{"metadata":{"kernelspec":{"language":"python","display_name":"Python 3","name":"python3"},"language_info":{"name":"python","version":"3.11.11","mimetype":"text/x-python","codemirror_mode":{"name":"ipython","version":3},"pygments_lexer":"ipython3","nbconvert_exporter":"python","file_extension":".py"},"kaggle":{"accelerator":"nvidiaTeslaT4","dataSources":[{"sourceId":91844,"databundleVersionId":11361821,"sourceType":"competition"}],"dockerImageVersionId":31041,"isInternetEnabled":true,"language":"python","sourceType":"notebook","isGpuEnabled":true}},"nbformat_minor":4,"nbformat":4,"cells":[{"cell_type":"code","source":"!pip install timm","metadata":{"_uuid":"8f2839f25d086af736a60e9eeb907d3b93b6e0e5","_cell_guid":"b1076dfc-b9ad-4769-8c92-a6c4dae69d19","trusted":true,"execution":{"iopub.status.busy":"2025-05-17T10:59:31.234512Z","iopub.execute_input":"2025-05-17T10:59:31.234717Z","iopub.status.idle":"2025-05-17T11:00:41.381849Z","shell.execute_reply.started":"2025-05-17T10:59:31.234694Z","shell.execute_reply":"2025-05-17T11:00:41.381183Z"}},"outputs":[{"name":"stdout","text":"Requirement already satisfied: timm in /usr/local/lib/python3.11/dist-packages (1.0.15)\nRequirement already satisfied: torch in /usr/local/lib/python3.11/dist-packages (from timm) (2.6.0+cu124)\nRequirement already satisfied: torchvision in /usr/local/lib/python3.11/dist-packages (from timm) (0.21.0+cu124)\nRequirement already satisfied: pyyaml in /usr/local/lib/python3.11/dist-packages (from timm) (6.0.2)\nRequirement already satisfied: huggingface_hub in /usr/local/lib/python3.11/dist-packages (from timm) (0.31.1)\nRequirement already satisfied: safetensors in /usr/local/lib/python3.11/dist-packages (from timm) (0.5.3)\nRequirement already satisfied: filelock in /usr/local/lib/python3.11/dist-packages (from huggingface_hub->timm) (3.18.0)\nRequirement already satisfied: fsspec>=2023.5.0 in /usr/local/lib/python3.11/dist-packages (from huggingface_hub->timm) (2025.3.2)\nRequirement already satisfied: packaging>=20.9 in /usr/local/lib/python3.11/dist-packages (from huggingface_hub->timm) (25.0)\nRequirement already satisfied: requests in /usr/local/lib/python3.11/dist-packages (from huggingface_hub->timm) (2.32.3)\nRequirement already satisfied: tqdm>=4.42.1 in /usr/local/lib/python3.11/dist-packages (from huggingface_hub->timm) (4.67.1)\nRequirement already satisfied: typing-extensions>=3.7.4.3 in /usr/local/lib/python3.11/dist-packages (from huggingface_hub->timm) (4.13.2)\nRequirement already satisfied: hf-xet<2.0.0,>=1.1.0 in /usr/local/lib/python3.11/dist-packages (from huggingface_hub->timm) (1.1.0)\nRequirement already satisfied: networkx in /usr/local/lib/python3.11/dist-packages (from torch->timm) (3.4.2)\nRequirement already satisfied: jinja2 in /usr/local/lib/python3.11/dist-packages (from torch->timm) (3.1.6)\nRequirement already satisfied: nvidia-cuda-nvrtc-cu12==12.4.127 in /usr/local/lib/python3.11/dist-packages (from torch->timm) (12.4.127)\nRequirement already satisfied: nvidia-cuda-runtime-cu12==12.4.127 in /usr/local/lib/python3.11/dist-packages (from torch->timm) (12.4.127)\nRequirement already satisfied: nvidia-cuda-cupti-cu12==12.4.127 in /usr/local/lib/python3.11/dist-packages (from torch->timm) (12.4.127)\nCollecting nvidia-cudnn-cu12==9.1.0.70 (from torch->timm)\n Downloading nvidia_cudnn_cu12-9.1.0.70-py3-none-manylinux2014_x86_64.whl.metadata (1.6 kB)\nCollecting nvidia-cublas-cu12==12.4.5.8 (from torch->timm)\n Downloading nvidia_cublas_cu12-12.4.5.8-py3-none-manylinux2014_x86_64.whl.metadata (1.5 kB)\nCollecting nvidia-cufft-cu12==11.2.1.3 (from torch->timm)\n Downloading nvidia_cufft_cu12-11.2.1.3-py3-none-manylinux2014_x86_64.whl.metadata (1.5 kB)\nCollecting nvidia-curand-cu12==10.3.5.147 (from torch->timm)\n Downloading nvidia_curand_cu12-10.3.5.147-py3-none-manylinux2014_x86_64.whl.metadata (1.5 kB)\nCollecting nvidia-cusolver-cu12==11.6.1.9 (from torch->timm)\n Downloading nvidia_cusolver_cu12-11.6.1.9-py3-none-manylinux2014_x86_64.whl.metadata (1.6 kB)\nCollecting nvidia-cusparse-cu12==12.3.1.170 (from torch->timm)\n Downloading nvidia_cusparse_cu12-12.3.1.170-py3-none-manylinux2014_x86_64.whl.metadata (1.6 kB)\nRequirement already satisfied: nvidia-cusparselt-cu12==0.6.2 in /usr/local/lib/python3.11/dist-packages (from torch->timm) (0.6.2)\nRequirement already satisfied: nvidia-nccl-cu12==2.21.5 in /usr/local/lib/python3.11/dist-packages (from torch->timm) (2.21.5)\nRequirement already satisfied: nvidia-nvtx-cu12==12.4.127 in /usr/local/lib/python3.11/dist-packages (from torch->timm) (12.4.127)\nCollecting nvidia-nvjitlink-cu12==12.4.127 (from torch->timm)\n Downloading nvidia_nvjitlink_cu12-12.4.127-py3-none-manylinux2014_x86_64.whl.metadata (1.5 kB)\nRequirement already satisfied: triton==3.2.0 in /usr/local/lib/python3.11/dist-packages (from torch->timm) (3.2.0)\nRequirement already satisfied: sympy==1.13.1 in /usr/local/lib/python3.11/dist-packages (from torch->timm) (1.13.1)\nRequirement already satisfied: mpmath<1.4,>=1.1.0 in /usr/local/lib/python3.11/dist-packages (from sympy==1.13.1->torch->timm) (1.3.0)\nRequirement already satisfied: numpy in /usr/local/lib/python3.11/dist-packages (from torchvision->timm) (1.26.4)\nRequirement already satisfied: pillow!=8.3.*,>=5.3.0 in /usr/local/lib/python3.11/dist-packages (from torchvision->timm) (11.1.0)\nRequirement already satisfied: MarkupSafe>=2.0 in /usr/local/lib/python3.11/dist-packages (from jinja2->torch->timm) (3.0.2)\nRequirement already satisfied: mkl_fft in /usr/local/lib/python3.11/dist-packages (from numpy->torchvision->timm) (1.3.8)\nRequirement already satisfied: mkl_random in /usr/local/lib/python3.11/dist-packages (from numpy->torchvision->timm) (1.2.4)\nRequirement already satisfied: mkl_umath in /usr/local/lib/python3.11/dist-packages (from numpy->torchvision->timm) (0.1.1)\nRequirement already satisfied: mkl in /usr/local/lib/python3.11/dist-packages (from numpy->torchvision->timm) (2025.1.0)\nRequirement already satisfied: tbb4py in /usr/local/lib/python3.11/dist-packages (from numpy->torchvision->timm) (2022.1.0)\nRequirement already satisfied: mkl-service in /usr/local/lib/python3.11/dist-packages (from numpy->torchvision->timm) (2.4.1)\nRequirement already satisfied: charset-normalizer<4,>=2 in /usr/local/lib/python3.11/dist-packages (from requests->huggingface_hub->timm) (3.4.2)\nRequirement already satisfied: idna<4,>=2.5 in /usr/local/lib/python3.11/dist-packages (from requests->huggingface_hub->timm) (3.10)\nRequirement already satisfied: urllib3<3,>=1.21.1 in /usr/local/lib/python3.11/dist-packages (from requests->huggingface_hub->timm) (2.4.0)\nRequirement already satisfied: certifi>=2017.4.17 in /usr/local/lib/python3.11/dist-packages (from requests->huggingface_hub->timm) (2025.4.26)\nRequirement already satisfied: intel-openmp<2026,>=2024 in /usr/local/lib/python3.11/dist-packages (from mkl->numpy->torchvision->timm) (2024.2.0)\nRequirement already satisfied: tbb==2022.