750 lines
166 KiB
Plaintext
750 lines
166 KiB
Plaintext
{
|
||
"cells": [
|
||
{
|
||
"cell_type": "code",
|
||
"execution_count": 1,
|
||
"metadata": {},
|
||
"outputs": [],
|
||
"source": [
|
||
"import os, pickle, sys\n",
|
||
"import matplotlib.pyplot as plt\n",
|
||
"from scipy import stats\n",
|
||
"import numpy as np\n",
|
||
"import glob\n",
|
||
"from prettytable import PrettyTable\n",
|
||
"from tqdm import tqdm"
|
||
]
|
||
},
|
||
{
|
||
"cell_type": "markdown",
|
||
"metadata": {},
|
||
"source": [
|
||
"## Table 1: Spearman ρ of zero-cost proxies on NAS-Bench-201."
|
||
]
|
||
},
|
||
{
|
||
"cell_type": "code",
|
||
"execution_count": 2,
|
||
"metadata": {},
|
||
"outputs": [
|
||
{
|
||
"name": "stdout",
|
||
"output_type": "stream",
|
||
"text": [
|
||
"../results_release/nasbench2/nb2_cf10_seed42_dlrandom_dlinfo1_initwnone_initbnone.p 15625\n",
|
||
"../results_release/nasbench2/nb2_cf100_seed42_dlrandom_dlinfo1_initwnone_initbnone.p 15625\n",
|
||
"../results_release/nasbench2/nb2_im120_seed42_dlrandom_dlinfo1_initwnone_initbnone.p 15625\n",
|
||
"+----------------+-----------+-------+-------+--------+---------+-----------+\n",
|
||
"| Dataset | grad_norm | snip | grasp | fisher | synflow | jacob_cov |\n",
|
||
"+----------------+-----------+-------+-------+--------+---------+-----------+\n",
|
||
"| CIFAR10 | 0.594 | 0.596 | 0.514 | 0.36 | 0.737 | 0.731 |\n",
|
||
"| CIFAR100 | 0.637 | 0.637 | 0.547 | 0.385 | 0.763 | 0.704 |\n",
|
||
"| ImageNet16-120 | 0.578 | 0.578 | 0.549 | 0.327 | 0.751 | 0.701 |\n",
|
||
"+----------------+-----------+-------+-------+--------+---------+-----------+\n"
|
||
]
|
||
}
|
||
],
|
||
"source": [
|
||
"t=None\n",
|
||
"all_ds = {}\n",
|
||
"all_acc = {}\n",
|
||
"allc = {}\n",
|
||
"all_metrics = {}\n",
|
||
"all_runs = {}\n",
|
||
"metric_names = ['grad_norm', 'snip', 'grasp', 'fisher', 'synflow', 'jacob_cov']\n",
|
||
"for fname,rname in [('../results_release/nasbench2/nb2_cf10_seed42_dlrandom_dlinfo1_initwnone_initbnone.p','CIFAR10'),\n",
|
||
" ('../results_release/nasbench2/nb2_cf100_seed42_dlrandom_dlinfo1_initwnone_initbnone.p','CIFAR100'),\n",
|
||
" ('../results_release/nasbench2/nb2_im120_seed42_dlrandom_dlinfo1_initwnone_initbnone.p','ImageNet16-120')]:\n",
|
||
" runs=[]\n",
|
||
" f = open(fname,'rb')\n",
|
||
" while(1):\n",
|
||
" try:\n",
|
||
" runs.append(pickle.load(f))\n",
|
||
" except EOFError:\n",
|
||
" break\n",
|
||
" f.close()\n",
|
||
" print(fname, len(runs))\n",
|
||
" \n",
|
||
" all_runs[fname]=runs\n",
|
||
" all_ds[fname] = {}\n",
|
||
" metrics={}\n",
|
||
" for k in metric_names:\n",
|
||
" metrics[k] = []\n",
|
||
" acc = []\n",
|
||
" \n",
|
||
" if t is None:\n",
|
||
" hl=['Dataset']\n",
|
||
" hl.extend(metric_names)\n",
|
||
" t = PrettyTable(hl)\n",
|
||
" \n",
|
||
" for r in runs:\n",
|
||
" for k,v in r['logmeasures'].items():\n",
|
||
" if k in metrics:\n",
|
||
" metrics[k].append(v)\n",
|
||
" acc.append(r['testacc'])\n",
|
||
" \n",
|
||
" all_ds[fname]['metrics'] = metrics\n",
|
||
" all_ds[fname]['acc'] = acc\n",
|
||
" \n",
|
||
" res = []\n",
|
||
" crs = {}\n",
|
||
" for k in hl:\n",
|
||
" if k=='Dataset':\n",
|
||
" continue\n",
|
||
" v = metrics[k]\n",
|
||
" cr = abs(stats.spearmanr(acc,v,nan_policy='omit').correlation)\n",
|
||
" #print(f'{k} = {cr}')\n",
|
||
" res.append(round(cr,3))\n",
