minicpmo-utils 0.1.0__py3-none-any.whl

minicpmo_utils-0.1.0.dist-info/METADATA

@@ -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.dist-info/RECORD

@@ -0,0 +1,148 @@
+cosyvoice/__init__.py,sha256=yXNTKqiEBedjE7e_Qav__IUpv9XfG9-em4yJxmoH5Zs,530
+cosyvoice/bin/average_model.py,sha256=82LeGlvQh8xFHA_T9fJJDBTtDhJ_UzukJJcFRjyjc9Y,3202
+cosyvoice/bin/export_jit.py,sha256=36EvvRkOE621pMqDL3Il69hXdLXsJtSGtDO3_r13nmA,3906
+cosyvoice/bin/export_onnx.py,sha256=nUHWmNTkGMowvHuXI92OR7R6v1doVsDxvxEzP5GWiQk,4661
+cosyvoice/bin/inference_deprecated.py,sha256=d4x-8eER00VwuOEHnObd9kd7cOriv_xbXZOAEXG60fk,6126
+cosyvoice/bin/train.py,sha256=1bpuU0d8fwc5ygrY65FbT9kfWpdZkfsXX83NBkNAMLc,8071
+cosyvoice/cli/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+cosyvoice/cli/cosyvoice.py,sha256=WCNOWg-_CJ-y9HDxL1tFrCirnB2a030jyN5EaELke6s,12808
+cosyvoice/cli/frontend.py,sha256=8z3yYsmjTpi8olTD2DNooPuW0nFnwDKSJIzPETvqv4c,13827
+cosyvoice/cli/model.py,sha256=X3Za8Ak5LEeRUldLalztoSnu6UsCP9WcYQFvu9MDacQ,24008
+cosyvoice/dataset/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+cosyvoice/dataset/dataset.py,sha256=QlINnr5R2kyDQQudr9jqdK_z63VQME2WJo8UBX8sVTo,4804
+cosyvoice/dataset/processor.py,sha256=RBpJ8QlP00_e8qkjEo86h2nBvYcuttNVtp_HIixx-wY,15876
+cosyvoice/flow/decoder.py,sha256=717Oudt_Y93aHVvKa_prKLjqEWVsSx-cEJ0o9lbLvyk,19866
+cosyvoice/flow/flow.py,sha256=qEl_61gzbnVmsfCF0RrP-ctPGiSUmr0sJE-_l8dvm20,12156
+cosyvoice/flow/flow_matching.py,sha256=sa1nH-N_hywDS96PdcwZwbiHWNVON1-itU4UoIit3-Y,10499
+cosyvoice/flow/length_regulator.py,sha256=srvavaBIUN8Mk0Vi35WyN8og-n6P6J0E2bgnqZ1nQRs,3137
+cosyvoice/hifigan/discriminator.py,sha256=PbWxtVhMYAyZbujFPqAhNwqK2cYuP1oo1c8l8Dq5-c8,8617
+cosyvoice/hifigan/f0_predictor.py,sha256=uoymCIodAtYIb-oh0E1p7pgXvdjqICiZgJnLCXGhhmk,2065
+cosyvoice/hifigan/generator.py,sha256=90YB5v6-tBCpYejtiTG0QHTThd7X9vd-6RigKbPUJiY,22754
+cosyvoice/hifigan/hifigan.py,sha256=6QDaL15-wwLbmm6m0rcZkOq9iCPgfllBm32MUt99wNY,3240
+cosyvoice/llm/llm.py,sha256=aYR8ZUgeYrXba-zoS7hONqbCIMsWVlHqt65Ug1uw6uQ,30566
+cosyvoice/tokenizer/tokenizer.py,sha256=lDQPx83ycMaaOutjKQxSQQROIHFOAf6nNvNh-eWlbfI,7456
+cosyvoice/tokenizer/assets/multilingual_zh_ja_yue_char_del.tiktoken,sha256=dHl5Yx6BMZNDaqvP98HCNdN96Al7ccVj7Itjt6UVxxg,907395
+cosyvoice/transformer/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+cosyvoice/transformer/activation.py,sha256=pKlsrqn3sFERKG3l6nYL39-cTlNEj1NCCFfcBKUEQMI,3089
+cosyvoice/transformer/attention.py,sha256=QdJpstXjo5UsClOPRkgK_4Vwdn64joBFLzZ0Ns72KLE,14389
+cosyvoice/transformer/convolution.py,sha256=619B8ySpciXHO5xDCvi7IxvXc4bvGEULsP0yn0aatOE,5230
+cosyvoice/transformer/decoder.py,sha256=2wQscn4OZTrJJHM7H7FeaXkv_YDJ089iJIN0VV1Yocw,16580
+cosyvoice/transformer/decoder_layer.py,sha256=uVZiq3LsawsPUMOhX77PFvrLeG0yO0rKHQY7nCHA1k4,4807
+cosyvoice/transformer/embedding.py,sha256=tQwwvlxmw5yV4NsQFYFMVF_NBiqTbA6nrUwBUIslldw,11777
+cosyvoice/transformer/encoder.py,sha256=J_nXSZcgNy--Z3TQkLif8GPH7PiPk6TXWye7GtspGKU,21434