* in /usr/local/lib/python3.11/dist-packages (from mkl->numpy->torchvision->timm) (2022.1.0)\nRequirement already satisfied: tcmlib==1.* in /usr/local/lib/python3.11/dist-packages (from tbb==2022.*->mkl->numpy->torchvision->timm) (1.3.0)\nRequirement already satisfied: intel-cmplr-lib-rt in /usr/local/lib/python3.11/dist-packages (from mkl_umath->numpy->torchvision->timm) (2024.2.0)\nRequirement already satisfied: intel-cmplr-lib-ur==2024.2.0 in /usr/local/lib/python3.11/dist-packages (from intel-openmp<2026,>=2024->mkl->numpy->torchvision->timm) (2024.2.0)\nDownloading nvidia_cublas_cu12-12.4.5.8-py3-none-manylinux2014_x86_64.whl (363.4 MB)\n\u001b[2K \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m363.4/363.4 MB\u001b[0m \u001b[31m4.2 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m0:00:01\u001b[0m00:01\u001b[0m\n\u001b[?25hDownloading nvidia_cudnn_cu12-9.1.0.70-py3-none-manylinux2014_x86_64.whl (664.8 MB)\n\u001b[2K \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m664.8/664.8 MB\u001b[0m \u001b[31m2.0 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m0:00:01\u001b[0m00:01\u001b[0m\n\u001b[?25hDownloading nvidia_cufft_cu12-11.2.1.3-py3-none-manylinux2014_x86_64.whl (211.5 MB)\n\u001b[2K \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m211.5/211.5 MB\u001b[0m \u001b[31m8.1 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m0:00:01\u001b[0m00:01\u001b[0m\n\u001b[?25hDownloading nvidia_curand_cu12-10.3.5.147-py3-none-manylinux2014_x86_64.whl (56.3 MB)\n\u001b[2K \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m56.3/56.3 MB\u001b[0m \u001b[31m32.0 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m:00:01\u001b[0m00:01\u001b[0m\n\u001b[?25hDownloading nvidia_cusolver_cu12-11.6.1.9-py3-none-manylinux2014_x86_64.whl (127.9 MB)\n\u001b[2K \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m127.9/127.9 MB\u001b[0m \u001b[31m13.7 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m:00:01\u001b[0m00:01\u001b[0m\n\u001b[?25hDownloading nvidia_cusparse_cu12-12.3.1.170-py3-none-manylinux2014_x86_64.whl (207.5 MB)\n\u001b[2K \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m207.5/207.5 MB\u001b[0m \u001b[31m8.3 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m0:00:01\u001b[0m00:01\u001b[0m\n\u001b[?25hDownloading nvidia_nvjitlink_cu12-12.4.127-py3-none-manylinux2014_x86_64.whl (21.1 MB)\n\u001b[2K \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m21.1/21.1 MB\u001b[0m \u001b[31m72.3 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m:00:01\u001b[0m00:01\u001b[0m\n\u001b[?25hInstalling collected packages: nvidia-nvjitlink-cu12, nvidia-curand-cu12, nvidia-cufft-cu12, nvidia-cublas-cu12, nvidia-cusparse-cu12, nvidia-cudnn-cu12, nvidia-cusolver-cu12\n Attempting uninstall: nvidia-nvjitlink-cu12\n Found existing installation: nvidia-nvjitlink-cu12 12.9.41\n Uninstalling nvidia-nvjitlink-cu12-12.9.41:\n Successfully uninstalled nvidia-nvjitlink-cu12-12.9.41\n Attempting uninstall: nvidia-curand-cu12\n Found existing installation: nvidia-curand-cu12 10.3.10.19\n Uninstalling nvidia-curand-cu12-10.3.10.19:\n Successfully uninstalled nvidia-curand-cu12-10.3.10.19\n Attempting uninstall: nvidia-cufft-cu12\n Found existing installation: nvidia-cufft-cu12 11.4.0.6\n Uninstalling nvidia-cufft-cu12-11.4.0.6:\n Successfully uninstalled nvidia-cufft-cu12-11.4.0.6\n Attempting uninstall: nvidia-cublas-cu12\n Found existing installation: nvidia-cublas-cu12 12.9.0.13\n Uninstalling nvidia-cublas-cu12-12.9.0.13:\n Successfully uninstalled nvidia-cublas-cu12-12.9.0.13\n Attempting uninstall: nvidia-cusparse-cu12\n Found existing installation: nvidia-cusparse-cu12 12.5.9.5\n Uninstalling nvidia-cusparse-cu12-12.5.9.5:\n Successfully uninstalled nvidia-cusparse-cu12-12.5.9.5\n Attempting uninstall: nvidia-cudnn-cu12\n Found existing installation: nvidia-cudnn-cu12 9.3.0.75\n Uninstalling nvidia-cudnn-cu12-9.3.0.75:\n Successfully uninstalled nvidia-cudnn-cu12-9.3.0.75\n Attempting uninstall: nvidia-cusolver-cu12\n Found existing installation: nvidia-cusolver-cu12 11.7.4.40\n Uninstalling nvidia-cusolver-cu12-11.7.4.40:\n Successfully uninstalled nvidia-cusolver-cu12-11.7.4.40\nSuccessfully installed nvidia-cublas-cu12-12.4.5.8 nvidia-cudnn-cu12-9.1.0.70 nvidia-cufft-cu12-11.2.1.3 nvidia-curand-cu12-10.3.5.147 nvidia-cusolver-cu12-11.6.1.9 nvidia-cusparse-cu12-12.3.1.170 nvidia-nvjitlink-cu12-12.4.127\n","output_type":"stream"}],"execution_count":1},{"cell_type":"code","source":"","metadata":{"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"import os\nimport gc\nimport warnings\nimport logging\nimport time\nimport math\nimport cv2\nfrom pathlib import Path\nimport numpy as np\nimport pandas as pd\nimport librosa\nimport torch\nimport torch.nn as nn\nimport torch.nn.functional as F\nimport timm\nfrom tqdm.auto import tqdm\n\n# Suppress warnings and limit logging output\nwarnings.filterwarnings(\"ignore\")\nlogging.basicConfig(level=logging.ERROR)","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2025-05-17T14:36:53.046666Z","iopub.execute_input":"2025-05-17T14:36:53.046926Z","iopub.status.idle":"2025-05-17T14:36:53.066368Z","shell.execute_reply.started":"2025-05-17T14:36:53.046906Z","shell.execute_reply":"2025-05-17T14:36:53.065314Z"}},"outputs":[{"traceback":["\u001b[0;31m---------------------------------------------------------------------------\u001b[0m","\u001b[0;31mAttributeError\u001b[0m Traceback (most recent call last)","\u001b[0;32m/tmp/ipykernel_35/325821887.py\u001b[0m in \u001b[0;36m\u001b[0;34m()\u001b[0m\n\u001b[1;32m 13\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mtorch\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mnn\u001b[0m \u001b[0;32mas\u001b[0m \u001b[0mnn\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 14\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mtorch\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mnn\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mfunctional\u001b[0m \u001b[0;32mas\u001b[0m \u001b[0mF\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 15\u001b[0;31m \u001b[0;32mimport\u001b[0m \u001b[0mtimm\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 16\u001b[0m \u001b[0;32mfrom\u001b[0m \u001b[0mtqdm\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mauto\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mtqdm\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 17\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n","\u001b[0;32m/usr/local/lib/python3.11/dist-packages/timm/__init__.py\u001b[0m in \u001b[0;36m\u001b[0;34m\u001b[0m\n\u001b[1;32m 1\u001b[0m \u001b[0;32mfrom\u001b[0m \u001b[0;34m.