minicpmo-utils 0.0.5__tar.gz → 0.1.0__tar.gz

minicpmo_utils-0.1.0/PKG-INFO

@@ -0,0 +1,72 @@
+Metadata-Version: 2.4
+Name: minicpmo-utils
+Version: 0.1.0
+Summary: Unified utilities package for MiniCPM-o: includes cosyvoice + stepaudio2 and extensible utils.
+Author: MiniCPM-o Utils Maintainers
+License: Apache-2.0
+Keywords: minicpmo,audio,tts,utils,cosyvoice,stepaudio2
+Classifier: Development Status :: 4 - Beta
+Classifier: Intended Audience :: Developers
+Classifier: License :: OSI Approved :: Apache Software License
+Classifier: Programming Language :: Python :: 3
+Classifier: Programming Language :: Python :: 3.10
+Classifier: Programming Language :: Python :: 3.11
+Classifier: Programming Language :: Python :: 3.12
+Requires-Python: >=3.10
+Description-Content-Type: text/markdown
+Requires-Dist: torch>=2.3.0
+Requires-Dist: torchaudio>=2.3.0
+Requires-Dist: transformers>=4.49.0
+Requires-Dist: numpy
+Requires-Dist: hyperpyyaml
+Requires-Dist: modelscope
+Requires-Dist: openai-whisper
+Requires-Dist: tqdm
+Requires-Dist: tiktoken
+Requires-Dist: inflect
+Requires-Dist: omegaconf
+Requires-Dist: einops
+Requires-Dist: librosa
+Requires-Dist: onnxruntime>=1.18.0
+Requires-Dist: diffusers
+Provides-Extra: gpu
+Requires-Dist: onnxruntime-gpu>=1.18.0; sys_platform == "linux" and extra == "gpu"
+
+## minicpmo-utils
+
+一个统一安装的工具包（一个 PyPI 分发包），把仓库里的 `cosyvoice` 与 `stepaudio2` 一起打进同一个 wheel，并预留 `minicpmo` 作为后续扩展 utils 的统一入口。
+
+### 安装方式
+
+- 从源码本地安装（开发态，可编辑）：
+```bash
+cd minicpmo-utils
+pip install -e .
+```
+
+- 构建并安装 wheel（推荐分发）：
+```bash
+cd minicpmo-utils
+python -m build        # 生成 dist/*.whl
+pip install dist/minicpmo_utils-0.1.0-py3-none-any.whl
+```
+
+### 导入方式
+
+包会暴露以下顶层模块，安装后可直接使用：
+- `import cosyvoice`
+- `import stepaudio2`
+- `import matcha`
+- `import minicpmo`
+
+也支持通过统一入口导入子包：
+```python
+from minicpmo import cosyvoice, stepaudio2, matcha
+```
+
+以及通过统一的 utils 入口使用通用工具函数，例如：
+
+```python
+from minicpmo.utils import get_video_frame_audio_segments
+```
+

{minicpmo_utils-0.0.5 → minicpmo_utils-0.1.0}/README.md

@@ -4,32 +4,17 @@
 
 ### 安装方式
 
-- 从源码本地安装（开发态，可编辑，默认只装公共依赖）：
+- 从源码本地安装（开发态，可编辑）：
 ```bash
 cd minicpmo-utils
 pip install -e .
 ```
 
-- 如果只想安装 cosyvoice 相关依赖（TTS）：
-```bash
-pip install -e .[tts]
-```
-
-- 如果只想安装 stepaudio2 / streaming 相关依赖：
-```bash
-pip install -e .[streaming]
-```
-
-- 同时安装 cosyvoice + stepaudio2 相关依赖：
-```bash
-pip install -e .[tts,streaming]
-```
-
 - 构建并安装 wheel（推荐分发）：
 ```bash
 cd minicpmo-utils
 python -m build        # 生成 dist/*.whl
-pip install \"dist/minicpmo_utils-0.1.0-py3-none-any.whl[tts,streaming]\"
+pip install dist/minicpmo_utils-0.1.0-py3-none-any.whl
 ```
 