+cosyvoice/transformer/encoder_layer.py,sha256=GSBYK-LJt894Nee1ORGOweudqPLHEcYlf4WYs3kpUbk,9602
+cosyvoice/transformer/label_smoothing_loss.py,sha256=24gEzxwg4a-_bDPeSDZYmxlH2IF5fQLVB8KoqNT0D90,3459
+cosyvoice/transformer/positionwise_feed_forward.py,sha256=boA447zIyght3KUI-5udQL86uYvrq89clJNdAyMp0Pg,4219
+cosyvoice/transformer/subsampling.py,sha256=MfwDR6hRq8EgXf1M9oCZwMQWWJw-maB7JQ6GMM7OGdA,12666
+cosyvoice/transformer/upsample_encoder.py,sha256=qAA8ISzmRpfOQwAfd2kC7mBpajt-Nzk1R5Ap3Mr31Wk,14149
+cosyvoice/utils/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+cosyvoice/utils/class_utils.py,sha256=ddaXf3V0ME-EZM21DIiVZMpVepmhEkGZGLKYHw6Nz8I,3321
+cosyvoice/utils/common.py,sha256=YWGo2Q176wdm9tLeZ8zONadUr0OTDCivN8TRm2Xu-Jc,6792
+cosyvoice/utils/executor.py,sha256=LqeifOT3xpwno9oy97jBBi1nuyCMZVsJsVeJh6n5vp4,8835
+cosyvoice/utils/file_utils.py,sha256=mV_0mRdhZSTODqVBgopxrjeLGbCWf-VLGVtHfgNcN_8,5461
+cosyvoice/utils/frontend_utils.py,sha256=DQypTgz1GeLRf3LNHcq2yweuoN5I5-eSgmNiKE6hjTA,4273
+cosyvoice/utils/losses.py,sha256=XND3_XjOViLBm7kRZRa3BWmeSMCPIXubiqESfyv5XBA,2121
+cosyvoice/utils/mask.py,sha256=hSxuSxQgGiOKsHY5bbWZwVey7lpaKqzw8nfdzoxkiXY,9728
+cosyvoice/utils/scheduler.py,sha256=lEfquE_Lcer2VG2zUVa0n-UxgvJEdEodyT66so-h6jQ,24920
+cosyvoice/utils/train_utils.py,sha256=yUFlYHUZRtrqQBx7kBJiYPbx2ArtAPKWzAxHQk4vsoo,16581
+cosyvoice/vllm/cosyvoice2.py,sha256=csgta5DvFgOjo_D6lrBPBdc66b5pBAwAfk_6SPTyb4k,4056
+matcha/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+matcha/app.py,sha256=bBOG1VM5L8X64aF8rH4iddu45dje9lmkyr5ubAq_4-4,13992
+matcha/cli.py,sha256=Iw_SITi4QrYGb_HZnxBNZ_ivnZ_zvXSJJEbyQ4clad8,15519
+matcha/train.py,sha256=eKFZFkaSvmdfMNhOAcvp63kp-Wvj_9IE47mAp4sHGxg,4613
+matcha/hifigan/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+matcha/hifigan/config.py,sha256=cxS_YTK_UoO2uma_Ue8QRFtNA6ldSEt1EL36u2-NgCc,779
+matcha/hifigan/denoiser.py,sha256=Q2juI2a3QTqce6fuvqWQnShoV671kVW2JsQyoqwwvLI,2644
+matcha/hifigan/env.py,sha256=QthDmqTWWAIo2tSI-arVteRVxnKUnm9mkjlQVnXbmDc,429
+matcha/hifigan/meldataset.py,sha256=4Ps0NXA3Yv0oV2PyCdHcptc4lZ43tp978m3BM2NYWxc,6786
+matcha/hifigan/models.py,sha256=Ln7J7YzeN4h33tBKQHhlPcoNfmX-_-kwBkTZIYJY5q8,11668
+matcha/hifigan/xutils.py,sha256=aNtYu1SyQaGtbN_NnSfLGhTi-r_y3vYKMIf7EYGaDOA,1396
+matcha/models/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+matcha/models/baselightningmodule.py,sha256=45aD1xV6AunZ7_O6RU6-B0OaUMMNO_gFKUWprs3EfCc,7003
+matcha/models/matcha_tts.py,sha256=zbpw9noGwR_fG9yBKxxtMhmHgnxS7PPE1_ft9KK-jPo,10059
+matcha/models/components/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+matcha/models/components/decoder.py,sha256=fqlKYc08lnA6oQf9zRuC3i8vat4go8X0NLYAUxgXams,14459
+matcha/models/components/flow_matching.py,sha256=pvKy7somAHSdV13ahiTYZNkEIn4qn3ZKbLK4kpBEXmw,4657
+matcha/models/components/text_encoder.py,sha256=QEQynUB7PaBsdlJ-e1TUBAe9rL2LBnmk5bAkr_vCTek,14845
+matcha/models/components/transformer.py,sha256=Zv8gktl0qZslG52OXAi7zszq7fr0HivYrytLx6Vty-o,13237
