Created
May 21, 2018 23:37
-
-
Save tbenst/476f26dc96c0ab2d9b0d6c13b9fa6d43 to your computer and use it in GitHub Desktop.
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
{ | |
"cells": [ | |
{ | |
"cell_type": "code", | |
"execution_count": 1, | |
"metadata": {}, | |
"outputs": [], | |
"source": [ | |
"import torch as T\n", | |
"import torch\n", | |
"import torch.nn.functional as F\n", | |
"import torch.nn as nn\n", | |
"from torch.utils.data import DataLoader, Dataset\n", | |
"import numpy as np\n", | |
"from __future__ import print_function\n", | |
"import gc" | |
] | |
}, | |
{ | |
"cell_type": "markdown", | |
"metadata": {}, | |
"source": [ | |
"$x_{t+1} = (A + pB)x_t + Cu_{t+1}$" | |
] | |
}, | |
{ | |
"cell_type": "code", | |
"execution_count": 2, | |
"metadata": {}, | |
"outputs": [], | |
"source": [ | |
"class MyData(Dataset): \n", | |
" def __init__(self, u, p, x,n_future_steps=1):\n", | |
" self.x = nn.Parameter(x,requires_grad=False)\n", | |
" self.p = nn.Parameter(p,requires_grad=False)\n", | |
" self.u = nn.Parameter(u,requires_grad=False)\n", | |
" self.nfeatures = x.shape[1]\n", | |
" self.n_future_steps = n_future_steps\n", | |
" \n", | |
" def __len__(self):\n", | |
" return len(self.x)-self.n_future_steps\n", | |
"\n", | |
" def __getitem__(self, idx):\n", | |
" indices = slice(idx,idx+self.n_future_steps)\n", | |
" x_true_indices = slice(idx+1,idx+self.n_future_steps+1)\n", | |
" return (self.u[indices], self.p[indices],\n", | |
" self.x[indices], self.x[x_true_indices])\n", | |
"\n", | |
"class Dynamics(nn.Module):\n", | |
" def __init__(self, nfeatures):\n", | |
" super(Dynamics, self).__init__()\n", | |
" self.A = nn.Parameter(T.normal(T.zeros(nfeatures,nfeatures),0.5),requires_grad=True)\n", | |
" self.B = nn.Parameter(T.normal(T.zeros(nfeatures,nfeatures),0.5),requires_grad=True)\n", | |
" self.C = nn.Parameter(T.normal(T.zeros(nfeatures),0.5),requires_grad=True)\n", | |
"\n", | |
" def forward(self, u, p, x):\n", | |
" return (x[:,0] + (T.matmul((self.A + p[:,0,None,None]*self.B), x[:,0,:,None]).squeeze()) + u[:,0,None] * self.C)[:,None]\n", | |
"\n", | |
"\n", | |
"def train(model,data,nepochs=10, lambdaA=1e-8, lambdaB=1e-6, lr=0.001):\n", | |
" dataloader = DataLoader(data, batch_size=batch_size, shuffle=True)\n", | |
" optimizer = T.optim.Adam(model.parameters(),lr=lr)\n", | |
" og_mem = T.cuda.memory_allocated() / 1024**2\n", | |
" print(\"Allocated Memory: {} MB\".format(og_mem))\n", | |
" for e in range(nepochs):\n", | |
" for batch_data in dataloader:\n", | |
" U,P,X, X_true = batch_data\n", | |
" X_pred = model(U,P,X)\n", | |
" mse_loss = F.mse_loss(X_pred,X_true)\n", | |
" l1_B = model.B.norm(1)\n", | |
" l1_A = model.A.norm(1)\n", | |
" \n", | |
" loss = mse_loss + lambdaA*l1_A + lambdaB*l1_B\n", | |
" optimizer.zero_grad()\n", | |
" loss.backward()\n", | |
" optimizer.step()\n", | |
" \n", | |
" del X_pred, U,P,X, X_true, mse_loss, l1_A, l1_B, loss\n", | |
" gc.collect()\n", | |
" torch.cuda.empty_cache()\n", | |
" \n", | |
" mem = T.cuda.memory_allocated() / 1024**2\n", | |
" print(\"New allocations: {} MB\".format(mem-og_mem))\n", | |
" og_mem = mem\n" | |
] | |
}, | |
{ | |
"cell_type": "code", | |
"execution_count": 3, | |
"metadata": {}, | |
"outputs": [], | |
"source": [ | |
"ntime = 2826\n", | |
"nstim = 30\n", | |
"nfeatures = 15888\n", | |
"\n", | |
"u_train = T.rand(ntime).cuda()\n", | |
"p_train = T.rand(ntime).cuda()\n", | |
"time_train = T.from_numpy(np.arange(ntime)).cuda()\n", | |
"x_train = T.rand(ntime,nfeatures).cuda()" | |
] | |
}, | |
{ | |
"cell_type": "code", | |
"execution_count": 4, | |
"metadata": {}, | |
"outputs": [ | |
{ | |
"data": { | |
"text/plain": [ | |
"Dynamics()" | |
] | |
}, | |
"execution_count": 4, | |
"metadata": {}, | |
"output_type": "execute_result" | |
} | |
], | |
"source": [ | |
"n_future_steps = 1\n", | |
"batch_size = 1\n", | |
"\n", | |
"data = MyData(u_train,p_train,x_train,n_future_steps)\n", | |
"model = Dynamics(data.nfeatures)\n", | |
"model.to(\"cuda\")" | |
