publish 0.0.9

docs/source/contact.rst

@@ -1,4 +1,4 @@
 Contact
 ============
 
-This project is mainly developed by Yu Pan etc.. Feel free to contact me at iperryuu@gmail.com.
+This project is mainly developed by Yu Pan(iperryuu@gmail.com) and Maolin Wang(morin.w98@gmail.com), etc. Feel free to contact us.

docs/source/quick_start.ipynb

@@ -0,0 +1,107 @@
+{
+ "cells": [
+  {
+   "cell_type": "markdown",
+   "metadata": {
+    "colab_type": "text",
+    "id": "gqwrcHFAVwgs"
+   },
+   "source": [
+    "# Quick Start"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {
+    "colab_type": "text",
+    "id": "oz-X6ZrGVwgu"
+   },
+   "source": [
+    "In this quick starting guide we show the basics of working with t3f library. The main concept of the library is a TensorTrain object – a compact (factorized) representation of a tensor (=multidimensional array). This is generalization of the matrix low-rank decomposition."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 1,
+   "metadata": {
+    "colab": {
+     "base_uri": "https://localhost:8080/",
+     "height": 34
+    },
+    "colab_type": "code",
+    "id": "ri9QCNEAVwgw",
+    "outputId": "efd8c64b-836e-449b-a385-dd3a63a5b4a2"
+   },
+   "outputs": [],
+   "source": [
+    "\n",
+    "import tednet as tdt\n",
+    "import tednet.tnn.tensor_ring as tr"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 2,
+   "metadata": {
+    "colab": {},
+    "colab_type": "code",
+    "id": "5Awp7wdwVwg3"
+   },
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "compression_ration is:  0.3968253968253968\n",
+      "compression_ration is:  14.17233560090703\n",
+      "compression_ration is:  241.54589371980677\n",
+      "compression_ration is:  2.867383512544803\n"
+     ]
+    }
+   ],
+   "source": [
+    "# Define a TR-LeNet5\n",
+    "model = tr.TRLeNet5(10, [6, 6, 6, 6])"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {
+    "colab_type": "text",
+    "id": "i2QmXYqeVwhQ",
+    "pycharm": {
+     "name": "#%% md\n"
+    }
+   },
+   "source": [
+    "It is easy to define a tensor ring model.\n"
+   ]
+  }
+ ],
+ "metadata": {
+  "colab": {
+   "collapsed_sections": [],
+   "name": "riemannian.ipynb",
+   "provenance": []
+  },
+  "kernelspec": {
+   "display_name": "Python 3",
+   "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": "3.7.9"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 1
+}

ext/tests/tt_opt.py

@@ -0,0 +1,13 @@
+# -*- coding: UTF-8 -*-
+
+import tednet
+
+import torch
+
+a = torch.Tensor(6, 8)
+b = tednet.to_numpy(a)
+c = tednet.to_tensor(b)
+d = c.cuda()
+e = tednet.to_numpy(d)
+f = tednet.to_tensor(e)
+

tednet/__init__.py

@@ -2,8 +2,8 @@
 
 
 __author__ = "Perry"
-__version__ = "0.0.8"
+__version__ = "0.0.9"
 
 from ._ops import *
 
-__all__ = ["hard_sigmoid", "eye"]
+__all__ = ["hard_sigmoid", "eye", "to_numpy", "to_tensor"]

tednet/_ops.py

@@ -1,13 +1,16 @@
 # -*- coding: UTF-8 -*-
 
+from collections import namedtuple
+
+import numpy as np
+
 import torch
 import torch.nn as nn
 from torch.nn import Parameter
 import torch.nn.functional as F
-from collections import namedtuple
 
 
-def hard_sigmoid(tensor: torch.Tensor):
+def hard_sigmoid(tensor: torch.Tensor) -> torch.Tensor:
     """Computes element-wise hard sigmoid of x.
     See e.g. https://github.com/Theano/Theano/blob/master/theano/tensor/nnet/sigm.py#L279
 
@@ -27,7 +30,7 @@ def hard_sigmoid(tensor: torch.Tensor):
     return tensor
 
 
-def eye(n: int, m: int, device: torch.device = "cpu", requires_grad: bool=False):
+def eye(n: int, m: int, device: torch.device = "cpu", requires_grad: bool=False) -> torch.Tensor:
     """Returns a 2-D tensor with ones on the diagonal and zeros elsewhere.
 
     Parameters
@@ -44,31 +47,45 @@ def eye(n: int, m: int, device: torch.device = "cpu", requires_grad: bool=False)
     Returns
     -------
     torch.Tensor
-        2-D tensor :math:`\in \mathbb{R}^{{i_1} \\times {i_2}}`
+        2-D tensor :math:`\in \mathbb{R}^{{n} \\times {m}}`
     """
     return torch.eye(n=n, m=m, device=device, requires_grad=requires_grad)
 
 
-def ones(x):
-    # TODO:
-    pass
+def to_numpy(tensor: torch.Tensor) -> np.ndarray:
+    """Convert torch.Tensor to numpy variable.
 
+    Parameters
+    ----------
+    tensor : torch.Tensor
+             tensor :math:`\in \mathbb{R}^{{i_1} \\times \dots \\times {i_n}}`
 
-def prod_all(x):
-    # TODO:
-    pass
+    Returns
+    -------
+    numpy.ndarray
+        arr :math:`\in \mathbb{R}^{{i_1} \\times \dots \\times {i_n}}`
+    """
+    if tensor.device.type == "cpu":
+        arr = tensor.numpy()
+    else:
+        arr = tensor.cpu().numpy()
 
+    return arr
 
-def diag(x):
-    # TODO:
-    pass
 
+def to_tensor(arr: np.ndarray) -> torch.Tensor:
+    """Convert numpy variable to torch.Tensor.
 
-def tensordot(x):
-    # TODO:
-    pass
+    Parameters
+    ----------
+    arr : numpy.ndarray
+             arr :math:`\in \mathbb{R}^{{i_1} \\times \dots \\times {i_n}}`
 
+    Returns
+    -------
+    torch.Tensor
+        tensor :math:`\in \mathbb{R}^{{i_1} \\times \dots \\times {i_n}}`
+    """
+    tensor = torch.from_numpy(arr)
 
-def from_numpy(x):
-    # TODO:
-    pass
+    return tensor

tednet/tnn/tn_module.py

@@ -8,7 +8,7 @@
 
 import numpy as np
 
-__all__ = ["_TNBase"]
+__all__ = ["_TNBase", "LambdaLayer"]
 
 
 class _TNBase(nn.Module, metaclass=abc.ABCMeta):
@@ -139,7 +139,7 @@ def recover(self):
 
 class LambdaLayer(nn.Module):
     def __init__(self, lambd):
-        """Tensor Ring Block.
+        """A layer consists of Lambda function.
 
         Parameters
         ----------