|
||
" crs[k]=cr\n",
|
||
" \n",
|
||
" ds = rname\n",
|
||
" all_acc[ds]=acc\n",
|
||
" allc[ds]=crs\n",
|
||
" t.add_row([ds]+res)\n",
|
||
" \n",
|
||
" all_metrics[ds] = metrics\n",
|
||
"print(t)"
|
||
]
|
||
},
|
||
{
|
||
"cell_type": "markdown",
|
||
"metadata": {},
|
||
"source": [
|
||
"### Voting between 3 metrics could improve rank correlation"
|
||
]
|
||
},
|
||
{
|
||
"cell_type": "code",
|
||
"execution_count": 3,
|
||
"metadata": {},
|
||
"outputs": [
|
||
{
|
||
"name": "stderr",
|
||
"output_type": "stream",
|
||
"text": [
|
||
"100%|██████████| 15625/15625 [08:13<00:00, 31.65it/s]\n",
|
||
"100%|██████████| 15625/15625 [08:17<00:00, 31.41it/s]\n",
|
||
"100%|██████████| 15625/15625 [08:20<00:00, 31.20it/s]"
|
||
]
|
||
},
|
||
{
|
||
"name": "stdout",
|
||
"output_type": "stream",
|
||
"text": [
|
||
"votes correlation: {'cifar10': 0.8170822831897641, 'cifar100': 0.8323757385510576, 'ImageNet16-120': 0.8131110314104887}\n"
|
||
]
|
||
},
|
||
{
|
||
"name": "stderr",
|
||
"output_type": "stream",
|
||
"text": [
|
||
"\n"
|
||
]
|
||
}
|
||
],
|
||
"source": [
|
||
"from tqdm import tqdm\n",
|
||
"votes = {}\n",
|
||
"def vote(mets, gt):\n",
|
||
" numpos = 0\n",
|
||
" for m in mets:\n",
|
||
" numpos += 1 if m > 0 else 0\n",
|
||
" if numpos >= len(mets)/2:\n",
|
||
" sign = +1\n",
|
||
" else:\n",
|
||
" sign = -1\n",
|
||
" return sign*gt\n",
|
||
"\n",
|
||
"for ds in all_acc.keys():\n",
|
||
" num_pts = 15625\n",
|
||
" #num_pts = 1000\n",
|
||
" tot=0\n",
|
||
" right=0\n",
|
||
" for i in tqdm(range(num_pts)):\n",
|
||
" for j in range(num_pts):\n",
|
||
" if i!=j:\n",
|
||
" diff = all_acc[ds][i] - all_acc[ds][j]\n",
|
||
" if diff == 0:\n",
|
||
" continue\n",
|
||
" diffsyn = []\n",
|
||
" for m in ['synflow', 'jacob_cov', 'snip']:\n",
|
||
" diffsyn.append(all_metrics[ds][m][i] - all_metrics[ds][m][j])\n",
|
||
" same_sign = vote(diffsyn, diff)\n",
|
||
" right += 1 if same_sign > 0 else 0\n",
|
||
" tot += 1\n",
|
||
" votes[ds.lower() if 'CIFAR' in ds else ds] = right/tot\n",
|
||
"print('votes correlation: ', votes)"
|
||
]
|
||
},
|
||
{
|
||
"cell_type": "markdown",
|
||
"metadata": {},
|
||
"source": [
|
||
"## Figure 1: Evaluation of different econas proxies on NAS-Bench-201 CIFAR-10"
|
||
]
|
||
},
|
||
{
|
||
"cell_type": "code",
|
||
"execution_count": 4,
|
||
"metadata": {},
|
||
"outputs": [
|
||
{
|
||
"name": "stdout",
|
||
"output_type": "stream",
|
||
"text": [
|
||
"try to create the NAS-Bench-201 api from ../data/NAS-Bench-201-v1_0-e61699.pth\n"
|
||
]
|
||
}
|
||
],
|
||
"source": [
|
||
"from nas_201_api import NASBench201API as API\n",
|
||
"api = API('../data/NAS-Bench-201-v1_0-e61699.pth')\n",
|
||
"api.verbose = False"
|
||
]
|
||
},
|
||
{
|
||
"cell_type": "code",
|
||
"execution_count": 5,
|
||
"metadata": {
|
||
"scrolled": true
|
||
},
|
||
"outputs": [
|
||
{
|
||
"name": "stderr",
|
||
"output_type": "stream",
|
||
"text": [
|
||