 ### 导入方式

{minicpmo_utils-0.0.5 → minicpmo_utils-0.1.0}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "minicpmo-utils"
-version = "0.0.5"
+version = "0.1.0"
 description = "Unified utilities package for MiniCPM-o: includes cosyvoice + stepaudio2 and extensible utils."
 readme = "README.md"
 requires-python = ">=3.10"
@@ -31,68 +31,34 @@ classifiers = [
 #   - s3tokenizer (来自 S3Tokenizer-main)
 #   - minicpmo (本项目扩展 utils 的统一入口：from minicpmo.utils import ...)
 dependencies = [
-  "numpy",
-  "pillow==10.4.0",
-  "librosa==0.9.0",
-  "decord==0.6.0",
-  "moviepy==2.1.2",
-  "numba==0.61.2",
-]
-
-[project.optional-dependencies]
-# cosyvoice TTS 相关依赖
-tts = [
+  # Core ML frameworks (align to stepaudio2's requirements)
   "torch>=2.3.0",
   "torchaudio>=2.3.0",
-  "transformers>=4.51.0,<4.53.0",  # 4.52+ 有兼容性问题
-  "onnxruntime>=1.18.0,<=1.21.0",
-  "onnx",
+  "transformers>=4.49.0",
+
+  # Shared / common
+  "numpy",
   "hyperpyyaml",
-  "openai-whisper==20231117",
+
+  # cosyvoice side
+  "modelscope",
+  "openai-whisper",
   "tqdm",
   "tiktoken",
   "inflect",
-  "omegaconf>=2.0.6",
-  "conformer==0.3.2",
-  "einops==0.8.1",
-  "hydra-core",
-  "lightning==2.2.4",
-  "rich",
-  "gdown==5.2.0",
-  "matplotlib",
-  "wget",
-  "pyarrow",
-  "pyworld",
-  # 新增依赖
-  "scipy",
-  "pyyaml",
-  "regex",
-  "soundfile==0.12.1",
-  "diffusers==0.29.0"
-]
-
-# stepaudio2 基础依赖（token2wav 等）
-streaming = [
-  "minicpmo-utils[tts]",  # streaming 依赖 tts
-]
+  "omegaconf",
+  "einops",
 
-# stepaudio2 Flash 推理引擎依赖（flashcosyvoice.engine 模块需要）
-streaming-flash = [
-  "minicpmo-utils[streaming]",
-  "flash-attn>=2.6.0; sys_platform == 'linux'",
-  "triton>=2.3.0; sys_platform == 'linux'",
-  "safetensors",
-  "pynvml",
-  "xxhash",
+  # stepaudio2 side
+  "librosa",
+  "onnxruntime>=1.18.0",
+  "diffusers",
 ]
 
-# Linux GPU onnxruntime 可以很重，且与环境强相关，保留为可选 extra
+[project.optional-dependencies]
+# Linux GPU onnxruntime can be heavy and environment-specific; keep as an opt-in extra.
 gpu = [
-  "onnxruntime-gpu>=1.18.0,<=1.23.2; sys_platform == 'linux'",
-]
-
-all = [
-  "minicpmo-utils[tts,streaming,gpu]",
+  "onnxruntime-gpu>=1.18.0; sys_platform == 'linux'",
 ]
 
 [tool.setuptools]

