From 132ac71696010ed46a51d1e5b8fc72dd2e3c0254 Mon Sep 17 00:00:00 2001
From: Guilherme <86894155+ShiromiyaG@users.noreply.github.com>
Date: Tue, 27 Aug 2024 08:43:14 -0300
Subject: [PATCH 1/6] Add more information to the model

---
 assets/i18n/languages/en_US.json       |  5 +++-
 core.py                                | 10 +++++++
 rvc/train/process/extract_model.py     | 19 +++++++++++++-
 rvc/train/process/model_information.py |  8 ++++++
 rvc/train/train.py                     | 36 ++++++++++++++++++++++++++
 tabs/extra/extra.py                    |  5 ++--
 tabs/extra/processing/processing.py    |  2 +-
 tabs/train/train.py                    |  9 +++++++
 8 files changed, 88 insertions(+), 6 deletions(-)

diff --git a/assets/i18n/languages/en_US.json b/assets/i18n/languages/en_US.json
index 8b3bb321..2545b488 100644
--- a/assets/i18n/languages/en_US.json
+++ b/assets/i18n/languages/en_US.json
@@ -301,5 +301,8 @@
   "Folder Name": "Folder Name",
   "Upload .bin": "Upload .bin",
   "Upload .json": "Upload .json",
-  "Move files to custom embedder folder": "Move files to custom embedder folder"
+  "Move files to custom embedder folder": "Move files to custom embedder folder",
+  "model information": "model information",
+  "Model Creator": "Model Creator",
+  "Name of the model creator.": "Name of the model creator."
 }
diff --git a/core.py b/core.py
index bca001c5..af33f44b 100644
--- a/core.py
+++ b/core.py
@@ -443,6 +443,7 @@ def run_train_script(
     custom_pretrained: bool = False,
     g_pretrained_path: str = None,
     d_pretrained_path: str = None,
+    model_creator: str = None,
 ):
 
     if pretrained == True:
@@ -484,6 +485,7 @@ def run_train_script(
                 overtraining_detector,
                 overtraining_threshold,
                 sync_graph,
+                model_creator,
             ],
         ),
     ]
@@ -526,6 +528,7 @@ def run_model_extract_script(
 # Model information
 def run_model_information_script(pth_path: str):
     print(model_information(pth_path))
+    return model_information(pth_path)
 
 
 # Model blender
@@ -1351,6 +1354,12 @@ def parse_arguments():
         help="Enable graph synchronization for distributed training.",
         default=False,
     )
+    train_parser.add_argument(
+        "--model_creator",
+        type=str,
+        help="Model creator name.",
+        default=None,
+    )
     train_parser.add_argument(
         "--cache_data_in_gpu",
         type=lambda x: bool(strtobool(x)),
@@ -1655,6 +1664,7 @@ def main():
                 pretrained=args.pretrained,
                 custom_pretrained=args.custom_pretrained,
                 sync_graph=args.sync_graph,
+                model_creator=args.model_creator,
                 index_algorithm=args.index_algorithm,
                 cache_data_in_gpu=args.cache_data_in_gpu,
                 g_pretrained_path=args.g_pretrained_path,
diff --git a/rvc/train/process/extract_model.py b/rvc/train/process/extract_model.py
index b0940e34..23536d8a 100644
--- a/rvc/train/process/extract_model.py
+++ b/rvc/train/process/extract_model.py
@@ -18,7 +18,20 @@ def replace_keys_in_dict(d, old_key_part, new_key_part):
     return updated_dict
 
 
-def extract_model(ckpt, sr, pitch_guidance, name, model_dir, epoch, step, version, hps):
+def extract_model(
+    ckpt,
+    sr,
+    pitch_guidance,
+    name,
+    model_dir,
+    epoch,
+    step,
+    version,
+    hps,
+    model_creator,
+    overtrain_info,
+    dataset_lenght,
+):
     try:
         print(f"Saved model '{model_dir}' (epoch {epoch} and step {step})")
 
