"
+ ]
+ },
+ "execution_count": 17,
+ "metadata": {},
+ "output_type": "execute_result"
+ }
+ ],
+ "source": [
+ "import IPython.display as ipd\n",
+ "import numpy as np\n",
+ "import random\n",
+ "\n",
+ "rand_int = random.randint(0, len(common_voice_train)-1)\n",
+ "\n",
+ "print(\"Target text:\", common_voice_train[rand_int][\"sentence\"])\n",
+ "print(\"Input array shape:\", common_voice_train[rand_int][\"audio\"][\"array\"].shape)\n",
+ "print(\"Sampling rate:\", common_voice_train[rand_int][\"audio\"][\"sampling_rate\"])\n",
+ "ipd.Audio(data=common_voice_train[rand_int][\"audio\"][\"array\"], autoplay=False, rate=16000)"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 18,
+ "id": "07cc904e",
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "# This does not prepare the input for the Transformer model.\n",
+ "# This will resample the data and convert the sentence into indices\n",
+ "# Batch here is just for one entry (row)\n",
+ "def prepare_dataset(batch):\n",
+ " audio = batch[\"audio\"]\n",
+ " \n",
+ " # batched output is \"un-batched\"\n",
+ " batch[\"input_values\"] = processor(audio[\"array\"], sampling_rate=audio[\"sampling_rate\"]).input_values[0]\n",
+ " batch[\"input_length\"] = len(batch[\"input_values\"])\n",
+ " \n",
+ " with processor.as_target_processor():\n",
+ " batch[\"labels\"] = processor(batch[\"sentence\"]).input_ids\n",
+ " return batch"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 19,
+ "id": "0a85ba00",
+ "metadata": {
+ "collapsed": true,
+ "jupyter": {
+ "outputs_hidden": true
+ }
+ },
+ "outputs": [
+ {
+ "name": "stderr",
+ "output_type": "stream",
+ "text": [
+ "Loading cached processed dataset at /workspace/.cache/huggingface/datasets/csv/default-decaf49f8e8b5be8/0.0.0/6b9057d9e23d9d8a2f05b985917a0da84d70c5dae3d22ddd8a3f22fb01c69d9e/cache-a8ad7f3bec152712.arrow\n",
+ "Loading cached processed dataset at /workspace/.cache/huggingface/datasets/csv/default-decaf49f8e8b5be8/0.0.0/6b9057d9e23d9d8a2f05b985917a0da84d70c5dae3d22ddd8a3f22fb01c69d9e/cache-5802da9af6ac9ac7.arrow\n",
+ "Loading cached processed dataset at /workspace/.cache/huggingface/datasets/csv/default-decaf49f8e8b5be8/0.0.0/6b9057d9e23d9d8a2f05b985917a0da84d70c5dae3d22ddd8a3f22fb01c69d9e/cache-418585d4baf07152.arrow\n",
+ "Loading cached processed dataset at /workspace/.cache/huggingface/datasets/csv/default-decaf49f8e8b5be8/0.0.0/6b9057d9e23d9d8a2f05b985917a0da84d70c5dae3d22ddd8a3f22fb01c69d9e/cache-c7e5028c91005615.arrow\n",
+ "Loading cached processed dataset at /workspace/.cache/huggingface/datasets/csv/default-decaf49f8e8b5be8/0.0.0/6b9057d9e23d9d8a2f05b985917a0da84d70c5dae3d22ddd8a3f22fb01c69d9e/cache-b96a6332cc5af3be.arrow\n",
+ "Loading cached processed dataset at /workspace/.cache/huggingface/datasets/csv/default-decaf49f8e8b5be8/0.0.0/6b9057d9e23d9d8a2f05b985917a0da84d70c5dae3d22ddd8a3f22fb01c69d9e/cache-6a1544fcabe8e1c5.arrow\n",
+ "Loading cached processed dataset at /workspace/.cache/huggingface/datasets/csv/default-decaf49f8e8b5be8/0.0.0/6b9057d9e23d9d8a2f05b985917a0da84d70c5dae3d22ddd8a3f22fb01c69d9e/cache-4926c7991e987d55.arrow\n",
+ "Loading cached processed dataset at /workspace/.cache/huggingface/datasets/csv/default-decaf49f8e8b5be8/0.0.0/6b9057d9e23d9d8a2f05b985917a0da84d70c5dae3d22ddd8a3f22fb01c69d9e/cache-b607d477202e12db.arrow\n",
+ "Loading cached processed dataset at /workspace/.cache/huggingface/datasets/csv/default-decaf49f8e8b5be8/0.0.0/6b9057d9e23d9d8a2f05b985917a0da84d70c5dae3d22ddd8a3f22fb01c69d9e/cache-a04833a515432724.arrow\n",
+ "Loading cached processed dataset at /workspace/.cache/huggingface/datasets/csv/default-decaf49f8e8b5be8/0.0.0/6b9057d9e23d9d8a2f05b985917a0da84d70c5dae3d22ddd8a3f22fb01c69d9e/cache-bee3e39e2b69f652.arrow\n",
