From 9ca7a7cbca68233bc4ecac9bc403ae4f0fa6ad23 Mon Sep 17 00:00:00 2001 From: billwuhao Date: Tue, 25 Mar 2025 18:48:41 +0800 Subject: [PATCH] =?UTF-8?q?path=20to=20=F0=9F=8E=A4MW/,=20fix=20bug?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- DiffRhythmNode.py | 50 +------------------ MWAudioRecorderDR.py | 11 ++-- {dataset => dr_dataset}/__init__.py | 0 {dataset => dr_dataset}/dataset.py | 0 {dataset => dr_dataset}/latent/2626046476.pt | Bin {dataset => dr_dataset}/latent/28528706.pt | Bin {dataset => dr_dataset}/latent/547512549.pt | Bin {dataset => dr_dataset}/latent/549162241.pt | Bin {dataset => dr_dataset}/lrc/2626046476.pt | Bin {dataset => dr_dataset}/lrc/28528706.pt | Bin {dataset => dr_dataset}/lrc/547512549.pt | Bin {dataset => dr_dataset}/lrc/549162241.pt | Bin {dataset => dr_dataset}/style/2626046476.pt | Bin {dataset => dr_dataset}/style/28528706.pt | Bin {dataset => dr_dataset}/style/547512549.pt | Bin {dataset => dr_dataset}/style/549162241.pt | Bin {dataset => dr_dataset}/train.scp | 0 model/trainer.py | 3 +- pyproject.toml | 2 +- 19 files changed, 10 insertions(+), 56 deletions(-) rename {dataset => dr_dataset}/__init__.py (100%) rename {dataset => dr_dataset}/dataset.py (100%) rename {dataset => dr_dataset}/latent/2626046476.pt (100%) rename {dataset => dr_dataset}/latent/28528706.pt (100%) rename {dataset => dr_dataset}/latent/547512549.pt (100%) rename {dataset => dr_dataset}/latent/549162241.pt (100%) rename {dataset => dr_dataset}/lrc/2626046476.pt (100%) rename {dataset => dr_dataset}/lrc/28528706.pt (100%) rename {dataset => dr_dataset}/lrc/547512549.pt (100%) rename {dataset => dr_dataset}/lrc/549162241.pt (100%) rename {dataset => dr_dataset}/style/2626046476.pt (100%) rename {dataset => dr_dataset}/style/28528706.pt (100%) rename {dataset => dr_dataset}/style/547512549.pt (100%) rename {dataset => dr_dataset}/style/549162241.pt (100%) rename {dataset => dr_dataset}/train.scp (100%) diff --git a/DiffRhythmNode.py b/DiffRhythmNode.py index b55c012..7469ed6 100644 --- a/DiffRhythmNode.py +++ b/DiffRhythmNode.py @@ -122,7 +122,7 @@ class MultiLinePrompt: }, } - CATEGORY = "MW/MW-DiffRhythm" + CATEGORY = "🎤MW/MW-DiffRhythm" RETURN_TYPES = ("STRING",) RETURN_NAMES = ("prompt",) FUNCTION = "promptgen" @@ -166,7 +166,7 @@ class DiffRhythmRun: }, } - CATEGORY = "MW/MW-DiffRhythm" + CATEGORY = "🎤MW/MW-DiffRhythm" RETURN_TYPES = ("AUDIO",) RETURN_NAMES = ("audio",) FUNCTION = "diffrhythmgen" @@ -295,52 +295,6 @@ class DiffRhythmRun: return text_emb - - # @torch.no_grad() - # def get_style_prompt(self, model, audio=None, prompt=None): - # mulan = model - - # if prompt is not None: - # return mulan(texts=prompt).half() - - # if audio is None: - # raise ValueError("Audio data or style prompt must be provided") - - # waveform = audio["waveform"] - # sample_rate = audio["sample_rate"] - - # # Ensure waveform has correct shape - # if len(waveform.shape) == 3: # [1, channels, samples] - # waveform = waveform.squeeze(0) - # if waveform.shape[0] > 1: # If stereo, convert to mono - # waveform = waveform.mean(0, keepdim=True) - - # # Calculate audio length (seconds) - # audio_len = waveform.shape[-1] / sample_rate - - # if audio_len < 10: - # raise ValueError(f"Audio too short ({audio_len:.2f}s), minimum 10 seconds required.") - - # # Extract middle 10-second segment - # mid_time = int((audio_len // 2) * sample_rate) - # start_sample = mid_time - int(5 * sample_rate) - # end_sample = start_sample + int(10 * sample_rate) - # wav_segment = waveform[..., start_sample:end_sample] - - # # Resample to 24kHz - # if sample_rate != 24000: - # wav_segment = torchaudio.transforms.Resample(sample_rate, 24000)(wav_segment) - - # # Ensure correct shape and device - # wav = wav_segment.to(model.device) - # if len(wav.shape) == 1: - # wav = wav.unsqueeze(0) - - # with torch.no_grad(): - # audio_emb = mulan(wavs=wav) # [1, 512] - - # audio_emb = audio_emb.half() - # return audio_emb def prepare_model(self, model, device, unload_model=False): # prepare tokenizer diff --git a/MWAudioRecorderDR.py b/MWAudioRecorderDR.py index d87ffed..92330f7 100644 --- a/MWAudioRecorderDR.py +++ b/MWAudioRecorderDR.py @@ -17,7 +17,6 @@ class AudioRecorderDR: "record_sec": ("INT", { "default": 5, "min": 1, - "max": 