files with 226 additions and 158 deletions.
diff --git a/requirements.txt b/requirements.txt
@@ -1,6 +1,5 @@
-torch
-torchvision
-torchaudio
+torch==2.1.1
+torchaudio==2.1.1
 deep-translator==1.11.1
 notify-py==0.3.42
 loguru
@@ -15,13 +14,10 @@ requests==2.31.0
 scipy==1.11.3
 sounddevice==0.4.6
 soundfile==0.12.1
-webrtcvad==2.0.10
+webrtcvad @ git+https://github.com/wiseman/py-webrtcvad.git@e283ca41df3a84b0e87fb1f5cb9b21580a286b09
 darkdetect==0.8.0
-arabic-reshaper==3.0.0
-python-bidi==0.4.2
 matplotlib==3.8.0
 onnxruntime==1.16.1
 demucs==4.0.1
-stable-ts @ git+https://github.com/jianfch/stable-ts.git@b336735ff784bb59690eec8f9f706b0151dda74c
-openai-whisper==20231106
-faster-whisper @ git+https://github.com/guillaumekln/faster-whisper.git@1fab6eee59b268f89ff3912a4b19c82751e6bb26
+stable-ts @ git+https://github.com/jianfch/stable-ts.git@71b9f1fcbd1268f8bfe95bba6a394a2bc2e7339b
+faster-whisper @ git+https://github.com/SYSTRAN/faster-whisper.git@e1a218fab1ab02d637b79565995bf1a9c4c83a09
diff --git a/speech_translate/_logging.py b/speech_translate/_logging.py
@@ -35,10 +35,10 @@ class StreamStderrToLogger(object):
     """
     def __init__(self, level):
         self.level = level
-        # tqdm use stderr to print, so we should consider it as info
+        # tqdm use stderr to print, so we can consider it as info
         self.considered_info = [
-            "Downloading", "Fetching", "run_threaded", "Estimating duration from bitrate, this may be inaccurate",
-            "Transcribe", "Translate", "Refine", "Align", "Running", "done", "Using cache found in"
+            "Downloading", "Fetching", "run_threaded", "Estimating duration from bitrate", "Transcribe", "Translate",
+            "Translating", "Refine", "Align", "Running", "done", "Using cache found in", "%|#", "0%|"
         ]
 
     def write(self, buf):

diff --git a/speech_translate/ui/custom/dialog.py b/speech_translate/ui/custom/dialog.py
@@ -81,6 +81,7 @@ def __init__(
         self.mode = mode
         self.data_list = []
         self.headers = headers
+        self.interact_disabled = False
 
         self.root = Toplevel(self.master)
         self.root.geometry("+400+250")
@@ -201,11 +202,30 @@ def submit(self):
         self.root.destroy()
 
     def on_close(self):
+        if self.interact_disabled:
+            return
+
         if not messagebox.askyesno("Cancel", "Are you sure you want to cancel?", parent=self.root):
             return
 
         self.root.destroy()
 
+    def disable_interactions(self):
+        self.interact_disabled = True
+        self.cb_model.configure(state="disabled")
+        self.btn_add.configure(state="disabled")
+        self.btn_delete.configure(state="disabled")
+        self.btn_start.configure(state="disabled")
+        self.btn_cancel.configure(state="disabled")
+
+    def enable_interactions(self):
+        self.interact_disabled = False
+        self.cb_model.configure(state="readonly")
+        self.btn_add.configure(state="normal")
+        self.btn_delete.configure(state="normal")
+        self.btn_start.configure(state="normal")
+        self.btn_cancel.configure(state="normal")
+
 
 class FileImportDialog(FileOperationDialog):
     def __init__(self, master, title: str, submit_func, theme: str, **kwargs):
@@ -375,6 +395,22 @@ def submit(self):
         if status:  # if status is True, meaning process thread is successfully started, then close the window
             self.root.destroy()
 
+    def disable_interactions(self):
+        super().disable_interactions()
+        self.cb_engine.configure(state="disabled")
+        self.cb_source_lang.configure(state="disabled")
+        self.cb_target_lang.configure(state="disabled")
+        self.cbtn_transcribe.configure(state="disabled")
+        self.cbtn_translate.configure(state="disabled")
+
+    def enable_interactions(self):
+        super().enable_interactions()
+        self.cb_engine.configure(state="readonly")
+        self.cb_source_lang.configure(state="readonly")
+        self.cb_target_lang.configure(state="readonly")
+        self.cbtn_transcribe.configure(state="normal")
+        self.cbtn_translate.configure(state="normal")
+
 
 class TranslateResultDialog(FileOperationDialog):
     def __init__(self, master, title: str, submit_func, theme: str, **kwargs):
@@ -455,6 +491,16 @@ def submit(self):
         self.submit_func(self.var_engine.get(), self.var_target_lang.get().lower(), [x[0] for x in self.data_list])
         self.root.destroy()
 
