From 3ec7e5fed5c765384add0efa3d16c956750a28db Mon Sep 17 00:00:00 2001 From: jlevine18 Date: Sun, 22 Sep 2019 23:05:49 -0500 Subject: [PATCH] added cuda to cudaregress notebook --- data analysis/cudaRegressTesting.ipynb | 75 +++++++++++++++++--------- 1 file changed, 49 insertions(+), 26 deletions(-) diff --git a/data analysis/cudaRegressTesting.ipynb b/data analysis/cudaRegressTesting.ipynb index 4a7b481e..a684148f 100644 --- a/data analysis/cudaRegressTesting.ipynb +++ b/data analysis/cudaRegressTesting.ipynb @@ -6,12 +6,13 @@ "metadata": {}, "outputs": [], "source": [ - "import torch" + "import torch\n", + "device='cuda:0' if torch.cuda.is_available() else 'cpu'" ] }, { "cell_type": "code", - "execution_count": 2, + "execution_count": 29, "metadata": {}, "outputs": [], "source": [ @@ -29,13 +30,16 @@ " out *= i\n", " return out\n", "\n", + "def set_device(new_device):\n", + " device=new_device\n", + "\n", "class LinearRegKernel():\n", " parameters= []\n", " weights=None\n", " bias=None\n", " def __init__(self, num_vars):\n", - " self.weights=torch.rand(num_vars, requires_grad=True)\n", - " self.bias=torch.rand(1, requires_grad=True)\n", + " self.weights=torch.rand(num_vars, requires_grad=True, device=device)\n", + " self.bias=torch.rand(1, requires_grad=True, device=device)\n", " self.parameters=[self.weights,self.bias]\n", " def forward(self,mtx):\n", " long_bias=self.bias.repeat([1,mtx.size()[1]])\n", @@ -47,8 +51,8 @@ " bias=None\n", " sigmoid=torch.nn.Sigmoid()\n", " def __init__(self, num_vars):\n", - " self.weights=torch.rand(num_vars)\n", - " self.bias=torch.rand(1)\n", + " self.weights=torch.rand(num_vars, requires_grad=True, device=device)\n", + " self.bias=torch.rand(1, requires_grad=True, device=device)\n", " self.parameters=[self.weights,self.bias]\n", " def forward(self,mtx):\n", " long_bias=self.bias.repeat([1,mtx.size()[1]])\n", @@ -59,8 +63,8 @@ " weights=None\n", " bias=None\n", " def __init__(self, num_vars):\n", - " self.weights=torch.rand(num_vars)\n", - " self.bias=torch.rand(1)\n", + " self.weights=torch.rand(num_vars, requires_grad=True, device=device)\n", + " self.bias=torch.rand(1, requires_grad=True, device=device)\n", " self.parameters=[self.weights,self.bias]\n", " def forward(self,mtx):\n", " long_bias=self.bias.repeat([1,mtx.size()[1]])\n", @@ -71,8 +75,8 @@ " weights=None\n", " bias=None\n", " def __init__(self, num_vars):\n", - " self.weights=torch.rand(num_vars)\n", - " self.bias=torch.rand(1)\n", + " self.weights=torch.rand(num_vars, requires_grad=True, device=device)\n", + " self.bias=torch.rand(1, requires_grad=True, device=device)\n", " self.parameters=[self.weights,self.bias]\n", " def forward(self,mtx):\n", " long_bias=self.bias.repeat([1,mtx.size()[1]])\n", @@ -86,8 +90,8 @@ " def __init__(self, num_vars, power):\n", " self.power=power\n", " num_terms=int(factorial(num_vars+power-1) / factorial(power) / factorial(num_vars-1))\n", - " self.weights=torch.rand(num_terms)\n", - " self.bias=torch.rand(1)\n", + " self.weights=torch.rand(num_terms, requires_grad=True, device=device)\n", + " self.bias=torch.rand(1, requires_grad=True, device=device)\n", " self.parameters=[self.weights,self.bias]\n", " def