+matcha/onnx/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+matcha/onnx/export.py,sha256=VUAviRbyh9WhXaxLPQjdQLwaVK_ZZNgaWPMFA2UKh4c,5386
+matcha/onnx/infer.py,sha256=ACgjtfMdvzuU4XmalNLrebmkBD2W8VLcLi8blXnjUmU,6293
+matcha/text/__init__.py,sha256=6dxZsGZ25Fvk4_b72yfA7NVcNInCJLSjpvN9Mc66uBI,1696
+matcha/text/cleaners.py,sha256=t4-wsMBQdJT6J12mHiXME23tmpOi8pmV-GkejDVcNf8,3560
+matcha/text/numbers.py,sha256=Nwp-HmqaqstAELVqqlMu_7bj4qiWQzMXFmw56nBq1h0,2249
+matcha/text/symbols.py,sha256=hw78I8JtBmkri5HiDUQC05i2N9MFkxWJoHujIxJFP1c,635
+matcha/utils/__init__.py,sha256=YNs91rjzeUcVSs3aZjv9thma42I81Jksy9RF3nquvOo,326
+matcha/utils/audio.py,sha256=L3QQZL_MlIXRmkI1ZJbYriTeyAw9bvXKf06Kv7-cYco,2282
+matcha/utils/generate_data_statistics.py,sha256=E5Z9Tpa6ZFXfSLLQcAwhg7YYqBgLZDkIqBoGoOZhEjA,3269
+matcha/utils/instantiators.py,sha256=QxqOnRVabCSeurpPODXqh3JSZ-E7m8jKSsxy1WvTX-I,1828
+matcha/utils/logging_utils.py,sha256=glOI_JG8_YBKHWwD5RRKKkCez3N7DJlH27Vz48yAEtU,1711
+matcha/utils/model.py,sha256=UViKHaV89_IeaKJFww1xHV_RTXqv0YvfQWqwOtnzQ-I,2935
+matcha/utils/pylogger.py,sha256=YbC8Ym5HZrJcDBIsQO6jSnuyY5CLZQR13E_oAS9SYZQ,720
+matcha/utils/rich_utils.py,sha256=Oj5jrkz5s1b3RJL6m_8EXj85LY079FWClMIzf_Gwvcc,3279
+matcha/utils/utils.py,sha256=-XaKibvqf6mXCoGPkNcFKB8Ynq_bK6_S8-ntUhSR1WE,7159
+matcha/utils/monotonic_align/__init__.py,sha256=_s_INV7vL_N9mhYtZADAk11CsSGP8kykn0fEyyM73ts,646
+matcha/utils/monotonic_align/setup.py,sha256=bf0d0cvGRaACC22qq0sqgZEBBhz4qzDlMqBxOyTKC2g,207
+minicpmo/__init__.py,sha256=wyo8jLG2ATqFcTUrcTcvIzzaf37AYcbHdOAK15PLTjE,556
+minicpmo/utils.py,sha256=XJ717gRjAto7KytP8LFiwBfyFaom6fYtVG9oOMOKkiw,22445
+minicpmo/version.py,sha256=p4MxVO8ih2-hnh7INVBkrqMKBDGTkMso1LEbM6mx3jg,23
+s3tokenizer/__init__.py,sha256=8fg7P72mJjjvkyieA6Y2thBj2eNV_9hBxJp8aW3gcVg,5048
+s3tokenizer/cli.py,sha256=qUUA-5Ro-ILeQd-6YXVuItpYikEdaphIUfabpUTlIh4,5942
+s3tokenizer/model.py,sha256=v-DfJg11h9ztX1gHlUOLcpx9xq4ZZT5aqHOKNtG7Hzw,19542
+s3tokenizer/model_v2.py,sha256=x3hNCIT4UaHb715HojZ3AbEBy-wcPDWR-EohJiY4Aw0,21806
+s3tokenizer/utils.py,sha256=baAg5C7yi0AZx7caF0hJBU-pkj3coMPKM8QrLqgk-9k,14140
+s3tokenizer/assets/BAC009S0764W0121.wav,sha256=RtvJmMnR1IERJnxAdB3TIA8uW89AdfjEyX9EURYNzlA,134570
+s3tokenizer/assets/BAC009S0764W0122.wav,sha256=V2DHrOCSOkmfYF03MBHHqvZY4wSgmZpJPGNmnHVjhOE,131724
+s3tokenizer/assets/mel_filters.npz,sha256=dFCucHI6XvnTQePO5ijHywF382zkLES37SvzMl8PbUw,4271
+stepaudio2/__init__.py,sha256=ahfLETm7fp_A320ERK6j2xT4nTyecy8OA8Z-syVLuQQ,1547
+stepaudio2/stepaudio2.py,sha256=hf1EZl5odF8wFqs0VqC-B_DBioHrfpzWgZN501o1-YA,9265
+stepaudio2/token2wav.py,sha256=eAhDCCXjf9R7F5TsNfbNP5Se256Kqq9gKIrKFh0I5qE,11642
+stepaudio2/utils.py,sha256=p83TqxnLAu2OyAq3eCnESlkbU9kkNdH5Tnh1Da-J7Fg,3384
+stepaudio2/cosyvoice2/__init__.py,sha256=VFv7gYSAXPuP6d6arx-EqTy8W1WbPM4lGmdmPENMWT8,44
+stepaudio2/cosyvoice2/flow/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+stepaudio2/cosyvoice2/flow/decoder_dit.py,sha256=eQxV2JjZLI2HSWHltHtfnU3VwGfsIMpepmgzsyzisV0,22418
+stepaudio2/cosyvoice2/flow/flow.py,sha256=-zF1KSmD8PouUvJ6Bg_V_2RwwotBXbqFjE_oALoXJ4Q,8284
