diff --git a/api-examples/nn/nn.bceloss.ipynb b/api-examples/nn/nn.bceloss.ipynb new file mode 100644 index 0000000..aea0394 --- /dev/null +++ b/api-examples/nn/nn.bceloss.ipynb @@ -0,0 +1,93 @@ +{ + "cells": [ + { + "cell_type": "code", + "execution_count": 10, + "metadata": {}, + "outputs": [], + "source": [ + "import numpy as np\n", + "import torch\n", + "from torch import Tensor\n", + "from torch import nn" + ] + }, + { + "cell_type": "code", + "execution_count": 13, + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "tensor(0.8536, grad_fn=)\n" + ] + } + ], + "source": [ + "# BCELoss\n", + "m = nn.Sigmoid()\n", + "input_np = np.random.rand(3,2)\n", + "target_np = np.random.rand(3,2)\n", + "output = nn.BCELoss()(m(Tensor(input)), Tensor(target))\n", + "print(output)" + ] + }, + { + "cell_type": "code", + "execution_count": 14, + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "0.71798706\n" + ] + } + ], + "source": [ + "import mindspore as ms\n", + "from mindspore import nn\n", + "from mindspore import Tensor\n", + "\n", + "m = nn.Sigmoid()\n", + "loss = nn.BCELoss(weight=None, reduction='mean')\n", + "input = ms.Tensor(input_np, ms.float32)\n", + "target = ms.Tensor(target_np, ms.float32)\n", + "output = loss(m(input), target)\n", + "print(output)" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "- 默认reduction=mean下输出与PyTorch似乎不一致\n", + "- mint.nn.BCELoss待验证，CPU上不支持" + ] + } + ], + "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": "3.11.11" + } + }, + "nbformat": 4, + "nbformat_minor": 4 +} diff --git a/api-examples/nn/nn.celldict.ipynb b/api-examples/nn/nn.celldict.ipynb new file mode 100644 index 0000000..26cfffd --- /dev/null +++ b/api-examples/nn/nn.celldict.ipynb @@ -0,0 +1,131 @@ +{ + "cells": [ + { + "cell_type": "code", + "execution_count": 1, + "metadata": {}, + "outputs": [], + "source": [ + "import numpy as np\n", + "import torch\n", + "from torch import Tensor\n", + "from torch import nn" + ] + }, + { + "cell_type": "code", + "execution_count": 2, + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "ModuleDict(\n", + " (cond2d): Conv2d(10, 10, kernel_size=(3, 3), stride=(1, 1))\n", + ")\n", + "ModuleDict(\n", + " (relu): ReLU()\n", + " (dict1): ModuleDict(\n", + " (cond2d): Conv2d(10, 10, kernel_size=(3, 3), stride=(1, 1))\n", + " )\n", + ")\n" + ] + } + ], + "source": [ + "# moduledict\n", + "moduledict1 = nn.ModuleDict({'cond2d': nn.Conv2d(10,10,3)})\n", + "moduledict2 = nn.ModuleDict({'relu': nn.ReLU(), 'dict1': moduledict1})\n", + "print(moduledict1)\n", + "print(moduledict2)" + ] + }, + { + "cell_type": "code", + "execution_count": 6, + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "CellDict<\n", + " (cond2d): Conv2d, bias_init=None, format=NCHW>\n", + " >\n" + ] + } + ], + "source": [ + "import mindspore as ms\n", + "from mindspore import nn\n", + "celldict1 = nn.CellDict({'cond2d': nn.Conv2d(10,10,3)})\n", + "print(celldict1)" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "**MindSpore支持CellDict(Cell)**\n", + "- 打印出的结构没有缩进；weight_init过于详细；\n", + "- 建议默认参数不打印，提供选择开关（print_all）" + ] + }, + { + "cell_type": "code", + "execution_count": 7, + "metadata": {}, + "outputs": [ + { + "ename": "TypeError", + "evalue": "For 'CellDict', the type of cell can not be CellDict, CellList or SequentialCell, but got CellDict.", + "output_type": "error", + "traceback": [ + "\u001b[1;31m---------------------------------------------------------------------------\u001b[0m", + "\u001b[1;31mTypeError\u001b[0m Traceback (most recent call last)", + "Cell \u001b[1;32mIn[7], line 1\u001b[0m\n\u001b[1;32m----> 1\u001b[0m celldict2 \u001b[38;5;241m=\u001b[39m nn\u001b[38;5;241m.\u001b[39mCellDict({\u001b[38;5;124m'\u001b[39m\u001b[38;5;124mrelu\u001b[39m\u001b[38;5;124m'\u001b[39m: nn\u001b[38;5;241m.\u001b[39mReLU(), \u001b[38;5;124m'\u001b[39m\u001b[38;5;124mdict1\u001b[39m\u001b[38;5;124m'\u001b[39m: celldict1})\n\u001b[0;32m 2\u001b[0m \u001b[38;5;28mprint\u001b[39m(celldict2)\n", + "File \u001b[1;32m~\\anaconda3\\envs\\tutorial\\Lib\\site-packages\\mindspore\\nn\\layer\\container.py:583\u001b[0m, in \u001b[0;36mCellDict.__init__\u001b[1;34m(self, *args, **kwargs)\u001b[0m\n\u001b[0;32m 581\u001b[0m Cell\u001b[38;5;241m.