05_anneal.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "%load_ext autoreload\n",
    "%autoreload 2\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "#export\n",
    "from exp.nb_04 import *"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Initial setup"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "x_train, y_train, x_valid, y_valid = get_data()\n",
    "train_ds, valid_ds = Dataset(x_train, y_train), Dataset(x_valid, y_valid)\n",
    "nh, bs = 50, 512\n",
    "c = y_train.max().item() + 1\n",
    "loss_func = F.cross_entropy"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "data = DataBunch(*get_dls(train_ds, valid_ds, bs), c)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "#export\n",
    "def create_learner(model_func, loss_func, data):\n",
    "    return Learner(*model_func(data), loss_func, data)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train: [0.6647103125, tensor(0.8097)]\n",
      "valid: [0.3747031982421875, tensor(0.8839)]\n",
      "train: [0.30409294921875, tensor(0.9114)]\n",
      "valid: [0.2523560791015625, tensor(0.9284)]\n",
      "train: [0.24483654296875, tensor(0.9298)]\n",
      "valid: [0.2151006591796875, tensor(0.9420)]\n"
     ]
    }
   ],
   "source": [
    "learn = create_learner(get_model, loss_func, data) # default lr is 0.5 in get_model\n",
    "run = Runner([AvgStatsCallback([accuracy])])\n",
    "\n",
    "run.fit(3, learn)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train: [0.747547109375, tensor(0.8089)]\n",
      "valid: [0.352033447265625, tensor(0.9011)]\n",
      "train: [0.340920703125, tensor(0.9025)]\n",
      "valid: [0.27916748046875, tensor(0.9220)]\n",
      "train: [0.29106853515625, tensor(0.9168)]\n",
      "valid: [0.258016357421875, tensor(0.9297)]\n"
     ]
    }
   ],
   "source": [
    "learn = create_learner(partial(get_model, lr=0.3), loss_func, data)\n",
    "run = Runner([AvgStatsCallback([accuracy])])\n",
    "\n",
    "run.fit(3, learn)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "#export\n",
    "def get_model_func(lr=0.5): return partial(get_model, lr=lr)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Annealing"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We define two new callbacks: the Recorder to save track of the loss and our scheduled learning rate, and a ParamScheduler that can schedule any hyperparameter as long as it's registered in the state_dict of the optimizer. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "code_folding": []
   },
   "outputs": [],
   "source": [
    "#export\n",
    "class Recorder(Callback):\n",
    "    def begin_fit(self): self.lrs, self.losses = [], []\n",
    "        \n",
    "    def after_batch(self):\n",
    "        if not self.in_train: return\n",
    "        self.lrs.append(self.opt.param_groups[-1]['lr'])# In fastai we tend to use the lr of the final layer_group as we have a lot of diff lrs for different layer_groups.\n",
    "        self.losses.append(self.loss.detach().cpu())\n",
    "        \n",
    "    def plot_lr(self): plt.plot(self.lrs)\n",
    "        \n",
    "    def plot_loss(self): plt.plot(self.losses)\n",
    "class ParamScheduler(Callback):\n",
    "    _order = 1\n",
    "    def __init__(self, pname, sched_func): self.pname, self.sched_func = pname, sched_func\n",
    "    \n",
    "    def set_param(self):\n",
    "        for pg in self.opt.param_groups:# param_groups same as layer_groups in fastai\n",
    "            pg[self.pname] = self.sched_func(self.n_epochs/self.epochs)\n",
    "            \n",
    "    def begin_batch(self):\n",
    "        if self.in_train: self.set_param() "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We create a Recorder class as its pretty hard to know that something's working unless you can see it."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let's start with a simple linear schedule going from start to end. It returns a function that takes a `pos` argument (going from 0 to 1) such that this function goes from `start` (at `pos=0`) to `end` (at `pos=1`) in a linear fashion."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "def sched_lin(start, end):\n",
    "    def _inner(start, end, pos): return start + pos*(end-start)\n",
    "    return partial(_inner, start, end)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Refactoring it with a decorator"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "#export\n",
    "def annealer(f):\n",
    "    def _inner(start, end): return partial(f, start, end)\n",
    "    return _inner\n",
    "\n",
    "@annealer\n",
    "def sched_lin(start, end, pos):\n",
    "    return start + pos* (end-start)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "# sched_lin()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "f = sched_lin(1,2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(1.3, 1.5)"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "f(0.3), f(0.5)# what should lr be 30 & 50% of the way through training"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "And here are other scheduler functions:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "#export\n",