\u001b[0m\u001b[0mversion\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0m__version__\u001b[0m \u001b[0;32mas\u001b[0m \u001b[0m__version__\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 2\u001b[0;31m from .layers import (\n\u001b[0m\u001b[1;32m 3\u001b[0m \u001b[0mis_scriptable\u001b[0m \u001b[0;32mas\u001b[0m \u001b[0mis_scriptable\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 4\u001b[0m \u001b[0mis_exportable\u001b[0m \u001b[0;32mas\u001b[0m \u001b[0mis_exportable\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 5\u001b[0m \u001b[0mset_scriptable\u001b[0m \u001b[0;32mas\u001b[0m \u001b[0mset_scriptable\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n","\u001b[0;32m/usr/local/lib/python3.11/dist-packages/timm/layers/__init__.py\u001b[0m in \u001b[0;36m\u001b[0;34m\u001b[0m\n\u001b[1;32m 6\u001b[0m \u001b[0;32mfrom\u001b[0m \u001b[0;34m.\u001b[0m\u001b[0mattention_pool2d\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mAttentionPool2d\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mRotAttentionPool2d\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mRotaryEmbedding\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 7\u001b[0m \u001b[0;32mfrom\u001b[0m \u001b[0;34m.\u001b[0m\u001b[0mblur_pool\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mBlurPool2d\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mcreate_aa\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 8\u001b[0;31m \u001b[0;32mfrom\u001b[0m \u001b[0;34m.\u001b[0m\u001b[0mclassifier\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mcreate_classifier\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mClassifierHead\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mNormMlpClassifierHead\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mClNormMlpClassifierHead\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 9\u001b[0m \u001b[0;32mfrom\u001b[0m \u001b[0;34m.\u001b[0m\u001b[0mcond_conv2d\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mCondConv2d\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mget_condconv_initializer\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 10\u001b[0m \u001b[0;32mfrom\u001b[0m \u001b[0;34m.\u001b[0m\u001b[0mconfig\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mis_exportable\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mis_scriptable\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mis_no_jit\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0muse_fused_attn\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;31m \u001b[0m\u001b[0;31m\\\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n","\u001b[0;32m/usr/local/lib/python3.11/dist-packages/timm/layers/classifier.py\u001b[0m in \u001b[0;36m\u001b[0;34m\u001b[0m\n\u001b[1;32m 13\u001b[0m \u001b[0;32mfrom\u001b[0m \u001b[0;34m.\u001b[0m\u001b[0madaptive_avgmax_pool\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mSelectAdaptivePool2d\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 14\u001b[0m \u001b[0;32mfrom\u001b[0m \u001b[0;34m.\u001b[0m\u001b[0mcreate_act\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mget_act_layer\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 15\u001b[0;31m \u001b[0;32mfrom\u001b[0m \u001b[0;34m.\u001b[0m\u001b[0mcreate_norm\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mget_norm_layer\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 16\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 17\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n","\u001b[0;32m/usr/local/lib/python3.11/dist-packages/timm/layers/create_norm.py\u001b[0m in \u001b[0;36m\u001b[0;34m\u001b[0m\n\u001b[1;32m 12\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 13\u001b[0m \u001b[0;32mfrom\u001b[0m \u001b[0;34m.\u001b[0m\u001b[0mnorm\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mGroupNorm\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mGroupNorm1\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mLayerNorm\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mLayerNorm2d\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mRmsNorm\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mRmsNorm2d\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mSimpleNorm\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mSimpleNorm2d\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 14\u001b[0;31m \u001b[0;32mfrom\u001b[0m \u001b[0mtorchvision\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mops\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mmisc\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mFrozenBatchNorm2d\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 15\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 16\u001b[0m _NORM_MAP = dict(\n","\u001b[0;32m/usr/local/lib/python3.11/dist-packages/torchvision/__init__.py\u001b[0m in \u001b[0;36m\u001b[0;34m\u001b[0m\n\u001b[1;32m 8\u001b[0m \u001b[0;31m# .extensions) before entering _meta_registrations.\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 9\u001b[0m \u001b[0;32mfrom\u001b[0m \u001b[0;34m.