\u001b[39m\u001b[38;5;21m__init__\u001b[39m(\u001b[38;5;28mself\u001b[39m, auto_prefix)\n\u001b[0;32m 582\u001b[0m \u001b[38;5;28;01mif\u001b[39;00m \u001b[38;5;28mlen\u001b[39m(args) \u001b[38;5;241m==\u001b[39m \u001b[38;5;241m1\u001b[39m:\n\u001b[1;32m--> 583\u001b[0m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39mupdate(args[\u001b[38;5;241m0\u001b[39m])\n", + "File \u001b[1;32m~\\anaconda3\\envs\\tutorial\\Lib\\site-packages\\mindspore\\nn\\layer\\container.py:720\u001b[0m, in \u001b[0;36mCellDict.update\u001b[1;34m(self, cells)\u001b[0m\n\u001b[0;32m 718\u001b[0m \u001b[38;5;28;01mif\u001b[39;00m \u001b[38;5;28misinstance\u001b[39m(cells, (OrderedDict, CellDict, abc\u001b[38;5;241m.\u001b[39mMapping)):\n\u001b[0;32m 719\u001b[0m \u001b[38;5;28;01mfor\u001b[39;00m key, cell \u001b[38;5;129;01min\u001b[39;00m cells\u001b[38;5;241m.\u001b[39mitems():\n\u001b[1;32m--> 720\u001b[0m \u001b[38;5;28mself\u001b[39m[key] \u001b[38;5;241m=\u001b[39m cell\n\u001b[0;32m 721\u001b[0m \u001b[38;5;28;01melse\u001b[39;00m:\n\u001b[0;32m 722\u001b[0m \u001b[38;5;28;01mfor\u001b[39;00m \u001b[38;5;28mid\u001b[39m, k_v \u001b[38;5;129;01min\u001b[39;00m \u001b[38;5;28menumerate\u001b[39m(cells):\n", + "File \u001b[1;32m~\\anaconda3\\envs\\tutorial\\Lib\\site-packages\\mindspore\\nn\\layer\\container.py:590\u001b[0m, in \u001b[0;36mCellDict.__setitem__\u001b[1;34m(self, key, cell)\u001b[0m\n\u001b[0;32m 588\u001b[0m \u001b[38;5;28;01mdef\u001b[39;00m \u001b[38;5;21m__setitem__\u001b[39m(\u001b[38;5;28mself\u001b[39m, key, cell):\n\u001b[0;32m 589\u001b[0m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_validate_key(key)\n\u001b[1;32m--> 590\u001b[0m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_validate_cell_type(cell)\n\u001b[0;32m 591\u001b[0m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_update_cell_para_name(key, cell)\n\u001b[0;32m 592\u001b[0m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_cells[key] \u001b[38;5;241m=\u001b[39m cell\n", + "File \u001b[1;32m~\\anaconda3\\envs\\tutorial\\Lib\\site-packages\\mindspore\\nn\\layer\\container.py:631\u001b[0m, in \u001b[0;36mCellDict._validate_cell_type\u001b[1;34m(self, cell)\u001b[0m\n\u001b[0;32m 628\u001b[0m \u001b[38;5;28;01mraise\u001b[39;00m \u001b[38;5;167;01mTypeError\u001b[39;00m(\u001b[38;5;124mf\u001b[39m\u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mFor \u001b[39m\u001b[38;5;124m'\u001b[39m\u001b[38;5;132;01m{\u001b[39;00mcls_name\u001b[38;5;132;01m}\u001b[39;00m\u001b[38;5;124m'\u001b[39m\u001b[38;5;124m, the type of cell should be Cell, \u001b[39m\u001b[38;5;124m\"\u001b[39m\n\u001b[0;32m 629\u001b[0m \u001b[38;5;124mf\u001b[39m\u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mbut got \u001b[39m\u001b[38;5;132;01m{\u001b[39;00m\u001b[38;5;28mtype\u001b[39m(cell)\u001b[38;5;241m.\u001b[39m\u001b[38;5;18m__name__\u001b[39m\u001b[38;5;132;01m}\u001b[39;00m\u001b[38;5;124m.\u001b[39m\u001b[38;5;124m\"\u001b[39m)\n\u001b[0;32m 630\u001b[0m \u001b[38;5;28;01mif\u001b[39;00m \u001b[38;5;28misinstance\u001b[39m(cell, (CellDict, CellList, SequentialCell)):\n\u001b[1;32m--> 631\u001b[0m \u001b[38;5;28;01mraise\u001b[39;00m \u001b[38;5;167;01mTypeError\u001b[39;00m(\u001b[38;5;124mf\u001b[39m\u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mFor \u001b[39m\u001b[38;5;124m'\u001b[39m\u001b[38;5;132;01m{\u001b[39;00mcls_name\u001b[38;5;132;01m}\u001b[39;00m\u001b[38;5;124m'\u001b[39m\u001b[38;5;124m, the type of cell can not be CellDict, CellList or SequentialCell, \u001b[39m\u001b[38;5;124m\"\u001b[39m\n\u001b[0;32m 632\u001b[0m \u001b[38;5;124mf\u001b[39m\u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mbut got \u001b[39m\u001b[38;5;132;01m{\u001b[39;00m\u001b[38;5;28mtype\u001b[39m(cell)\u001b[38;5;241m.\u001b[39m\u001b[38;5;18m__name__\u001b[39m\u001b[38;5;132;01m}\u001b[39;00m\u001b[38;5;124m.