@@ -70,6 +83,10 @@ def extract_model(ckpt, sr, pitch_guidance, name, model_dir, epoch, step, versio
         hash_input = f"{str(ckpt)} {epoch} {step} {datetime.datetime.now().isoformat()}"
         model_hash = hashlib.sha256(hash_input.encode()).hexdigest()
         opt["model_hash"] = model_hash
+        opt["model_name"] = name
+        opt["model_creator"] = model_creator
+        opt["overtrain_info"] = overtrain_info
+        opt["dataset_lenght"] = dataset_lenght
 
         torch.save(opt, os.path.join(model_dir_path, pth_file))
 
diff --git a/rvc/train/process/model_information.py b/rvc/train/process/model_information.py
index be74269e..6c150b22 100644
--- a/rvc/train/process/model_information.py
+++ b/rvc/train/process/model_information.py
@@ -17,24 +17,32 @@ def model_information(path):
 
     print(f"Loaded model from {path}")
 
+    model_name = model_data.get("model_name", "None")
+    model_creator = model_data.get("model_creator", "None")
     epochs = model_data.get("epoch", "None")
     steps = model_data.get("step", "None")
     sr = model_data.get("sr", "None")
     f0 = model_data.get("f0", "None")
+    dataset_lenght = model_data.get("dataset_lenght", "None")
     version = model_data.get("version", "None")
     creation_date = model_data.get("creation_date", "None")
     model_hash = model_data.get("model_hash", None)
+    overtrain_info = model_data.get("overtrain_info", "None")
 
     pitch_guidance = "True" if f0 == 1 else "False"
 
     creation_date_str = prettify_date(creation_date) if creation_date else "None"
 
     return (
+        f"Model Name: {model_name}\n"
+        f"Model Creator: {model_creator}\n"
         f"Epochs: {epochs}\n"
         f"Steps: {steps}\n"
         f"RVC Version: {version}\n"
         f"Sampling Rate: {sr}\n"
         f"Pitch Guidance: {pitch_guidance}\n"
+        f"Dataset Length: {dataset_lenght}\n"
         f"Creation Date: {creation_date_str}\n"
         f"Hash (ID): {model_hash}"
+        f"Overtrain Info: {overtrain_info}"
     )
diff --git a/rvc/train/train.py b/rvc/train/train.py
index 2adbcee7..a3e06d68 100644
--- a/rvc/train/train.py
+++ b/rvc/train/train.py
@@ -28,6 +28,7 @@
 from torch.utils.tensorboard import SummaryWriter
 import torch.distributed as dist
 import torch.multiprocessing as mp
+from pydub import AudioSegment
 
 now_dir = os.getcwd()
 sys.path.append(os.path.join(now_dir))
@@ -72,10 +73,12 @@
 overtraining_detector = strtobool(sys.argv[14])
 overtraining_threshold = int(sys.argv[15])
 sync_graph = strtobool(sys.argv[16])
+model_creator = sys.argv[17]
 
 current_dir = os.getcwd()
 experiment_dir = os.path.join(current_dir, "logs", model_name)
 config_save_path = os.path.join(experiment_dir, "config.json")
+dataset_path = os.path.join(experiment_dir, "sliced_audios")
 
 with open(config_save_path, "r") as f:
     config = json.load(f)
@@ -97,6 +100,8 @@
 smoothed_loss_disc_history = []
 lowest_value = {"step": 0, "value": float("inf"), "epoch": 0}
 training_file_path = os.path.join(experiment_dir, "training_data.json")
+dataset_duration = 0
+overtrain_info = None
 
 import logging
 
@@ -124,6 +129,24 @@ def record(self):
         return f"time={current_time} | training_speed={elapsed_time_str}"
 