\u001b[0m\u001b[0mextension\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0m_HAS_OPS\u001b[0m \u001b[0;31m# usort:skip\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 10\u001b[0;31m \u001b[0;32mfrom\u001b[0m \u001b[0mtorchvision\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0m_meta_registrations\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mdatasets\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mio\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mmodels\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mops\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mtransforms\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mutils\u001b[0m \u001b[0;31m# usort:skip\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 11\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 12\u001b[0m \u001b[0;32mtry\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n","\u001b[0;32m/usr/local/lib/python3.11/dist-packages/torchvision/_meta_registrations.py\u001b[0m in \u001b[0;36m\u001b[0;34m\u001b[0m\n\u001b[1;32m 23\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 24\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 25\u001b[0;31m \u001b[0;34m@\u001b[0m\u001b[0mregister_meta\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m\"roi_align\"\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 26\u001b[0m \u001b[0;32mdef\u001b[0m \u001b[0mmeta_roi_align\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0minput\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mrois\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mspatial_scale\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mpooled_height\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mpooled_width\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0msampling_ratio\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0maligned\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 27\u001b[0m \u001b[0mtorch\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_check\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mrois\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0msize\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;36m1\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;34m==\u001b[0m \u001b[0;36m5\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;32mlambda\u001b[0m\u001b[0;34m:\u001b[0m \u001b[0;34m\"rois must have shape as Tensor[K, 5]\"\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n","\u001b[0;32m/usr/local/lib/python3.11/dist-packages/torchvision/_meta_registrations.py\u001b[0m in \u001b[0;36mwrapper\u001b[0;34m(fn)\u001b[0m\n\u001b[1;32m 16\u001b[0m \u001b[0;32mdef\u001b[0m \u001b[0mregister_meta\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mop_name\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0moverload_name\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;34m\"default\"\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 17\u001b[0m \u001b[0;32mdef\u001b[0m \u001b[0mwrapper\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mfn\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 18\u001b[0;31m \u001b[0;32mif\u001b[0m \u001b[0mtorchvision\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mextension\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_has_ops\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 19\u001b[0m \u001b[0mget_meta_lib\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mimpl\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mgetattr\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mgetattr\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mtorch\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mops\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mtorchvision\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mop_name\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0moverload_name\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mfn\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 20\u001b[0m \u001b[0;32mreturn\u001b[0m \u001b[0mfn\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n","\u001b[0;31mAttributeError\u001b[0m: partially initialized module 'torchvision' has no attribute 'extension' (most likely due to a circular import)"],"ename":"AttributeError","evalue":"partially initialized module 'torchvision' has no attribute 'extension' (most likely due to a circular import)","output_type":"error"}],"execution_count":16},{"cell_type":"code","source":"","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2025-05-17T13:21:21.486025Z","iopub.execute_input":"2025-05-17T13:21:21.48664Z","iopub.status.idle":"2025-05-17T13:22:22.4792Z","shell.execute_reply.started":"2025-05-17T13:21:21.486607Z","shell.execute_reply":"2025-05-17T13:22:22.478501Z"}},"outputs":[{"name":"stdout","text":"Found existing installation: torch 2.6.0+cu124\nUninstalling torch-2.6.0+cu124:\n Successfully uninstalled torch-2.6.0+cu124\nFound existing installation: torchvision 0.21.0+cu124\nUninstalling torchvision-0.21.0+cu124:\n Successfully uninstalled torchvision-0.21.0+cu124\nFound existing installation: torchaudio 2.6.0+cu124\nUninstalling torchaudio-2.6.0+cu124:\n Successfully uninstalled torchaudio-2.6.0+cu124\nLooking in indexes: https://download.pytorch.org/whl/cpu\nCollecting torch\n Downloading https://download.pytorch.org/whl/cpu/torch-2.7.0%2Bcpu-cp311-cp311-manylinux_2_28_x86_64.whl.metadata (27 kB)\nCollecting torchvision\n Downloading https://download.pytorch.org/whl/cpu/torchvision-0.22.0%2Bcpu-cp311-cp311-manylinux_2_28_x86_64.whl.metadata (6.1 kB)\nCollecting torchaudio\n Downloading https://download.pytorch.org/whl/cpu/torchaudio-2.7.0%2Bcpu-cp311-cp311-manylinux_2_28_x86_64.whl.metadata (6.6 kB)\nRequirement already satisfied: filelock in /usr/local/lib/python3.11/dist-packages (from torch) (3.18.0)\nRequirement already satisfied: typing-extensions>=4.10.0 in /usr/local/lib/python3.11/dist-packages (from torch) (4.13.2)\nCollecting sympy>=1.13.3 (from torch)\n Downloading https://download.pytorch.org/whl/sympy-1.13.3-py3-none-any.whl.metadata (12 kB)\nRequirement already satisfied: networkx in /usr/local/lib/python3.11/dist-packages (from torch) (3.4.2)\nRequirement already satisfied: jinja2 in /usr/local/lib/python3.11/dist-packages (from torch) (3.1.6)\nRequirement already satisfied: fsspec in /usr/local/lib/python3.11/dist-packages (from torch) (2025.3.2)\nRequirement already satisfied: numpy in /usr/local/lib/python3.11/dist-packages (from torchvision) (1.26.4)\nRequirement already satisfied: pillow!=8.3.