{minicpmo_utils-0.0.5 → minicpmo_utils-0.1.0}/src/cosyvoice/cli/cosyvoice.py

@@ -16,6 +16,7 @@ import time
 from typing import Generator
 from tqdm import tqdm
 from hyperpyyaml import load_hyperpyyaml
+from modelscope import snapshot_download
 import torch
 from cosyvoice.cli.frontend import CosyVoiceFrontEnd
 from cosyvoice.cli.model import CosyVoiceModel, CosyVoice2Model
@@ -29,6 +30,8 @@ class CosyVoice:
         self.instruct = True if '-Instruct' in model_dir else False
         self.model_dir = model_dir
         self.fp16 = fp16
+        if not os.path.exists(model_dir):
+            model_dir = snapshot_download(model_dir)
         hyper_yaml_path = '{}/cosyvoice.yaml'.format(model_dir)
         if not os.path.exists(hyper_yaml_path):
             raise ValueError('{} not found!'.format(hyper_yaml_path))
@@ -151,6 +154,8 @@ class CosyVoice2(CosyVoice):
         self.instruct = True if '-Instruct' in model_dir else False
         self.model_dir = model_dir
         self.fp16 = fp16
+        if not os.path.exists(model_dir):
+            model_dir = snapshot_download(model_dir)
         
         if config_path is None:
             config_path = f'{model_dir}/cosyvoice2.yaml'

{minicpmo_utils-0.0.5 → minicpmo_utils-0.1.0}/src/cosyvoice/utils/file_utils.py

@@ -20,8 +20,7 @@ import torch
 import torchaudio
 import logging
 logging.getLogger('matplotlib').setLevel(logging.WARNING)
-logging.getLogger('numba').setLevel(logging.WARNING)
-logging.basicConfig(level=logging.INFO,
+logging.basicConfig(level=logging.DEBUG,
                     format='%(asctime)s %(levelname)s %(message)s')
 
 

{minicpmo_utils-0.0.5 → minicpmo_utils-0.1.0}/src/minicpmo/__init__.py

@@ -12,3 +12,13 @@
 """
 
 from .version import __version__
+
+# Eager re-exports to allow:
+#   from minicpmo import cosyvoice, stepaudio2, matcha
+# 而不需要懒加载。
+import cosyvoice as cosyvoice
+import stepaudio2 as stepaudio2
+import matcha as matcha
+
+__all__ = ["__version__", "cosyvoice", "stepaudio2", "matcha"]
+

{minicpmo_utils-0.0.5 → minicpmo_utils-0.1.0}/src/minicpmo/utils.py

@@ -26,84 +26,6 @@ MAX_NUM_FRAMES = int(os.getenv("MAX_NUM_FRAMES", 64))
 VIDEO_MME_DURATION = os.getenv("VIDEO_MME_DURATION", "ALL")
 