+ "Loading cached processed dataset at /workspace/.cache/huggingface/datasets/csv/default-decaf49f8e8b5be8/0.0.0/6b9057d9e23d9d8a2f05b985917a0da84d70c5dae3d22ddd8a3f22fb01c69d9e/cache-ad833922c61a3f31.arrow\n",
+ "Loading cached processed dataset at /workspace/.cache/huggingface/datasets/csv/default-decaf49f8e8b5be8/0.0.0/6b9057d9e23d9d8a2f05b985917a0da84d70c5dae3d22ddd8a3f22fb01c69d9e/cache-ecf0c779c655274d.arrow\n",
+ "Loading cached processed dataset at /workspace/.cache/huggingface/datasets/csv/default-decaf49f8e8b5be8/0.0.0/6b9057d9e23d9d8a2f05b985917a0da84d70c5dae3d22ddd8a3f22fb01c69d9e/cache-2ba17aaff236f685.arrow\n",
+ "Loading cached processed dataset at /workspace/.cache/huggingface/datasets/csv/default-decaf49f8e8b5be8/0.0.0/6b9057d9e23d9d8a2f05b985917a0da84d70c5dae3d22ddd8a3f22fb01c69d9e/cache-5ab7f29ea26c63ef.arrow\n",
+ "Loading cached processed dataset at /workspace/.cache/huggingface/datasets/csv/default-decaf49f8e8b5be8/0.0.0/6b9057d9e23d9d8a2f05b985917a0da84d70c5dae3d22ddd8a3f22fb01c69d9e/cache-240348bb32bdbb06.arrow\n",
+ "Loading cached processed dataset at /workspace/.cache/huggingface/datasets/csv/default-decaf49f8e8b5be8/0.0.0/6b9057d9e23d9d8a2f05b985917a0da84d70c5dae3d22ddd8a3f22fb01c69d9e/cache-38e1126fb0ce94f8.arrow\n"
+ ]
+ }
+ ],
+ "source": [
+ "common_voice_train = common_voice_train.map(prepare_dataset, remove_columns=common_voice_train.column_names, num_proc=16)\n",
+ "common_voice_valid = common_voice_valid.map(prepare_dataset, remove_columns=common_voice_valid.column_names, num_proc=16)"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 20,
+ "id": "76d9219c",
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "# In case the dataset is too long which can lead to OOM. We should filter them out.\n",
+ "# max_input_length_in_sec = 5.0\n",
+ "# common_voice_train = common_voice_train.filter(lambda x: x < max_input_length_in_sec * processor.feature_extractor.sampling_rate, input_columns=[\"input_length\"])"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 21,
+ "id": "1d89d3df",
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "import torch\n",
+ "\n",
+ "from dataclasses import dataclass, field\n",
+ "from typing import Any, Dict, List, Optional, Union\n",
+ "\n",
+ "@dataclass\n",
+ "class DataCollatorCTCWithPadding:\n",
+ " \"\"\"\n",
+ " Data collator that will dynamically pad the inputs received.\n",
+ " Args:\n",
+ " processor (:class:`~transformers.Wav2Vec2Processor`)\n",
+ " The processor used for proccessing the data.\n",
+ " padding (:obj:`bool`, :obj:`str` or :class:`~transformers.tokenization_utils_base.PaddingStrategy`, `optional`, defaults to :obj:`True`):\n",
+ " Select a strategy to pad the returned sequences (according to the model's padding side and padding index)\n",
+ " among:\n",
+ " * :obj:`True` or :obj:`'longest'`: Pad to the longest sequence in the batch (or no padding if only a single\n",
+ " sequence if provided).\n",
+ " * :obj:`'max_length'`: Pad to a maximum length specified with the argument :obj:`max_length` or to the\n",
+ " maximum acceptable input length for the model if that argument is not provided.\n",
+ " * :obj:`False` or :obj:`'do_not_pad'` (default): No padding (i.e., can output a batch with sequences of\n",
+ " different lengths).\n",
+ " \"\"\"\n",
+ "\n",
+ " processor: Wav2Vec2Processor\n",
+ " padding: Union[bool, str] = True\n",
+ "\n",
+ " def __call__(self, features: List[Dict[str, Union[List[int], torch.Tensor]]]) -> Dict[str, torch.Tensor]:\n",
+ " # split inputs and labels since they have to be of different lenghts and need\n",
+ " # different padding methods\n",
+ " input_features = [{\"input_values\": feature[\"input_values\"]} for feature in features]\n",
