Merge pull request #80 from lucasdelimanogueira/tmp

Tmp

README.md

@@ -7,7 +7,7 @@ Project details explanations can also be found on [medium](https://towardsdatasc
 **PyNorch** is a deep learning framework constructed using C/C++, CUDA and Python. This is a personal project with educational purpose only! `Norch` means **NOT** PyTorch, and we have **NO** claims to rivaling the already established PyTorch. The main objective of **PyNorch** was to give a brief understanding of how a deep learning framework works internally. It implements the Tensor object, GPU support and an automatic differentiation system. 
 
 # 2 - Installation
-Install this package from PyPi (you can test on Colab!)
+Install this package from PyPi (you can test on Colab! Also tested on AWS p3 instances ami-00f1d513c2bb78c75)
 
 ```css
 $ pip install norch
@@ -15,11 +15,9 @@ $ pip install norch
 
 or from cloning this repository
 ```css
-$ sudo apt install nvidia-cuda-toolkit
 $ git clone https://github.com/lucasdelimanogueira/PyNorch.git
-$ cd build
-$ make
-$ cd ..
+$ cd PyNorch
+$ pip install . -v
 ```
 
 # 3 - Get started
@@ -66,7 +64,7 @@ class MyModel(nn.Module):
 ```python
 import norch
 from norch.utils.data.dataloader import DataLoader
-from norch.norchvision import transforms
+from norch.norchvision import transforms as T
 import norch
 import norch.nn as nn
 import norch.optim as optim
@@ -77,16 +75,16 @@ BATCH_SIZE = 32
 device = "cuda" #cpu
 epochs = 10
 
-transform = transforms.Compose(
+transform = T.Compose(
     [
-        transforms.ToTensor(),
-        transforms.Reshape([-1, 784, 1])
+        T.ToTensor(),
+        T.Reshape([-1, 784, 1])
     ]
 )
 
-target_transform = transforms.Compose(
+target_transform = T.Compose(
     [
-        transforms.ToTensor()
+        T.ToTensor()
     ]
 )
 

build/Makefile

@@ -13,7 +13,7 @@ CUDAFLAGS = -arch=sm_75
 CUDALIBS = -I/usr/local/cuda/include -L/usr/local/cuda/lib64 -lcudart -lcuda
 
 # MPI flags and libraries
-DISTRIBUTEDLIBS = -lmpi_cxx -lnccl
+DISTRIBUTEDLIBS = -lmpi_cxx -lnccl -lmpi
 
 # Directories
 SRCDIR = ../norch/csrc

examples/train.ipynb

@@ -34,7 +34,7 @@
     "import norch.nn as nn\n",
     "import norch.optim as optim\n",
     "from norch.utils.data.dataloader import DataLoader\n",
-    "from norch.norchvision import transforms\n",
+    "from norch.norchvision import transforms as T\n",
     "import numpy as np\n",
     "import matplotlib.pyplot as plt\n",
     "import random\n",
@@ -102,16 +102,16 @@
     "device = \"cuda\" #cpu\n",
     "epochs = 10\n",
     "\n",
-    "transform = transforms.Compose(\n",
+    "transform = T.Compose(\n",
     "    [\n",
-    "        transforms.ToTensor(),\n",
-    "        transforms.Reshape([-1, 784, 1])\n",
+    "        T.ToTensor(),\n",
+    "        T.Reshape([-1, 784, 1])\n",
     "    ]\n",
     ")\n",
     "\n",
-    "target_transform = transforms.Sequential(\n",
+    "target_transform = T.Compose(\n",
     "    [\n",
-    "        transforms.ToTensor()\n",
+    "        T.ToTensor()\n",
     "    ]\n",
     ")\n",
     "\n",

norch/csrc/tensor.cpp

@@ -66,9 +66,12 @@ extern "C" {
     void to_device(Tensor* tensor, char* target_device) {
         int device_id = 0;
         char* endptr;
+
         long num = strtol(target_device, &endptr, 10);
         if (*endptr == '\0') {
             device_id = (int)num;
+            target_device = new char[strlen("cuda") + 1];
+            strcpy(target_device, "cuda");
         }
 
         if ((strcmp(target_device, "cuda") == 0) && (strcmp(tensor->device, "cpu") == 0)) {
@@ -78,6 +81,10 @@ extern "C" {
         else if ((strcmp(target_device, "cpu") == 0) && (strcmp(tensor->device, "cuda") == 0)) {
             cuda_to_cpu(tensor);
         }
+
+        else {
+            printf("Could not send tensor to device %d", device_id);
+        }
     }
 
     Tensor* add_tensor(Tensor* tensor1, Tensor* tensor2) {