+stepaudio2/cosyvoice2/flow/flow_matching.py,sha256=CdXiwC78e2ViEre35wP2pVxlz6Xvtmv8bEcHrfuRYOM,8385
+stepaudio2/cosyvoice2/transformer/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+stepaudio2/cosyvoice2/transformer/attention.py,sha256=Yy3xmqfnEjE9DxsGeJ7NKKh5YLu7nsLSLcCgmieUo90,14314
+stepaudio2/cosyvoice2/transformer/embedding.py,sha256=2xDZKyM0P26YG2ZBevr-G3ZkyVukZEKjetho0yobSo4,4766
+stepaudio2/cosyvoice2/transformer/encoder_layer.py,sha256=3lRTqu3m7KhJJhqc6B4M5Oqgvakz3junmX2TTYtVXEE,6838
+stepaudio2/cosyvoice2/transformer/positionwise_feed_forward.py,sha256=r49V1dFDxbnGocDqCO-B8OXhb3Z4oykhRD5DmBMghAo,1887
+stepaudio2/cosyvoice2/transformer/subsampling.py,sha256=uFjanKdf855iAlq3sv4YhpbOQoInRI2Q7xADLIfIFqc,2529
+stepaudio2/cosyvoice2/transformer/upsample_encoder_v2.py,sha256=YVqzOMmNj1-GqJCbDXzzX6XS_WgZD2laYF8LIuNgbAQ,18854
+stepaudio2/cosyvoice2/utils/__init__.py,sha256=J3q6oPwKcQzvgZAdW8pZ58wpCRgx73ZDa12YdsggeTw,35
+stepaudio2/cosyvoice2/utils/class_utils.py,sha256=dGLa5anz6qeHKMAcPMh_z2gNGp2SuFLbjT30YN5gGp0,1377
+stepaudio2/cosyvoice2/utils/common.py,sha256=WIuAd4n0cGwwk1NyD0oeU1oK-5csW76kXL-xz1graII,3334
+stepaudio2/cosyvoice2/utils/mask.py,sha256=CkCYG5pGBHLH-Fdt5fEeQqnc3ytXFYA5g3dhsD5RtpY,1703
+stepaudio2/flashcosyvoice/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+stepaudio2/flashcosyvoice/cli.py,sha256=eSDEAFtGbgUv91vTzx4KXFRfAtx6eIgXg_5Ut3EtxUs,19863
+stepaudio2/flashcosyvoice/config.py,sha256=4VoEScKra4qgHxbcFuTycYAPMYYijt-Jjf6GEeKzCdU,2619
+stepaudio2/flashcosyvoice/cosyvoice2.py,sha256=7t2pRvJRxxVHrLRAOMMXwQjsOTHnxnHGGG3ciBk6xmk,7729
+stepaudio2/flashcosyvoice/cosyvoice3.py,sha256=UMxsD-RBudOyiUc4zLNwveluKfwwP4AcrrDOWRrSXLk,57
+stepaudio2/flashcosyvoice/engine/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+stepaudio2/flashcosyvoice/engine/block_manager.py,sha256=Q6Fd3GbUwYlXUJlFdjVh_t682C0N0yCAjUVSrAeQb3w,4027
+stepaudio2/flashcosyvoice/engine/llm_engine.py,sha256=j16UKgoPVxmO-f4oMQ4cR6CxmR4OqRQvlHjm2GFnhHM,5430
+stepaudio2/flashcosyvoice/engine/model_runner.py,sha256=fzYoPoUkcr1epAtaDeKVTTypWTcpkDj87MD4-f88opU,14036
+stepaudio2/flashcosyvoice/engine/scheduler.py,sha256=4o8M0XhFCrkaFtoGQHgz0BitYA_YpZP_mvTrfo0PwM4,3134
+stepaudio2/flashcosyvoice/engine/sequence.py,sha256=pwYdtCST9ZHqxPK62JCPRxt-G885nAbYh26oOuLHuao,2835
+stepaudio2/flashcosyvoice/modules/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+stepaudio2/flashcosyvoice/modules/flow.py,sha256=INAXZTIkGCxdosq7LBSNxb6BNQmWqDGyQV2Xshh228s,8292
+stepaudio2/flashcosyvoice/modules/hifigan.py,sha256=5wrEedjKzE8D6m1JNcaKxyLkvH2cgI0tBbU_1OWFzoo,9991
+stepaudio2/flashcosyvoice/modules/qwen2.py,sha256=L4Kx5s0SdqUmOYv6X6UCkcbRIctG3Dl8IPCNmiuB0Y0,3346
+stepaudio2/flashcosyvoice/modules/sampler.py,sha256=xhRQW6yP9e5mHjY8npypwZY1oe9mZ5CLH_IG5Oz1EAw,10914
+stepaudio2/flashcosyvoice/modules/flow_components/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+stepaudio2/flashcosyvoice/modules/flow_components/estimator.py,sha256=N7DmynOzkItXY2jTZl6eqaSH5IsXlkumpDxyOP-__5g,38148