+ " label_features = [{\"input_ids\": feature[\"labels\"]} for feature in features]\n",
+ "\n",
+ " batch = self.processor.pad(\n",
+ " input_features,\n",
+ " padding=self.padding,\n",
+ " return_tensors=\"pt\",\n",
+ " )\n",
+ "\n",
+ " with self.processor.as_target_processor():\n",
+ " labels_batch = self.processor.pad(\n",
+ " label_features,\n",
+ " padding=self.padding,\n",
+ " return_tensors=\"pt\",\n",
+ " )\n",
+ "\n",
+ " # replace padding with -100 to ignore loss correctly\n",
+ " labels = labels_batch[\"input_ids\"].masked_fill(labels_batch.attention_mask.ne(1), -100)\n",
+ "\n",
+ " batch[\"labels\"] = labels\n",
+ "\n",
+ " return batch"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 22,
+ "id": "d3b3ba57",
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "data_collator = DataCollatorCTCWithPadding(processor=processor, padding=True)"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 23,
+ "id": "6c33263a",
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "wer_metric = load_metric(\"wer\")\n",
+ "# cer_metric = load_metric(\"cer\")"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 24,
+ "id": "7d2e7755",
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "def compute_metrics(pred):\n",
+ " pred_logits = pred.predictions\n",
+ " pred_ids = np.argmax(pred_logits, axis=-1)\n",
+ "\n",
+ " pred.label_ids[pred.label_ids == -100] = tokenizer.pad_token_id\n",
+ "\n",
+ " pred_str = tokenizer.batch_decode(pred_ids)\n",
+ " label_str = tokenizer.batch_decode(pred.label_ids, group_tokens=False)\n",
+ " \n",
+ " wer = wer_metric.compute(predictions=pred_str, references=label_str)\n",
+ "\n",
+ " return {\"wer\": wer}"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 25,
+ "id": "8f05c6fd",
+ "metadata": {},
+ "outputs": [
+ {
+ "name": "stderr",
+ "output_type": "stream",
+ "text": [
+ "Some weights of the model checkpoint at facebook/wav2vec2-xls-r-300m were not used when initializing Wav2Vec2ForCTC: ['project_hid.bias', 'quantizer.codevectors', 'quantizer.weight_proj.weight', 'project_q.weight', 'project_hid.weight', 'quantizer.weight_proj.bias', 'project_q.bias']\n",
+ "- This IS expected if you are initializing Wav2Vec2ForCTC from the checkpoint of a model trained on another task or with another architecture (e.g. initializing a BertForSequenceClassification model from a BertForPreTraining model).\n",
+ "- This IS NOT expected if you are initializing Wav2Vec2ForCTC from the checkpoint of a model that you expect to be exactly identical (initializing a BertForSequenceClassification model from a BertForSequenceClassification model).\n",
+ "Some weights of Wav2Vec2ForCTC were not initialized from the model checkpoint at facebook/wav2vec2-xls-r-300m and are newly initialized: ['lm_head.weight', 'lm_head.bias']\n",
+ "You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.\n"
+ ]
+ }
+ ],
+ "source": [
+ "from transformers import Wav2Vec2ForCTC\n",
+ "\n",
+ "model = Wav2Vec2ForCTC.from_pretrained(\n",
+ " \"facebook/wav2vec2-xls-r-300m\", \n",
+ " attention_dropout=0.1,\n",
+ " layerdrop=0.0,\n",
+ " feat_proj_dropout=0.0,\n",
+ " mask_time_prob=0.75, \n",
+ " mask_time_length=10,\n",
+ " mask_feature_prob=0.25,\n",
+ " mask_feature_length=64,\n",
+ " ctc_loss_reduction=\"mean\",\n",
+ " pad_token_id=processor.tokenizer.pad_token_id,\n",
+ " vocab_size=len(processor.tokenizer)\n",
+ ")"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 26,
+ "id": "ab8e1dfa",
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "model.freeze_feature_encoder()"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 27,
+ "id": "b46286e7",
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "from transformers import TrainingArguments\n",
+ "\n",
+ "training_args = TrainingArguments(\n",