\u001b[39m\u001b[38;5;124m\"\u001b[39m)\n", + "\u001b[1;31mTypeError\u001b[0m: For 'CellDict', the type of cell can not be CellDict, CellList or SequentialCell, but got CellDict." + ] + } + ], + "source": [ + "celldict2 = nn.CellDict({'relu': nn.ReLU(), 'dict1': celldict1})\n", + "print(celldict2)" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "**MindSpore不支持CellDict(CellDict)**\n", + "- PyTorch支持CellDict/SequentialCell/CellList嵌套" + ] + } + ], + "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": "3.11.11" + } + }, + "nbformat": 4, + "nbformat_minor": 4 +} diff --git a/api-examples/nn/nn.dense.ipynb b/api-examples/nn/nn.dense.ipynb new file mode 100644 index 0000000..e4bb35f --- /dev/null +++ b/api-examples/nn/nn.dense.ipynb @@ -0,0 +1,113 @@ +{ + "cells": [ + { + "cell_type": "code", + "execution_count": 18, + "metadata": {}, + "outputs": [], + "source": [ + "import numpy as np\n", + "import torch\n", + "from torch import Tensor\n", + "from torch import nn" + ] + }, + { + "cell_type": "code", + "execution_count": 20, + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "torch.Size([128, 30])\n" + ] + } + ], + "source": [ + "# Linear\n", + "input = np.random.randn(128, 20)\n", + "output = nn.Linear(20,30)(Tensor(input))\n", + "print(output.size())" + ] + }, + { + "cell_type": "code", + "execution_count": 32, + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "(128, 30)\n" + ] + } + ], + "source": [ + "import jax\n", + "import flax.linen as nn\n", + "from jax import numpy as jnp\n", + "from jax import random\n", + "\n", + "input = jnp.ones((128,20))\n", + "dense = nn.Dense(features=30)\n", + "variables = dense.init(random.key(42), input)\n", + "output = dense.apply(variables, input)\n", + "print(output.shape)" + ] + }, + { + "cell_type": "code", + "execution_count": 24, + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "(128, 30)\n" + ] + } + ], + "source": [ + "import mindspore as ms\n", + "from mindspore import nn\n", + "from mindspore import Tensor\n", + "output = nn.Dense(20,30)(Tensor(input, dtype=ms.float32))\n", + "print(output.shape)" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "**MindSpore Dense用法与输出与PyTorch一致**\n", + "- MindSpore的命名与jax保持一致（Dense）\n", + "- jax需要额外定义参数" + ] + } + ], + "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": "3.11.11" + } + }, + "nbformat": 4, + "nbformat_minor": 4 +} diff --git a/api-examples/nn/nn.embedding.ipynb b/api-examples/nn/nn.embedding.ipynb new file mode 100644 index 0000000..69871f1 --- /dev/null +++ b/api-examples/nn/nn.embedding.ipynb @@ -0,0 +1,151 @@ +{ + "cells": [ + { + "cell_type": "code", + "execution_count": 12, + "metadata": {}, + "outputs": [], + "source": [ + "import numpy as np\n", + "import torch\n", + "from torch import Tensor\n", + "from torch import nn\n", + "import time" + ] + }, + { + "cell_type": "code", + "execution_count": 13, + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "tensor([[[ 1.1268, -0.9513, 0.6823],\n", + " [-0.3245, -0.8497, -0.7579],\n", + " [-0.6075, 2.7415, -2.0689],\n", + " [-1.4612, -1.0289, -0.4242]],\n", + "\n", + " [[-0.6075, 2.7415, -2.0689],\n", + " [ 2.1071, 0.1938, 0.0255],\n", + " [-0.3245, -0.8497, -0.7579],\n", + " [-0.4831, -1.7133, 0.5985]]], grad_fn=)\n", + "tensor([[[ 0.0000, 0.0000, 0.0000],\n", + " [-1.0849, 1.1815, 1.0977],\n", + " [ 0.9290, 1.8014, -0.3725],\n", + " [ 0.0787, 1.9454, 1.1112]],\n", + "\n", + " [[ 1.7580, -0.3959, 0.9574],\n", + " [ 1.7580, -0.3959, 0.9574],\n", + " [-1.0849, 1.1815, 1.0977],\n", + " [ 0.3496, 1.7060, -0.4346]]], grad_fn=)\n", + "0.004005908966064453\n" + ] + } + ], + "source": [ + "start = time.time()\n", + "# an Embedding module containing 10 tensors of size 3\n", + "embedding = nn.Embedding(10, 3)\n", + "# a batch of 2 samples of 4 indices each\n", + "input = torch.LongTensor([[1, 2, 4, 5], [4, 3, 2, 9]])\n", + "print(embedding(input))\n", + "\n", + "# example with padding_idx\n", + "embedding = nn.Embedding(10, 3, padding_idx=0)\n", + "input = torch.LongTensor([[0, 2, 4, 5], [3, 3, 2, 9]])\n", + "print(embedding(input))\n", + "end = time.time()\n", + "print(end-start)\n", + "\n", + "# # example of changing `pad` vector\n", + "# padding_idx = 0\n", + "# embedding = nn.Embedding(3, 3, padding_idx=padding_idx)\n", + "# print(embedding.weight)\n", + "# with torch.no_grad():\n", + "# embedding.weight[padding_idx] = torch.ones(3)\n", + "# print(embedding.weight)" + ] + }, + { + "cell_type": "code", + "execution_count": 15, + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "[[[-0.0065871 0.01535338 0.01434076]\n", + " [ 0.00331824 0.00791553 -0.02024699]\n", + " [ 0.01239492 0.00385856 0.00478753]\n", + " [-0.01227701 0.00011452 0.0142579 ]]\n", + "\n", + " [[ 0.01239492 0.00385856 0.00478753]\n", + " [-0.00323425 0.00501913 -0.00874054]\n", + " [ 0.00331824 0.00791553 -0.02024699]\n", + " [ 0.01642694 -0.0065531 0.00681009]]]\n", + "[[[ 0. 