"100%|██████████| 15625/15625 [00:04<00:00, 3694.77it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:04<00:00, 3677.47it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:04<00:00, 3666.86it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:04<00:00, 3668.30it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:04<00:00, 3647.79it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:04<00:00, 3645.00it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:04<00:00, 3633.39it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:04<00:00, 3639.16it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:04<00:00, 3630.43it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:04<00:00, 3618.66it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:04<00:00, 3606.21it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:04<00:00, 3591.64it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 6387.22it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 6375.77it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 6371.41it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 6363.89it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 6352.84it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 6370.86it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 6365.73it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 6330.58it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 6358.99it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 6357.03it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 6369.05it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 6348.56it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 6362.67it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 6343.42it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 6341.09it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 6350.24it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 6364.69it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 6329.40it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 6323.95it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 6292.45it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 6297.17it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 6354.38it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 6375.51it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 6360.75it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 6350.71it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 6312.46it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 6349.09it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 6367.49it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 6360.85it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 5456.53it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 5444.93it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 5434.87it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 5437.96it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 5429.18it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 5397.78it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 5414.16it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 5470.91it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 5439.19it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 5432.42it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 5428.60it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 5416.43it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 5435.99it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 5432.32it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 5435.25it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 5429.95it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 5430.57it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 5434.24it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 5515.30it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 5447.61it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 5494.53it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 5402.25it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 5421.46it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 5409.47it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 5428.66it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 5416.57it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 5416.47it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 5421.63it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 5408.41it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 5441.42it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 5421.06it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 5415.65it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 5414.25it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 5407.01it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 5408.28it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 5380.57it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 5371.12it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 5381.57it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 5413.72it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 5430.10it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 5438.85it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 5464.67it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 5454.57it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 5454.08it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 5439.15it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 5452.53it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 5438.12it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 5437.44it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 5412.62it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 5432.36it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 5406.65it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 5410.13it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 5404.87it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 5453.51it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 5434.85it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 5441.20it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 5439.81it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 5449.05it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 5451.43it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 5445.37it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 5443.61it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 5425.72it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 5437.65it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 5430.78it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 5434.60it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 5433.80it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 5435.14it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 5430.05it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 5404.00it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 5428.38it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 