+ " output_dir='.',\n",
+ " group_by_length=True,\n",
+ " per_device_train_batch_size=8,\n",
+ " gradient_accumulation_steps=4,\n",
+ " evaluation_strategy=\"steps\",\n",
+ " gradient_checkpointing=True,\n",
+ " fp16=True,\n",
+ " num_train_epochs=100,\n",
+ " save_steps=400,\n",
+ " eval_steps=400,\n",
+ " logging_steps=100,\n",
+ " learning_rate=5e-5,\n",
+ " warmup_steps=1000,\n",
+ " save_total_limit=3,\n",
+ " load_best_model_at_end=True\n",
+ ")"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 29,
+ "id": "8545cd28",
+ "metadata": {},
+ "outputs": [
+ {
+ "name": "stderr",
+ "output_type": "stream",
+ "text": [
+ "Using amp half precision backend\n"
+ ]
+ }
+ ],
+ "source": [
+ "from transformers import Trainer\n",
+ "\n",
+ "trainer = Trainer(\n",
+ " model=model,\n",
+ " data_collator=data_collator,\n",
+ " args=training_args,\n",
+ " compute_metrics=compute_metrics,\n",
+ " train_dataset=common_voice_train,\n",
+ " eval_dataset=common_voice_valid,\n",
+ " tokenizer=processor.feature_extractor,\n",
+ ")"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 30,
+ "id": "f7b39403",
+ "metadata": {},
+ "outputs": [
+ {
+ "name": "stderr",
+ "output_type": "stream",
+ "text": [
+ "The following columns in the training set don't have a corresponding argument in `Wav2Vec2ForCTC.forward` and have been ignored: input_length.\n",
+ "/opt/conda/lib/python3.8/site-packages/transformers/optimization.py:306: FutureWarning: This implementation of AdamW is deprecated and will be removed in a future version. Use thePyTorch implementation torch.optim.AdamW instead, or set `no_deprecation_warning=True` to disable this warning\n",
+ " warnings.warn(\n",
+ "***** Running training *****\n",
+ " Num examples = 2353\n",
+ " Num Epochs = 100\n",
+ " Instantaneous batch size per device = 8\n",
+ " Total train batch size (w. parallel, distributed & accumulation) = 32\n",
+ " Gradient Accumulation steps = 4\n",
+ " Total optimization steps = 7300\n"
+ ]
+ },
+ {
+ "data": {
+ "text/html": [
+ "\n",
+ " \n",
+ " \n",
+ "
\n",
+ " [7300/7300 4:05:50, Epoch 99/100]\n",
+ "
\n",
+ " \n",
+ " \n",
+ " | Step | \n",
+ " Training Loss | \n",
+ " Validation Loss | \n",
+ " Wer | \n",
+ "
\n",
+ " \n",
+ " \n",
+ " \n",
+ " | 400 | \n",
+ " 5.079500 | \n",
+ " 4.412057 | \n",
+ " 1.000000 | \n",
+ "
\n",
+ " \n",
+ " | 800 | \n",
+ " 3.565800 | \n",
+ " 3.520337 | \n",
+ " 1.000000 | \n",
+ "
\n",
+ " \n",
+ " | 1200 | \n",
+ " 3.368900 | \n",
+ " 2.898376 | \n",
+ " 0.999596 | \n",
+ "
\n",
+ " \n",
+ " | 1600 | \n",
+ " 2.010000 | \n",
+ " 1.004076 | \n",
+ " 0.728814 | \n",
+ "
\n",
+ " \n",
+ " | 2000 | \n",
+ " 1.678300 | \n",
+ " 0.694064 | \n",
+ " 0.598870 | \n",
+ "
\n",
+ " \n",
+ " | 2400 | \n",
+ " 1.527000 | \n",
+ " 0.559943 | \n",
+ " 0.528249 | \n",
+ "
\n",
+ " \n",
+ " | 2800 | \n",
+ " 1.427800 | \n",
+ " 0.482676 | \n",
+ " 0.480630 | \n",
+ "
\n",
+ " \n",
+ " | 3200 | \n",
+ " 1.345800 | \n",
+ " 0.442869 | \n",
+ " 0.453188 | \n",
+ "
\n",
+ " \n",
+ " | 3600 | \n",
+ " 1.289300 | \n",
+ " 0.415609 | \n",
+ " 0.433010 | \n",
+ "
\n",
+ " \n",
+ " | 4000 | \n",
+ " 1.244100 | \n",
+ " 0.402046 | \n",
+ " 0.403955 | \n",
+ "
\n",
+ " \n",
+ " | 4400 | \n",
+ " 1.188000 | \n",
+ " 0.377747 | \n",
+ " 0.386602 | \n",
+ "
\n",
+ " \n",
+ " | 4800 | \n",
+ " 1.162800 | \n",
+ " 0.360746 | \n",
+ " 0.385795 | \n",
+ "
\n",
+ " \n",
+ " | 5200 | \n",
+ " 1.132400 | \n",
+ " 0.353398 | \n",
+ " 0.360371 | \n",
+ "
\n",
+ " \n",
+ " | 5600 | \n",
+ " 1.096900 | \n",
+ " 0.342845 | \n",
+ " 0.362389 | \n",
+ "
\n",
+ " \n",
+ " | 6000 | \n",
+ " 1.089700 | \n",
+ " 0.338743 | \n",
+ " 0.356739 | \n",
+ "
\n",
+ " \n",
+ " | 6400 | \n",
+ " 1.062500 | \n",
+ " 0.333874 | \n",