*,>=5.3.0 in /usr/local/lib/python3.11/dist-packages (from torchvision) (11.1.0)\nRequirement already satisfied: mpmath<1.4,>=1.1.0 in /usr/local/lib/python3.11/dist-packages (from sympy>=1.13.3->torch) (1.3.0)\nRequirement already satisfied: MarkupSafe>=2.0 in /usr/local/lib/python3.11/dist-packages (from jinja2->torch) (3.0.2)\nRequirement already satisfied: mkl_fft in /usr/local/lib/python3.11/dist-packages (from numpy->torchvision) (1.3.8)\nRequirement already satisfied: mkl_random in /usr/local/lib/python3.11/dist-packages (from numpy->torchvision) (1.2.4)\nRequirement already satisfied: mkl_umath in /usr/local/lib/python3.11/dist-packages (from numpy->torchvision) (0.1.1)\nRequirement already satisfied: mkl in /usr/local/lib/python3.11/dist-packages (from numpy->torchvision) (2025.1.0)\nRequirement already satisfied: tbb4py in /usr/local/lib/python3.11/dist-packages (from numpy->torchvision) (2022.1.0)\nRequirement already satisfied: mkl-service in /usr/local/lib/python3.11/dist-packages (from numpy->torchvision) (2.4.1)\nRequirement already satisfied: intel-openmp<2026,>=2024 in /usr/local/lib/python3.11/dist-packages (from mkl->numpy->torchvision) (2024.2.0)\nRequirement already satisfied: tbb==2022.* in /usr/local/lib/python3.11/dist-packages (from mkl->numpy->torchvision) (2022.1.0)\nRequirement already satisfied: tcmlib==1.* in /usr/local/lib/python3.11/dist-packages (from tbb==2022.*->mkl->numpy->torchvision) (1.3.0)\nRequirement already satisfied: intel-cmplr-lib-rt in /usr/local/lib/python3.11/dist-packages (from mkl_umath->numpy->torchvision) (2024.2.0)\nRequirement already satisfied: intel-cmplr-lib-ur==2024.2.0 in /usr/local/lib/python3.11/dist-packages (from intel-openmp<2026,>=2024->mkl->numpy->torchvision) (2024.2.0)\nDownloading https://download.pytorch.org/whl/cpu/torch-2.7.0%2Bcpu-cp311-cp311-manylinux_2_28_x86_64.whl (176.0 MB)\n\u001b[2K \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m176.0/176.0 MB\u001b[0m \u001b[31m10.1 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m:00:01\u001b[0m00:01\u001b[0m\n\u001b[?25hDownloading https://download.pytorch.org/whl/cpu/torchvision-0.22.0%2Bcpu-cp311-cp311-manylinux_2_28_x86_64.whl (2.0 MB)\n\u001b[2K \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m2.0/2.0 MB\u001b[0m \u001b[31m66.7 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n\u001b[?25hDownloading https://download.pytorch.org/whl/cpu/torchaudio-2.7.0%2Bcpu-cp311-cp311-manylinux_2_28_x86_64.whl (1.8 MB)\n\u001b[2K \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m1.8/1.8 MB\u001b[0m \u001b[31m61.8 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n\u001b[?25hDownloading https://download.pytorch.org/whl/sympy-1.13.3-py3-none-any.whl (6.2 MB)\n\u001b[2K \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m6.2/6.2 MB\u001b[0m \u001b[31m94.8 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0mta \u001b[36m0:00:01\u001b[0m\n\u001b[?25hInstalling collected packages: sympy, torch, torchaudio, torchvision\n Attempting uninstall: sympy\n Found existing installation: sympy 1.13.1\n Uninstalling sympy-1.13.1:\n Successfully uninstalled sympy-1.13.1\n\u001b[31mERROR: pip's dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts.\nfastai 2.7.19 requires torch<2.7,>=1.10, but you have torch 2.7.0+cpu which is incompatible.\u001b[0m\u001b[31m\n\u001b[0mSuccessfully installed sympy-1.13.3 torch-2.7.0+cpu torchaudio-2.7.0+cpu torchvision-0.22.0+cpu\n","output_type":"stream"}],"execution_count":9},{"cell_type":"code","source":"import os\nimport pandas as pd\nimport numpy as np\nimport torch\nimport torchaudio\nfrom torch.utils.data import Dataset, DataLoader\nfrom sklearn.preprocessing import MultiLabelBinarizer\nfrom sklearn.metrics import roc_auc_score\nimport torch.nn as nn\nimport torch.optim as optim\nfrom tqdm import tqdm\n\n\n# === CONFIG ===\nclass CFG:\n sr = 32000\n duration = 5 # seconds\n n_mels = 128\n batch_size = 16\n epochs = 10\n lr = 1e-3\n audio_dir = '/kaggle/input/birdclef-2025/train_audio/'\n train_csv = '/kaggle/input/birdclef-2025/train.csv'\n model_path = 'best_model.pth'\n num_classes = None\n use_cuda = torch.cuda.is_available()\n device = 'cuda' if use_cuda else 'cpu'\n mixup_alpha = 0.5\n\n# === AUGMENTATION UTILS ===\ndef add_noise(waveform, noise_level=0.005):\n return waveform + noise_level * torch.randn_like(waveform)\n\ndef time_stretch(waveform, rate=1.1):\n return torchaudio.functional.phase_vocoder(waveform, rate=rate, phase_advance=torch.linspace(0, np.pi * rate, waveform.shape[1]))\n\n# === DATASET ===\nclass BirdDataset(Dataset):\n def __init__(self, df, mlb, cfg):\n self.df = df\n self.cfg = cfg\n self.mlb = mlb\n\n def __len__(self):\n return len(self.df)\n\n def __getitem__(self, idx):\n row = self.df.iloc[idx]\n audio_path = os.path.join(self.cfg.audio_dir, row['filename'])\n\n waveform, sr = torchaudio.load(audio_path)\n waveform = torchaudio.transforms.Resample(orig_freq=sr, new_freq=self.cfg.sr)(waveform)\n\n # Pad or truncate\n target_len = self.cfg.sr * self.cfg.duration\n waveform = waveform[:, :target_len]\n pad_len = target_len - waveform.shape[1]\n if pad_len > 0:\n waveform = torch.nn.functional.pad(waveform, (0, pad_len))\n\n # Original mel\n mel_transform = torchaudio.transforms.MelSpectrogram(sample_rate=self.cfg.sr, n_mels=self.cfg.n_mels)\n original = mel_transform(waveform)\n\n # Noise-augmented mel\n noise_waveform = add_noise(waveform)\n noise = mel_transform(noise_waveform)\n\n # Time-stretched mel\n try:\n stretched_waveform = time_stretch(waveform, rate=1.1)\n stretch = mel_transform(stretched_waveform)\n except:\n stretch = original.clone()\n\n stacked = torch.stack([original, noise, stretch]) # Shape: [3, 128, T]\n\n # Multi-label binary target\n labels = row['primary_label'].split(';') # Assuming multi-label is semi-colon separated\n y = self.mlb.transform([labels])[0]\n return stacked, torch.tensor(y, dtype=torch.float32)\n\n# === MODEL ===\nclass BirdEffNet(nn.Module):\n def __init__(self, num_classes):\n super().