diff --git a/ACL_PyTorch/built-in/audio/CosyVoice/CosyVoice2/300I/diff_CosyVoice_300I.patch b/ACL_PyTorch/built-in/audio/CosyVoice2/300I/diff_CosyVoice_300I.patch similarity index 100% rename from ACL_PyTorch/built-in/audio/CosyVoice/CosyVoice2/300I/diff_CosyVoice_300I.patch rename to ACL_PyTorch/built-in/audio/CosyVoice2/300I/diff_CosyVoice_300I.patch diff --git a/ACL_PyTorch/built-in/audio/CosyVoice/CosyVoice2/300I/modeling_qwen2.py b/ACL_PyTorch/built-in/audio/CosyVoice2/300I/modeling_qwen2.py similarity index 100% rename from ACL_PyTorch/built-in/audio/CosyVoice/CosyVoice2/300I/modeling_qwen2.py rename to ACL_PyTorch/built-in/audio/CosyVoice2/300I/modeling_qwen2.py diff --git a/ACL_PyTorch/built-in/audio/CosyVoice/CosyVoice2/800I/diff_CosyVoice_800I.patch b/ACL_PyTorch/built-in/audio/CosyVoice2/800I/diff_CosyVoice_800I.patch similarity index 100% rename from ACL_PyTorch/built-in/audio/CosyVoice/CosyVoice2/800I/diff_CosyVoice_800I.patch rename to ACL_PyTorch/built-in/audio/CosyVoice2/800I/diff_CosyVoice_800I.patch diff --git a/ACL_PyTorch/built-in/audio/CosyVoice/CosyVoice2/800I/modeling_qwen2.py b/ACL_PyTorch/built-in/audio/CosyVoice2/800I/modeling_qwen2.py similarity index 100% rename from ACL_PyTorch/built-in/audio/CosyVoice/CosyVoice2/800I/modeling_qwen2.py rename to ACL_PyTorch/built-in/audio/CosyVoice2/800I/modeling_qwen2.py diff --git a/ACL_PyTorch/built-in/audio/CosyVoice/CosyVoice2/README.md b/ACL_PyTorch/built-in/audio/CosyVoice2/README.md old mode 100755 new mode 100644 similarity index 92% rename from ACL_PyTorch/built-in/audio/CosyVoice/CosyVoice2/README.md rename to ACL_PyTorch/built-in/audio/CosyVoice2/README.md index b418438adf65815cf58a74e63b988a7a321d1e87..833da1e741a1ff1760ceabf9c7ad84117537f8f9 --- a/ACL_PyTorch/built-in/audio/CosyVoice/CosyVoice2/README.md +++ b/ACL_PyTorch/built-in/audio/CosyVoice2/README.md @@ -30,11 +30,11 @@ | 配套 | 版本 | 环境准备指导 | | ------------------------------------------------------------ | ------ | ------------------------------------------------------------ | - | 固件与驱动 | 24.0.RC3 | [Pytorch框架推理环境准备](https://www.hiascend.com/document/detail/zh/ModelZoo/pytorchframework/pies) | - | CANN | 8.0.RC3 | 包含kernels包和toolkit包 | - | Python | 3.8 | - | - | PyTorch | 2.4.0 | - | - | Ascend Extension PyTorch | 2.4.0.post2 | - | + | 固件与驱动 | 25.2.0 | [Pytorch框架推理环境准备](https://www.hiascend.com/document/detail/zh/ModelZoo/pytorchframework/pies) | + | CANN | 8.2.RC1 | 包含kernels包和toolkit包 | + | Python | 3.11 | - | + | PyTorch | 2.3.1 | - | + | Ascend Extension PyTorch | 2.3.1.post6 | - | | 说明：Atlas 800I A2 推理卡和Atlas 300I DUO 推理卡请以CANN版本选择实际固件与驱动版本。 | \ | \ | @@ -59,7 +59,6 @@ cd ModelZoo-PyTorch/ACL_PyTorch/built-in/audio/CosyVoice/CosyVoice2 # 将infer.py复制到CosyVoice中 cp ../infer.py ./ # 获取Transformer源码 - cd .. git clone https://github.com/huggingface/transformers.git cd transformers git checkout v4.37.0 @@ -192,9 +191,9 @@ cd ModelZoo-PyTorch/ACL_PyTorch/built-in/audio/CosyVoice/CosyVoice2 * 非流式输入：将推理结果保存在`sft_i.wav`中，并打屏性能数据：实时率(rtf)，指的是平均1s时长的音频需要多少时间处理。 * 流式输入：将推理结果保存在`stream_input_out_i.wav`文件中，并打屏性能数据：实时率(rtf) - 3. 如因为意外操作导致torchair编译失败，需将已生成的.torchair_cache路径删除，避免使用编译错误的图导致的前向出错。 + 2. 如因为意外操作导致torchair编译失败，需将已生成的.torchair_cache路径删除，避免使用编译错误的图导致的前向出错。 -### 3 性能数据 +### 3 性能数据(SFT场景) | 模型 |芯片|rtf(实时率)| |-----------|------|------| diff --git a/ACL_PyTorch/built-in/audio/CosyVoice/CosyVoice2/infer.py b/ACL_PyTorch/built-in/audio/CosyVoice2/infer.py similarity index 100% rename from ACL_PyTorch/built-in/audio/CosyVoice/CosyVoice2/infer.py rename to ACL_PyTorch/built-in/audio/CosyVoice2/infer.py diff --git a/ACL_PyTorch/built-in/audio/CosyVoice/CosyVoice2/modify_onnx.py b/ACL_PyTorch/built-in/audio/CosyVoice2/modify_onnx.py old mode 100755 new mode 100644 similarity index 100% rename from ACL_PyTorch/built-in/audio/CosyVoice/CosyVoice2/modify_onnx.py rename to ACL_PyTorch/built-in/audio/CosyVoice2/modify_onnx.py diff --git a/ACL_PyTorch/built-in/audio/CosyVoice/CosyVoice2/requirements.txt b/ACL_PyTorch/built-in/audio/CosyVoice2/requirements.txt old mode 100755 new mode 100644 similarity index 88% rename from ACL_PyTorch/built-in/audio/CosyVoice/CosyVoice2/requirements.txt rename to ACL_PyTorch/built-in/audio/CosyVoice2/requirements.txt index 7afcddafb269269a08c4b19c0f1ef3059c09bd22..fb8c778f6249f1649090f99c6fce4565371e8a61 --- a/ACL_PyTorch/built-in/audio/CosyVoice/CosyVoice2/requirements.txt +++ b/ACL_PyTorch/built-in/audio/CosyVoice2/requirements.txt @@ -24,8 +24,8 @@ pyworld==0.3.4 rich==13.7.1 soundfile==0.12.1 tensorboard==2.14.0 -torch==2.4.0 -torch_npu==2.4.0.post2 +torch==2.3.1 +torch_npu==2.3.1.post6 torchaudio==2.4.0 uvicorn==0.30.0 wget==3.2