+ " 0.349879 | \n",
+ "
\n",
+ " \n",
+ " | 6800 | \n",
+ " 1.060100 | \n",
+ " 0.328815 | \n",
+ " 0.344633 | \n",
+ "
\n",
+ " \n",
+ " | 7200 | \n",
+ " 1.047400 | \n",
+ " 0.328094 | \n",
+ " 0.346247 | \n",
+ "
\n",
+ " \n",
+ "
"
+ ],
+ "text/plain": [
+ ""
+ ]
+ },
+ "metadata": {},
+ "output_type": "display_data"
+ },
+ {
+ "name": "stderr",
+ "output_type": "stream",
+ "text": [
+ "The following columns in the evaluation set don't have a corresponding argument in `Wav2Vec2ForCTC.forward` and have been ignored: input_length.\n",
+ "***** Running Evaluation *****\n",
+ " Num examples = 262\n",
+ " Batch size = 8\n",
+ "Saving model checkpoint to ./checkpoint-400\n",
+ "Configuration saved in ./checkpoint-400/config.json\n",
+ "Model weights saved in ./checkpoint-400/pytorch_model.bin\n",
+ "Configuration saved in ./checkpoint-400/preprocessor_config.json\n",
+ "The following columns in the evaluation set don't have a corresponding argument in `Wav2Vec2ForCTC.forward` and have been ignored: input_length.\n",
+ "***** Running Evaluation *****\n",
+ " Num examples = 262\n",
+ " Batch size = 8\n",
+ "Saving model checkpoint to ./checkpoint-800\n",
+ "Configuration saved in ./checkpoint-800/config.json\n",
+ "Model weights saved in ./checkpoint-800/pytorch_model.bin\n",
+ "Configuration saved in ./checkpoint-800/preprocessor_config.json\n",
+ "Deleting older checkpoint [checkpoint-3200] due to args.save_total_limit\n",
+ "The following columns in the evaluation set don't have a corresponding argument in `Wav2Vec2ForCTC.forward` and have been ignored: input_length.\n",
+ "***** Running Evaluation *****\n",
+ " Num examples = 262\n",
+ " Batch size = 8\n",
+ "Saving model checkpoint to ./checkpoint-1200\n",
+ "Configuration saved in ./checkpoint-1200/config.json\n",
+ "Model weights saved in ./checkpoint-1200/pytorch_model.bin\n",
+ "Configuration saved in ./checkpoint-1200/preprocessor_config.json\n",
+ "Deleting older checkpoint [checkpoint-3600] due to args.save_total_limit\n",
+ "The following columns in the evaluation set don't have a corresponding argument in `Wav2Vec2ForCTC.forward` and have been ignored: input_length.\n",
+ "***** Running Evaluation *****\n",
+ " Num examples = 262\n",
+ " Batch size = 8\n",
+ "Saving model checkpoint to ./checkpoint-1600\n",
+ "Configuration saved in ./checkpoint-1600/config.json\n",
+ "Model weights saved in ./checkpoint-1600/pytorch_model.bin\n",
+ "Configuration saved in ./checkpoint-1600/preprocessor_config.json\n",
+ "Deleting older checkpoint [checkpoint-400] due to args.save_total_limit\n",
+ "The following columns in the evaluation set don't have a corresponding argument in `Wav2Vec2ForCTC.forward` and have been ignored: input_length.\n",
+ "***** Running Evaluation *****\n",
+ " Num examples = 262\n",
+ " Batch size = 8\n",
+ "Saving model checkpoint to ./checkpoint-2000\n",
+ "Configuration saved in ./checkpoint-2000/config.json\n",
+ "Model weights saved in ./checkpoint-2000/pytorch_model.bin\n",
+ "Configuration saved in ./checkpoint-2000/preprocessor_config.json\n",
+ "Deleting older checkpoint [checkpoint-800] due to args.save_total_limit\n",
+ "The following columns in the evaluation set don't have a corresponding argument in `Wav2Vec2ForCTC.forward` and have been ignored: input_length.\n",
+ "***** Running Evaluation *****\n",
+ " Num examples = 262\n",
+ " Batch size = 8\n",
+ "Saving model checkpoint to ./checkpoint-2400\n",
+ "Configuration saved in ./checkpoint-2400/config.json\n",
+ "Model weights saved in ./checkpoint-2400/pytorch_model.bin\n",
+ "Configuration saved in ./checkpoint-2400/preprocessor_config.json\n",
+ "Deleting older checkpoint [checkpoint-1200] due to args.save_total_limit\n",
