Fixing errors and upgrading gradio

* suppress gradio and triton warnings * upgrade gradio and fix visual issues
rsxdalv · Jan 15, 2024 · 8b0f102 · 8b0f102
1 parent 06ab788
commit 8b0f102
Show file tree

Hide file tree

Showing 8 changed files with 101 additions and 74 deletions.
diff --git a/README.md b/README.md
@@ -40,7 +40,15 @@ https://rsxdalv.github.io/bark-speaker-directory/
 ## Info about managing models, caches and system space for AI projects
 https://github.com/rsxdalv/tts-generation-webui/discussions/186#discussioncomment-7291274
 
-## Changelog
+## Changelo
+Jan 15:
+* Upgraded Gradio to 3.48.0
+  * Several visual bugs have appeared, if they are critical, please report them or downgrade gradio.
+  * Gradio: Suppress useless warnings
+* Supress Triton warnings
+* Gradio-Bark: Fix "Use last generation as history" behavior, empty selection no longer errors
+* Improve extensions loader display
+
 Jan 14:
 * React UI: Fix missing directory errors
 
@@ -50,7 +58,7 @@ Jan 13:
 Jan 12:
 * React UI: Fix names for audio actions
 * Gradio: Fix multiple API warnings
-* Integration - React UI now is launched alongside Gradio, with a link to open it.
+* Integration - React UI now is launched alongside Gradio, with a link to open it
 
 Jan 11:
 * React UI: Make the build work without any errors

diff --git a/requirements.txt b/requirements.txt
@@ -1,7 +1,7 @@
 safetensors==0.3.1 # Until they fix it
 torch==2.0.0 # BSD 3-Clause - ensures torch is not reinstalled
 ffmpeg-python # Apache 2.0
-gradio==3.35.2
+gradio==3.48.0
 python-dotenv==1.0.0
 soundfile==0.12.1; sys_platform == 'win32' # torchaudio
 # sox # torchaudio for linux

diff --git a/server.py b/server.py
@@ -2,6 +2,24 @@
 import src.utils.setup_or_recover as setup_or_recover
 import src.utils.dotenv_init as dotenv_init
 import gradio as gr
+import warnings
+
+warnings.filterwarnings(
+    "ignore",
+    message="Using the update method is deprecated. Simply return a new object instead",
+)
+warnings.filterwarnings(
+    "ignore",
+    message="Trying to convert audio automatically from float32 to 16-bit int format.",
+)
+
+import logging
+
+# suppress warning from logging "A matching Triton is not available, some optimizations will not be enabled"
+# suppress warning from logging "Triton is not available, some optimizations will not be enabled."
+logging.getLogger("xformers").addFilter(
+    lambda record: "Triton is not available" not in record.getMessage()
+)
 
 from src.config.load_config import default_config
 from src.config.config import config
@@ -10,6 +28,7 @@
 from src.Joutai import Joutai
 from src.history_tab.collections_directories_atom import collections_directories_atom
 
+
 setup_or_recover.dummy()
 dotenv_init.init()
 
@@ -35,7 +54,9 @@ def reload_config_and_restart_ui():
     title="TTS Generation WebUI",
     analytics_enabled=False,  # it broke too many times
 ) as demo:
-    gr.Markdown("# TTS Generation WebUI (Bark, MusicGen + AudioGen, Tortoise, RVC) [NEW React UI (Beta)](http://localhost:3000)")
+    gr.Markdown(
+        "# TTS Generation WebUI (Bark, MusicGen + AudioGen, Tortoise, RVC) [NEW React UI (Beta)](http://localhost:3000)"
+    )
     with Joutai.singleton.tabs:
         from src.tortoise.generation_tab_tortoise import generation_tab_tortoise
         from src.settings_tab_gradio import settings_tab_gradio
@@ -122,7 +143,7 @@ def reload_config_and_restart_ui():
 
 
 def print_pretty_options(options):
-    print("Gradio interface options:")
+    print(" Gradio interface options:")
     max_key_length = max(len(key) for key in options.keys())
     for key, value in options.items():
         if key == "auth" and value is not None:
@@ -138,9 +159,8 @@ def print_pretty_options(options):
 
 
 print("Starting Gradio server...")
-if not gradio_interface_options["enable_queue"]:
-    print("Warning: Gradio server queue is disabled. Automatically enabling")
-    gradio_interface_options["enable_queue"] = True
+if "enable_queue" in gradio_interface_options:
+    del gradio_interface_options["enable_queue"]
 if gradio_interface_options["auth"] is not None:
     # split username:password into (username, password)
     gradio_interface_options["auth"] = tuple(
@@ -156,8 +176,8 @@ def start_server():
     ).launch(**gradio_interface_options)
 