5413.04it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 5427.40it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 5421.51it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 5435.61it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 5427.46it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 5442.03it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 5430.49it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 5435.78it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 5393.37it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 5399.36it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 5413.35it/s]\n",
|
||
"100%|██████████| 15625/15625 [00:02<00:00, 5410.74it/s]\n"
|
||
]
|
||
}
|
||
],
|
||
"source": [
|
||
"dallb={}\n",
|
||
"dallb_10={}\n",
|
||
"dallb_10f={}\n",
|
||
"for ds in ['cifar10', 'cifar100','ImageNet16-120']:\n",
|
||
" allb = {}\n",
|
||
" allb_10 = {}\n",
|
||
" allb_10f = {}\n",
|
||
"\n",
|
||
" for k in range(0,41):\n",
|
||
"\n",
|
||
" b=[]\n",
|
||
" b_10 = []\n",
|
||
" b_10f = []\n",
|
||
"\n",
|
||
" for i in tqdm(range(len(api))):\n",
|
||
" info = api.get_more_info(i, 'cifar10-valid' if ds=='cifar10' else ds, iepoch=None, hp='200', is_random=False)\n",
|
||
" info_10 = api.get_more_info(i, 'cifar10-valid' if ds=='cifar10' else ds, iepoch=k, hp='200', is_random=False)\n",
|
||
"\n",
|
||
" try:\n",
|
||
" info_10_fast = api.get_more_info(i, 'cifar10-valid' if ds=='cifar10' else ds, iepoch=k, hp='12', is_random=False)\n",
|
||
" testacc_10_fast = info_10_fast['valid-accuracy' if ds=='cifar10' else 'valtest-accuracy']\n",
|
||
" except Exception:\n",
|
||
" pass\n",
|
||
" \n",
|
||
" testacc = info['test-accuracy']\n",
|
||
" testacc_10 = info_10['valid-accuracy' if ds=='cifar10' else 'valtest-accuracy']\n",
|
||
"\n",
|
||
" b.append(testacc)\n",
|
||
" b_10.append(testacc_10) \n",
|
||
" b_10f.append(testacc_10_fast)\n",
|
||
"\n",
|
||
" allb[k] = b\n",
|
||
" allb_10[k] = b_10\n",
|
||
" allb_10f[k] = b_10f\n",
|
||
" dallb[ds]=allb\n",
|
||
" dallb_10[ds]=allb_10\n",
|
||
" dallb_10f[ds]=allb_10f"
|
||
]
|
||
},
|
||
{
|
||
"cell_type": "code",
|
||
"execution_count": 6,
|
||
"metadata": {
|
||
"scrolled": true
|
||
},
|
||
"outputs": [
|
||
{
|
||
"name": "stderr",
|
||
"output_type": "stream",
|
||
"text": [
|
||
"/home/SERILOCAL/mohamed1.a/anaconda3/envs/snip-torch/lib/python3.7/site-packages/numpy/lib/function_base.py:2534: RuntimeWarning: invalid value encountered in true_divide\n",
|
||
" c /= stddev[:, None]\n",
|
||
"/home/SERILOCAL/mohamed1.a/anaconda3/envs/snip-torch/lib/python3.7/site-packages/numpy/lib/function_base.py:2535: RuntimeWarning: invalid value encountered in true_divide\n",
|
||
" c /= stddev[None, :]\n",
|
||
"/home/SERILOCAL/mohamed1.a/anaconda3/envs/snip-torch/lib/python3.7/site-packages/scipy/stats/_distn_infrastructure.py:903: RuntimeWarning: invalid value encountered in greater\n",
|
||
" return (a < x) & (x < b)\n",
|
||
"/home/SERILOCAL/mohamed1.a/anaconda3/envs/snip-torch/lib/python3.7/site-packages/scipy/stats/_distn_infrastructure.py:903: RuntimeWarning: invalid value encountered in less\n",
|
||
" return (a < x) & (x < b)\n",
|
||
"/home/SERILOCAL/mohamed1.a/anaconda3/envs/snip-torch/lib/python3.7/site-packages/scipy/stats/_distn_infrastructure.py:1912: RuntimeWarning: invalid value encountered in less_equal\n",
|
||
" cond2 = cond0 & (x <= _a)\n"
|
||
]
|
||
}
|
||
],
|
||
"source": [
|
||
"dslow = {}\n",
|
||
"dfast = {}\n",
|
||
"for ds,allb in dallb.items():\n",
|
||
" dslow[ds] = []\n",
|
||
" dfast[ds] = []\n",
|
||
" t = PrettyTable(['Epoch', 'Normal Training (200 epochs)', 'Fast Training (12 Epochs)'])\n",
|
||
" for k,b in allb.items():\n",
|
||
" r = [k]\n",
|
||
" for v in [dallb_10[ds][k], dallb_10f[ds][k]]:\n",
|
||
" cr = abs(stats.spearmanr(b,v,nan_policy='omit').correlation)\n",
|
||
" r.append(round(cr,3))\n",
|
||
" t.add_row(r)\n",
|
||
" dslow[ds].append(r[1])\n",
|
||
" dfast[ds].append(r[2])"
|
||
]
|
||
},
|
||
{
|
||
"cell_type": "code",
|
||
"execution_count": 7,
|
||
"metadata": {},
|
||
"outputs": [
|
||
{
|
||
"data": {
|
||
"image/png": "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\n",
|
||
"text/plain": [
|
||
"<Figure size 770.4x259.2 with 3 Axes>"
|
||
]
|
||
},
|
||
"metadata": {
|
||
"needs_background": "light"
|
||
},
|
||
"output_type": "display_data"
|
||
}
|
||
],
|
||
"source": [
|
||
"with open('../results_release/nasbench2/nb2_fast_train.p','rb') as f:\n",
|
||
" truns = pickle.load(f)\n",
|
||
" \n",
|
||
"per_epochs=[]\n",
|
||
"for expname in ['nb2_fast_train', 'nb2_fast_train_ch8', 'nb2_fast_train_im16', 'nb2_fast_train_im16_ch8']:\n",
|
||
"\n",
|
||
" with open(f'../results_release/nasbench2/{expname}.p','rb') as f:\n",
|
||
" truns = pickle.load(f)\n",
|
||
"\n",
|
||
" #form array per epoch for fast training\n",
|
||
" per_epoch = {}\n",
|
||
" for r in truns:\n",
|
||
" for i,e in enumerate(r['logmeasures']):\n",
|
||
" if i not in per_epoch:\n",
|
||
" per_epoch[i] = []\n",
|
||
" per_epoch[i].append(e['val_acc'])\n",
|
||
" per_epochs.append(per_epoch)\n",
|
||
" \n",
|
||
"ds = 'cifar10'\n",
|
||
"econas = []\n",
|
||
"acc = dallb[ds][0]\n",
|
||
"for exp in per_epochs:\n",
|
||
" l = []\n",
|
||
" t = PrettyTable(['Epoch', 'Correlation of Proxy Training'])\n",
|
||
" for k,b in exp.items():\n",
|
||
" r = [k]\n",
|
||
" cr = abs(stats.spearmanr(b,acc,nan_policy='omit').correlation)\n",
|
||
" r.append(round(cr,3))\n",
|
||
" t.add_row(r)\n",
|
||
" l.append(cr)\n",
|
||
" econas.append(l)\n",
|
||
"\n",
|
||
"#WE ONLY IMPLEMENT ECONAS FOR CIFAR10\n",
|
||
"\n",
|
||
"ls = ['solid','dotted','dashed','dashdot',(0, (3, 5, 1, 5, 1, 5))]\n",
|
||
"\n",
|
||
"ds = 'cifar10'\n",
|
||
"slow=dslow[ds]\n",
|
||
"fig, axs = plt.subplots(1,3,figsize=(10.7,3.6), sharey=True)\n",
|
||
"\n",
|
||
"ax=axs[0]\n",
|
||
"\n",
|
||
"epx=41\n",
|
||
"\n",
|
||
"#regular training\n",
|
||
"x=range(0,epx,1)\n",
|
||
"ax.plot(x,slow[0:epx], label= 'baseline $r_{32}c_{16}$', linestyle=ls[0])\n",
|
||
"\n",
|
||
"#econas\n",
|
||
"x2=range(0,epx,1)\n",
|
||
"ax.plot(x2,econas[0][0:epx], label='econas $r_8c_4$', linestyle=ls[1])\n",
|
||
"ax.plot(x2,econas[1][0:epx], label='econas $r_8c_8$', linestyle=ls[2])\n",
|
||
"ax.plot(x2,econas[2][0:epx], label='econas $r_{16}c_4$', linestyle=ls[3])\n",
|
||
"ax.plot(x2,econas[3][0:epx], label='econas $r_{16}c_8$', linestyle=ls[4])\n",
|
||
"ax.grid()\n",
|
||
"ax.set_ylim(0.5,0.85)\n",
|
||
"ax.set_xlabel(\"Epoch\")\n",
|
||
"ax.set_ylabel('Spearman $\\\\rho$')\n",
|
||
"\n",
|
||
"#--------------------------------------------------------------------------------\n",
|
||
"ax=axs[1]\n",
|