+    def disable_interactions(self):
+        super().disable_interactions()
+        self.cb_engine.configure(state="disabled")
+        self.cb_target_lang.configure(state="disabled")
+
+    def enable_interactions(self):
+        super().enable_interactions()
+        self.cb_engine.configure(state="readonly")
+        self.cb_target_lang.configure(state="readonly")
+
 
 class RefinementDialog(FileOperationDialog):
     def __init__(self, master, title: str, submit_func, theme: str, **kwargs):

diff --git a/speech_translate/ui/custom/download.py b/speech_translate/ui/custom/download.py
@@ -86,7 +86,7 @@ def whisper_download_with_progress_gui(
     # flag
     paused = False
 
-    def pause_download():
+    def toggle_pause():
         nonlocal paused
         paused = not paused
         if paused:
@@ -115,7 +115,7 @@ def pause_download():
     lbl_status_text = ttk.Label(status_frame, text=f"Downloading {model_name} model")
     lbl_status_text.pack(side="left", padx=5, pady=5)
 
-    btn_pause = ttk.Button(btn_frame, text="Pause", command=pause_download)
+    btn_pause = ttk.Button(btn_frame, text="Pause", command=toggle_pause)
     btn_pause.pack(side="left", fill="x", padx=5, pady=5, expand=True)
 
     btn_cancel = ttk.Button(btn_frame, text="Cancel", command=cancel_func, style="Accent.TButton")
@@ -305,7 +305,7 @@ def snapshot_download(
 
     filtered_repo_files = list(
         huggingface_hub.utils.filter_repo_objects(
-            items=[f.rfilename for f in repo_info.siblings],
+            items=[f.rfilename for f in repo_info.siblings],  # type: ignore
             allow_patterns=allow_patterns,
             ignore_patterns=ignore_patterns,
         )
@@ -404,10 +404,6 @@ def faster_whisper_download_with_progress_gui(
     # clear recent_stderr
     recent_stderr.clear()
 
-    # add label that says downloading please wait
-    failed = False
-    msg = ""
-
     f1 = ttk.Frame(root)
     f1.pack(side="top", fill="x", expand=True)
 
@@ -426,6 +422,9 @@ def faster_whisper_download_with_progress_gui(
     btn_cancel = ttk.Button(f1, text="Cancel", command=cancel_func, style="Accent.TButton")
     btn_cancel.pack(side="right", padx=(5, 10), pady=(5, 0))
 
+    btn_pause = ttk.Button(f1, text="Pause", command=lambda: toggle_pause())
+    btn_pause.pack(side="right", padx=5, pady=(5, 0))
+
     # add progress bar that just goes back and forth
     progress = ttk.Progressbar(f2, orient="horizontal", length=200, mode="indeterminate")
     progress.pack(expand=True, fill="x", padx=10, pady=(2, 2))
@@ -450,8 +449,14 @@ def update_log():
         text_log.insert(1.0, content)
         text_log.see("end")  # scroll to the bottom
 
+    failed = False
+    msg = ""
+    finished = False
+    paused = False
+    killed = False
+
     def run_threaded():
-        nonlocal failed, msg
+        nonlocal failed, msg, finished, paused
 
         root.title("Verifying Model")
         lbl_status_text.configure(text=f"Verifying {model_name} model please wait...")
@@ -479,24 +484,52 @@ def run_threaded():
             failed = True
             msg = str(e)
 
+        finally:
+            if not paused:
+                finished = True
+
     threaded = Thread(target=run_threaded, daemon=True)
     threaded.start()
     start_time = time()
 
-    while threaded.is_alive():
+    def toggle_pause():
+        nonlocal paused, killed, threaded
+        paused = not paused
+        if paused:
+            logger.info("Download paused")
+            btn_pause["text"] = "Resume"
+            progress.stop()
+        else:
+            logger.info("Download resumed")
+            btn_pause["text"] = "Pause"
+            progress.start(15)
+            killed = False
+            threaded = Thread(target=run_threaded, daemon=True)
+            threaded.start()
+
+    while not finished:
+        if paused and not killed:
+            kill_thread(threaded)
+            killed = True
+            recent_stderr.append("Download paused")
+            update_log()
+
         if bc.cancel_dl:
             kill_thread(threaded)
+            finished = True  # mark as finished
             root.destroy()
             mbox("Download Cancelled", f"Downloading of {model_name} faster whisper model has been cancelled", 0, master)
             break
 
         # check if 2 second have passed. Means probably downloading from the hub
         if time() - start_time > 2:
-            root.title("Downloading Faster Whisper Model")
+            root.title(f"{'Downloading' if not paused else 'Paused downloading of'} Faster Whisper Model")
             lbl_status_text.configure(
-                text=f"Downloading {model_name} model, {get_file_amount(storage_folder + '/' + 'blobs')} files downloaded..."
+                text=
+                f"{'Downloading' if not paused else 'Paused downloading'} {model_name} model, {get_file_amount(storage_folder + '/' + 'blobs')} files downloaded..."
             )
-            update_log()
+            if not paused:
+                update_log()
         sleep(1)
 