 
-
 if __name__ == "__main__":
     import subprocess
+
     subprocess.Popen("npm start --prefix react-ui", shell=True)
     start_server()
diff --git a/src/bark/bark_css.py b/src/bark/bark_css.py
@@ -12,4 +12,8 @@
     height: 150px;
     min-height: 150px;
 }
+.tts-image > button {
+    width: 100%;
+    height: 100%;   
+}
 """
diff --git a/src/bark/generation_tab_bark.py b/src/bark/generation_tab_bark.py
@@ -298,22 +298,13 @@ def gen(
         if prompt is None or prompt == "":
             raise ValueError("Prompt is empty")
 
-        print(
-            "gen",
-            "old_generation_filename",
-            old_generation_filename,
-            "semantic",
-            history_prompt_semantic,
-        )
-        if history_prompt_semantic == "":
-            history_prompt_semantic = None
-        if history_prompt_semantic is not None:
-            history_prompt_semantic = load_npz(history_prompt_semantic)
-
         if history_setting == HistorySettings.NPZ_FILE:
-            if old_generation_filename is None:
-                raise ValueError("old_generation_filename is None")
-            history_prompt = load_npz(old_generation_filename)
+            history_prompt_semantic = (
+                load_npz(history_prompt_semantic) if history_prompt_semantic else None
+            )
+            history_prompt = (
+                load_npz(old_generation_filename) if old_generation_filename else None
+            )
 
         _original_history_prompt = history_prompt
 
@@ -481,8 +472,8 @@ def generation_tab_bark():
     with gr.Tab(label="Generation (Bark)", id="generation_bark"):
         with gr.Row():
             history_setting = gr.Radio(
-                HistorySettings.choices,
-                value="Empty history",
+                HistorySettings.choices, # type: ignore
+                value=HistorySettings.EMPTY,
                 type="value",
                 label="History Prompt (voice) setting:",
             )
@@ -519,19 +510,17 @@ def unload_models():
             outputs=[column],  # type: ignore
         )
 
-        with gr.Row():
-            (
-                history_prompt_semantic_dropdown,
-                copy_history_prompt_semantic_button,
-                reload_history_prompt_semantic_dropdown,
-            ) = old_generation_dropdown_ui("Semantic Voice (Optional)")
+        (
+            history_prompt_semantic_dropdown,
+            copy_history_prompt_semantic_button,
+            reload_history_prompt_semantic_dropdown,
+        ) = old_generation_dropdown_ui("Semantic Voice (Optional)")
 
-        with gr.Row():
-            (
-                old_generation_dropdown,
-                copy_old_generation_button,
-                reload_old_generation_dropdown,
-            ) = old_generation_dropdown_ui("Audio Voice")
+        (
+            old_generation_dropdown,
+            copy_old_generation_button,
+            reload_old_generation_dropdown,
+        ) = old_generation_dropdown_ui("Audio Voice")
 
         history_setting.change(
             fn=lambda choice: [
@@ -564,14 +553,14 @@ def unload_models():
         with gr.Row():
             with gr.Column():
                 long_prompt_radio = gr.Radio(
-                    PromptSplitSettings.choices,
+                    PromptSplitSettings.choices, # type: ignore
                     type="value",
                     label="Prompt type",
                     value=PromptSplitSettings.NONE,
                     show_label=False,
                 )
                 long_prompt_history_radio = gr.Radio(
-                    LongPromptHistorySettings.choices,
+                    LongPromptHistorySettings.choices, # type: ignore
                     type="value",
                     label="For each subsequent generation:",
                     value=LongPromptHistorySettings.CONTINUE,
@@ -741,40 +730,42 @@ def register_use_as_history_button(button, source):
 