60, "step": 1 }), "sample_rate": (["16000", "44100", "48000"], { @@ -31,14 +30,14 @@ class AudioRecorderDR: }), "sensitivity": ("FLOAT", { "default": 1.2, - "min": 0.5, + "min": 0.1, "max": 3.0, "step": 0.1 }), "smooth": ("INT", { - "default": 5, - "min": 1, - "max": 11, + "default": 1, + "min": 5, + "max": 7, "step": 2 }), "seed": ("INT", { @@ -52,7 +51,7 @@ class AudioRecorderDR: RETURN_TYPES = ("AUDIO",) RETURN_NAMES = ("audio",) FUNCTION = "record_and_clean" - CATEGORY = "MW/MW-DiffRhythm" + CATEGORY = "🎤MW/MW-DiffRhythm" def _stft(self, y, n_fft): hop = n_fft // 4 diff --git a/dataset/__init__.py b/dr_dataset/__init__.py similarity index 100% rename from dataset/__init__.py rename to dr_dataset/__init__.py diff --git a/dataset/dataset.py b/dr_dataset/dataset.py similarity index 100% rename from dataset/dataset.py rename to dr_dataset/dataset.py diff --git a/dataset/latent/2626046476.pt b/dr_dataset/latent/2626046476.pt similarity index 100% rename from dataset/latent/2626046476.pt rename to dr_dataset/latent/2626046476.pt diff --git a/dataset/latent/28528706.pt b/dr_dataset/latent/28528706.pt similarity index 100% rename from dataset/latent/28528706.pt rename to dr_dataset/latent/28528706.pt diff --git a/dataset/latent/547512549.pt b/dr_dataset/latent/547512549.pt similarity index 100% rename from dataset/latent/547512549.pt rename to dr_dataset/latent/547512549.pt diff --git a/dataset/latent/549162241.pt b/dr_dataset/latent/549162241.pt similarity index 100% rename from dataset/latent/549162241.pt rename to dr_dataset/latent/549162241.pt diff --git a/dataset/lrc/2626046476.pt b/dr_dataset/lrc/2626046476.pt similarity index 100% rename from dataset/lrc/2626046476.pt rename to dr_dataset/lrc/2626046476.pt diff --git a/dataset/lrc/28528706.pt b/dr_dataset/lrc/28528706.pt similarity index 100% rename from dataset/lrc/28528706.pt rename to dr_dataset/lrc/28528706.pt diff --git a/dataset/lrc/547512549.pt b/dr_dataset/lrc/547512549.pt similarity index 100% rename from dataset/lrc/547512549.pt rename to dr_dataset/lrc/547512549.pt diff --git a/dataset/lrc/549162241.pt b/dr_dataset/lrc/549162241.pt similarity index 100% rename from dataset/lrc/549162241.pt rename to dr_dataset/lrc/549162241.pt diff --git a/dataset/style/2626046476.pt b/dr_dataset/style/2626046476.pt similarity index 100% rename from dataset/style/2626046476.pt rename to dr_dataset/style/2626046476.pt diff --git a/dataset/style/28528706.pt b/dr_dataset/style/28528706.pt similarity index 100% rename from dataset/style/28528706.pt rename to dr_dataset/style/28528706.pt diff --git a/dataset/style/547512549.pt b/dr_dataset/style/547512549.pt similarity index 100% rename from dataset/style/547512549.pt rename to dr_dataset/style/547512549.pt diff --git a/dataset/style/549162241.pt b/dr_dataset/style/549162241.pt similarity index 100% rename from dataset/style/549162241.pt rename to dr_dataset/style/549162241.pt diff --git a/dataset/train.scp b/dr_dataset/train.scp similarity index 100% rename from dataset/train.scp rename to dr_dataset/train.scp diff --git a/model/trainer.py b/model/trainer.py index 4667621..8030afd 100644 --- a/model/trainer.py +++ b/model/trainer.py @@ -31,7 +31,8 @@ from torch.optim.lr_scheduler import LinearLR, SequentialLR, ConstantLR from accelerate import Accelerator from accelerate.utils import DistributedDataParallelKwargs -from dataset.dataset import DiffusionDataset + +from dr_dataset.dataset import DiffusionDataset from torch.utils.data import DataLoader diff --git a/pyproject.toml b/pyproject.toml index 8bb609d..8e0838d 100644 --- a/pyproject.toml +++ b/pyproject.toml @@ -1,7 +1,7 @@ [project] name = "diffrhythm_mw" description = "Blazingly Fast and Embarrassingly Simple End-to-End Full-Length Song Generation. A node for ComfyUI." -version = "2.1.0" +version = "2.1.1" license = {file = "LICENSE"} dependencies = ["# accelerate==1.4.0", "# torchdiffeq==0.2.5", "# torchaudio==2.6.0", "# transformers==4.49.0", "# librosa==0.10.2.post1", "# pyarrow==19.0.1", "# pandas==2.2.3", "# bitsandbytes", "# jieba==0.42.1", "# cn2an==0.5.23", "# pypinyin==0.53.0", "# onnxruntime", "LangSegment", "x-transformers", "pylance", "ema-pytorch", "prefigure", "muq", "mutagen", "pyopenjtalk", "pykakasi", "Unidecode", "phonemizer"]