{"metadata":{"kernelspec":{"language":"python","display_name":"Python 3","name":"python3"},"language_info":{"pygments_lexer":"ipython3","nbconvert_exporter":"python","version":"3.6.4","file_extension":".py","codemirror_mode":{"name":"ipython","version":3},"name":"python","mimetype":"text/x-python"}},"nbformat_minor":4,"nbformat":4,"cells":[{"cell_type":"markdown","source":"- These inference notebook is based on @nischaydnk's [notebook](https://www.kaggle.com/code/nischaydnk/bengali-finetuning-baseline-wav2vec2-inference). If it's helpful to you, please upvote his firstly.\n- Running this notebook you can score 0.338 on leaderboard.\n- This model is trained around 8 epochs with 1e-4 learning rate on the data of [Common Voice 13](http://www.kaggle.com/datasets/umongsain/common-voice-13-bengali-normalized) provided by @UMONG SAIN and the competition data 'valid' part.","metadata":{}},{"cell_type":"code","source":"!cp -r ../input/python-packages2 ./\n\n!tar xvfz ./python-packages2/jiwer.tgz\n!pip install ./jiwer/jiwer-2.3.0-py3-none-any.whl -f ./ --no-index\n!tar xvfz ./python-packages2/normalizer.tgz\n!pip install ./normalizer/bnunicodenormalizer-0.0.24.tar.gz -f ./ --no-index\n!tar xvfz ./python-packages2/pyctcdecode.tgz\n!pip install ./pyctcdecode/attrs-22.1.0-py2.py3-none-any.whl -f ./ --no-index --no-deps\n!pip install ./pyctcdecode/exceptiongroup-1.0.0rc9-py3-none-any.whl -f ./ --no-index --no-deps\n!pip install ./pyctcdecode/hypothesis-6.54.4-py3-none-any.whl -f ./ --no-index --no-deps\n!pip install ./pyctcdecode/numpy-1.21.6-cp37-cp37m-manylinux_2_12_x86_64.manylinux2010_x86_64.whl -f ./ --no-index --no-deps\n!pip install ./pyctcdecode/pygtrie-2.5.0.tar.gz -f ./ --no-index --no-deps\n!pip install ./pyctcdecode/sortedcontainers-2.4.0-py2.py3-none-any.whl -f ./ --no-index --no-deps\n!pip install ./pyctcdecode/pyctcdecode-0.4.0-py2.py3-none-any.whl -f ./ --no-index --no-deps\n\n!tar xvfz ./python-packages2/pypikenlm.tgz\n!pip install ./pypikenlm/pypi-kenlm-0.1.20220713.tar.gz -f ./ --no-index --no-deps","metadata":{"_kg_hide-output":true,"execution":{"iopub.status.busy":"2023-10-07T13:24:55.979151Z","iopub.execute_input":"2023-10-07T13:24:55.979557Z","iopub.status.idle":"2023-10-07T13:26:05.268441Z","shell.execute_reply.started":"2023-10-07T13:24:55.979523Z","shell.execute_reply":"2023-10-07T13:26:05.267295Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"rm -r python-packages2 jiwer normalizer pyctcdecode pypikenlm","metadata":{"execution":{"iopub.status.busy":"2023-10-07T13:26:05.270655Z","iopub.execute_input":"2023-10-07T13:26:05.271311Z","iopub.status.idle":"2023-10-07T13:26:06.26091Z","shell.execute_reply.started":"2023-10-07T13:26:05.271274Z","shell.execute_reply":"2023-10-07T13:26:06.259638Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"import typing as tp\nfrom pathlib import Path\nfrom functools import partial\nfrom dataclasses import dataclass, field\n\nimport pandas as pd\nimport pyctcdecode\nimport numpy as np\nfrom tqdm.notebook import tqdm\n\nimport librosa\n\nimport pyctcdecode\nimport kenlm\nimport torch\nfrom transformers import Wav2Vec2Processor, Wav2Vec2ProcessorWithLM, Wav2Vec2ForCTC\nfrom bnunicodenormalizer import Normalizer\n\nimport cloudpickle as cpkl","metadata":{"_uuid":"8f2839f25d086af736a60e9eeb907d3b93b6e0e5","_cell_guid":"b1076dfc-b9ad-4769-8c92-a6c4dae69d19","execution":{"iopub.status.busy":"2023-10-07T13:26:06.262627Z","iopub.execute_input":"2023-10-07T13:26:06.262985Z","iopub.status.idle":"2023-10-07T13:26:18.828408Z","shell.execute_reply.started":"2023-10-07T13:26:06.262951Z","shell.execute_reply":"2023-10-07T13:26:18.827527Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"ROOT = Path.cwd().parent\nINPUT = ROOT / \"input\"\nDATA = INPUT / \"bengaliai-speech\"\nTRAIN = DATA / \"train_mp3s\"\nTEST = DATA / \"test_mp3s\"\n\nSAMPLING_RATE = 16_000\nMODEL_PATH = INPUT / \"wave-2-vec-2-bengali-ai\"\nLM_PATH = INPUT / \"arijitx-full-model/wav2vec2-xls-r-300m-bengali/language_model\"","metadata":{"execution":{"iopub.status.busy":"2023-10-07T13:26:18.830606Z","iopub.execute_input":"2023-10-07T13:26:18.831254Z","iopub.status.idle":"2023-10-07T13:26:18.836835Z","shell.execute_reply.started":"2023-10-07T13:26:18.83122Z","shell.execute_reply":"2023-10-07T13:26:18.835882Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"### load model, processor, decoder","metadata":{}},{"cell_type":"code","source":"model = Wav2Vec2ForCTC.from_pretrained(MODEL_PATH)\nprocessor = Wav2Vec2Processor.from_pretrained(MODEL_PATH)","metadata":{"execution":{"iopub.status.busy":"2023-10-07T13:26:46.131105Z","iopub.execute_input":"2023-10-07T13:26:46.131442Z","iopub.status.idle":"2023-10-07T13:26:50.582841Z","shell.execute_reply.started":"2023-10-07T13:26:46.131415Z","shell.execute_reply":"2023-10-07T13:26:50.581908Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"vocab_dict = processor.tokenizer.get_vocab()\nsorted_vocab_dict = {k: v for k, v in sorted(vocab_dict.items(), key=lambda item: item[1])}\n\ndecoder = pyctcdecode.build_ctcdecoder(\n    list(sorted_vocab_dict.keys()),\n    str(LM_PATH / \"5gram.bin\"),\n)","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"processor_with_lm = Wav2Vec2ProcessorWithLM(\n    feature_extractor=processor.feature_extractor,\n    tokenizer=processor.tokenizer,\n    decoder=decoder\n)","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"## prepare dataloader","metadata":{}},{"cell_type":"code","source":"class BengaliSRTestDataset(torch.utils.data.Dataset):\n    \n    def __init__(\n        self,\n        audio_paths: list[str],\n        sampling_rate: int\n    ):\n        self.audio_paths = audio_paths\n        self.sampling_rate = sampling_rate\n        \n    def __len__(self,):\n        return len(self.audio_paths)\n    \n    def __getitem__(self, index: int):\n        audio_path = self.audio_paths[index]\n        sr = self.sampling_rate\n        w = librosa.load(audio_path, sr=sr, mono=False)[0]\n        \n        return w","metadata":{"execution":{"iopub.status.busy":"2023-10-07T01:31:00.828035Z","iopub.execute_input":"2023-10-07T01:31:00.828612Z","iopub.status.idle":"2023-10-07T01:31:00.83929Z","shell.execute_reply.started":"2023-10-07T01:31:00.82858Z","shell.execute_reply":"2023-10-07T01:31:00.838288Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"test = pd.read_csv(DATA / \"sample_submission.csv\", dtype={\"id\": str})\nprint(test.head())","metadata":{"execution":{"iopub.status.busy":"2023-10-07T01:31:00.841883Z","iopub.execute_input":"2023-10-07T01:31:00.842474Z","iopub.status.idle":"2023-10-07T01:31:00.864765Z","shell.execute_reply.started":"2023-10-07T01:31:00.842444Z","shell.execute_reply":"2023-10-07T01:31:00.863705Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"test_audio_paths = [str(TEST / f\"{aid}.mp3\") for aid in test[\"id\"].values]","metadata":{"execution":{"iopub.status.busy":"2023-10-07T01:31:00.865953Z","iopub.execute_input":"2023-10-07T01:31:00.866498Z","iopub.status.idle":"2023-10-07T01:31:00.871828Z","shell.execute_reply.started":"2023-10-07T01:31:00.866467Z","shell.execute_reply":"2023-10-07T01:31:00.870746Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"test_dataset = BengaliSRTestDataset(\n    test_audio_paths, SAMPLING_RATE\n)\n\ncollate_func = partial(\n    processor_with_lm.feature_extractor,\n    return_tensors=\"pt\", sampling_rate=SAMPLING_RATE,\n    padding=True,\n)\n\ntest_loader = torch.utils.data.DataLoader(\n    test_dataset, batch_size=16, shuffle=False,\n    num_workers=2, collate_fn=collate_func, drop_last=False,\n    