||
"\n",
|
||
"#regular training\n",
|
||
"x=[ff for ff in range(0,epx)]\n",
|
||
"ax.plot(x,slow[0:epx], label= 'baseline $r_{32}c_{16}$', linestyle=ls[0])\n",
|
||
"\n",
|
||
"#econas\n",
|
||
"d = 230\n",
|
||
"x2=[ff/d for ff in range(0,epx)]\n",
|
||
"ax.plot(x2,econas[0][0:epx], label='econas $r_8c_4$', linestyle=ls[1])\n",
|
||
"d = 59.5\n",
|
||
"x2=[ff/d for ff in range(0,epx)]\n",
|
||
"ax.plot(x2,econas[1][0:epx], label='econas $r_8c_8$', linestyle=ls[2])\n",
|
||
"ax.plot(x2,econas[2][0:epx], label='econas $r_{16}c_4$', linestyle=ls[3])\n",
|
||
"d = 15.4\n",
|
||
"x2=[ff/d for ff in range(0,epx)]\n",
|
||
"ax.plot(x2,econas[3][0:epx], label='econas $r_{16}c_8$', linestyle=ls[4])\n",
|
||
"\n",
|
||
"ax.grid()\n",
|
||
"ax.set_ylim(0.5,0.85)\n",
|
||
"#plt.xlim(0,0.01)\n",
|
||
"ax.set_xlabel(\"Normalized FLOPS\")\n",
|
||
"ax.set_xscale('log', basex=2)\n",
|
||
"\n",
|
||
"from fractions import Fraction\n",
|
||
"labels = (\"1\", \"1\", '1/64', \"1/8\", \"1\", \"8\")\n",
|
||
"ax.set_xticklabels(labels)\n",
|
||
"\n",
|
||
"#--------------------------------------------------------------------------------\n",
|
||
"ax=axs[2]\n",
|
||
"\n",
|
||
"#regular training\n",
|
||
"x=[ff for ff in range(0,epx)]\n",
|
||
"ax.plot(x,slow[0:epx], label= 'baseline $r_{32}c_{16}$', linestyle=ls[0])\n",
|
||
"\n",
|
||
"#econas\n",
|
||
"d = 4\n",
|
||
"x2=[ff/d for ff in range(0,epx)]\n",
|
||
"ax.plot(x2,econas[0][0:epx], label='econas $r_8c_4$', linestyle=ls[1])\n",
|
||
"d = 4\n",
|
||
"x2=[ff/d for ff in range(0,epx)]\n",
|
||
"ax.plot(x2,econas[1][0:epx], label='econas $r_8c_8$', linestyle=ls[2])\n",
|
||
"ax.plot(x2,econas[2][0:epx], label='econas $r_{16}c_4$', linestyle=ls[3])\n",
|
||
"d = 3.3\n",
|
||
"x2=[ff/d for ff in range(0,epx)]\n",
|
||
"ax.plot(x2,econas[3][0:epx], label='econas $r_{16}c_8$', linestyle=ls[4])\n",
|
||
"\n",
|
||
"p=15\n",
|
||
"ax.scatter(p/d, econas[3][p], marker='o', color='purple')\n",
|
||
"ax.annotate(f'({round(p/d,1)}, {round(econas[3][p],2)})',(p/d-2, econas[3][p]+0.01), horizontalalignment='left', color='purple')\n",
|
||
"ax.annotate(f'econas+',(p/d-2, econas[3][p]+0.03), horizontalalignment='left', color='purple')\n",
|
||
"\n",
|
||
"p=20\n",
|
||
"ax.scatter(p/4, econas[0][p], marker='p', color='orange')\n",
|
||
"ax.annotate(f'({round(p/4,1)}, {round(econas[0][p],2)})',(p/d-1.3, econas[0][p]-0.03), horizontalalignment='left', color='chocolate')\n",
|
||
"ax.annotate(f'econas',(p/d-1.3, econas[0][p]-0.05), horizontalalignment='left', color='chocolate')\n",
|
||
"\n",
|
||
"ax.grid()\n",
|
||
"ax.set_ylim(0.5,0.85)\n",
|
||
"ax.set_xlim(0,10)\n",
|
||
"ax.set_xlabel(\"Normalized GPU Runtime\")\n",
|
||
"#ax.set_xscale('log', basex=10)\n",
|
||
"\n",
|
||
"fig.tight_layout(pad=0.3)\n",
|
||
"plt.legend(bbox_to_anchor=(1,0.7))\n",
|
||
"plt.tight_layout()\n",
|
||
"plt.savefig('econas.pdf')\n",
|
||
"plt.show()"
|
||
]
|
||
},
|
||
{
|
||
"cell_type": "markdown",
|
||
"metadata": {},
|
||
"source": [
|
||
"## Figure 2: Correlation of validation accuracy to final test accuracy during the first 12 epochs training for three datasets on the NAS-Bench-201 search space. Zero-cost and EcoNAS proxies are also labeled for comparison.\n"
|
||
]
|
||
},
|
||
{
|
||
"cell_type": "code",
|
||
"execution_count": 8,
|
||
"metadata": {
|
||
"scrolled": false
|
||
},
|
||