+ "The following columns in the evaluation set don't have a corresponding argument in `Wav2Vec2ForCTC.forward` and have been ignored: input_length.\n",
+ "***** Running Evaluation *****\n",
+ " Num examples = 262\n",
+ " Batch size = 8\n",
+ "Saving model checkpoint to ./checkpoint-2800\n",
+ "Configuration saved in ./checkpoint-2800/config.json\n",
+ "Model weights saved in ./checkpoint-2800/pytorch_model.bin\n",
+ "Configuration saved in ./checkpoint-2800/preprocessor_config.json\n",
+ "Deleting older checkpoint [checkpoint-1600] due to args.save_total_limit\n",
+ "The following columns in the evaluation set don't have a corresponding argument in `Wav2Vec2ForCTC.forward` and have been ignored: input_length.\n",
+ "***** Running Evaluation *****\n",
+ " Num examples = 262\n",
+ " Batch size = 8\n",
+ "Saving model checkpoint to ./checkpoint-3200\n",
+ "Configuration saved in ./checkpoint-3200/config.json\n",
+ "Model weights saved in ./checkpoint-3200/pytorch_model.bin\n",
+ "Configuration saved in ./checkpoint-3200/preprocessor_config.json\n",
+ "Deleting older checkpoint [checkpoint-2000] due to args.save_total_limit\n",
+ "The following columns in the evaluation set don't have a corresponding argument in `Wav2Vec2ForCTC.forward` and have been ignored: input_length.\n",
+ "***** Running Evaluation *****\n",
+ " Num examples = 262\n",
+ " Batch size = 8\n",
+ "Saving model checkpoint to ./checkpoint-3600\n",
+ "Configuration saved in ./checkpoint-3600/config.json\n",
+ "Model weights saved in ./checkpoint-3600/pytorch_model.bin\n",
+ "Configuration saved in ./checkpoint-3600/preprocessor_config.json\n",
+ "Deleting older checkpoint [checkpoint-2400] due to args.save_total_limit\n",
+ "The following columns in the evaluation set don't have a corresponding argument in `Wav2Vec2ForCTC.forward` and have been ignored: input_length.\n",
+ "***** Running Evaluation *****\n",
+ " Num examples = 262\n",
+ " Batch size = 8\n",
+ "Saving model checkpoint to ./checkpoint-4000\n",
+ "Configuration saved in ./checkpoint-4000/config.json\n",
+ "Model weights saved in ./checkpoint-4000/pytorch_model.bin\n",
+ "Configuration saved in ./checkpoint-4000/preprocessor_config.json\n",
+ "Deleting older checkpoint [checkpoint-2800] due to args.save_total_limit\n",
+ "The following columns in the evaluation set don't have a corresponding argument in `Wav2Vec2ForCTC.forward` and have been ignored: input_length.\n",
+ "***** Running Evaluation *****\n",
+ " Num examples = 262\n",
+ " Batch size = 8\n",
+ "Saving model checkpoint to ./checkpoint-4400\n",
+ "Configuration saved in ./checkpoint-4400/config.json\n",
+ "Model weights saved in ./checkpoint-4400/pytorch_model.bin\n",
+ "Configuration saved in ./checkpoint-4400/preprocessor_config.json\n",
+ "Deleting older checkpoint [checkpoint-3200] due to args.save_total_limit\n",
+ "The following columns in the evaluation set don't have a corresponding argument in `Wav2Vec2ForCTC.forward` and have been ignored: input_length.\n",
+ "***** Running Evaluation *****\n",
+ " Num examples = 262\n",
+ " Batch size = 8\n",
+ "Saving model checkpoint to ./checkpoint-4800\n",
+ "Configuration saved in ./checkpoint-4800/config.json\n",
+ "Model weights saved in ./checkpoint-4800/pytorch_model.bin\n",
+ "Configuration saved in ./checkpoint-4800/preprocessor_config.json\n",
+ "Deleting older checkpoint [checkpoint-3600] due to args.save_total_limit\n",
+ "The following columns in the evaluation set don't have a corresponding argument in `Wav2Vec2ForCTC.forward` and have been ignored: input_length.\n",
+ "***** Running Evaluation *****\n",
+ " Num examples = 262\n",
+ " Batch size = 8\n",
+ "Saving model checkpoint to ./checkpoint-5200\n",
+ "Configuration saved in ./checkpoint-5200/config.json\n",