+stepaudio2/flashcosyvoice/modules/flow_components/upsample_encoder.py,sha256=B2G2NM9xPeIx6r3v8n6Y49qh07PGBo55YZajL_m1Xjw,41265
+stepaudio2/flashcosyvoice/modules/hifigan_components/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+stepaudio2/flashcosyvoice/modules/hifigan_components/layers.py,sha256=jEb8ZuPZZ6Q2Nx5CKBrmCkAeTLR-Hlw7k1qxuUJ4kD0,16353
+stepaudio2/flashcosyvoice/modules/qwen2_components/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+stepaudio2/flashcosyvoice/modules/qwen2_components/layers.py,sha256=34Z0U3Qus11b6iE8I-k42lGq0lHz0KYDnVjUt6f0f84,21366
+stepaudio2/flashcosyvoice/utils/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+stepaudio2/flashcosyvoice/utils/audio.py,sha256=afxVAIkV_QliIzAwyOW1QygyjZxBUXjLwz4rSCbBs-c,2162
+stepaudio2/flashcosyvoice/utils/context.py,sha256=y_u92r100Wp7xjs4tQd70eB00iaqUOa92qkb5sw-8ho,821
+stepaudio2/flashcosyvoice/utils/loader.py,sha256=ETQwg7E6HxndH-hWSwTcSg2Q2MWrPDTEZtb2ebZaxS0,6030
+stepaudio2/flashcosyvoice/utils/memory.py,sha256=ciip3CEUfiSGe6qOFOfsYsGi6yYFcNcZVoAjnABVp28,597
+minicpmo_utils-0.1.0.dist-info/METADATA,sha256=dVIxQpW56luyHDeDXMrGMnyKNgUo9VBEo61Zi1a5YCA,2120
+minicpmo_utils-0.1.0.dist-info/WHEEL,sha256=wUyA8OaulRlbfwMtmQsvNngGrxQHAvkKcvRmdizlJi0,92
+minicpmo_utils-0.1.0.dist-info/top_level.txt,sha256=yIeAnGCB_ihapVKakC0wqRtAwCjezcDPCC-z6Vjlt0s,49
+minicpmo_utils-0.1.0.dist-info/RECORD,,

minicpmo_utils-0.1.0.dist-info/WHEEL

@@ -0,0 +1,5 @@
+Wheel-Version: 1.0
+Generator: setuptools (80.10.2)
+Root-Is-Purelib: true
+Tag: py3-none-any
+

minicpmo_utils-0.1.0.dist-info/top_level.txt

@@ -0,0 +1,5 @@
+cosyvoice
+matcha
+minicpmo
+s3tokenizer
+stepaudio2

s3tokenizer/__init__.py

@@ -0,0 +1,153 @@
+# Copyright (c) 2023 OpenAI. (authors: Whisper Team)
+#               2024 Tsinghua Univ. (authors: Xingchen Song)
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+"""Modified from
+    https://github.com/openai/whisper/blob/main/whisper/__init__.py
+"""
+
+import hashlib
+import os
+import urllib
+import warnings
+from typing import List, Union
+
+from tqdm import tqdm
+
+from s3tokenizer.model_v2 import S3TokenizerV2
+
+from .model import S3Tokenizer
+from .utils import (load_audio, log_mel_spectrogram, make_non_pad_mask,
+                    mask_to_bias, onnx2torch, padding, merge_tokenized_segments)
+
+__all__ = [
+    'load_audio', 'log_mel_spectrogram', 'make_non_pad_mask', 'mask_to_bias',
+    'onnx2torch', 'padding', 'merge_tokenized_segments'
+]
+_MODELS = {
+    "speech_tokenizer_v1":
+    "https://www.modelscope.cn/models/iic/cosyvoice-300m/"
+    "resolve/master/speech_tokenizer_v1.onnx",
+    "speech_tokenizer_v1_25hz":
+    "https://www.modelscope.cn/models/iic/CosyVoice-300M-25Hz/"
+    "resolve/master/speech_tokenizer_v1.onnx",
+    "speech_tokenizer_v2_25hz":
+    "https://www.modelscope.cn/models/iic/CosyVoice2-0.5B/"
+    "resolve/master/speech_tokenizer_v2.onnx",
+}
+
+_SHA256S = {
+    "speech_tokenizer_v1":
+    "23b5a723ed9143aebfd9ffda14ac4c21231f31c35ef837b6a13bb9e5488abb1e",