 
 def old_generation_dropdown_ui(label):
-    old_generation_dropdown = gr.Dropdown(
-        label=label,
-        choices=get_npz_files(),
-        type="value",
-        value=None,
-        allow_custom_value=True,
-        visible=False,
-        container=False,
-    )
-    copy_old_generation_button = gr.Button(
-        "save",
-        visible=False,
-        elem_classes="btn-sm material-symbols-outlined",
-        size="sm",
-    )
-    copy_old_generation_button.click(
-        fn=lambda x: [
-            shutil.copy(x, os.path.join("voices", os.path.basename(x))),
-        ],
-        inputs=[old_generation_dropdown],
-    )
+    with gr.Row():
+        old_generation_dropdown = gr.Dropdown(
+            label=label,
+            choices=get_npz_files(), # type: ignore
+            type="value",
+            value=None,
+            allow_custom_value=True,
+            visible=False,
+            # container=False,
+            show_label=True,
+        )
+        copy_old_generation_button = gr.Button(
+            "save",
+            visible=False,
+            elem_classes="btn-sm material-symbols-outlined",
+            size="sm",
+        )
+        copy_old_generation_button.click(
+            fn=lambda x: [
+                shutil.copy(x, os.path.join("voices", os.path.basename(x))),
+            ],
+            inputs=[old_generation_dropdown],
+        )
 
-    reload_old_generation_dropdown = gr.Button(
-        "refresh",
-        visible=False,
-        elem_classes=ICON_ELEM_CLASS,
-        size="sm",
-    )
+        reload_old_generation_dropdown = gr.Button(
+            "refresh",
+            visible=False,
+            elem_classes=ICON_ELEM_CLASS,
+            size="sm",
+        )
 
-    reload_old_generation_dropdown.click(
-        fn=lambda: gr.Dropdown.update(choices=get_npz_files()),
-        outputs=old_generation_dropdown,
-        api_name=f"reload_old_generation_dropdown{ '' if label == 'Audio Voice' else '_semantic'}",
-    )
+        reload_old_generation_dropdown.click(
+            fn=lambda: gr.Dropdown.update(choices=get_npz_files()), # type: ignore
+            outputs=old_generation_dropdown,
+            api_name=f"reload_old_generation_dropdown{ '' if label == 'Audio Voice' else '_semantic'}",
+        )
 
     return (
         old_generation_dropdown,
@@ -798,7 +789,7 @@ def setup_bark_voice_prompt_ui():
 
         speaker_ids = ["0", "1", "2", "3", "4", "5", "6", "7", "8", "9"]
         speakerIdRadio = gr.Radio(
-            speaker_ids, type="value", label="Speaker ID", value="0"
+            speaker_ids, type="value", label="Speaker ID", value="0" # type: ignore
         )
 
         voice_inputs = [useV2, languageRadio, speakerIdRadio]

diff --git a/src/extensions_loader/ext_callback_save_generation.py b/src/extensions_loader/ext_callback_save_generation.py
@@ -18,6 +18,7 @@
 for file_name in extension_files:
     if file_name.endswith(".py"):
         module_name = file_name[:-3]
+        print("  ", end="")
         try:
             spec = importlib.util.spec_from_file_location(
                 module_name, os.path.join(extensions_folder, file_name)
@@ -48,8 +49,9 @@
                 f"Module {module_name} does not contain the function 'callback_save_generation'"
             )
 
-# print(f"Loaded {len(callbacks_save_generation)} extensions.")
+print("  ", end="")
 print(f"Loaded {len(callbacks_save_generation)} callback_save_generation extensions.")
+print("  ", end="")
 print(
     f"Loaded {len(callbacks_save_generation_musicgen)} callback_save_generation_musicgen extensions."
 )

diff --git a/src/history_tab/main.py b/src/history_tab/main.py
@@ -107,6 +107,7 @@ def history_content(
                 max_cols=len(datatypes),
                 datatype=datatypes,
                 headers=headers,
+                height=800,
             )
 
         with gr.Column():

diff --git a/src/history_tab/voices_tab.py b/src/history_tab/voices_tab.py
@@ -54,6 +54,7 @@ def voices_tab(register_use_as_history_button, directory="voices"):
                 col_count=len(datatypes),
                 max_cols=len(datatypes),
                 headers=headers,
+                height=800,
                 #  elem_classes="file-list"
             )
         with gr.Column():