__init__()\n self.backbone = create_model('tf_efficientnet_b0', pretrained=True, in_chans=3)\n self.backbone.classifier = nn.Identity()\n self.drop = nn.Dropout(0.3)\n self.fc = nn.Linear(1280, num_classes)\n\n def forward(self, x):\n x = self.backbone(x)\n x = self.drop(x)\n return self.fc(x)\n\n# === TRAINING ===\ndef run_training(cfg):\n df = pd.read_csv(cfg.train_csv)\n df = df[df['primary_label'].notnull() & df['filename'].notnull()]\n\n # Multi-label binarizer\n df['primary_label'] = df['primary_label'].astype(str)\n all_labels = list(set(\";\".join(df['primary_label']).split(\";\")))\n mlb = MultiLabelBinarizer(classes=sorted(all_labels))\n mlb.fit([all_labels])\n cfg.num_classes = len(mlb.classes_)\n\n dataset = BirdDataset(df, mlb, cfg)\n dataloader = DataLoader(dataset, batch_size=cfg.batch_size, shuffle=True, num_workers=2)\n\n model = BirdEffNet(cfg.num_classes).to(cfg.device)\n criterion = nn.BCEWithLogitsLoss()\n optimizer = optim.Adam(model.parameters(), lr=cfg.lr)\n\n best_loss = float('inf')\n for epoch in range(cfg.epochs):\n model.train()\n running_loss = 0\n preds, targets = [], []\n for x, y in tqdm(dataloader, desc=f\"Epoch {epoch+1}/{cfg.epochs}\"):\n x, y = x.to(cfg.device), y.to(cfg.device)\n outputs = model(x)\n loss = criterion(outputs, y)\n optimizer.zero_grad()\n loss.backward()\n optimizer.step()\n running_loss += loss.item()\n\n preds.append(torch.sigmoid(outputs).detach().cpu().numpy())\n targets.append(y.cpu().numpy())\n\n avg_loss = running_loss / len(dataloader)\n print(f\"Epoch {epoch+1} Loss: {avg_loss:.4f}\")\n\n # AUC metric\n preds = np.vstack(preds)\n targets = np.vstack(targets)\n auc = roc_auc_score(targets, preds, average='macro')\n print(f\"Epoch {epoch+1} AUC: {auc:.4f}\")\n\n if avg_loss < best_loss:\n best_loss = avg_loss\n torch.save(model.state_dict(), cfg.model_path)\n print(\"✅ Model saved.\")\n\n return model, mlb\n\n# === RUN IT ===\ncfg = CFG()\nmodel, label_binarizer = run_training(cfg)\n\n ","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2025-05-17T14:40:34.691032Z","iopub.execute_input":"2025-05-17T14:40:34.691568Z","iopub.status.idle":"2025-05-17T14:40:34.841586Z","shell.execute_reply.started":"2025-05-17T14:40:34.691547Z","shell.execute_reply":"2025-05-17T14:40:34.840597Z"}},"outputs":[{"traceback":["\u001b[0;31m---------------------------------------------------------------------------\u001b[0m","\u001b[0;31mNameError\u001b[0m Traceback (most recent call last)","\u001b[0;32m/tmp/ipykernel_35/3219974542.py\u001b[0m in \u001b[0;36m\u001b[0;34m()\u001b[0m\n\u001b[1;32m 149\u001b[0m \u001b[0;31m# === RUN IT ===\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 150\u001b[0m \u001b[0mcfg\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mCFG\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 151\u001b[0;31m \u001b[0mmodel\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mlabel_binarizer\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mrun_training\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mcfg\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 152\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 153\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n","\u001b[0;32m/tmp/ipykernel_35/3219974542.py\u001b[0m in \u001b[0;36mrun_training\u001b[0;34m(cfg)\u001b[0m\n\u001b[1;32m 110\u001b[0m \u001b[0mdataloader\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mDataLoader\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mdataset\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mbatch_size\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mcfg\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mbatch_size\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mshuffle\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;32mTrue\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mnum_workers\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;36m2\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 111\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 112\u001b[0;31m \u001b[0mmodel\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mBirdEffNet\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mcfg\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mnum_classes\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mto\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mcfg\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mdevice\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 113\u001b[0m \u001b[0mcriterion\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mnn\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mBCEWithLogitsLoss\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 114\u001b[0m \u001b[0moptimizer\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0moptim\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mAdam\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mmodel\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mparameters\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mlr\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mcfg\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mlr\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n","\u001b[0;32m/tmp/ipykernel_35/3219974542.py\u001b[0m in \u001b[0;36m__init__\u001b[0;34m(self, num_classes)\u001b[0m\n\u001b[1;32m 85\u001b[0m \u001b[0;32mdef\u001b[0m \u001b[0m__init__\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mnum_classes\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 86\u001b[0m \u001b[0msuper\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m__init__\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 87\u001b[0;31m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mbackbone\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mcreate_model\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m'tf_efficientnet_b0'\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mpretrained\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;32mTrue\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0min_chans\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;36m3\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 88\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mbackbone\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mclassifier\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mnn\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mIdentity\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 89\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mdrop\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mnn\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mDropout\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;36m0.3\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n","\u001b[0;31mNameError\u001b[0m: name 'create_model' is not defined"],"ename":"NameError","evalue":"name 'create_model' is not defined","output_type":"error"}],"execution_count":21},{"cell_type":"code","source":"class CFG:\n test_soundscapes = '/kaggle/input/birdclef-2025/test_soundscapes' \n submission_csv = '/kaggle/input/birdclef-2025/sample_submission.csv' \n taxonomy_csv = '/kaggle/input/birdclef-2025/taxonomy.csv' \n FS = 32000 \n WINDOW_SIZE = 5 \n \n N_FFT = 1024 \n HOP_LENGTH = 512\n N_MELS = 128\n FMIN = 50 \n FMAX = 14000 \n TARGET_SHAPE = (256, 256) \n model_name = 'regnety_008' \n in_channels = 1 \n device = 'cpu' \n use_tta = False \n tta_count = 3 \n threshold = 0.7 \n \n \n use_specific_folds = False \n folds = [0, 1] \n debug = False \n debug_count = 5 \n\n if debug:\n test_soundscapes = '/kaggle/input/birdclef-2025/train_soundscapes' ","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2025-05-17T11:41:49.191504Z","iopub.execute_input":"2025-05-17T11:41:49.191769Z","iopub.status.idle":"2025-05-17T11:41:49.19652Z","shell.execute_reply.started":"2025-05-17T11:41:49.19175Z","shell.execute_reply":"2025-05-17T11:41:49.195865Z"}},"outputs":[],"execution_count":22},{"cell_type":"code","source":"class BirdCLEF2025Pipeline:\n\n class BirdCLEFModel(nn.Module):\n def __init__(self, cfg, num_classes):\n super().