pin_memory=True,\n)","metadata":{"execution":{"iopub.status.busy":"2023-10-07T01:31:00.873231Z","iopub.execute_input":"2023-10-07T01:31:00.874082Z","iopub.status.idle":"2023-10-07T01:31:00.882947Z","shell.execute_reply.started":"2023-10-07T01:31:00.87405Z","shell.execute_reply":"2023-10-07T01:31:00.881975Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"## Inference","metadata":{}},{"cell_type":"code","source":"if not torch.cuda.is_available():\n    device = torch.device(\"cpu\")\nelse:\n    device = torch.device(\"cuda\")\nprint(device)","metadata":{"execution":{"iopub.status.busy":"2023-10-07T01:31:00.884359Z","iopub.execute_input":"2023-10-07T01:31:00.884999Z","iopub.status.idle":"2023-10-07T01:31:00.924079Z","shell.execute_reply.started":"2023-10-07T01:31:00.884968Z","shell.execute_reply":"2023-10-07T01:31:00.923137Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"model = model.to(device)\nmodel = model.eval()\nmodel = model.half()","metadata":{"execution":{"iopub.status.busy":"2023-10-07T01:31:00.925445Z","iopub.execute_input":"2023-10-07T01:31:00.926361Z","iopub.status.idle":"2023-10-07T01:31:05.880582Z","shell.execute_reply.started":"2023-10-07T01:31:00.926328Z","shell.execute_reply":"2023-10-07T01:31:05.879638Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"pred_sentence_list = []\n\nwith torch.no_grad():\n    for batch in tqdm(test_loader):\n        x = batch[\"input_values\"]\n        x = x.to(device, non_blocking=True)\n        with torch.cuda.amp.autocast(True):\n            y = model(x).logits\n        y = y.detach().cpu().numpy()\n        \n        for l in y:  \n            sentence = processor_with_lm.decode(l, beam_width=512).text\n            pred_sentence_list.append(sentence)","metadata":{"execution":{"iopub.status.busy":"2023-10-07T01:31:05.881862Z","iopub.execute_input":"2023-10-07T01:31:05.882681Z","iopub.status.idle":"2023-10-07T01:31:20.100672Z","shell.execute_reply.started":"2023-10-07T01:31:05.882649Z","shell.execute_reply":"2023-10-07T01:31:20.099533Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"## Make Submission","metadata":{}},{"cell_type":"code","source":"bnorm = Normalizer()\n\ndef postprocess(sentence):\n    period_set = set([\".\", \"?\", \"!\", \"।\"])\n    _words = [bnorm(word)['normalized']  for word in sentence.split()]\n    sentence = \" \".join([word for word in _words if word is not None])\n    try:\n        if sentence[-1] not in period_set:\n            sentence+=\"।\"\n    except:\n        # print(sentence)\n        sentence = \"।\"\n    return sentence","metadata":{"execution":{"iopub.status.busy":"2023-10-07T01:31:20.103974Z","iopub.execute_input":"2023-10-07T01:31:20.104769Z","iopub.status.idle":"2023-10-07T01:31:20.111288Z","shell.execute_reply.started":"2023-10-07T01:31:20.104731Z","shell.execute_reply":"2023-10-07T01:31:20.110378Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"pp_pred_sentence_list = [postprocess(s) for s in tqdm(pred_sentence_list)]","metadata":{"execution":{"iopub.status.busy":"2023-10-07T01:31:20.112453Z","iopub.execute_input":"2023-10-07T01:31:20.113252Z","iopub.status.idle":"2023-10-07T01:31:20.15577Z","shell.execute_reply.started":"2023-10-07T01:31:20.113221Z","shell.execute_reply":"2023-10-07T01:31:20.154895Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"test[\"sentence\"] = pp_pred_sentence_list\n\ntest.to_csv(\"submission.csv\", index=False)\n\ndisplay(test.head())","metadata":{"execution":{"iopub.status.busy":"2023-10-07T01:31:20.15689Z","iopub.execute_input":"2023-10-07T01:31:20.157671Z","iopub.status.idle":"2023-10-07T01:31:20.176516Z","shell.execute_reply.started":"2023-10-07T01:31:20.157639Z","shell.execute_reply":"2023-10-07T01:31:20.175586Z"},"trusted":true},"execution_count":null,"outputs":[]}]}