{
"cells": [
{
"cell_type": "markdown",
"metadata": {},
"source": [
"# Classification"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"\n",
"\n",
"This tutorial is available as an IPython notebook at [Malaya/example/zeroshot-classification](https://github.com/huseinzol05/Malaya/tree/master/example/zeroshot-classification).\n",
" \n",
"
"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"\n",
"\n",
"This module trained on both standard and local (included social media) language structures, so it is save to use for both.\n",
" \n",
"
"
]
},
{
"cell_type": "code",
"execution_count": 1,
"metadata": {},
"outputs": [],
"source": [
"import os\n",
"\n",
"os.environ['CUDA_VISIBLE_DEVICES'] = ''\n",
"os.environ['TF_FORCE_GPU_ALLOW_GROWTH'] = 'true'"
]
},
{
"cell_type": "code",
"execution_count": 2,
"metadata": {},
"outputs": [
{
"name": "stderr",
"output_type": "stream",
"text": [
"/home/husein/.local/lib/python3.8/site-packages/bitsandbytes/cextension.py:34: UserWarning: The installed version of bitsandbytes was compiled without GPU support. 8-bit optimizers, 8-bit multiplication, and GPU quantization are unavailable.\n",
" warn(\"The installed version of bitsandbytes was compiled without GPU support. \"\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"/home/husein/.local/lib/python3.8/site-packages/bitsandbytes/libbitsandbytes_cpu.so: undefined symbol: cadam32bit_grad_fp32\n",
"CPU times: user 3.03 s, sys: 2.5 s, total: 5.53 s\n",
"Wall time: 2.74 s\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"/home/husein/dev/malaya/malaya/tokenizer.py:214: FutureWarning: Possible nested set at position 3397\n",
" self.tok = re.compile(r'({})'.format('|'.join(pipeline)))\n",
"/home/husein/dev/malaya/malaya/tokenizer.py:214: FutureWarning: Possible nested set at position 3927\n",
" self.tok = re.compile(r'({})'.format('|'.join(pipeline)))\n"
]
}
],
"source": [
"%%time\n",
"import malaya"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### what is zero-shot classification\n",
"\n",
"Commonly we supervised a machine learning on specific labels, negative / positive for sentiment, anger / happy / sadness for emotion and etc. The model cannot give an output if we want to know how much percentage of 'jealous' in emotion analysis model because supported labels are only {anger, happy, sadness}. Imagine, for example, trying to identify a text without ever having seen one 'jealous' label before, impossible. **So, zero-shot trying to solve this problem.**\n",
"\n",
"zero-shot learning refers to the process by which a machine learns how to recognize objects (image, text, any features) without any labeled training data to help in the classification.\n",
"\n",
"[Yin et al. (2019)](https://arxiv.org/abs/1909.00161) stated in his paper, any pretrained language model finetuned on text similarity actually can acted as an out-of-the-box zero-shot text classifier."
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"So, we are going to use transformer models from `malaya.similarity.semantic.huggingface` with a little tweaks."