__init__()\n self.cfg = cfg\n self.backbone = timm.create_model(\n cfg.model_name,\n pretrained=False, \n in_chans=cfg.in_channels,\n drop_rate=0.0, \n drop_path_rate=0.0\n )\n \n if 'efficientnet' in cfg.model_name:\n backbone_out = self.backbone.classifier.in_features\n self.backbone.classifier = nn.Identity()\n elif 'resnet' in cfg.model_name:\n backbone_out = self.backbone.fc.in_features\n self.backbone.fc = nn.Identity()\n else:\n backbone_out = self.backbone.get_classifier().in_features\n self.backbone.reset_classifier(0, '')\n \n self.pooling = nn.AdaptiveAvgPool2d(1) \n self.feat_dim = backbone_out\n self.classifier = nn.Linear(backbone_out, num_classes) \n \n def forward(self, x):\n features = self.backbone(x) \n if isinstance(features, dict):\n features = features['features']\n if len(features.shape) == 4:\n features = self.pooling(features)\n features = features.view(features.size(0), -1)\n logits = self.classifier(features) \n return logits\n def __init__(self, cfg):\n self.cfg = cfg\n self.taxonomy_df = None\n self.species_ids = []\n self.models = []\n self._load_taxonomy() \n\n def _load_taxonomy(self):\n self.taxonomy_df = pd.read_csv(self.cfg.taxonomy_csv)\n self.species_ids = self.taxonomy_df['primary_label'].tolist() \n def audio2melspec(self, audio_data):\n if np.isnan(audio_data).any():\n mean_signal = np.nanmean(audio_data)\n audio_data = np.nan_to_num(audio_data, nan=mean_signal)\n \n mel_spec = librosa.feature.melspectrogram(\n y=audio_data,\n sr=self.cfg.FS,\n n_fft=self.cfg.N_FFT,\n hop_length=self.cfg.HOP_LENGTH,\n n_mels=self.cfg.N_MELS,\n fmin=self.cfg.FMIN,\n fmax=self.cfg.FMAX,\n power=2.0\n )\n mel_spec_db = librosa.power_to_db(mel_spec, ref=np.max)\n mel_spec_norm = (mel_spec_db - mel_spec_db.min()) / (mel_spec_db.max() - mel_spec_db.min() + 1e-8)\n return mel_spec_norm\n def process_audio_segment(self, audio_data):\n \n if len(audio_data) < self.cfg.FS * self.cfg.WINDOW_SIZE:\n audio_data = np.pad(\n audio_data,\n (0, self.cfg.FS * self.cfg.WINDOW_SIZE - len(audio_data)),\n mode='constant'\n )\n \n mel_spec = self.audio2melspec(audio_data) \n if mel_spec.shape != self.cfg.TARGET_SHAPE:\n mel_spec = cv2.resize(mel_spec, self.cfg.TARGET_SHAPE, interpolation=cv2.INTER_LINEAR)\n return mel_spec.astype(np.float32)\n def find_model_files(self):\n model_files = []\n model_dir = Path(self.cfg.model_path)\n for path in model_dir.glob('**/*.pth'):\n model_files.append(str(path))\n return model_files\n def load_models(self):\n #self.models = []\n #model_files = self.find_model_files() \n #if not model_files:\n # print(self.cfg.model_path)\n # return self.models\n\n #print(len(model_files))\n \n \n if self.cfg.use_specific_folds:\n filtered_files = []\n for fold in self.cfg.folds:\n fold_files = [f for f in model_files if f\"fold{fold}\" in f]\n filtered_files.extend(fold_files)\n model_files = filtered_files\n for model_path in model_files:\n try:\n print(f\": {model_path}\")\n checkpoint = torch.load(model_path, map_location=torch.device(self.cfg.device))\n model = self.BirdCLEFModel(self.cfg, len(self.species_ids))\n model.load_state_dict(checkpoint['model_state_dict'])\n model = model.to(self.cfg.device)\n model.eval() # 推論モード\n self.models.append(model)\n except Exception as e:\n print(e)\n \n return self.models\n def apply_tta(self, spec, tta_idx):\n if tta_idx == 0:\n return spec\n elif tta_idx == 1:\n return np.flip(spec, axis=1)\n elif tta_idx == 2:\n return np.flip(spec, axis=0)\n else:\n return spec\n def predict_on_spectrogram(self, audio_path):\n predictions = []\n row_ids = []\n soundscape_id = Path(audio_path).stem\n \n try:\n audio_data, _ = librosa.load(audio_path, sr=self.cfg.FS)\n total_segments = int(len(audio_data) / (self.cfg.FS * self.cfg.WINDOW_SIZE))\n \n for segment_idx in range(total_segments):\n start_sample = segment_idx * self.cfg.FS * self.cfg.WINDOW_SIZE\n end_sample = start_sample + self.cfg.FS * self.cfg.WINDOW_SIZE\n segment_audio = audio_data[start_sample:end_sample]\n \n end_time_sec = (segment_idx + 1) * self.cfg.WINDOW_SIZE\n row_id = f\"{soundscape_id}_{end_time_sec}\"\n row_ids.append(row_id)\n if self.cfg.use_tta:\n all_preds = []\n for tta_idx in range(self.cfg.tta_count):\n mel_spec = self.process_audio_segment(segment_audio)\n mel_spec = self.apply_tta(mel_spec, tta_idx)\n mel_spec_tensor = torch.tensor(mel_spec, dtype=torch.float32).unsqueeze(0).unsqueeze(0)\n mel_spec_tensor = mel_spec_tensor.to(self.cfg.device)\n\n if len(self.models) == 1:\n with torch.no_grad():\n outputs = self.models[0](mel_spec_tensor)\n probs = torch.sigmoid(outputs).cpu().numpy().squeeze()\n all_preds.append(probs)\n else:\n segment_preds = []\n for model in self.models:\n with torch.no_grad():\n outputs = model(mel_spec_tensor)\n probs = torch.sigmoid(outputs).cpu().numpy().squeeze()\n segment_preds.append(probs)\n avg_preds = np.mean(segment_preds, axis=0)\n all_preds.append(avg_preds)\n final_preds = np.mean(all_preds, axis=0)\n else:\n mel_spec = self.process_audio_segment(segment_audio)\n mel_spec_tensor = torch.tensor(mel_spec, dtype=torch.float32).unsqueeze(0).unsqueeze(0)\n mel_spec_tensor = mel_spec_tensor.to(self.cfg.device)\n \n if len(self.models) == 1:\n with torch.no_grad():\n outputs = self.models[0](mel_spec_tensor)\n final_preds = torch.sigmoid(outputs).cpu().numpy().squeeze()\n else:\n segment_preds = []\n for model in self.models:\n with torch.no_grad():\n outputs = model(mel_spec_tensor)\n probs = torch.sigmoid(outputs).cpu().numpy().squeeze()\n segment_preds.append(probs)\n final_preds = np.mean(segment_preds, axis=0)\n \n predictions.append(final_preds)\n except Exception as e:\n print(f\"{audio_path}:{e}\")\n \n return row_ids, predictions\n def run_inference(self):\n test_files = list(Path(self.cfg.test_soundscapes).glob('*.ogg')) \n if self.cfg.debug:\n print(f\"{self.cfg.debug_count}\")\n test_files = test_files[:self.cfg.debug_count]\n print(f\"{len(test_files)}\")\n\n all_row_ids = []\n all_predictions = []\n\n for audio_path in tqdm(test_files):\n row_ids, predictions = self.predict_on_spectrogram(str(audio_path))\n all_row_ids.extend(row_ids)\n all_predictions.extend(predictions)\n \n return all_row_ids, all_predictions\n def create_submission(self, row_ids, predictions):\n submission_dict = {'row_id': row_ids}\n for i, species in enumerate(self.species_ids):\n submission_dict[species] = [pred[i] for pred in predictions]\n\n submission_df = pd.DataFrame(submission_dict)\n submission_df.set_index('row_id', inplace=True)\n\n sample_sub = pd.read_csv(self.cfg.submission_csv, index_col='row_id')\n missing_cols = set(sample_sub.columns) - set(submission_df.columns)\n if missing_cols:\n print(f\": {len(missing_cols)}\")\n for col in missing_cols:\n submission_df[col] = 0.0\n\n submission_df = submission_df[sample_sub.columns] \n submission_df = submission_df.reset_index()\n \n return submission_df\n def smooth_submission(self, submission_path):\n sub = pd.read_csv(submission_path)\n cols = sub.columns[1:]\n groups = sub['row_id'].str.rsplit('_', n=1).str[0].values\n unique_groups = np.unique(groups)\n \n for group in unique_groups:\n idx = np.where(groups == group)[0]\n sub_group = sub.iloc[idx].copy()\n predictions = sub_group[cols].values\n new_predictions = predictions.copy()\n \n if predictions.shape[0] > 1:\n new_predictions[0] = (predictions[0] * 0.8) + (predictions[1] * 0.2)\n new_predictions[-1] = (predictions[-1] * 0.8) + (predictions[-2] * 0.2)\n for i in range(1, predictions.shape[0]-1):\n new_predictions[i] = (predictions[i-1] * 0.2) + (predictions[i] * 0.6) + (predictions[i+1] * 0.2)\n sub.iloc[idx, 1:] = new_predictions\n sub.to_csv(submission_path, index=False)\n print(f\"{submission_path}\")\n def run(self):\n start_time = time.time()\n print(\"BirdCLEF-2025 \")\n print(f\": {self.cfg.use_tta} (: {self.cfg.tta_count if self.cfg.use_tta else 0})\")\n \n self.load_models()\n if not self.models:\n return\n row_ids, predictions = self.run_inference()\n submission_df = self.create_submission(row_ids, predictions)\n \n submission_path = 'submission.csv'\n submission_df.to_csv(submission_path, index=False)\n print( {submission_path})\n \n self.smooth_submission(submission_path)\n \n end_time = time.time()\n print(f\": {(end_time - start_time) / 60:.2f}\")","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2025-05-17T11:52:20.165966Z","iopub.execute_input":"2025-05-17T11:52:20.16644Z","iopub.status.idle":"2025-05-17T11:52:20.192653Z","shell.execute_reply.started":"2025-05-17T11:52:20.166417Z","shell.execute_reply":"2025-05-17T11:52:20.191851Z"}},"outputs":[],"execution_count":28},{"cell_type":"code","source":"if __name__ == \"__main__\":\n cfg = CFG()\n print(f\"Using device: {cfg.device}\")\n pipeline = BirdCLEF2025Pipeline(cfg)\n pipeline.run()","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2025-05-17T11:52:26.594429Z","iopub.execute_input":"2025-05-17T11:52:26.594949Z","iopub.status.idle":"2025-05-17T11:52:26.633323Z","shell.execute_reply.started":"2025-05-17T11:52:26.594904Z","shell.execute_reply":"2025-05-17T11:52:26.632441Z"}},"outputs":[{"name":"stdout","text":"Using device: cpu\nBirdCLEF-2025 \n: False (: 0)\n","output_type":"stream"},{"traceback":["\u001b[0;31m---------------------------------------------------------------------------\u001b[0m","\u001b[0;31mUnboundLocalError\u001b[0m Traceback (most recent call last)","\u001b[0;32m/tmp/ipykernel_35/2611710126.py\u001b[0m in \u001b[0;36m\u001b[0;34m()\u001b[0m\n\u001b[1;32m 3\u001b[0m \u001b[0mprint\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34mf\"Using device: {cfg.device}\"\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 4\u001b[0m \u001b[0mpipeline\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mBirdCLEF2025Pipeline\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mcfg\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 5\u001b[0;31m \u001b[0mpipeline\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mrun\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m","\u001b[0;32m/tmp/ipykernel_35/3156311966.py\u001b[0m in \u001b[0;36mrun\u001b[0;34m(self)\u001b[0m\n\u001b[1;32m 244\u001b[0m \u001b[0mprint\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34mf\": {self.cfg.use_tta} (: {self.cfg.tta_count if self.cfg.use_tta else 0})\"\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 245\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 246\u001b[0;31m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mload_models\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 247\u001b[0m \u001b[0;32mif\u001b[0m \u001b[0;32mnot\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mmodels\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 248\u001b[0m \u001b[0;32mreturn\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n","\u001b[0;32m/tmp/ipykernel_35/3156311966.py\u001b[0m in \u001b[0;36mload_models\u001b[0;34m(self)\u001b[0m\n\u001b[1;32m 99\u001b[0m \u001b[0mfiltered_files\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mextend\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mfold_files\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 100\u001b[0m \u001b[0mmodel_files\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mfiltered_files\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 101\u001b[0;31m \u001b[0;32mfor\u001b[0m \u001b[0mmodel_path\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mmodel_files\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 102\u001b[0m \u001b[0;32mtry\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 103\u001b[0m \u001b[0mprint\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34mf\": {model_path}\"\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n","\u001b[0;31mUnboundLocalError\u001b[0m: cannot access local variable 'model_files' where it is not associated with a value"],"ename":"UnboundLocalError","evalue":"cannot access local variable 'model_files' where it is not associated with a value","output_type":"error"}],"execution_count":29}]}