 
+def ms_to_min_sec(ms):
+    seconds = ms // 1000
+    minutes = seconds // 60
+    seconds = seconds % 60
+    return f"{minutes}:{seconds:02}"
+
+
+def get_audio_durations(dataset_path):
+    durations = []
+    for filename in os.listdir(dataset_path):
+        if filename.endswith(".wav"):  # Assumindo que os arquivos de áudio são .wav
+            audio_path = os.path.join(dataset_path, filename)
+            audio = AudioSegment.from_wav(audio_path)
+            duration_ms = len(audio)
+            durations.append(ms_to_min_sec(duration_ms))
+    return durations
+
+
 def main():
     """
     Main function to start the training process.
@@ -203,6 +226,8 @@ def continue_overtrain_detector(training_file_path):
         print("GPU not detected, reverting to CPU (not recommended)")
         n_gpus = 1
 
+    dataset_duration = get_audio_durations(dataset_path)
+
     if sync_graph == True:
         print(
             "Sync graph is now activated! With sync graph enabled, the model undergoes a single epoch of training. Once the graphs are synchronized, training proceeds for the previously specified number of epochs."
@@ -821,6 +846,9 @@ def train_and_evaluate(
                 step=global_step,
                 version=version,
                 hps=hps,
+                model_creator=model_creator,
+                overtrain_info=overtrain_info,
+                dataset_lenght=dataset_duration,
             )
 
     def check_overtraining(smoothed_loss_history, threshold, epsilon=0.004):
@@ -917,6 +945,8 @@ def save_to_json(
             consecutive_increases_gen += 1
         else:
             consecutive_increases_gen = 0
+
+        overtrain_info = f"Smoothed loss_g {smoothed_value_gen:.3f} and loss_d {smoothed_value_disc:.3f}"
         # Save the data in the JSON file if the epoch is divisible by save_every_epoch
         if epoch % save_every_epoch == 0:
             save_to_json(
@@ -965,6 +995,9 @@ def save_to_json(
                 step=global_step,
                 version=version,
                 hps=hps,
+                model_creator=model_creator,
+                overtrain_info=overtrain_info,
+                dataset_lenght=dataset_duration,
             )
 
     # Print training progress
@@ -1025,6 +1058,9 @@ def save_to_json(
                 step=global_step,
                 version=version,
                 hps=hps,
+                model_creator=model_creator,
+                overtrain_info=overtrain_info,
+                dataset_lenght=dataset_duration,
             )
         sleep(1)
         os._exit(2333333)
diff --git a/tabs/extra/extra.py b/tabs/extra/extra.py
index 40bf778a..eeb25649 100644
--- a/tabs/extra/extra.py
+++ b/tabs/extra/extra.py
@@ -21,12 +21,11 @@ def extra_tab():
             "This section contains some extra utilities that often may be in experimental phases."
         )
     )
+    with gr.TabItem(i18n("Model information")):
+        processing_tab()
 
     with gr.TabItem(i18n("F0 Curve")):
         f0_extractor_tab()
 
-    with gr.TabItem(i18n("Processing")):
-        processing_tab()
-
     with gr.TabItem(i18n("Audio Analyzer")):
         analyzer_tab()
diff --git a/tabs/extra/processing/processing.py b/tabs/extra/processing/processing.py
index dbc9e7cb..1eb037ac 100644
--- a/tabs/extra/processing/processing.py
+++ b/tabs/extra/processing/processing.py
@@ -27,7 +27,7 @@ def processing_tab():
             label=i18n("Output Information"),
             info=i18n("The output information will be displayed here."),
             value="",
-            max_lines=8,
+            max_lines=11,
         )
         model_view_button = gr.Button(i18n("View"), variant="primary")
         model_view_button.click(
diff --git a/tabs/train/train.py b/tabs/train/train.py
index 211b159b..60eab759 100644
--- a/tabs/train/train.py
+++ b/tabs/train/train.py
@@ -311,6 +311,14 @@ def train_tab():
             interactive=True,
             allow_custom_value=True,
         )
+        model_creator = gr.Dropdown(
+            label=i18n("Model Creator"),
+            info=i18n("Name of the model creator."),
+            value="",
+            interactive=True,
+            visible=True,
+            allow_custom_value=True,
+        )
         sampling_rate = gr.Radio(
             label=i18n("Sampling Rate"),
             info=i18n("The sampling rate of the audio files."),
@@ -752,6 +760,7 @@ def train_tab():
                     custom_pretrained,
                     g_pretrained_path,
                     d_pretrained_path,
+                    model_creator,
                 ],
                 outputs=[train_output_info],
                 api_name="start_training",