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### List available HuggingFace models"
]
},
{
"cell_type": "code",
"execution_count": 3,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"{'mesolitica/finetune-mnli-t5-super-tiny-standard-bahasa-cased': {'Size (MB)': 50.7,\n",
" 'macro precision': 0.74562,\n",
" 'macro recall': 0.74574,\n",
" 'macro f1-score': 0.74501},\n",
" 'mesolitica/finetune-mnli-t5-tiny-standard-bahasa-cased': {'Size (MB)': 139,\n",
" 'macro precision': 0.76584,\n",
" 'macro recall': 0.76565,\n",
" 'macro f1-score': 0.76542},\n",
" 'mesolitica/finetune-mnli-t5-small-standard-bahasa-cased': {'Size (MB)': 242,\n",
" 'macro precision': 0.78067,\n",
" 'macro recall': 0.78063,\n",
" 'macro f1-score': 0.7801},\n",
" 'mesolitica/finetune-mnli-t5-base-standard-bahasa-cased': {'Size (MB)': 892,\n",
" 'macro precision': 0.78903,\n",
" 'macro recall': 0.79064,\n",
" 'macro f1-score': 0.78918}}"
]
},
"execution_count": 3,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"malaya.zero_shot.classification.available_huggingface"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Load HuggingFace model\n",
"\n",
"```python\n",
"def huggingface(\n",
" model: str = 'mesolitica/finetune-mnli-t5-small-standard-bahasa-cased',\n",
" force_check: bool = True,\n",
" **kwargs,\n",
"):\n",
" \"\"\"\n",
" Load HuggingFace model to zeroshot text classification.\n",
"\n",
" Parameters\n",
" ----------\n",
" model: str, optional (default='mesolitica/finetune-mnli-t5-small-standard-bahasa-cased')\n",
" Check available models at `malaya.zero_shot.classification.available_huggingface()`.\n",
" force_check: bool, optional (default=True)\n",
" Force check model one of malaya model.\n",
" Set to False if you have your own huggingface model.\n",
"\n",
" Returns\n",
" -------\n",
" result: malaya.torch_model.huggingface.ZeroShotClassification\n",
" \"\"\"\n",
"```"
]
},
{
"cell_type": "code",
"execution_count": 4,
"metadata": {},
"outputs": [
{
"name": "stderr",
"output_type": "stream",
"text": [
"Loading the tokenizer from the `special_tokens_map.json` and the `added_tokens.json` will be removed in `transformers 5`, it is kept for forward compatibility, but it is recommended to update your `tokenizer_config.json` by uploading it again. You will see the new `added_tokens_decoder` attribute that will store the relevant information.\n",
"You are using the default legacy behaviour of the . If you see this, DO NOT PANIC! This is expected, and simply means that the `legacy` (previous) behavior will be used so nothing changes for you. If you want to use the new behaviour, set `legacy=False`. This should only be set if you understand what it means, and thouroughly read the reason why this was added as explained in https://github.com/huggingface/transformers/pull/24565\n",
"Some weights of the model checkpoint at mesolitica/finetune-mnli-t5-small-standard-bahasa-cased were not used when initializing T5ForSequenceClassification: ['classification_head.dense.weight', 'classification_head.out_proj.weight', 'classification_head.dense.bias', 'classification_head.out_proj.bias']\n",
"- This IS expected if you are initializing T5ForSequenceClassification 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 T5ForSequenceClassification 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 T5ForSequenceClassification were not initialized from the model checkpoint at mesolitica/finetune-mnli-t5-small-standard-bahasa-cased and are newly initialized: ['classifier.weight', 'classifier.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": [
"model = malaya.zero_shot.classification.huggingface()"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### predict batch\n",
"\n",
"```python\n",
"def predict_proba(\n",
" self,\n",
" strings: List[str],\n",
" labels: List[str],\n",
" prefix: str = 'ayat ini berkaitan tentang',\n",
" multilabel: bool = True,\n",
"):\n",
" \"\"\"\n",
" classify list of strings and return probability.\n",
"\n",
" Parameters\n",
" ----------\n",
" strings: List[str]\n",
" labels: List[str]\n",
" prefix: str, optional (default='ayat ini berkaitan tentang')\n",
" prefix of labels to zero shot. Playing around with prefix can get better results.\n",
" multilabel: bool, optional (default=True)\n",
" probability of labels can be more than 1.0\n",
"```\n",
"\n",
"Because it is a zero-shot, we need to give labels for the model."
]
},
{
"cell_type": "code",
"execution_count": 5,
"metadata": {},
"outputs": [],
"source": [
"# copy from twitter\n",
"\n",
"string = 'gov macam bengong, kami nk pilihan raya, gov backdoor, sakai'"
]
},
{
"cell_type": "code",
"execution_count": 6,
"metadata": {},
"outputs": [
{
"name": "stderr",
"output_type": "stream",
"text": [
"You're using a T5TokenizerFast tokenizer. Please note that with a fast tokenizer, using the `__call__` method is faster than using a method to encode the text followed by a call to the `pad` method to get a padded encoding.\n"
]
},
{
"data": {
"text/plain": [
"[{'najib razak': 0.47769544,\n",
" 'mahathir': 0.49602416,\n",
" 'kerajaan': 0.49770266,\n",
" 'PRU': 0.5020965,\n",
" 'anarki': 0.47935393}]"
]
},
"execution_count": 6,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"model.predict_proba([string], labels = ['najib razak', 'mahathir', 'kerajaan', 'PRU', 'anarki'])"
]
},
{
"cell_type": "code",
"execution_count": 7,
"metadata": {},
"outputs": [],
"source": [
"string = 'tolong order foodpanda jab, lapar'"
]
},
{
"cell_type": "code",
"execution_count": 8,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"[{'makan': 0.49923217,\n",
" 'makanan': 0.50025105,\n",
" 'novel': 0.50996864,\n",
" 'buku': 0.5179709,\n",
" 'kerajaan': 0.52829444,\n",
" 'food delivery': 0.5014325}]"
]
},
"execution_count": 8,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"model.predict_proba([string], labels = ['makan', 'makanan', 'novel', 'buku', 'kerajaan', 'food delivery'])"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"the model understood `order foodpanda` got close relationship with `makan`, `makanan` and `food delivery`."
]
},
{
"cell_type": "code",
"execution_count": 9,
"metadata": {},
"outputs": [],
"source": [
"string = 'kerajaan sebenarnya sangat prihatin dengan rakyat, bagi duit bantuan'"
]
},
{
"cell_type": "code",
"execution_count": 10,
"metadata": {
"scrolled": true
},
"outputs": [
{
"data": {
"text/plain": [
"[{'makan': 0.4649984,\n",
" 'makanan': 0.4640362,\n",
" 'novel': 0.513372,\n",
" 'buku': 0.50357056,\n",
" 'kerajaan': 0.52359533,\n",
" 'food delivery': 0.49170837,\n",
" 'kerajaan jahat': 0.51742524,\n",
" 'kerajaan prihatin': 0.5301894,\n",
" 'bantuan rakyat': 0.5329738}]"
]
},
"execution_count": 10,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"model.predict_proba([string], labels = ['makan', 'makanan', 'novel', 'buku', 'kerajaan', 'food delivery',\n",
" 'kerajaan jahat', 'kerajaan prihatin', 'bantuan rakyat'])"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### able to infer for mixed MS and EN"
]
},
{
"cell_type": "code",
"execution_count": 11,
"metadata": {},
"outputs": [],
"source": [
"string = 'Hi guys! I noticed semalam & harini dah ramai yang dapat cookies ni kan. So harini i nak share some post mortem of our first batch:'"
]
},
{
"cell_type": "code",
"execution_count": 12,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"[{'makan': 0.446774,\n",
" 'makanan': 0.44988176,\n",
" 'novel': 0.46197286,\n",
" 'buku': 0.46333978,\n",
" 'kerajaan': 0.47140214,\n",
" 'food delivery': 0.4580575,\n",
" 'kerajaan jahat': 0.45878497,\n",
" 'kerajaan prihatin': 0.46423724,\n",
" 'bantuan rakyat': 0.44994086,\n",
" 'biskut': 0.4561887,\n",
" 'very helpful': 0.4278154,\n",
" 'sharing experiences': 0.458337,\n",
" 'sharing session': 0.4593555}]"
]
},
"execution_count": 12,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"model.predict_proba([string], labels = ['makan', 'makanan', 'novel', 'buku', 'kerajaan', 'food delivery',\n",
" 'kerajaan jahat', 'kerajaan prihatin', 'bantuan rakyat',\n",
" 'biskut', 'very helpful', 'sharing experiences',\n",
" 'sharing session'])"
]
},
{
"cell_type": "code",
"execution_count": 13,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"[{'makan': 0.44845203,\n",
" 'makanan': 0.45382816,\n",
" 'novel': 0.46462435,\n",
" 'buku': 0.4645531,\n",
" 'kerajaan': 0.47166649,\n",
" 'food delivery': 0.460771,\n",
" 'kerajaan jahat': 0.46060804,\n",
" 'kerajaan prihatin': 0.46640876,\n",
" 'bantuan rakyat': 0.4555358,\n",
" 'biskut': 0.45959476,\n",
" 'very helpful': 0.4333261,\n",
" 'sharing experiences': 0.4627727,\n",
" 'sharing session': 0.4624747}]"
]
},
"execution_count": 13,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"model.predict_proba([string], labels = ['makan', 'makanan', 'novel', 'buku', 'kerajaan', 'food delivery',\n",
" 'kerajaan jahat', 'kerajaan prihatin', 'bantuan rakyat',\n",
" 'biskut', 'very helpful', 'sharing experiences',\n",
" 'sharing session'],\n",
" prefix = 'teks ini berkaitan tentang')"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Multiclasses but not multilabel\n",
"\n",
"Sum of probability equal to 1.0, so to do that, set `multilabel=False`."