forward(self,mtx):\n", " #TODO: Vectorize the last part\n", @@ -100,13 +104,15 @@ "\n", "def SGDTrain(kernel, data, ground, loss=torch.nn.MSELoss(), iterations=1000, learning_rate=.1, return_losses=False):\n", " optim=torch.optim.SGD(kernel.parameters, lr=learning_rate)\n", + " data_cuda=data.to(device)\n", + " ground_cuda=ground.to(device)\n", " if (return_losses):\n", " losses=[]\n", " for i in range(iterations):\n", " with torch.set_grad_enabled(True):\n", " optim.zero_grad()\n", - " pred=kernel.forward(data)\n", - " ls=loss(pred,ground)\n", + " pred=kernel.forward(data_cuda)\n", + " ls=loss(pred,ground_cuda)\n", " losses.append(ls.item())\n", " ls.backward()\n", " optim.step()\n", @@ -115,13 +121,15 @@ " for i in range(iterations):\n", " with torch.set_grad_enabled(True):\n", " optim.zero_grad()\n", - " pred=kernel.forward(data)\n", - " ls=loss(pred,ground)\n", + " pred=kernel.forward(data_cuda)\n", + " ls=loss(pred,ground_cuda)\n", " ls.backward()\n", " optim.step() \n", " return kernel\n", "\n", "def CustomTrain(kernel, optim, data, ground, loss=torch.nn.MSELoss(), iterations=1000, return_losses=False):\n", + " data_cuda=data.to(device)\n", + " ground_cuda=ground.to(device)\n", " if (return_losses):\n", " losses=[]\n", " for i in range(iterations):\n", @@ -137,8 +145,8 @@ " for i in range(iterations):\n", " with torch.set_grad_enabled(True):\n", " optim.zero_grad()\n", - " pred=kernel.forward(data)\n", - " ls=loss(pred,ground)\n", + " pred=kernel.forward(data_cuda)\n", + " ls=loss(pred,ground_cuda)\n", " ls.backward()\n", " optim.step() \n", " return kernel" @@ -146,31 +154,46 @@ }, { "cell_type": "code", - "execution_count": 3, + "execution_count": 31, "metadata": {}, "outputs": [ { "data": { "text/plain": [ - "tensor([[1.0000, 2.0000]], grad_fn=)" + "tensor([[1.0000, 2.0000]], device='cuda:0', grad_fn=)" ] }, - "execution_count": 3, + "execution_count": 31, "metadata": {}, "output_type": "execute_result" } ], "source": [ - "model=SGDTrain(LinearRegKernel(3),torch.tensor([[1,2],[3,4],[5,6]]).to(torch.float),torch.tensor([[1,2]]).to(torch.float),iterations=10000, learning_rate=.01, return_losses=True)\n", - "model[0].forward(torch.tensor([[1,2],[3,4],[5,6]]).to(torch.float))" + "model=SGDTrain(LinearRegKernel(3),torch.tensor([[1,2],[3,4],[5,6]]).to(torch.float).cuda(),torch.tensor([[1,2]]).to(torch.float).cuda(),iterations=10000, learning_rate=.01, return_losses=True)\n", + "model[0].forward(torch.tensor([[1,2],[3,4],[5,6]]).to(torch.float).cuda())" ] }, { "cell_type": "code", - "execution_count": null, + "execution_count": 30, "metadata": {}, - "outputs": [], - "source": [] + "outputs": [ + { + "data": { + "text/plain": [ + "[tensor([0.2347, 0.4494, 0.3156], device='cuda:0', requires_grad=True),\n", + " tensor([0.9541], device='cuda:0', requires_grad=True)]" + ] + }, + "execution_count": 30, + "metadata": {}, + "output_type": "execute_result" + } + ], + "source": [ + "kernel=LinearRegKernel(3)\n", + "kernel.parameters\n" + ] }, { "cell_type": "code",