] | |
}, | |
{ | |
"cell_type": "code", | |
"execution_count": 5, | |
"metadata": {}, | |
"outputs": [ | |
{ | |
"name": "stdout", | |
"output_type": "stream", | |
"text": [ | |
"Allocated Memory: 2098 MB\n", | |
"New allocations: 0 MB\n", | |
"New allocations: 0 MB\n", | |
"New allocations: 0 MB\n", | |
"New allocations: 0 MB\n", | |
"New allocations: 0 MB\n", | |
"New allocations: 0 MB\n", | |
"New allocations: 0 MB\n", | |
"New allocations: 0 MB\n", | |
"New allocations: 0 MB\n", | |
"New allocations: 0 MB\n" | |
] | |
} | |
], | |
"source": [ | |
"og_mem = T.cuda.memory_allocated() / 1024**2\n", | |
"print(\"Allocated Memory: {} MB\".format(og_mem))\n", | |
"with T.no_grad():\n", | |
" for t in range(10):\n", | |
" x_pred = T.squeeze(model(u_train[[t],None], p_train[[t],None], \n", | |
" x_train[[t],None]))\n", | |
" mem = T.cuda.memory_allocated() / 1024**2\n", | |
" print(\"New allocations: {} MB\".format(mem-og_mem))" | |
] | |
}, | |
{ | |
"cell_type": "code", | |
"execution_count": 6, | |
"metadata": {}, | |
"outputs": [ | |
{ | |
"name": "stdout", | |
"output_type": "stream", | |
"text": [ | |
"Allocated Memory: 2098 MB\n", | |
"New allocations: 5778 MB\n" | |
] | |
}, | |
{ | |
"ename": "RuntimeError", | |
"evalue": "cuda runtime error (2) : out of memory at /opt/conda/conda-bld/pytorch_1524577523076/work/aten/src/THC/generic/THCStorage.cu:58", | |
"output_type": "error", | |
"traceback": [ | |
"\u001b[0;31m---------------------------------------------------------------------------\u001b[0m", | |
"\u001b[0;31mRuntimeError\u001b[0m Traceback (most recent call last)", | |
"\u001b[0;32m<ipython-input-6-f76860546b30>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[0;32m----> 1\u001b[0;31m \u001b[0mtrain\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mmodel\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0mdata\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;36m1\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;36m1e-4\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;36m1e-6\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0mlr\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;36m0.001\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m", | |
"\u001b[0;32m<ipython-input-2-d5d0d1975ac8>\u001b[0m in \u001b[0;36mtrain\u001b[0;34m(model, data, nepochs, lambdaA, lambdaB, lr)\u001b[0m\n\u001b[1;32m 42\u001b[0m \u001b[0mloss\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mmse_loss\u001b[0m \u001b[0;34m+\u001b[0m \u001b[0mlambdaA\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0ml1_A\u001b[0m \u001b[0;34m+\u001b[0m \u001b[0mlambdaB\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0ml1_B\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 43\u001b[0m \u001b[0moptimizer\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mzero_grad\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 44\u001b[0;31m \u001b[0mloss\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mbackward\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 45\u001b[0m \u001b[0moptimizer\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mstep\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 46\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n", | |
"\u001b[0;32m/home/ubuntu/anaconda3/envs/pytorch_p27/lib/python2.7/site-packages/torch/tensor.pyc\u001b[0m in \u001b[0;36mbackward\u001b[0;34m(self, gradient, retain_graph, create_graph)\u001b[0m\n\u001b[1;32m 91\u001b[0m \u001b[0mproducts\u001b[0m\u001b[0;34m.\u001b[0m \u001b[0mDefaults\u001b[0m \u001b[0mto\u001b[0m \u001b[0;34m`\u001b[0m\u001b[0;34m`\u001b[0m\u001b[0mFalse\u001b[0m\u001b[0;34m`\u001b[0m\u001b[0;34m`\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 92\u001b[0m \"\"\"\n\u001b[0;32m---> 93\u001b[0;31m \u001b[0mtorch\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mautograd\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mbackward\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mgradient\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mretain_graph\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mcreate_graph\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 94\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 95\u001b[0m \u001b[0;32mdef\u001b[0m \u001b[0mregister_hook\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mhook\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n", | |