]
},
{
"cell_type": "code",
"execution_count": 14,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"[{'makan': 0.07241088,\n",
" 'makanan': 0.07458564,\n",
" 'novel': 0.07798509,\n",
" 'buku': 0.07774425,\n",
" 'kerajaan': 0.07856386,\n",
" 'food delivery': 0.07746716,\n",
" 'kerajaan jahat': 0.076677,\n",
" 'kerajaan prihatin': 0.07793205,\n",
" 'bantuan rakyat': 0.08176315,\n",
" 'biskut': 0.07559921,\n",
" 'very helpful': 0.07805643,\n",
" 'sharing experiences': 0.07687527,\n",
" 'sharing session': 0.07434013}]"
]
},
"execution_count": 14,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"string = 'kerajaan sebenarnya sangat prihatin dengan rakyat, bagi duit bantuan'\n",
"\n",
"model.predict_proba([string], labels = ['makan', 'makanan', 'novel', 'buku', 'kerajaan', 'food delivery',\n",
" 'kerajaan jahat', 'kerajaan prihatin', 'bantuan rakyat',\n",
" 'biskut', 'very helpful', 'sharing experiences',\n",
" 'sharing session'], multilabel = False)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Stacking models\n",
"\n",
"More information, you can read at https://malaya.readthedocs.io/en/latest/Stack.html\n",
"\n",
"If you want to stack zero-shot classification models, you need to pass labels using keyword parameter,\n",
"\n",
"```python\n",
"malaya.stack.predict_stack([model1, model2], List[str], labels = List[str])\n",
"```\n",
"\n",
"We will passed `labels` as `**kwargs`."
]
},
{
"cell_type": "code",
"execution_count": 15,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"[{'makan': 0.46499833,\n",
" 'makanan': 0.4640362,\n",
" 'novel': 0.513372,\n",
" 'buku': 0.50357056,\n",
" 'kerajaan': 0.52359533,\n",
" 'food delivery': 0.49170837,\n",
" 'kerajaan jahat': 0.51742524,\n",
" 'kerajaan prihatin': 0.53018934,\n",
" 'bantuan rakyat': 0.53297377,\n",
" 'comel': 0.49191207,\n",
" 'kerajaan syg sgt kepada rakyat': 0.5212374}]"
]
},
"execution_count": 15,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"string = 'kerajaan sebenarnya sangat prihatin dengan rakyat, bagi duit bantuan'\n",
"labels = ['makan', 'makanan', 'novel', 'buku', 'kerajaan', 'food delivery', \n",
" 'kerajaan jahat', 'kerajaan prihatin', 'bantuan rakyat', 'comel', 'kerajaan syg sgt kepada rakyat']\n",
"malaya.stack.predict_stack([model, model, model], [string], \n",
" labels = labels)"
]
}
],
"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": "3.8.10"
},
"varInspector": {
"cols": {
"lenName": 16,
"lenType": 16,
"lenVar": 40
},
"kernels_config": {
"python": {
"delete_cmd_postfix": "",
"delete_cmd_prefix": "del ",
"library": "var_list.py",
"varRefreshCmd": "print(var_dic_list())"
},
"r": {
"delete_cmd_postfix": ") ",
"delete_cmd_prefix": "rm(",
"library": "var_list.r",
"varRefreshCmd": "cat(var_dic_list()) "
}
},
"types_to_exclude": [
"module",
"function",
"builtin_function_or_method",
"instance",
"_Feature"
],
"window_display": false
}
},
"nbformat": 4,
"nbformat_minor": 2
}