    "@annealer\n",
    "def sched_cos(start, end, pos): return start +(1+ math.cos(math.pi*(1-pos))) * (end-start) / 2\n",
    "\n",
    "@annealer\n",
    "def sched_exp(start, end, pos): return start * (end/start) ** pos\n",
    "\n",
    "@annealer\n",
    "def sched_no(start, end, pos): return start\n",
    "\n",
    "# This monkey patch is there to be able to plot tensors.\n",
    "# ndim tells matplotlib how many dimensions there are.\n",
    "torch.Tensor.ndim = property(lambda x: len(x.shape))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['NO', 'LINEAR', 'COS', 'EXP']"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "annealings = \"NO LINEAR COS EXP\". split()\n",
    "annealings"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor([0.0100, 0.1200, 0.2300, 0.3400, 0.4500, 0.5600, 0.6700, 0.7800, 0.8900,\n",
       "        1.0000])"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "p = torch.linspace(0.01, 1, 10)\n",
    "p"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "a = torch.arange(0,100)\n",
    "p = torch.linspace(0.01, 1, 100)\n",
    "\n",
    "fns = [sched_no, sched_lin, sched_cos, sched_exp]\n",
    "for fn, t in zip(fns, annealings):\n",
    "    f = fn(2, 1e-2)\n",
    "    plt.plot(a, [f(o) for o in p], label= t)\n",
    "    plt.legend()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "In practice, we'll often want to combine different schedulers, the following function does that: it uses `scheds[i]` for `pcts[i]` of the training."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "#export\n",
    "def combine_scheds(pcts, scheds):\n",
    "    assert sum(pcts) == 1.\n",
    "    pcts = tensor([0] + listify(pcts))\n",
    "    assert torch.all(pcts>= 0)\n",
    "    pcts = torch.cumsum(pcts, 0)\n",
    "    def _inner(pos):\n",
    "        idx = (pos>= pcts).nonzero().max()\n",
    "        actual_pos = (pos-pcts[idx])/ (pcts[idx+1]- pcts[idx])\n",
    "        return scheds[idx](actual_pos)\n",
    "    return _inner        "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Here is an example: use 30% of the budget(inputs/batches) to go from 0.3 to 0.6 following a cosine, then the last 70% of the budget to go from 0.6 to 0.2, still following a cosine."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "sched = combine_scheds([0.3, 0.7], [sched_cos(0.3, 0.6), sched_cos(0.6, 0.2)]) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x201da60bfd0>]"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(a, [sched(o) for o in p])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We can use it for training quite easily..."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "cbfs = [Recorder,\n",
    "       partial(AvgStatsCallback, accuracy),\n",
    "       partial(ParamScheduler, 'lr', sched)]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [],
   "source": [
    "learn = create_learner(get_model_func(0.3), loss_func, data)\n",
    "run = Runner(cb_funcs=cbfs)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train: [0.770986328125, tensor(0.7930)]\n",
      "valid: [0.4053299072265625, tensor(0.8682)]\n",
      "train: [0.2997938671875, tensor(0.9120)]\n",
      "valid: [0.2445391845703125, tensor(0.9301)]\n",
      "train: [0.24155150390625, tensor(0.9308)]\n",
      "valid: [0.2158880615234375, tensor(0.9372)]\n"
     ]
    }
   ],
   "source": [
    "run.fit(3, learn)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "run.recorder.plot_lr()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "run.recorder.plot_loss()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "heading_collapsed": true
   },
   "source": [
    "## Export"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {
    "hidden": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Converted 05_anneal.ipynb to exp\\nb_05.py\n"
     ]
    }
   ],
   "source": [
    "!python notebook2script.py 05_anneal.ipynb"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "hidden": true
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "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": "3.7.3"
  },
  "toc": {
   "base_numbering": 1,
   "nav_menu": {},
   "number_sections": true,
   "sideBar": true,
   "skip_h1_title": false,
   "title_cell": "Table of Contents",
   "title_sidebar": "Contents",
   "toc_cell": false,
   "toc_position": {},
   "toc_section_display": true,
   "toc_window_display": false
  },
  "varInspector": {
   "cols": {
    "lenName": 16,
    "lenType": 16,
    "lenVar": 40
   },
   "kernels_config": {
    "python": {
     "delete_cmd_postfix": "",
     "delete_cmd_prefix": "del ",
     "library": "var_list.py",
     "varRefreshCmd": "print(var_dic_list())"
    },
    "r": {
     "delete_cmd_postfix": ") ",
     "delete_cmd_prefix": "rm(",
     "library": "var_list.r",
     "varRefreshCmd": "cat(var_dic_list()) "
    }
   },
   "types_to_exclude": [
    "module",
    "function",
    "builtin_function_or_method",
    "instance",
    "_Feature"
   ],
   "window_display": false
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}