 
-def find_cjk_font():
-    """
-    查找支持中文的字体。
-    按优先级返回字体名称或路径，如果找不到则返回 None。
-    """
-    # 常见的中文字体名称（按优先级排列）
-    font_names = [
-        # Noto CJK 字体（推荐）
-        "Noto Sans CJK SC",
-        "Noto Sans CJK",
-        "NotoSansCJK-Regular",
-        # 文泉驿字体
-        "WenQuanYi Zen Hei",
-        "WenQuanYi Micro Hei",
-        "文泉驿正黑",
-        "文泉驿微米黑",
-        # 思源字体
-        "Source Han Sans SC",
-        "Source Han Sans CN",
-        # 其他常见中文字体
-        "SimHei",
-        "Microsoft YaHei",
-        "PingFang SC",
-        "Hiragino Sans GB",
-        "STHeiti",
-        "AR PL UMing CN",
-        "AR PL UKai CN",
-    ]
-
-    # 常见的中文字体文件路径
-    font_paths = [
-        # Noto CJK
-        "/usr/share/fonts/opentype/noto/NotoSansCJK-Regular.ttc",
-        "/usr/share/fonts/noto-cjk/NotoSansCJK-Regular.ttc",
-        "/usr/share/fonts/google-noto-cjk/NotoSansCJK-Regular.ttc",
-        "/usr/share/fonts/truetype/noto/NotoSansCJK-Regular.ttc",
-        # 文泉驿
-        "/usr/share/fonts/truetype/wqy/wqy-zenhei.ttc",
-        "/usr/share/fonts/truetype/wqy/wqy-microhei.ttc",
-        "/usr/share/fonts/wenquanyi/wqy-zenhei/wqy-zenhei.ttc",
-        "/usr/share/fonts/wenquanyi/wqy-microhei/wqy-microhei.ttc",
-        # macOS
-        "/System/Library/Fonts/PingFang.ttc",
-        "/Library/Fonts/Arial Unicode.ttf",
-        # Windows (WSL)
-        "/mnt/c/Windows/Fonts/msyh.ttc",
-        "/mnt/c/Windows/Fonts/simhei.ttf",
-    ]
-
-    # 首先尝试使用 fc-list 查找已安装的中文字体
-    try:
-        result = subprocess.run(
-            ["fc-list", ":lang=zh", "-f", "%{family}\n"],
-            capture_output=True,
-            text=True,
-            timeout=5,
-        )
-        if result.returncode == 0:
-            installed_fonts = set(result.stdout.strip().split("\n"))
-            for font_name in font_names:
-                for installed in installed_fonts:
-                    if font_name.lower() in installed.lower():
-                        logger.info(f"Found CJK font via fc-list: {installed}")
-                        return installed.split(",")[0]  # 取第一个名称
-    except Exception as e:
-        logger.debug(f"fc-list failed: {e}")
-
-    # 然后检查常见的字体文件路径
-    for path in font_paths:
-        if os.path.exists(path):
-            logger.info(f"Found CJK font file: {path}")
-            return path
-
-    logger.warning("No CJK font found. Chinese subtitles may display as boxes/garbled text.")
-    logger.warning("Install Chinese fonts with: sudo apt-get install fonts-noto-cjk")
-    return None
-
-
 def concat_images(images, bg_color=(255, 255, 255), cell_size=None, line_color=(0, 0, 0), line_width=6):
     """
     images: List[PIL.Image.Image]
@@ -649,18 +571,9 @@ def generate_duplex_video(
 
     if has_subtitles:
         srt_path_escaped = srt_path.replace("\\", "\\\\").replace("'", "'\\''").replace(":", "\\:")
-
-        # 查找支持中文的字体
-        cjk_font = find_cjk_font()
-        font_style = ""
-        if cjk_font:
-            # 转义字体路径/名称中的特殊字符
-            font_escaped = cjk_font.replace("\\", "\\\\").replace("'", "'\\''").replace(":", "\\:")
-            font_style = f"FontName={font_escaped},"
-
         subtitle_filter = (
             f"subtitles='{srt_path_escaped}':"
-            f"force_style='{font_style}FontSize=28,"
+            f"force_style='FontSize=28,"
             f"PrimaryColour=&H00FFFFFF,"
             f"OutlineColour=&H00000000,"
             f"BorderStyle=3,"