From 3da9187bac26c7f831bd7acadac7229524d854ed Mon Sep 17 00:00:00 2001
From: Guilherme <86894155+ShiromiyaG@users.noreply.github.com>
Date: Tue, 27 Aug 2024 08:57:50 -0300
Subject: [PATCH 2/6] Make the author's name compatible with Fumiama's fork

---
 rvc/train/process/extract_model.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/rvc/train/process/extract_model.py b/rvc/train/process/extract_model.py
index 23536d8a..76964fc7 100644
--- a/rvc/train/process/extract_model.py
+++ b/rvc/train/process/extract_model.py
@@ -84,7 +84,7 @@ def extract_model(
         model_hash = hashlib.sha256(hash_input.encode()).hexdigest()
         opt["model_hash"] = model_hash
         opt["model_name"] = name
-        opt["model_creator"] = model_creator
+        opt["author"] = model_creator
         opt["overtrain_info"] = overtrain_info
         opt["dataset_lenght"] = dataset_lenght
 

From e88a953f2a540df679471d7b15c66cf29d9b73ec Mon Sep 17 00:00:00 2001
From: Guilherme <86894155+ShiromiyaG@users.noreply.github.com>
Date: Tue, 27 Aug 2024 08:59:00 -0300
Subject: [PATCH 3/6] oops

---
 rvc/train/process/model_information.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/rvc/train/process/model_information.py b/rvc/train/process/model_information.py
index 6c150b22..88214f95 100644
--- a/rvc/train/process/model_information.py
+++ b/rvc/train/process/model_information.py
@@ -18,7 +18,7 @@ def model_information(path):
     print(f"Loaded model from {path}")
 
     model_name = model_data.get("model_name", "None")
-    model_creator = model_data.get("model_creator", "None")
+    model_creator = model_data.get("author", "None")
     epochs = model_data.get("epoch", "None")
     steps = model_data.get("step", "None")
     sr = model_data.get("sr", "None")

From a8d238e3a663f069055f48c143a95d65add3cb40 Mon Sep 17 00:00:00 2001
From: Guilherme <86894155+ShiromiyaG@users.noreply.github.com>
Date: Tue, 27 Aug 2024 09:12:28 -0300
Subject: [PATCH 4/6] Make the creator's name optional

---
 core.py                            | 1 +
 rvc/train/process/extract_model.py | 2 ++
 tabs/extra/model_information.py    | 2 +-
 tabs/train/train.py                | 2 +-
 4 files changed, 5 insertions(+), 2 deletions(-)