0. 0. ]\n", + " [ 0.00261477 0.01124485 0.00828796]\n", + " [-0.00046545 0.00450285 0.00737254]\n", + " [-0.01147073 0.00254146 0.00321731]]\n", + "\n", + " [[-0.00656487 -0.00497597 -0.00669393]\n", + " [-0.00656487 -0.00497597 -0.00669393]\n", + " [ 0.00261477 0.01124485 0.00828796]\n", + " [ 0.00989333 0.02307425 0.01142865]]]\n", + "0.00896143913269043\n" + ] + } + ], + "source": [ + "import mindspore as ms\n", + "from mindspore import nn\n", + "from mindspore import Tensor\n", + "\n", + "start = time.time()\n", + "embedding = nn.Embedding(10, 3)\n", + "# a batch of 2 samples of 4 indices each\n", + "input = Tensor([[1, 2, 4, 5], [4, 3, 2, 9]])\n", + "print(embedding(input))\n", + "\n", + "# example with padding_idx\n", + "embedding = nn.Embedding(10, 3, padding_idx=0)\n", + "input = Tensor([[0, 2, 4, 5], [3, 3, 2, 9]])\n", + "print(embedding(input))\n", + "end = time.time()\n", + "print(end-start)" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "**MindSpore Embedding用法与输出与PyTorch一致**\n", + "- mint.nn.embedding待验证，CPU不支持" + ] + } + ], + "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": "3.11.11" + } + }, + "nbformat": 4, + "nbformat_minor": 4 +} diff --git a/api-examples/nn/nn.identity.ipynb b/api-examples/nn/nn.identity.ipynb new file mode 100644 index 0000000..699b21e --- /dev/null +++ b/api-examples/nn/nn.identity.ipynb @@ -0,0 +1,87 @@ +{ + "cells": [ + { + "cell_type": "code", + "execution_count": 4, + "metadata": {}, + "outputs": [], + "source": [ + "import numpy as np\n", + "import torch\n", + "from torch import Tensor\n", + "from torch import nn" + ] + }, + { + "cell_type": "code", + "execution_count": 5, + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "torch.Size([128, 20])\n" + ] + } + ], + "source": [ + "input = np.random.rand(128, 20)\n", + "\n", + "m = nn.Identity(54, unused_argument1=0.1, unused_argument2=False)\n", + "output = nn.Identity(54, unused_argument1=0.1, unused_argument2=False)(Tensor(input))\n", + "print(output.size())" + ] + }, + { + "cell_type": "code", + "execution_count": 6, + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "(128, 20)\n" + ] + } + ], + "source": [ + "import mindspore as ms\n", + "from mindspore import nn\n", + "from mindspore import Tensor\n", + "output = nn.Identity()(Tensor(input, dtype=ms.float32))\n", + "print(output.shape)" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "**MindSpore identity用法与输出与PyTorch一致**\n", + "- MindSpore的size/shape打印比PyTorch简洁一点" + ] + } + ], + "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": "3.11.11" + } + }, + "nbformat": 4, + "nbformat_minor": 4 +} diff --git a/api-examples/nn/nn.multiheadattention.ipynb b/api-examples/nn/nn.multiheadattention.ipynb new file mode 100644 index 0000000..5ab3e11 --- /dev/null +++ b/api-examples/nn/nn.multiheadattention.ipynb @@ -0,0 +1,160 @@ +{ + "cells": [ + { + "cell_type": "code", + "execution_count": 10, + "metadata": {}, + "outputs": [], + "source": [ + "import numpy as np\n", + "import torch\n", + "from torch import Tensor\n", + "from torch import nn" + ] + }, + { + "cell_type": "code", + "execution_count": 11, + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "tensor([[[ 0.0109, 0.3142, 0.3060, -0.0961, -0.2895, 0.2136, -0.1708,\n", + " -0.0997, 0.0912, -0.0031, -0.1519, 0.0259, 0.0421, 0.1709,\n", + " 0.2150, -0.1000, 0.3393, 0.2651, -0.0085, -0.0135, 0.0730,\n", + " 0.0116, 0.1105, -0.1352, -0.0823, 0.3899, 0.1814, 0.1009,\n", + " -0.2097, -0.3307, 0.1143, 0.2149]],\n", + "\n", + " [[ 0.0072, 0.3131, 0.3063, -0.0994, -0.2912, 0.2116, -0.1724,\n", + " -0.1033, 0.0965, 0.0024, -0.1512, 0.0233, 0.0406, 0.1782,\n", + " 0.2066, -0.0972, 0.3377, 0.2652, -0.0051, -0.0142, 0.0742,\n", + " 0.0048, 0.1180, -0.1363, -0.0826, 0.3924, 0.1792, 0.0996,\n", + " -0.2099, -0.3327, 0.1132, 0.2122]]], grad_fn=)\n", + "torch.Size([2, 1, 32])\n" + ] + } + ], + "source": [ + "multihead_attn = nn.MultiheadAttention(32, 8)\n", + "query = np.random.rand(2, 1, 32)\n", + "key = np.random.rand(2, 1, 32)\n", + "value = np.random.rand(2, 1, 32)\n", + "attn_output, attn_output_weights = multihead_attn(Tensor(query), Tensor(key), Tensor(value))\n", + "print(attn_output)\n", + "print(attn_output.shape)" + ] + }, + { + "cell_type": "code", + "execution_count": 12, + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "(2, 1, 32)\n", + "[[[ 0.21025246 -1.3815589 1.1856667 0.5420509 -0.03878744\n", + " 0.36193654 -0.30533046 -1.1700163 -0.57054275 0.46141142\n", + " -0.30795845 0.39367586 -0.54798263 0.19380882 -0.5910839\n", + " -0.4666907 -0.97481775 0.45212376 -0.10193141 -0.6575114\n", + " 0.25120044 0.122086 -1.1403954 -0.42054623 -0.61952204\n", + " -0.6566897 0.73287785 -0.35822588 -0.5958375 -0.32572877\n", + " 1.2399867 -0.22128716]]\n", + "\n", + " [[ 0.03517252 -0.6995417 0.97694 0.07682744 0.24728103\n", + " 0.08959051 0.01169628 -0.8235637 -0.2518189 0.59149474\n", + " -0.40409014 -0.5977189 0.09025317 0.08469346 -0.29417813\n", + " -0.34428942 -0.61110383 0.7962756 0.02414452 -0.67949474\n", + " 0.6065721 -0.24253327 -0.3226804 0.27078897 -1.3271445\n", + " 0.25307137 0.3808884 0.18227924 -0.38555723 -0.42521483\n", + " 1.2446275 -0.62895083]]]\n" + ] + } + ], + "source": [ + "import jax\n", + "import flax.linen as nn\n", + "from jax import numpy as jnp\n", + "from jax import random\n", + "\n", + "input = jnp.ones((2,1,32))\n", + "layer = nn.MultiHeadAttention(num_heads=8, qkv_features=32)\n", + "variables = layer.init(random.key(42), input)\n", + "output = layer.apply(variables, Tensor(query), Tensor(key), Tensor(value))\n", + "print(output.shape)\n", + "print(output)" + ] + }, + { + "cell_type": "code", + "execution_count": 13, + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "[[[ 0.00069493 -0.03911137 -0.17373352 -0.08817849 0.00585917\n", + " 0.0038011 0.13423112 0.12640306 -0.13031635 -0.10004204\n", + " -0.22439755 -0.07857274 -0.31675625 -0.07869545 0.18808702\n", + " 0.18783928 0.21366805 0.16640219 -0.2740562 0.02806476\n", + " -0.16295458 0.29520345 0.17730789 0.13442674 0.35357815\n", + " 0.15538093 0.08111625 -0.09462467 -0.03054206 0.1470299\n", + " 0.03899852 0.20147526]]\n", + "\n", + " [[-0.00058219 -0.03654152 -0.17798103 -0.08727801 0.00280233\n", + " 0.00116073 0.1349529 0.12789461 -0.13350655 -0.10066106\n", + " -0.2256635 -0.07729645 -0.31817535 -0.0761395 0.19332808\n", + " 0.19098276 0.21913132 0.16640013 -0.27903354 0.03269039\n", + " -0.16899411 0.2947319 0.18080507 0.13609359 0.34764683\n", + " 0.16034965 0.08051767 -0.09571798 -0.03570132 0.15065208\n", + " 0.03331698 0.19594805]]]\n", + "(2, 1, 32)\n" + ] + } + ], + "source": [ + "import mindspore as ms\n", + "from mindspore import nn\n", + "from mindspore import Tensor\n", + "\n", + "multihead_attn = nn.MultiheadAttention(32, 8)\n", + "attn_output, attn_output_weights = multihead_attn(Tensor(query, ms.float32), Tensor(key, ms.float32), Tensor(value, ms.float32))\n", + "print(attn_output)\n", + "print(attn_output.shape)" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "**MindSpore MHA用法与输出与PyTorch一致**\n", + "- MindSpore的Tensor()没有像torch.Tensor()一样内置类型转换，传入float64会报错" + ] + } + ], + "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": "3.11.11" + } + }, + "nbformat": 4, + "nbformat_minor": 4 +} diff --git a/api-examples/nn/nn.relu.ipynb b/api-examples/nn/nn.relu.ipynb new file mode 100644 index 0000000..c112926 --- /dev/null +++ b/api-examples/nn/nn.relu.ipynb @@ -0,0 +1,109 @@ +{ + "cells": [ + { + "cell_type": "code", + "execution_count": 1, + "metadata": {}, + "outputs": [], + "source": [ + "import numpy as np\n", + "import torch\n", + "from torch import Tensor\n", + "from torch import nn" + ] + }, + { + "cell_type": "code", + "execution_count": 4, + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "tensor([0.7546, 0.1698, 0.7326, 0.5446, 0.9181, 0.1017, 0.3542, 0.0074, 0.8615,\n", + " 0.0965])\n" + ] + } + ], + "source": [ + "# ReLU\n", + "input_np = np.random.rand(10)\n", + "output = nn.ReLU()(Tensor(input_np))\n", + "print(output)" + ] + }, + { + "cell_type": "code", + "execution_count": 6, + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "[0.75460404 0.16979711 0.7325533 0.544558 0.9180574 0.1017127\n", + " 0.35418442 0.00744265 0.86148655 0.09653632]\n" + ] + } + ], + "source": [ + "import jax\n", + "output = jax.nn.relu(input_np)\n", + "print(output)" + ] + }, + { + "cell_type": "code", + "execution_count": 5, + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "[0.75460406 0.16979711 0.7325533 0.544558 0.91805736 0.1017127\n", + " 0.35418441 0.00744265 0.86148653 0.09653632]\n" + ] + } + ], + "source": [ + "import mindspore as ms\n", + "from mindspore import nn\n", + "from mindspore import Tensor\n", + "output = nn.ReLU()(Tensor(input_np))\n", + "print(output)" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "**MindSpore ReLU用法与输出与PyTorch一致**\n", + "- MindSpore的输出打印为list格式，与jax一致；PyTorch打印了tensor类信息，元素之间有逗号分隔\n", + "- jax可以直接输入numpy向量" + ] + } + ], + "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": "3.11.11" + } + }, + "nbformat": 4, + "nbformat_minor": 4 +} diff --git a/api-examples/nn/nn.transformer.ipynb b/api-examples/nn/nn.transformer.ipynb new file mode 100644 index 0000000..3ec48af --- /dev/null +++ b/api-examples/nn/nn.transformer.ipynb @@ -0,0 +1,108 @@ +{ + "cells": [ + { + "cell_type": "code", + "execution_count": 8, + "metadata": {}, + "outputs": [], + "source": [ + "import numpy as np\n", + "import torch\n", + "from torch import Tensor\n", + "from torch import nn\n", + "\n", + "import time" + ] + }, + { + "cell_type": "code", + "execution_count": 9, + "metadata": {}, + "outputs": [ + { + "name": "stderr", + "output_type": "stream", + "text": [ + "C:\\Users\\wangzeyangyi\\anaconda3\\envs\\tutorial\\Lib\\site-packages\\torch\\nn\\modules\\transformer.py:379: UserWarning: enable_nested_tensor is True, but self.use_nested_tensor is False because encoder_layer.self_attn.batch_first was not True(use batch_first for better inference performance)\n", + " warnings.warn(\n" + ] + }, + { + "name": "stdout", + "output_type": "stream", + "text": [ + "torch.Size([20, 32, 512])\n", + "2.4219019412994385\n" + ] + } + ], + "source": [ + "start = time.time()\n", + "transformer_model = nn.Transformer(nhead=16, num_encoder_layers=12)\n", + "src = np.random.rand(10, 32, 512)\n", + "tgt = np.random.rand(20, 32, 512)\n", + "out = transformer_model(Tensor(src), Tensor(tgt))\n", + "print(out.size())\n", + "end = time.time()\n", + "print(end-start)" + ] + }, + { + "cell_type": "code", + "execution_count": 10, + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "(20, 32, 512)\n", + "13.059895038604736\n" + ] + } + ], + "source": [ + "import mindspore as ms\n", + "from mindspore import nn\n", + "from mindspore import Tensor\n", + "\n", + "start = time.time()\n", + "transformer_model = nn.Transformer(nhead=16, num_encoder_layers=12)\n", + "out = transformer_model(Tensor(src, ms.float32), Tensor(tgt, ms.float32))\n", + "print(out.shape)\n", + "end = time.time()\n", + "print(end-start)" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "**MindSpore transformer用法与输出与PyTorch一致**\n", + "- MindSpore transformer在CPU上的性能待优化（~6x PyTorch）\n", + "- jax没有官方transformer API" + ] + } + ], + "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": "3.11.11" + } + }, + "nbformat": 4, + "nbformat_minor": 4 +} diff --git a/api-examples/nn/nn.upsample.ipynb b/api-examples/nn/nn.upsample.ipynb new file mode 100644 index 0000000..794c7b5 --- /dev/null +++ b/api-examples/nn/nn.upsample.ipynb @@ -0,0 +1,136 @@ +{ + "cells": [ + { + "cell_type": "code", + "execution_count": 1, + "metadata": {}, + "outputs": [], + "source": [ + "import numpy as np\n", + "import torch\n", + "from torch import Tensor\n", + "from torch import nn" + ] + }, + { + "cell_type": "code", + "execution_count": 4, + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "[[[[1 1]\n", + " [1 0]]]]\n", + "tensor([[[[1., 1., 1., 1.],\n", + " [1., 1., 1., 1.],\n", + " [1., 1., 0., 0.],\n", + " [1., 1., 0., 0.]]]])\n", + "tensor([[[[1.0000, 1.0000, 1.0000, 1.0000],\n", + " [1.0000, 0.9375, 0.8125, 0.7500],\n", + " [1.0000, 0.8125, 0.4375, 0.2500],\n", + " [1.0000, 0.7500, 0.2500, 0.0000]]]])\n", + "tensor([[[[1.0000, 1.0000, 1.0000, 1.0000],\n", + " [1.0000, 0.8889, 0.7778, 0.6667],\n", + " [1.0000, 0.7778, 0.5556, 0.3333],\n", + " [1.0000, 0.6667, 0.3333, 0.0000]]]])\n" + ] + } + ], + "source": [ + "# input = torch.arange(1, 5, dtype=torch.float32).view(1, 1, 2, 2)\n", + "input = np.random.randint(2, size=(1,1,2,2))\n", + "print(input)\n", + "\n", + "m = nn.Upsample(scale_factor=2, mode='nearest')\n", + "print(m(Tensor(input)))\n", + "\n", + "m = nn.Upsample(scale_factor=2, mode='bilinear') # align_corners=False\n", + "print(m(Tensor(input)))\n", + "\n", + "m = nn.Upsample(scale_factor=2, mode='bilinear', align_corners=True)\n", + "print(m(Tensor(input)))\n", + "\n", + "# # Try scaling the same data in