     # if cancel button is pressed, return

diff --git a/speech_translate/ui/frame/setting/general.py b/speech_translate/ui/frame/setting/general.py
@@ -187,22 +187,6 @@ def __init__(self, root: Toplevel, master_frame: Union[ttk.Frame, Frame]):
         )
         self.btn_log_config.pack(side="left", padx=5, pady=5)
 
-        # self.lbl_ignore_stdout = ttk.Label(self.f_logging_2, text="Ignore stdout", width=16)
-        # self.lbl_ignore_stdout.pack(side="left", padx=5)
-        # tk_tooltip(self.lbl_ignore_stdout, "Collection to ignore stdout / print from the console.")
-        # self.entry_ignore_stdout = ttk.Entry(self.f_logging_2)
-        # self.entry_ignore_stdout.pack(side="left", padx=5, fill="x", expand=True)
-        # self.entry_ignore_stdout.insert(0, ', '.join(sj.cache["ignore_stdout"]))
-        # self.entry_ignore_stdout.bind("<FocusOut>", lambda e: self.save_ignore_stdout())
-        # self.entry_ignore_stdout.bind("<Return>", lambda e: self.save_ignore_stdout())
-        # tk_tooltip(
-        #     self.entry_ignore_stdout,
-        #     "Collection to ignore stdout / print from the console with its input separated by comma.\n\n"
-        #     "This is useful if you want to ignore some of the stdout / print from the console.\n\n"
-        #     "Example: `Predicting silences(s) with VAD..., Predicted silences(s) with VAD`",
-        #     wrapLength=500,
-        # )
-
         self.menu_config_log = Menu(self.master, tearoff=0)
         self.menu_config_log.add_command(
             label="Open", image=self.open_emoji, compound="left", command=lambda: start_file(dir_log)
@@ -274,7 +258,7 @@ def __init__(self, root: Toplevel, master_frame: Union[ttk.Frame, Frame]):
         tk_tooltip(
             self.cbtn_debug_realtime_record,
             "Show some debugging process of the realtime record.\n\n"
-            "Enabling will probably slow down the app.",
+            "Enabling could slow down the app.",
         )
 
         self.cbtn_debug_recorded_audio = CustomCheckButton(
@@ -550,7 +534,7 @@ def failed_func():
         kwargs = {
             "after_func": after_func,
             "use_faster_whisper": use_faster_whisper,
-            "cancel_func": lambda: self.cancel_model_download(model, btn),
+            "cancel_func": lambda: self.cancel_model_download(model, btn, use_faster_whisper),
             "failed_func": failed_func,
         }
 
@@ -580,7 +564,7 @@ def failed_func():
             )
             mbox("Download error", f"Err details: {e}", 0, self.root)
 
-    def cancel_model_download(self, model: str, btn: ttk.Button) -> None:
+    def cancel_model_download(self, model: str, btn: ttk.Button, use_faster_whisper) -> None:
         """
         Cancel whisper model download.
 
@@ -589,7 +573,7 @@ def cancel_model_download(self, model: str, btn: ttk.Button) -> None:
         if not mbox("Cancel confirmation", "Are you sure you want to cancel downloading?", 3, self.root):
             return
 
-        btn.configure(text="Download", command=lambda: self.model_download(model, btn, False), state="normal")
+        btn.configure(text="Download", command=lambda: self.model_download(model, btn, use_faster_whisper), state="normal")
         bc.cancel_dl = True  # Raise flag to stop
 
     def model_btn_checker(self, model: str, btn: ttk.Button, faster_whisper: bool = False, on_start=False) -> None:
@@ -901,10 +885,3 @@ def path_default(self, key: str, element: ttk.Entry, default_path: str, save=Tru
         element.configure(state="readonly")
         if save:
             sj.save_key(key, "auto")
-
-    # def save_ignore_stdout(self):
-    #     _input = self.entry_ignore_stdout.get().split(",")
-    #     _input = [i.strip() for i in _input if i.strip() != ""]  # remove any empty string or space
-
-    #     sj.save_key("ignore_stdout", _input)
-    #     update_stdout_ignore_list(_input)
diff --git a/speech_translate/ui/template/detached.py b/speech_translate/ui/template/detached.py
@@ -59,7 +59,7 @@ def __init__(self, master: Tk, title: str, winType: Literal["tc", "tl"]):
             wrapLength=250,
         )
 
-        self.menuDropdown = Menu(self.root, tearoff=0, fg="white")
+        self.menuDropdown = Menu(self.root, tearoff=0)
         self.menuDropdown.add_command(label=self.title, command=self.open_menu, image=self.title_emoji, compound="left")
         self.menuDropdown.add_command(label="Help", command=self.show_help, image=self.help_emoji, compound="left")
         self.menuDropdown.add_command(