minicpmo_utils-0.1.0/src/minicpmo_utils.egg-info/PKG-INFO

@@ -0,0 +1,72 @@
+Metadata-Version: 2.4
+Name: minicpmo-utils
+Version: 0.1.0
+Summary: Unified utilities package for MiniCPM-o: includes cosyvoice + stepaudio2 and extensible utils.
+Author: MiniCPM-o Utils Maintainers
+License: Apache-2.0
+Keywords: minicpmo,audio,tts,utils,cosyvoice,stepaudio2
+Classifier: Development Status :: 4 - Beta
+Classifier: Intended Audience :: Developers
+Classifier: License :: OSI Approved :: Apache Software License
+Classifier: Programming Language :: Python :: 3
+Classifier: Programming Language :: Python :: 3.10
+Classifier: Programming Language :: Python :: 3.11
+Classifier: Programming Language :: Python :: 3.12
+Requires-Python: >=3.10
+Description-Content-Type: text/markdown
+Requires-Dist: torch>=2.3.0
+Requires-Dist: torchaudio>=2.3.0
+Requires-Dist: transformers>=4.49.0
+Requires-Dist: numpy
+Requires-Dist: hyperpyyaml
+Requires-Dist: modelscope
+Requires-Dist: openai-whisper
+Requires-Dist: tqdm
+Requires-Dist: tiktoken
+Requires-Dist: inflect
+Requires-Dist: omegaconf
+Requires-Dist: einops
+Requires-Dist: librosa
+Requires-Dist: onnxruntime>=1.18.0
+Requires-Dist: diffusers
+Provides-Extra: gpu
+Requires-Dist: onnxruntime-gpu>=1.18.0; sys_platform == "linux" and extra == "gpu"
+
+## minicpmo-utils
+
+一个统一安装的工具包（一个 PyPI 分发包），把仓库里的 `cosyvoice` 与 `stepaudio2` 一起打进同一个 wheel，并预留 `minicpmo` 作为后续扩展 utils 的统一入口。
+
+### 安装方式
+
+- 从源码本地安装（开发态，可编辑）：
+```bash
+cd minicpmo-utils
+pip install -e .
+```
+
+- 构建并安装 wheel（推荐分发）：
+```bash
+cd minicpmo-utils
+python -m build        # 生成 dist/*.whl
+pip install dist/minicpmo_utils-0.1.0-py3-none-any.whl
+```
+
+### 导入方式
+
+包会暴露以下顶层模块，安装后可直接使用：
+- `import cosyvoice`
+- `import stepaudio2`
+- `import matcha`
+- `import minicpmo`
+
+也支持通过统一入口导入子包：
+```python
+from minicpmo import cosyvoice, stepaudio2, matcha
+```
+
+以及通过统一的 utils 入口使用通用工具函数，例如：
+
+```python
+from minicpmo.utils import get_video_frame_audio_segments
+```
+

minicpmo_utils-0.1.0/src/minicpmo_utils.egg-info/requires.txt

@@ -0,0 +1,20 @@
+torch>=2.3.0
+torchaudio>=2.3.0
+transformers>=4.49.0
+numpy
+hyperpyyaml
+modelscope
+openai-whisper
+tqdm
+tiktoken
+inflect
+omegaconf
+einops
+librosa
+onnxruntime>=1.18.0
+diffusers
+
+[gpu]
+
+[gpu:sys_platform == "linux"]
+onnxruntime-gpu>=1.18.0

{minicpmo_utils-0.0.5 → minicpmo_utils-0.1.0}/src/stepaudio2/token2wav.py

@@ -7,6 +7,7 @@ import torchaudio
 import s3tokenizer
 import onnxruntime
 import numpy as np
+from copy import deepcopy
 
 import torchaudio.compliance.kaldi as kaldi
 from stepaudio2.flashcosyvoice.modules.hifigan import HiFTGenerator