+    "speech_tokenizer_v1_25hz":
+    "56285ddd4a83e883ee0cb9f8d69c1089b53a94b1f78ff7e4a0224a27eb4cb486",
+    "speech_tokenizer_v2_25hz":
+    "d43342aa12163a80bf07bffb94c9de2e120a8df2f9917cd2f642e7f4219c6f71",
+}
+
+
+def _download(name: str, root: str) -> Union[bytes, str]:
+    os.makedirs(root, exist_ok=True)
+
+    expected_sha256 = _SHA256S[name]
+    url = _MODELS[name]
+    download_target = os.path.join(root, f"{name}.onnx")
+
+    if os.path.exists(download_target) and not os.path.isfile(download_target):
+        raise RuntimeError(
+            f"{download_target} exists and is not a regular file")
+
+    if os.path.isfile(download_target):
+        with open(download_target, "rb") as f:
+            model_bytes = f.read()
+        if hashlib.sha256(model_bytes).hexdigest() == expected_sha256:
+            return download_target
+        else:
+            warnings.warn(
+                f"{download_target} exists, but the SHA256 checksum does not"
+                " match; re-downloading the file")
+
+    with urllib.request.urlopen(url) as source, open(download_target,
+                                                     "wb") as output:
+        with tqdm(
+                total=int(source.info().get("Content-Length")),
+                ncols=80,
+                unit="iB",
+                unit_scale=True,
+                unit_divisor=1024,
+                desc="Downloading onnx checkpoint",
+        ) as loop:
+            while True:
+                buffer = source.read(8192)
+                if not buffer:
+                    break
+
+                output.write(buffer)
+                loop.update(len(buffer))
+
+    model_bytes = open(download_target, "rb").read()
+    if hashlib.sha256(model_bytes).hexdigest() != expected_sha256:
+        raise RuntimeError(
+            "Model has been downloaded but the SHA256 checksum does not not"
+            " match. Please retry loading the model.")
+
+    return download_target
+
+
+def available_models() -> List[str]:
+    """Returns the names of available models"""
+    return list(_MODELS.keys())
+
+
+def load_model(
+    name: str,
+    download_root: str = None,
+) -> S3Tokenizer:
+    """
+    Load a S3Tokenizer ASR model
+
+    Parameters
+    ----------
+    name : str
+        one of the official model names listed by
+        `s3tokenizer.available_models()`, or path to a model checkpoint
+         containing the model dimensions and the model state_dict.
+    download_root: str
+        path to download the model files; by default,
+        it uses "~/.cache/s3tokenizer"
+
+    Returns
+    -------
+    model : S3Tokenizer
+        The S3Tokenizer model instance
+    """
+
+    if download_root is None:
+        default = os.path.join(os.path.expanduser("~"), ".cache")
+        download_root = os.path.join(os.getenv("XDG_CACHE_HOME", default),
+                                     "s3tokenizer")
+
+    if name in _MODELS:
+        checkpoint_file = _download(name, download_root)