diff --git a/core.py b/core.py
index af33f44b..7b4c5d70 100644
--- a/core.py
+++ b/core.py
@@ -1359,6 +1359,7 @@ def parse_arguments():
         type=str,
         help="Model creator name.",
         default=None,
+        required=False,
     )
     train_parser.add_argument(
         "--cache_data_in_gpu",
diff --git a/rvc/train/process/extract_model.py b/rvc/train/process/extract_model.py
index 76964fc7..17f2a544 100644
--- a/rvc/train/process/extract_model.py
+++ b/rvc/train/process/extract_model.py
@@ -84,6 +84,8 @@ def extract_model(
         model_hash = hashlib.sha256(hash_input.encode()).hexdigest()
         opt["model_hash"] = model_hash
         opt["model_name"] = name
+        if model_creator is None:
+            model_creator = "Unknown"
         opt["author"] = model_creator
         opt["overtrain_info"] = overtrain_info
         opt["dataset_lenght"] = dataset_lenght
diff --git a/tabs/extra/model_information.py b/tabs/extra/model_information.py
index 27e8f77e..de0fc73d 100644
--- a/tabs/extra/model_information.py
+++ b/tabs/extra/model_information.py
@@ -18,7 +18,7 @@ def model_information_tab():
             label=i18n("Output Information"),
             info=i18n("The output information will be displayed here."),
             value="",
-            max_lines=8,
+            max_lines=11,
             interactive=False,
         )
         model_information_button = gr.Button(i18n("See Model Information"))
diff --git a/tabs/train/train.py b/tabs/train/train.py
index 60eab759..73056629 100644
--- a/tabs/train/train.py
+++ b/tabs/train/train.py
@@ -314,7 +314,7 @@ def train_tab():
         model_creator = gr.Dropdown(
             label=i18n("Model Creator"),
             info=i18n("Name of the model creator."),
-            value="",
+            value=None,
             interactive=True,
             visible=True,
             allow_custom_value=True,

From 9a900845f1425ba8bfb34adc76f8bbaf19df67e8 Mon Sep 17 00:00:00 2001
From: Guilherme <86894155+ShiromiyaG@users.noreply.github.com>
Date: Tue, 27 Aug 2024 10:41:43 -0300
Subject: [PATCH 5/6] Fix TTS

---
 core.py | 15 +++++++++++++++
 1 file changed, 15 insertions(+)

diff --git a/core.py b/core.py
index 7b4c5d70..cb9ec88f 100644
--- a/core.py
+++ b/core.py
@@ -340,6 +340,21 @@ def run_tts_script(
         f0_file=f0_file,
         embedder_model=embedder_model,
         embedder_model_custom=embedder_model_custom,
+        formant_shifting=None,
+        formant_qfrency=None,
+        formant_timbre=None,
+        post_process=None,
+        reverb=None,
+        pitch_shift=None,
+        limiter=None,
+        gain=None,
+        distortion=None,
+        chorus=None,
+        bitcrush=None,
+        clipping=None,
+        compressor=None,
+        delay=None,
+        sliders=None,
     )
 
     return f"Text {tts_text} synthesized successfully.", output_rvc_path.replace(

From d890d6621d4492291500637b3d40c3913b3ff559 Mon Sep 17 00:00:00 2001
From: Guilherme <86894155+ShiromiyaG@users.noreply.github.com>
Date: Tue, 27 Aug 2024 11:59:57 -0300
Subject: [PATCH 6/6] Add a note

---
 assets/i18n/languages/en_US.json | 2 +-
 tabs/train/train.py              | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/assets/i18n/languages/en_US.json b/assets/i18n/languages/en_US.json
index 2545b488..2a7da46a 100644
--- a/assets/i18n/languages/en_US.json
+++ b/assets/i18n/languages/en_US.json
@@ -304,5 +304,5 @@
   "Move files to custom embedder folder": "Move files to custom embedder folder",
   "model information": "model information",
   "Model Creator": "Model Creator",
-  "Name of the model creator.": "Name of the model creator."
+  "Name of the model creator. (Default: Unknown)": "Name of the model creator. (Default: Unknown)"
 }
diff --git a/tabs/train/train.py b/tabs/train/train.py
index 73056629..2d5b4f6f 100644
--- a/tabs/train/train.py
+++ b/tabs/train/train.py
@@ -313,7 +313,7 @@ def train_tab():
         )
         model_creator = gr.Dropdown(
             label=i18n("Model Creator"),
-            info=i18n("Name of the model creator."),
+            info=i18n("Name of the model creator. (Default: Unknown)"),
             value=None,
             interactive=True,
             visible=True,