a larger tensor\n", + "# input_3x3 = torch.zeros(3, 3).view(1, 1, 3, 3)\n", + "# input_3x3[:, :, :2, :2].copy_(input)\n", + "# input_3x3\n", + "\n", + "# m = nn.Upsample(scale_factor=2, mode='bilinear') # align_corners=False\n", + "# # Notice that values in top left corner are the same with the small input (except at boundary)\n", + "# m(input_3x3)\n", + "\n", + "# m = nn.Upsample(scale_factor=2, mode='bilinear', align_corners=True)\n", + "# # Notice that values in top left corner are now changed\n", + "# m(input_3x3)" + ] + }, + { + "cell_type": "code", + "execution_count": 6, + "metadata": {}, + "outputs": [ + { + "ename": "ValueError", + "evalue": "For 'interpolate', 'scale_factor' option cannot currently be set with the mode = nearest and dim = 4D.", + "output_type": "error", + "traceback": [ + "\u001b[1;31m---------------------------------------------------------------------------\u001b[0m", + "\u001b[1;31mValueError\u001b[0m Traceback (most recent call last)", + "Cell \u001b[1;32mIn[6], line 6\u001b[0m\n\u001b[0;32m 3\u001b[0m \u001b[38;5;28;01mfrom\u001b[39;00m \u001b[38;5;21;01mmindspore\u001b[39;00m \u001b[38;5;28;01mimport\u001b[39;00m Tensor\n\u001b[0;32m 5\u001b[0m m \u001b[38;5;241m=\u001b[39m nn\u001b[38;5;241m.\u001b[39mUpsample(scale_factor\u001b[38;5;241m=\u001b[39m\u001b[38;5;241m2.0\u001b[39m, mode\u001b[38;5;241m=\u001b[39m\u001b[38;5;124m'\u001b[39m\u001b[38;5;124mnearest\u001b[39m\u001b[38;5;124m'\u001b[39m)\n\u001b[1;32m----> 6\u001b[0m \u001b[38;5;28mprint\u001b[39m(m(Tensor(\u001b[38;5;28minput\u001b[39m)))\n\u001b[0;32m 8\u001b[0m m \u001b[38;5;241m=\u001b[39m nn\u001b[38;5;241m.\u001b[39mUpsample(scale_factor\u001b[38;5;241m=\u001b[39m\u001b[38;5;241m2.0\u001b[39m, mode\u001b[38;5;241m=\u001b[39m\u001b[38;5;124m'\u001b[39m\u001b[38;5;124mbilinear\u001b[39m\u001b[38;5;124m'\u001b[39m) \u001b[38;5;66;03m# align_corners=False\u001b[39;00m\n\u001b[0;32m 9\u001b[0m \u001b[38;5;28mprint\u001b[39m(m(Tensor(\u001b[38;5;28minput\u001b[39m)))\n", + "File \u001b[1;32m~\\anaconda3\\envs\\tutorial\\Lib\\site-packages\\mindspore\\nn\\cell.py:733\u001b[0m, in \u001b[0;36mCell.__call__\u001b[1;34m(self, *args, **kwargs)\u001b[0m\n\u001b[0;32m 730\u001b[0m \u001b[38;5;28;01mif\u001b[39;00m \u001b[38;5;129;01mnot\u001b[39;00m (\u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39mrequires_grad \u001b[38;5;129;01mor\u001b[39;00m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_dynamic_shape_inputs \u001b[38;5;129;01mor\u001b[39;00m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39mmixed_precision_type):\n\u001b[0;32m 731\u001b[0m \u001b[38;5;28;01mif\u001b[39;00m \u001b[38;5;129;01mnot\u001b[39;00m (\u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_forward_pre_hook \u001b[38;5;129;01mor\u001b[39;00m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_forward_hook \u001b[38;5;129;01mor\u001b[39;00m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_backward_pre_hook \u001b[38;5;129;01mor\u001b[39;00m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_backward_hook \u001b[38;5;129;01mor\u001b[39;00m\n\u001b[0;32m 732\u001b[0m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_shard_fn \u001b[38;5;129;01mor\u001b[39;00m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_recompute_cell \u001b[38;5;129;01mor\u001b[39;00m (\u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39mhas_bprop \u001b[38;5;129;01mand\u001b[39;00m _pynative_executor\u001b[38;5;241m.\u001b[39mrequires_grad())):\n\u001b[1;32m--> 733\u001b[0m \u001b[38;5;28;01mreturn\u001b[39;00m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39mconstruct(\u001b[38;5;241m*\u001b[39margs, \u001b[38;5;241m*\u001b[39m\u001b[38;5;241m*\u001b[39mkwargs)\n\u001b[0;32m 735\u001b[0m \u001b[38;5;28;01mreturn\u001b[39;00m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_run_construct(\u001b[38;5;241m*\u001b[39margs, \u001b[38;5;241m*\u001b[39m\u001b[38;5;241m*\u001b[39mkwargs)\n\u001b[0;32m 737\u001b[0m \u001b[38;5;28;01mreturn\u001b[39;00m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_complex_call(\u001b[38;5;241m*\u001b[39margs, \u001b[38;5;241m*\u001b[39m\u001b[38;5;241m*\u001b[39mkwargs)\n", + "File \u001b[1;32m~\\anaconda3\\envs\\tutorial\\Lib\\site-packages\\mindspore\\nn\\layer\\basic.py:473\u001b[0m, in \u001b[0;36mUpsample.construct\u001b[1;34m(self, x)\u001b[0m\n\u001b[0;32m 472\u001b[0m \u001b[38;5;28;01mdef\u001b[39;00m \u001b[38;5;21mconstruct\u001b[39m(\u001b[38;5;28mself\u001b[39m, x):\n\u001b[1;32m--> 473\u001b[0m out \u001b[38;5;241m=\u001b[39m F\u001b[38;5;241m.