"\u001b[0;32m/home/ubuntu/anaconda3/envs/pytorch_p27/lib/python2.7/site-packages/torch/autograd/__init__.pyc\u001b[0m in \u001b[0;36mbackward\u001b[0;34m(tensors, grad_tensors, retain_graph, create_graph, grad_variables)\u001b[0m\n\u001b[1;32m 87\u001b[0m Variable._execution_engine.run_backward(\n\u001b[1;32m 88\u001b[0m \u001b[0mtensors\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mgrad_tensors\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mretain_graph\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mcreate_graph\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 89\u001b[0;31m allow_unreachable=True) # allow_unreachable flag\n\u001b[0m\u001b[1;32m 90\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 91\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n", | |
"\u001b[0;31mRuntimeError\u001b[0m: cuda runtime error (2) : out of memory at /opt/conda/conda-bld/pytorch_1524577523076/work/aten/src/THC/generic/THCStorage.cu:58" | |
] | |
} | |
], | |
"source": [ | |
"train(model,data,1,1e-4,1e-6,lr=0.001)" | |
] | |
}, | |
{ | |
"cell_type": "code", | |
"execution_count": 7, | |
"metadata": {}, | |
"outputs": [ | |
{ | |
"name": "stdout", | |
"output_type": "stream", | |
"text": [ | |
"<class 'torch.Tensor'> torch.Size([1, 1])\n", | |
"<class 'torch.Tensor'> torch.Size([1, 1])\n", | |
"<class 'torch.Tensor'> torch.Size([1, 1, 15888])\n", | |
"<class 'torch.Tensor'> torch.Size([1, 1, 15888])\n", | |
"<class 'torch.Tensor'> torch.Size([1, 1, 15888])\n", | |
"<class 'torch.Tensor'> torch.Size([])\n", | |
"<class 'torch.Tensor'> torch.Size([])\n", | |
"<class 'torch.Tensor'> torch.Size([])\n", | |
"<class 'torch.Tensor'> torch.Size([])\n", | |
"<class 'torch.Tensor'> torch.Size([])\n", | |
"<class 'torch.Tensor'> torch.Size([2826])\n", | |
"<class 'torch.Tensor'> torch.Size([2826])\n", | |
"<class 'torch.Tensor'> torch.Size([2826])\n", | |
"<class 'torch.Tensor'> torch.Size([2826, 15888])\n", | |
"<class 'torch.nn.parameter.Parameter'> torch.Size([2826])\n", | |
"<class 'torch.nn.parameter.Parameter'> torch.Size([2826])\n", | |
"<class 'torch.nn.parameter.Parameter'> torch.Size([15888, 15888])\n", | |
"<class 'torch.nn.parameter.Parameter'> torch.Size([15888, 15888])\n", | |
"<class 'torch.nn.parameter.Parameter'> torch.Size([15888])\n", | |
"<class 'torch.Tensor'> torch.Size([15888])\n", | |
"<class 'torch.Tensor'> torch.Size([15888, 15888])\n", | |
"<class 'torch.Tensor'> torch.Size([15888, 15888])\n", | |
"<class 'torch.Tensor'> torch.Size([15888])\n", | |
"<class 'torch.nn.parameter.Parameter'> torch.Size([2826, 15888])\n", | |
"<class 'torch.Tensor'> torch.Size([15888, 15888])\n", | |
"<class 'torch.Tensor'> torch.Size([15888, 15888])\n", | |
"<class 'torch.Tensor'> torch.Size([15888])\n" | |
] | |
} | |
], | |
"source": [ | |
"for obj in gc.get_objects():\n", | |
" if T.is_tensor(obj) or (hasattr(obj, 'data') and T.is_tensor(obj.data)):\n", | |
" print(type(obj), obj.size())" | |
] | |
}, | |
{ | |
"cell_type": "code", | |
"execution_count": null, | |
"metadata": {}, | |
"outputs": [], | |
"source": [] | |
} | |
], | |
"metadata": { | |
"kernelspec": { | |
"display_name": "Environment (conda_pytorch_p27)", | |
"language": "python", | |
"name": "conda_pytorch_p27" | |
}, | |
"language_info": { | |
"codemirror_mode": { | |
"name": "ipython", | |
"version": 2 | |
}, | |
"file_extension": ".py", | |
"mimetype": "text/x-python", | |
"name": "python", | |
"nbconvert_exporter": "python", | |
"pygments_lexer": "ipython2", | |
"version": "2.7.14" | |
} | |
}, | |
"nbformat": 4, | |
"nbformat_minor": 2 | |
} |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment