{ "cells": [ { "cell_type": "markdown", "metadata": {}, "source": [ "# 使用Trainer和Tester快速训练和测试" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "## 数据读入和处理" ] }, { "cell_type": "code", "execution_count": 1, "metadata": {}, "outputs": [ { "name": "stderr", "output_type": "stream", "text": [ "/remote-home/ynzheng/anaconda3/envs/now/lib/python3.8/site-packages/FastNLP-0.5.0-py3.8.egg/fastNLP/io/loader/classification.py:340: UserWarning: SST2's test file has no target.\n" ] }, { "name": "stdout", "output_type": "stream", "text": [ "In total 3 datasets:\n", "\ttest has 1821 instances.\n", "\ttrain has 67349 instances.\n", "\tdev has 872 instances.\n", "In total 2 vocabs:\n", "\twords has 16292 entries.\n", "\ttarget has 2 entries.\n", "\n", "+-----------------------------------+--------+-----------------------------------+---------+\n", "| raw_words | target | words | seq_len |\n", "+-----------------------------------+--------+-----------------------------------+---------+\n", "| hide new secretions from the p... | 1 | [4110, 97, 12009, 39, 2, 6843,... | 7 |\n", "+-----------------------------------+--------+-----------------------------------+---------+\n", "Vocabulary(['hide', 'new', 'secretions', 'from', 'the']...)\n" ] } ], "source": [ "from fastNLP.io import SST2Pipe\n", "\n", "pipe = SST2Pipe()\n", "databundle = pipe.process_from_file()\n", "vocab = databundle.get_vocab('words')\n", "print(databundle)\n", "print(databundle.get_dataset('train')[0])\n", "print(databundle.get_vocab('words'))" ] }, { "cell_type": "code", "execution_count": 2, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "4925 872 75\n" ] } ], "source": [ "train_data = databundle.get_dataset('train')[:5000]\n", "train_data, test_data = train_data.split(0.015)\n", "dev_data = databundle.get_dataset('dev')\n", "print(len(train_data),len(dev_data),len(test_data))" ] }, { "cell_type": "code", "execution_count": 3, "metadata": { "scrolled": false }, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "+-------------+-----------+--------+-------+---------+\n", "| field_names | raw_words | target | words | seq_len |\n", "+-------------+-----------+--------+-------+---------+\n", "| is_input | False | False | True | True |\n", "| is_target | False | True | False | False |\n", "| ignore_type | | False | False | False |\n", "| pad_value | | 0 | 0 | 0 |\n", "+-------------+-----------+--------+-------+---------+\n" ] }, { "data": { "text/plain": [ "" ] }, "execution_count": 3, "metadata": {}, "output_type": "execute_result" } ], "source": [ "train_data.print_field_meta()" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "## 使用内置模型训练" ] }, { "cell_type": "code", "execution_count": 4, "metadata": {}, "outputs": [], "source": [ "from fastNLP.models import CNNText\n", "\n", "#词嵌入的维度\n", "EMBED_DIM = 100\n", "\n", "#使用CNNText的时候第一个参数输入一个tuple,作为模型定义embedding的参数\n", "#还可以传入 kernel_nums, kernel_sizes, padding, dropout的自定义值\n", "model_cnn = CNNText((len(vocab),EMBED_DIM), num_classes=2, dropout=0.1)" ] }, { "cell_type": "code", "execution_count": 5, "metadata": {}, "outputs": [], "source": [ "from fastNLP import AccuracyMetric\n", "from fastNLP import Const\n", "\n", "# metrics=AccuracyMetric() 在本例中与下面这行代码等价\n", "metrics=AccuracyMetric(pred=Const.OUTPUT, target=Const.TARGET)" ] }, { "cell_type": "code", "execution_count": 6, "metadata": {}, "outputs": [], "source": [ "from fastNLP import CrossEntropyLoss\n", "\n", "# loss = CrossEntropyLoss() 在本例中与下面这行代码等价\n", "loss = CrossEntropyLoss(pred=Const.OUTPUT, target=Const.TARGET)" ] }, { "cell_type": "code", "execution_count": 7, "metadata": {}, "outputs": [], "source": [ "# 这表示构建了一个损失函数类，由func计算损失函数，其中将从模型返回值或者DataSet的target=True的field\n", "# 当中找到一个参数名为`pred`的参数传入func一个参数名为`input`的参数；找到一个参数名为`label`的参数\n", "# 传入func作为一个名为`target`的参数\n", "#下面自己构建了一个交叉熵函数，和之后直接使用fastNLP中的交叉熵函数是一个效果\n", "import torch\n", "from fastNLP import LossFunc\n", "func = torch.nn.functional.cross_entropy\n", "loss_func = LossFunc(func, input=Const.OUTPUT, target=Const.TARGET)" ] }, { "cell_type": "code", "execution_count": 8, "metadata": {}, "outputs": [], "source": [ "import torch.optim as optim\n", "\n", "#使用 torch.optim 定义优化器\n", "optimizer=optim.RMSprop(model_cnn.parameters(), lr=0.01, alpha=0.99, eps=1e-08, weight_decay=0, momentum=0, centered=False)" ] }, { "cell_type": "code", "execution_count": 9, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "input fields after batch(if batch size is 2):\n", "\twords: (1)type:torch.Tensor (2)dtype:torch.int64, (3)shape:torch.Size([2, 4]) \n", "\tseq_len: (1)type:torch.Tensor (2)dtype:torch.int64, (3)shape:torch.Size([2]) \n", "target fields after batch(if batch size is 2):\n", "\ttarget: (1)type:torch.Tensor (2)dtype:torch.int64, (3)shape:torch.Size([2]) \n", "\n", "training epochs started 2020-02-27-11-31-25\n" ] }, { "data": { "application/vnd.jupyter.widget-view+json": { "model_id": "", "version_major": 2, "version_minor": 0 }, "text/plain": [ "HBox(children=(FloatProgress(value=0.0, layout=Layout(flex='2'), max=3080.0), HTML(value='')), layout=Layout(d…" ] }, "metadata": {}, "output_type": "display_data" }, { "data": { "application/vnd.jupyter.widget-view+json": { "model_id": "", "version_major": 2, "version_minor": 0 }, "text/plain": [ "HBox(children=(FloatProgress(value=0.0, layout=Layout(flex='2'), max=55.0), HTML(value='')), layout=Layout(dis…" ] }, "metadata": {}, "output_type": "display_data" }, { "name": "stdout", "output_type": "stream", "text": [ "\r", "Evaluate data in 0.75 seconds!\n", "\r", "Evaluation on dev at Epoch 1/10. Step:308/3080: \n", "\r", "AccuracyMetric: acc=0.751147\n", "\n" ] }, { "data": { "application/vnd.jupyter.widget-view+json": { "model_id": "", "version_major": 2, "version_minor": 0 }, "text/plain": [ "HBox(children=(FloatProgress(value=0.0, layout=Layout(flex='2'), max=55.0), HTML(value='')), layout=Layout(dis…" ] }, "metadata": {}, "output_type": "display_data" }, { "name": "stdout", "output_type": "stream", "text": [ "\r", "Evaluate data in 0.83 seconds!\n", "\r", "Evaluation on dev at Epoch 2/10. Step:616/3080: \n", "\r", "AccuracyMetric: acc=0.755734\n", "\n" ] }, { "data": { "application/vnd.jupyter.widget-view+json": { "model_id": "", "version_major": 2, "version_minor": 0 }, "text/plain": [ "HBox(children=(FloatProgress(value=0.0, layout=Layout(flex='2'), max=55.0), HTML(value='')), layout=Layout(dis…" ] }, "metadata": {}, "output_type": "display_data" }, { "name": "stdout", "output_type": "stream", "text": [ "\r", "Evaluate data in 1.32 seconds!\n", "\r", "Evaluation on dev at Epoch 3/10. Step:924/3080: \n", "\r", "AccuracyMetric: acc=0.758028\n", "\n" ] }, { "data": { "application/vnd.jupyter.widget-view+json": { "model_id": "", "version_major": 2, "version_minor": 0 }, "text/plain": [ "HBox(children=(FloatProgress(value=0.0, layout=Layout(flex='2'), max=55.0), HTML(value='')), layout=Layout(dis…" ] }, "metadata": {}, "output_type": "display_data" }, { "name": "stdout", "output_type": "stream", "text": [ "\r", "Evaluate data in 0.88 seconds!\n", "\r", "Evaluation on dev at Epoch 4/10. Step:1232/3080: \n", "\r", "AccuracyMetric: acc=0.741972\n", "\n" ] }, { "data": { "application/vnd.jupyter.widget-view+json": { "model_id": "", "version_major": 2, "version_minor": 0 }, "text/plain": [ "HBox(children=(FloatProgress(value=0.0, layout=Layout(flex='2'), max=55.0), HTML(value='')), layout=Layout(dis…" ] }, "metadata": {}, "output_type": "display_data" }, { "name": "stdout", "output_type": "stream", "text": [ "\r", "Evaluate data in 0.96 seconds!\n", "\r", "Evaluation on dev at Epoch 5/10. Step:1540/3080: \n", "\r", "AccuracyMetric: acc=0.728211\n", "\n" ] }, { "data": { "application/vnd.jupyter.widget-view+json": { "model_id": "", "version_major": 2, "version_minor": 0 }, "text/plain": [ "HBox(children=(FloatProgress(value=0.0, layout=Layout(flex='2'), max=55.0), HTML(value='')), layout=Layout(dis…" ] }, "metadata": {}, "output_type": "display_data" }, { "name": "stdout", "output_type": "stream", "text": [ "\r", "Evaluate data in 0.87 seconds!