\u001b[39minterpolate(x, \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39msize, \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39mscale_factor, \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39mmode,\n\u001b[0;32m 474\u001b[0m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39malign_corners, \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39mrecompute_scale_factor)\n\u001b[0;32m 475\u001b[0m \u001b[38;5;28;01mreturn\u001b[39;00m out\n", + "File \u001b[1;32m~\\anaconda3\\envs\\tutorial\\Lib\\site-packages\\mindspore\\ops\\function\\nn_func.py:2613\u001b[0m, in \u001b[0;36minterpolate\u001b[1;34m(input, size, scale_factor, mode, align_corners, recompute_scale_factor)\u001b[0m\n\u001b[0;32m 2611\u001b[0m \u001b[38;5;28;01melse\u001b[39;00m:\n\u001b[0;32m 2612\u001b[0m \u001b[38;5;28;01mif\u001b[39;00m dim_unknown \u001b[38;5;129;01mis\u001b[39;00m \u001b[38;5;28;01mFalse\u001b[39;00m:\n\u001b[1;32m-> 2613\u001b[0m _interpolate_scale_factor_check(scale_factor, mode, rank,\n\u001b[0;32m 2614\u001b[0m supported_dict)\n\u001b[0;32m 2616\u001b[0m \u001b[38;5;66;03m# align_corners\u001b[39;00m\n\u001b[0;32m 2617\u001b[0m \u001b[38;5;28;01mif\u001b[39;00m align_corners \u001b[38;5;129;01mis\u001b[39;00m \u001b[38;5;129;01mnot\u001b[39;00m \u001b[38;5;28;01mNone\u001b[39;00m:\n", + "File \u001b[1;32m~\\anaconda3\\envs\\tutorial\\Lib\\site-packages\\mindspore\\ops\\function\\nn_func.py:2293\u001b[0m, in \u001b[0;36m_interpolate_scale_factor_check\u001b[1;34m(scale_factor, mode, rank, supported_dict)\u001b[0m\n\u001b[0;32m 2288\u001b[0m \u001b[38;5;250m\u001b[39m\u001b[38;5;124;03m\"\"\"\u001b[39;00m\n\u001b[0;32m 2289\u001b[0m \u001b[38;5;124;03mscale_factor check\u001b[39;00m\n\u001b[0;32m 2290\u001b[0m \u001b[38;5;124;03m\"\"\"\u001b[39;00m\n\u001b[0;32m 2291\u001b[0m \u001b[38;5;28;01mif\u001b[39;00m scale_factor \u001b[38;5;129;01mis\u001b[39;00m \u001b[38;5;129;01mnot\u001b[39;00m \u001b[38;5;28;01mNone\u001b[39;00m \u001b[38;5;129;01mand\u001b[39;00m \u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mscale_factor\u001b[39m\u001b[38;5;124m\"\u001b[39m \u001b[38;5;129;01mnot\u001b[39;00m \u001b[38;5;129;01min\u001b[39;00m supported_dict\u001b[38;5;241m.\u001b[39mget(\n\u001b[0;32m 2292\u001b[0m mode, {})\u001b[38;5;241m.\u001b[39mget(rank):\n\u001b[1;32m-> 2293\u001b[0m \u001b[38;5;28;01mraise\u001b[39;00m \u001b[38;5;167;01mValueError\u001b[39;00m(\n\u001b[0;32m 2294\u001b[0m \u001b[38;5;124mf\u001b[39m\u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mFor \u001b[39m\u001b[38;5;124m'\u001b[39m\u001b[38;5;124minterpolate\u001b[39m\u001b[38;5;124m'\u001b[39m\u001b[38;5;124m, \u001b[39m\u001b[38;5;124m'\u001b[39m\u001b[38;5;124mscale_factor\u001b[39m\u001b[38;5;124m'\u001b[39m\u001b[38;5;124m option cannot currently be set with the \u001b[39m\u001b[38;5;124m\"\u001b[39m\n\u001b[0;32m 2295\u001b[0m \u001b[38;5;124mf\u001b[39m\u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mmode = \u001b[39m\u001b[38;5;132;01m{\u001b[39;00mmode\u001b[38;5;132;01m}\u001b[39;00m\u001b[38;5;124m and dim = \u001b[39m\u001b[38;5;132;01m{\u001b[39;00mrank\u001b[38;5;132;01m}\u001b[39;00m\u001b[38;5;124mD.\u001b[39m\u001b[38;5;124m\"\u001b[39m)\n", + "\u001b[1;31mValueError\u001b[0m: For 'interpolate', 'scale_factor' option cannot currently be set with the mode = nearest and dim = 4D." + ] + } + ], + "source": [ + "import mindspore as ms\n", + "from mindspore import nn\n", + "from mindspore import Tensor\n", + "\n", + "m = nn.Upsample(scale_factor=2.0, mode='nearest')\n", + "print(m(Tensor(input)))\n", + "\n", + "m = nn.Upsample(scale_factor=2.0, mode='bilinear') # align_corners=False\n", + "print(m(Tensor(input)))\n", + "\n", + "m = nn.Upsample(scale_factor=2.0, mode='bilinear', align_corners=True)\n", + "print(m(Tensor(input)))" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "- jax没有upsample接口\n", + "- MindSpore的scale_factor参数必须为float\n", + "- MindSpore部分输入不支持" + ] + } + ], + "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": "3.11.11" + } + }, + "nbformat": 4, + "nbformat_minor": 4 +}