+    elif os.path.isfile(name):
+        checkpoint_file = name
+    else:
+        raise RuntimeError(
+            f"Model {name} not found; available models = {available_models()}")
+    if 'v2' in name:
+        model = S3TokenizerV2(name)
+    else:
+        model = S3Tokenizer(name)
+    model.init_from_onnx(checkpoint_file)
+
+    return model

s3tokenizer/cli.py

@@ -0,0 +1,183 @@
+# Copyright (c) 2024 Tsinghua Univ. (authors: Xingchen Song)
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+""" Example Usage
+cpu:
+
+s3tokenizer --wav_scp xxx.scp \
+            --device "cpu" \
+            --output_dir "./" \
+            --batch_size 32
+
+gpu:
+
+torchrun --nproc_per_node=8 --nnodes=1 \
+     --rdzv_id=2024 --rdzv_backend="c10d" --rdzv_endpoint="localhost:0" \
+    `which s3tokenizer` --wav_scp xxx.scp \
+                --device "cuda" \
+                --output_dir "./" \
+                --batch_size 32
+
+"""
+
+import argparse
+import json
+import os
+
+import torch
+import torch.distributed as dist
+from torch.utils.data import DataLoader, Dataset, DistributedSampler
+from tqdm import tqdm
+
+import s3tokenizer
+
+
+class AudioDataset(Dataset):
+
+    def __init__(self, wav_scp):
+        self.data = []
+        self.keys = []
+
+        with open(wav_scp, 'r', encoding='utf-8') as f:
+            for line in f:
+                key, file_path = line.strip().split()
+                self.data.append(file_path)
+                self.keys.append(key)
+
+    def __len__(self):
+        return len(self.data)
+
+    def __getitem__(self, idx):
+        file_path = self.data[idx]
+        key = self.keys[idx]
+        audio = s3tokenizer.load_audio(file_path)
+        mel = s3tokenizer.log_mel_spectrogram(audio)
+        return key, mel
+
+
+def collate_fn(batch):
+    keys = [item[0] for item in batch]
+    mels = [item[1] for item in batch]
+    mels, mels_lens = s3tokenizer.padding(mels)
+    return keys, mels, mels_lens
+
+
+def init_distributed():
+    world_size = int(os.environ.get('WORLD_SIZE', 1))
+    local_rank = int(os.environ.get('LOCAL_RANK', 0))
+    rank = int(os.environ.get('RANK', 0))
+    print('Inference on multiple gpus, this gpu {}'.format(local_rank) +
+          ', rank {}, world_size {}'.format(rank, world_size))
+    torch.cuda.set_device(local_rank)
+    dist.init_process_group("nccl")
+    return world_size, local_rank, rank
+
+
+def get_args():
+    parser = argparse.ArgumentParser(description='extract speech code')
+    parser.add_argument('--model',
+                        required=True,
+                        type=str,
+                        choices=[
+                            "speech_tokenizer_v1", "speech_tokenizer_v1_25hz",
+                            "speech_tokenizer_v2_25hz"
+                        ],
+                        help='model version')
+    parser.add_argument('--wav_scp',
+                        required=True,
+                        type=str,