+ "Model weights saved in ./checkpoint-5200/pytorch_model.bin\n",
+ "Configuration saved in ./checkpoint-5200/preprocessor_config.json\n",
+ "Deleting older checkpoint [checkpoint-4000] due to args.save_total_limit\n",
+ "The following columns in the evaluation set don't have a corresponding argument in `Wav2Vec2ForCTC.forward` and have been ignored: input_length.\n",
+ "***** Running Evaluation *****\n",
+ " Num examples = 262\n",
+ " Batch size = 8\n",
+ "Saving model checkpoint to ./checkpoint-5600\n",
+ "Configuration saved in ./checkpoint-5600/config.json\n",
+ "Model weights saved in ./checkpoint-5600/pytorch_model.bin\n",
+ "Configuration saved in ./checkpoint-5600/preprocessor_config.json\n",
+ "Deleting older checkpoint [checkpoint-4400] due to args.save_total_limit\n",
+ "The following columns in the evaluation set don't have a corresponding argument in `Wav2Vec2ForCTC.forward` and have been ignored: input_length.\n",
+ "***** Running Evaluation *****\n",
+ " Num examples = 262\n",
+ " Batch size = 8\n",
+ "Saving model checkpoint to ./checkpoint-6000\n",
+ "Configuration saved in ./checkpoint-6000/config.json\n",
+ "Model weights saved in ./checkpoint-6000/pytorch_model.bin\n",
+ "Configuration saved in ./checkpoint-6000/preprocessor_config.json\n",
+ "Deleting older checkpoint [checkpoint-4800] due to args.save_total_limit\n",
+ "The following columns in the evaluation set don't have a corresponding argument in `Wav2Vec2ForCTC.forward` and have been ignored: input_length.\n",
+ "***** Running Evaluation *****\n",
+ " Num examples = 262\n",
+ " Batch size = 8\n",
+ "Saving model checkpoint to ./checkpoint-6400\n",
+ "Configuration saved in ./checkpoint-6400/config.json\n",
+ "Model weights saved in ./checkpoint-6400/pytorch_model.bin\n",
+ "Configuration saved in ./checkpoint-6400/preprocessor_config.json\n",
+ "Deleting older checkpoint [checkpoint-5200] due to args.save_total_limit\n",
+ "The following columns in the evaluation set don't have a corresponding argument in `Wav2Vec2ForCTC.forward` and have been ignored: input_length.\n",
+ "***** Running Evaluation *****\n",
+ " Num examples = 262\n",
+ " Batch size = 8\n",
+ "Saving model checkpoint to ./checkpoint-6800\n",
+ "Configuration saved in ./checkpoint-6800/config.json\n",
+ "Model weights saved in ./checkpoint-6800/pytorch_model.bin\n",
+ "Configuration saved in ./checkpoint-6800/preprocessor_config.json\n",
+ "Deleting older checkpoint [checkpoint-5600] due to args.save_total_limit\n",
+ "The following columns in the evaluation set don't have a corresponding argument in `Wav2Vec2ForCTC.forward` and have been ignored: input_length.\n",
+ "***** Running Evaluation *****\n",
+ " Num examples = 262\n",
+ " Batch size = 8\n",
+ "Saving model checkpoint to ./checkpoint-7200\n",
+ "Configuration saved in ./checkpoint-7200/config.json\n",
+ "Model weights saved in ./checkpoint-7200/pytorch_model.bin\n",
+ "Configuration saved in ./checkpoint-7200/preprocessor_config.json\n",
+ "Deleting older checkpoint [checkpoint-6000] due to args.save_total_limit\n",
+ "\n",
+ "\n",
+ "Training completed. Do not forget to share your model on huggingface.co/models =)\n",
+ "\n",
+ "\n",
+ "Loading best model from ./checkpoint-7200 (score: 0.3280937969684601).\n"
+ ]
+ },
+ {
+ "data": {
+ "text/plain": [
+ "TrainOutput(global_step=7300, training_loss=2.0282830110314776, metrics={'train_runtime': 14754.0737, 'train_samples_per_second': 15.948, 'train_steps_per_second': 0.495, 'total_flos': 3.5572390287970673e+19, 'train_loss': 2.0282830110314776, 'epoch': 99.99})"
+ ]
+ },
+ "execution_count": 30,
+ "metadata": {},
+ "output_type": "execute_result"
+ }
+ ],
+ "source": [
+ "trainer.train()"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 31,