\n", "\r", "Evaluation on dev at Epoch 6/10. Step:1848/3080: \n", "\r", "AccuracyMetric: acc=0.755734\n", "\n" ] }, { "data": { "application/vnd.jupyter.widget-view+json": { "model_id": "", "version_major": 2, "version_minor": 0 }, "text/plain": [ "HBox(children=(FloatProgress(value=0.0, layout=Layout(flex='2'), max=55.0), HTML(value='')), layout=Layout(dis…" ] }, "metadata": {}, "output_type": "display_data" }, { "name": "stdout", "output_type": "stream", "text": [ "\r", "Evaluate data in 1.04 seconds!\n", "\r", "Evaluation on dev at Epoch 7/10. Step:2156/3080: \n", "\r", "AccuracyMetric: acc=0.732798\n", "\n" ] }, { "data": { "application/vnd.jupyter.widget-view+json": { "model_id": "", "version_major": 2, "version_minor": 0 }, "text/plain": [ "HBox(children=(FloatProgress(value=0.0, layout=Layout(flex='2'), max=55.0), HTML(value='')), layout=Layout(dis…" ] }, "metadata": {}, "output_type": "display_data" }, { "name": "stdout", "output_type": "stream", "text": [ "\r", "Evaluate data in 0.57 seconds!\n", "\r", "Evaluation on dev at Epoch 8/10. Step:2464/3080: \n", "\r", "AccuracyMetric: acc=0.747706\n", "\n" ] }, { "data": { "application/vnd.jupyter.widget-view+json": { "model_id": "", "version_major": 2, "version_minor": 0 }, "text/plain": [ "HBox(children=(FloatProgress(value=0.0, layout=Layout(flex='2'), max=55.0), HTML(value='')), layout=Layout(dis…" ] }, "metadata": {}, "output_type": "display_data" }, { "name": "stdout", "output_type": "stream", "text": [ "\r", "Evaluate data in 0.48 seconds!\n", "\r", "Evaluation on dev at Epoch 9/10. Step:2772/3080: \n", "\r", "AccuracyMetric: acc=0.732798\n", "\n" ] }, { "data": { "application/vnd.jupyter.widget-view+json": { "model_id": "", "version_major": 2, "version_minor": 0 }, "text/plain": [ "HBox(children=(FloatProgress(value=0.0, layout=Layout(flex='2'), max=55.0), HTML(value='')), layout=Layout(dis…" ] }, "metadata": {}, "output_type": "display_data" }, { "name": "stdout", "output_type": "stream", "text": [ "\r", "Evaluate data in 0.48 seconds!\n", "\r", "Evaluation on dev at Epoch 10/10. Step:3080/3080: \n", "\r", "AccuracyMetric: acc=0.740826\n", "\n", "\r\n", "In Epoch:3/Step:924, got best dev performance:\n", "AccuracyMetric: acc=0.758028\n", "Reloaded the best model.\n" ] }, { "data": { "text/plain": [ "{'best_eval': {'AccuracyMetric': {'acc': 0.758028}},\n", " 'best_epoch': 3,\n", " 'best_step': 924,\n", " 'seconds': 160.58}" ] }, "execution_count": 9, "metadata": {}, "output_type": "execute_result" } ], "source": [ "from fastNLP import Trainer\n", "\n", "#训练的轮数和batch size\n", "N_EPOCHS = 10\n", "BATCH_SIZE = 16\n", "\n", "#如果在定义trainer的时候没有传入optimizer参数，模型默认的优化器为torch.optim.Adam且learning rate为lr=4e-3\n", "#这里只使用了loss作为损失函数输入，感兴趣可以尝试其他损失函数（如之前自定义的loss_func）作为输入\n", "trainer = Trainer(model=model_cnn, train_data=train_data, dev_data=dev_data, loss=loss, metrics=metrics,\n", "optimizer=optimizer,n_epochs=N_EPOCHS, batch_size=BATCH_SIZE)\n", "trainer.train()" ] }, { "cell_type": "code", "execution_count": 10, "metadata": {}, "outputs": [ { "data": { "application/vnd.jupyter.widget-view+json": { "model_id": "", "version_major": 2, "version_minor": 0 }, "text/plain": [ "HBox(children=(FloatProgress(value=0.0, layout=Layout(flex='2'), max=5.0), HTML(value='')), layout=Layout(disp…" ] }, "metadata": {}, "output_type": "display_data" }, { "name": "stdout", "output_type": "stream", "text": [ "\r", "Evaluate data in 0.43 seconds!\n", "[tester] \n", "AccuracyMetric: acc=0.773333\n" ] }, { "data": { "text/plain": [ "{'AccuracyMetric': {'acc': 0.773333}}" ] }, "execution_count": 10, "metadata": {}, "output_type": "execute_result" } ], "source": [ "from fastNLP import Tester\n", "\n", "tester = Tester(test_data, model_cnn, metrics=AccuracyMetric())\n", "tester.test()" ] }, { "cell_type": "code", "execution_count": null, "metadata": {}, "outputs": [], "source": [] } ], "metadata": { "kernelspec": { "display_name": "Python Now", "language": "python", "name": "now" }, "language_info": { "codemirror_mode": { "name": "ipython", "version": 3 }, "file_extension": ".py", "mimetype": "text/x-python", "name": "python", "nbconvert_exporter": "python", "pygments_lexer": "ipython3", "version": "3.8.0" } }, "nbformat": 4, "nbformat_minor": 2 }