diff --git a/DiffRhythmNode.py b/DiffRhythmNode.py index c04e7e9..8d91b88 100644 --- a/DiffRhythmNode.py +++ b/DiffRhythmNode.py @@ -164,6 +164,7 @@ class DiffRhythmRun: self.cfm = None self.vae = None self.muq = None + self.tokenizer = None @classmethod def INPUT_TYPES(cls): @@ -213,9 +214,9 @@ class DiffRhythmRun: max_frames = 6144 if self.cfm is None: - self.cfm, tokenizer, self.muq, self.vae = self.prepare_model(model, self.device) + self.cfm, self.tokenizer, self.muq, self.vae = self.prepare_model(model, self.device) - lrc_prompt, start_time = get_lrc_token(max_frames, lyrics_prompt, tokenizer, self.device) + lrc_prompt, start_time = get_lrc_token(max_frames, lyrics_prompt, self.tokenizer, self.device) vocal_flag = False if style_audio: @@ -256,6 +257,7 @@ class DiffRhythmRun: self.cfm = None self.muq = None self.vae = None + self.tokenizer = None gc.collect() torch.cuda.empty_cache() diff --git a/pyproject.toml b/pyproject.toml index c10ba2a..6bec0c2 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 = "2.1.4" +version = "2.1.5" 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"]