+                        help='each line contains `wav_name wav_path`')
+    parser.add_argument('--device',
+                        required=True,
+                        type=str,
+                        choices=["cuda", "cpu"],
+                        help='device for inference')
+    parser.add_argument('--output_dir',
+                        required=True,
+                        type=str,
+                        help='dir to save result')
+    parser.add_argument('--batch_size',
+                        required=True,
+                        type=int,
+                        help='batch size (per-device) for inference')
+    parser.add_argument('--num_workers',
+                        type=int,
+                        default=4,
+                        help='workers for dataloader')
+    parser.add_argument('--prefetch',
+                        type=int,
+                        default=5,
+                        help='prefetch for dataloader')
+    args = parser.parse_args()
+    return args
+
+
+def main():
+    args = get_args()
+    os.makedirs(args.output_dir, exist_ok=True)
+
+    if args.device == "cuda":
+        assert (torch.cuda.is_available())
+        world_size, local_rank, rank = init_distributed()
+    else:
+        world_size, local_rank, rank = 1, 0, 0
+
+    device = torch.device(args.device)
+    model = s3tokenizer.load_model(args.model).to(device)
+    dataset = AudioDataset(args.wav_scp)
+
+    if args.device == "cuda":
+        model = torch.nn.parallel.DistributedDataParallel(
+            model, device_ids=[local_rank])
+        sampler = DistributedSampler(dataset,
+                                     num_replicas=world_size,
+                                     rank=rank)
+    else:
+        sampler = None
+
+    dataloader = DataLoader(dataset,
+                            batch_size=args.batch_size,
+                            sampler=sampler,
+                            shuffle=False,
+                            num_workers=args.num_workers,
+                            prefetch_factor=args.prefetch,
+                            collate_fn=collate_fn)
+
+    total_steps = len(dataset)
+
+    if rank == 0:
+        progress_bar = tqdm(total=total_steps, desc="Processing", unit="wavs")
+
+    writer = open(f"{args.output_dir}/part_{rank + 1}_of_{world_size}", "w")
+    for keys, mels, mels_lens in dataloader:
+        codes, codes_lens = model(mels.to(device), mels_lens.to(device))
+        for i, k in enumerate(keys):
+            code = codes[i, :codes_lens[i].item()].tolist()
+            writer.write(
+                json.dumps({
+                    "key": k,
+                    "code": code
+                }, ensure_ascii=False) + "\n")
+        if rank == 0:
+            progress_bar.update(world_size * len(keys))
+
+    if rank == 0:
+        progress_bar.close()
+    writer.close()
+    if args.device == "cuda":
+        dist.barrier()
+        dist.destroy_process_group()
+
+
+if __name__ == "__main__":
+    main()