+ "id": "9f476560",
+ "metadata": {},
+ "outputs": [
+ {
+ "data": {
+ "text/plain": [
+ "1"
+ ]
+ },
+ "execution_count": 31,
+ "metadata": {},
+ "output_type": "execute_result"
+ }
+ ],
+ "source": [
+ "1"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 32,
+ "id": "cc4997d8",
+ "metadata": {},
+ "outputs": [
+ {
+ "name": "stderr",
+ "output_type": "stream",
+ "text": [
+ "tokenizer config file saved in vitouphy/xls-r-300m-km/tokenizer_config.json\n",
+ "Special tokens file saved in vitouphy/xls-r-300m-km/special_tokens_map.json\n",
+ "added tokens file saved in vitouphy/xls-r-300m-km/added_tokens.json\n",
+ "To https://huggingface.co/vitouphy/xls-r-300m-km\n",
+ " 3ef5dfc..cb4f72c main -> main\n",
+ "\n"
+ ]
+ },
+ {
+ "data": {
+ "text/plain": [
+ "'https://huggingface.co/vitouphy/xls-r-300m-km/commit/cb4f72cb420eee8ca1f44b582a9d3cfbcd258f3d'"
+ ]
+ },
+ "execution_count": 32,
+ "metadata": {},
+ "output_type": "execute_result"
+ }
+ ],
+ "source": [
+ "tokenizer.push_to_hub('vitouphy/xls-r-300m-km')"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 34,
+ "id": "186d4678",
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "kwargs = {\n",
+ " \"finetuned_from\": \"facebook/wav2vec2-xls-r-300m\",\n",
+ " \"tasks\": \"speech-recognition\",\n",
+ " \"tags\": [\"automatic-speech-recognition\", \"openslr\", \"robust-speech-event\", \"km\"],\n",
+ " \"dataset_args\": f\"Config: km, Training split: train, Eval split: validation\",\n",
+ " \"dataset\": \"openslr\",\n",
+ " \"language\": \"km\"\n",
+ "}"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 35,
+ "id": "c71021ea",
+ "metadata": {},
+ "outputs": [
+ {
+ "name": "stderr",
+ "output_type": "stream",
+ "text": [
+ "Dropping the following result as it does not have all the necessary fields:\n",
+ "{}\n"
+ ]
+ }
+ ],
+ "source": [
+ "trainer.create_model_card(**kwargs)"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 36,
+ "id": "3afcbfb5",
+ "metadata": {},
+ "outputs": [
+ {
+ "name": "stderr",
+ "output_type": "stream",
+ "text": [
+ "Configuration saved in vitouphy/xls-r-300m-km/config.json\n",
+ "Model weights saved in vitouphy/xls-r-300m-km/pytorch_model.bin\n"
+ ]
+ },
+ {
+ "data": {
+ "application/vnd.jupyter.widget-view+json": {
+ "model_id": "69dc015463b64e3c946ccfbe017d1828",
+ "version_major": 2,
+ "version_minor": 0
+ },
+ "text/plain": [
+ "Upload file pytorch_model.bin: 0%| | 3.39k/1.18G [00:00 main\n",
+ "\n"
+ ]
+ },
+ {
+ "data": {
+ "text/plain": [
+ "'https://huggingface.co/vitouphy/xls-r-300m-km/commit/8fe88762a9fca1dce5e056605465042b5700b69e'"
+ ]
+ },
+ "execution_count": 36,
+ "metadata": {},
+ "output_type": "execute_result"
+ }
+ ],
+ "source": [
+ "model.push_to_hub('vitouphy/xls-r-300m-km')"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 38,
+ "id": "fdc2a063",
+ "metadata": {},
+ "outputs": [
+ {
+ "name": "stderr",
+ "output_type": "stream",
+ "text": [
+ "Saving model checkpoint to .\n",
+ "Configuration saved in ./config.json\n",
+ "Model weights saved in ./pytorch_model.bin\n",
+ "Configuration saved in ./preprocessor_config.json\n"
+ ]
+ }
+ ],
+ "source": [
+ "trainer.save_model()"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "id": "5eb9cc6c",
+ "metadata": {},
+ "outputs": [],
+ "source": []
+ }
+ ],
+ "metadata": {
+ "kernelspec": {
+ "display_name": "Python 3 (ipykernel)",
+ "language": "python",
+ "name": "python3"
+ },
+ "language_info": {
+ "codemirror_mode": {
+ "name": "ipython",
+ "version": 3
+ },
+ "file_extension": ".py",
+ "mimetype": "text/x-python",
+ "name": "python",
+ "nbconvert_exporter": "python",
+ "pygments_lexer": "ipython3",
+ "version": "3.8.8"
+ }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 5
+}