minicpmo_utils-0.0.5/PKG-INFO

@@ -1,116 +0,0 @@
-Metadata-Version: 2.4
-Name: minicpmo-utils
-Version: 0.0.5
-Summary: Unified utilities package for MiniCPM-o: includes cosyvoice + stepaudio2 and extensible utils.
-Author: MiniCPM-o Utils Maintainers
-License: Apache-2.0
-Keywords: minicpmo,audio,tts,utils,cosyvoice,stepaudio2
-Classifier: Development Status :: 4 - Beta
-Classifier: Intended Audience :: Developers
-Classifier: License :: OSI Approved :: Apache Software License
-Classifier: Programming Language :: Python :: 3
-Classifier: Programming Language :: Python :: 3.10
-Classifier: Programming Language :: Python :: 3.11
-Classifier: Programming Language :: Python :: 3.12
-Requires-Python: >=3.10
-Description-Content-Type: text/markdown
-Requires-Dist: numpy
-Requires-Dist: pillow==10.4.0
-Requires-Dist: librosa==0.9.0
-Requires-Dist: decord==0.6.0
-Requires-Dist: moviepy==2.1.2
-Requires-Dist: numba==0.61.2
-Provides-Extra: tts
-Requires-Dist: torch>=2.3.0; extra == "tts"
-Requires-Dist: torchaudio>=2.3.0; extra == "tts"
-Requires-Dist: transformers<4.53.0,>=4.51.0; extra == "tts"
-Requires-Dist: onnxruntime<=1.21.0,>=1.18.0; extra == "tts"
-Requires-Dist: onnx; extra == "tts"
-Requires-Dist: hyperpyyaml; extra == "tts"
-Requires-Dist: openai-whisper==20231117; extra == "tts"
-Requires-Dist: tqdm; extra == "tts"
-Requires-Dist: tiktoken; extra == "tts"
-Requires-Dist: inflect; extra == "tts"
-Requires-Dist: omegaconf>=2.0.6; extra == "tts"
-Requires-Dist: conformer==0.3.2; extra == "tts"
-Requires-Dist: einops==0.8.1; extra == "tts"
-Requires-Dist: hydra-core; extra == "tts"
-Requires-Dist: lightning==2.2.4; extra == "tts"
-Requires-Dist: rich; extra == "tts"
-Requires-Dist: gdown==5.2.0; extra == "tts"
-Requires-Dist: matplotlib; extra == "tts"
-Requires-Dist: wget; extra == "tts"
-Requires-Dist: pyarrow; extra == "tts"
-Requires-Dist: pyworld; extra == "tts"
-Requires-Dist: scipy; extra == "tts"
-Requires-Dist: pyyaml; extra == "tts"
-Requires-Dist: regex; extra == "tts"
-Requires-Dist: soundfile==0.12.1; extra == "tts"
-Requires-Dist: diffusers==0.29.0; extra == "tts"
-Provides-Extra: streaming
-Requires-Dist: minicpmo-utils[tts]; extra == "streaming"
-Provides-Extra: streaming-flash
-Requires-Dist: minicpmo-utils[streaming]; extra == "streaming-flash"
-Requires-Dist: flash-attn>=2.6.0; sys_platform == "linux" and extra == "streaming-flash"
-Requires-Dist: triton>=2.3.0; sys_platform == "linux" and extra == "streaming-flash"
-Requires-Dist: safetensors; extra == "streaming-flash"
-Requires-Dist: pynvml; extra == "streaming-flash"
-Requires-Dist: xxhash; extra == "streaming-flash"
-Provides-Extra: gpu
-Requires-Dist: onnxruntime-gpu<=1.23.2,>=1.18.0; sys_platform == "linux" and extra == "gpu"
-Provides-Extra: all
-Requires-Dist: minicpmo-utils[gpu,streaming,tts]; extra == "all"
-
-## minicpmo-utils
-
-一个统一安装的工具包（一个 PyPI 分发包），把仓库里的 `cosyvoice` 与 `stepaudio2` 一起打进同一个 wheel，并预留 `minicpmo` 作为后续扩展 utils 的统一入口。
-
-### 安装方式
-
-- 从源码本地安装（开发态，可编辑，默认只装公共依赖）：
-```bash
-cd minicpmo-utils
-pip install -e .
-```
-
-- 如果只想安装 cosyvoice 相关依赖（TTS）：
-```bash
-pip install -e .[tts]
-```
-
-- 如果只想安装 stepaudio2 / streaming 相关依赖：
-```bash
-pip install -e .[streaming]
-```
-
-- 同时安装 cosyvoice + stepaudio2 相关依赖：
-```bash
-pip install -e .[tts,streaming]
-```
-
-- 构建并安装 wheel（推荐分发）：
-```bash
-cd minicpmo-utils
-python -m build        # 生成 dist/*.whl
-pip install \"dist/minicpmo_utils-0.1.0-py3-none-any.whl[tts,streaming]\"
-```
-
-### 导入方式
-
-包会暴露以下顶层模块，安装后可直接使用：
-- `import cosyvoice`
-- `import stepaudio2`
-- `import matcha`
-- `import minicpmo`
-
-也支持通过统一入口导入子包：
-```python
-from minicpmo import cosyvoice, stepaudio2, matcha
-```
-
-以及通过统一的 utils 入口使用通用工具函数，例如：
-
-```python
-from minicpmo.utils import get_video_frame_audio_segments
-```
-