"outputs": [
|
||
{
|
||
"name": "stdout",
|
||
"output_type": "stream",
|
||
"text": [
|
||
"cifar10\n"
|
||
]
|
||
},
|
||
{
|
||
"data": {
|
||
"image/png": "iVBORw0KGgoAAAANSUhEUgAAARgAAAEYCAYAAACHjumMAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjEsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+j8jraAAAgAElEQVR4nOydd1hU19aH3w0MVVAQJWLFGguIYo293KtGozGJQZMYTbsxX3rxanrvvXhTNLYUNUWjMcbEhpJYQVGxN1SsYKW3Wd8fZyAoKDMDw1D2+zzzcM4+e++zzmFY7Lp+SkTQaDQaR+DibAM0Gk3VRTsYjUbjMLSD0Wg0DkM7GI1G4zC0g9FoNA5DOxiNRuMwtIPRlAqlVJ5SKq7QZ3IZ1t1EKRVfVvVpyh83ZxugqfRkiEi4s43QVEx0C0bjEJRSCUqpd5RS25VSG5VSzS3pTZRSK5VS25RSK5RSjSzpQUqpBUqprZbPdZaqXJVSU5VSO5RSfyqlvCz5H1FK7bTUM9dJj6kpAe1gNKXF67IuUmShaxdEJBT4DPjIkvYpMEtEwoDvgE8s6Z8Aq0WkPdAR2GFJbwFMEZG2wHngZkv6ZKCDpZ4Jjno4TelQequApjQopVJFpEYx6QlAfxE5qJQyASdFpLZSKhmoJyI5lvQTIhKolEoCGohIVqE6mgDLRKSF5XwSYBKR15RSS4FU4BfgFxFJdfCjauxAt2A0jkSucGwLWYWO8/hn3HAoMAWjtbNJKaXHEysg2sFoHElkoZ/rLMdrgdGW49uBaMvxCuABAKWUq1Kq5pUqVUq5AA1FZBUwCagJFGlFaZyP9vqa0uKllIordL5URPKnqv2VUtswWiFjLGkPAzOUUhOBJOAuS/qjwFdKqXswWioPACeucE9X4FuLE1LAJyJyvsyeSFNm6DEYjUOwjMF0EpFkZ9uicR66i6TRaByGbsFoNBqHoVswGo3GYVRJB6OUGqyU2qOU2l/avTFKqelKqdNlsSdGKdVQKbXKsgJ1h1Lq0VLU5WlZIbvVUtfLZWCfq1Jqi1JqcRnUlWBZxRunlIopZV21lFI/KaV2K6V2KaW621lPq8sWBV5USj1WStset7z/eKXUHKWUZynqetRSzw577Cruu6qUClBKLVNK7bP89C9FXaMstpmVUp2sMkpEqtQHY4bhANAUcAe2Am1KUV9vjLUW8WVgWz2go+XYF9hrr20Ysyc1LMcmYAPQrZT2PQF8Dywug2dNAALL6Hc6C7jXcuwO1Cqj78lJoHEp6qgPHAK8LOc/AOPtrKsdEA94Y8zuLgea21hHke8q8A4w2XI8GXi7FHW1BloBURgD+CXWUxVbMF2A/SJyUESygbnACHsrE5E1wNmyMExETojIZstxCrAL40tqT10i/6xeNVk+dg+oKaUaYCxem2ZvHY7AMhXdG/gaQESypWympAcAB0TkcCnrccOYqnfDcA7H7aynNbBBRNJFJBdYDdxkSwVX+K6OwHDQWH7eaG9dIrJLRPbYYlNVdDD1gaOFzhOx84/YkViWwXfAaHnYW4erZQ3KaYwl9XbXhbFX6L+AuRR1FEaAP5VSsUqp/5SinhCM9TIzLN23aUopnzKwbzQwpzQViMgx4D3gCMaanQsi8qed1cUDvZRStZVS3sD1QMPS2GchSETy1xOdBILKoE6rqYoOpsKjlKoB/Aw8JiIX7a1HRPLECJXQAOiilGpnpz3DgNMiEmuvLcXQU0Q6AkOAB5VSve2sxw2jqf65iHQA0jCa+najlHIHhgM/lrIef4wWQggQDPgope6wpy4R2QW8DfwJLAXiMBYclhli9HPKddq4KjqYY1zq+RtY0ioElg1+PwPficj8sqjT0mVYBQy2s4oewHDL4ri5QH+l1LeltOmY5edpYAFG19UeEoHEQq2znzAcTmkYAmwWkVOlrGcgcEhEkkQkB5gPXFdCmSsiIl+LSISI9AbOYYzRlZZTSql6AJafp8ugTqupig5mE9BCKRVi+U81GljkZJsAUEopjLGEXSLyQSnrqqOUqmU59gL+Bey2py4ReVpEGohIE4z3tVJE7PpPbLHHRynlm38M/BujC2CPbSeBo0qpVpakAcBOe22zMIZSdo8sHAG6KaW8Lb/bARjjanahlKpr+dkIY/zl+zKwcREwznI8DlhYBnVaj70j6BX5g9F/3Ysxm/RsKeuag9G/zsH4b3pPKerqidFE3YbRBI4DrrezrjBgi6WueOCFMnp3fSnlLBLGDN5Wy2dHGfwOwoEYy7P+AviXoi4f4AxQs4ze18sYjj0e+AbwKEVd0RjOcyswwI7yRb6rQG2MjaT7MGamAkpR10jLcRZwCvijpHr0Sl6NRuMwqmIXSaPRVBC0g9FoNA5DOxiNRuMwtIPRaDQOo1wdTEmbEJVSjSybAbcoQ47iekt6E6VURqFNal9Yca/SrB51aH0Vta6yrq+62Kaf8yqUxVSdldNeJW5CBL4CHrActwESLMdNsHGzIRBTxvaXWX0VtS5tm/Prqsi22VNXebZgrNmEKICf5bgm9m8c02g0FYDyDPpd3CbErpfleQljg9zDGAuiBha6FqKU2gJcBJ4TkejLyuY34fKbcRGenp5ltsjHzc2NsqqvotZV1vVVF9uqy3Nizz6msmzaldC8ugWYVuh8LPDZZXmeAJ60HHfHWNXoAngAtS3pERiOyu9q9/P29hZrWLVqlVX5KiLadudQmW0Xsd9+IE0qcBfJmk2I92AE7UFE1gGeGEGLskTkjCU9FmMsp6XDLdZoKjliFsx5ZRWBw3bK08FYswnxCMaGMZRSrTEcTJJlY5+rJb0phl7xwXKzXKMpY8x5ZnLSc4qkn44/TcLqhCLpB5Yd4IdbfuDbQd+y7sN1Ra7HTo3ljyf/KJIe80UMfzxeNH3T/zbxmudrvFXzLZZPXl7k+pYZW1j53Eorn+bKlNsYjIjkKqUeAv7AmFGaLiI7lFKvYIxOLwKeBKYqpR7H6O+NFxGxxBJ5RSmVgxEQaYKIlEmUOY3GERxaeYiNn20kNzOXkAEhXPfkpVEc4mbGcWrrKYZ8MuSS9IyzGaQcTylSX63GtWgb2RYPXw/8mxUNq9v+zvaIuegQSef/61ysfZ0mdCL8rnDysvJwcSvazmhzSxskr/RDN+Wq7CgiS4All6W9UOh4J0ZsksvL/YwRQ0WjcQrZadmknU7DP+TSP+6DKw6y77d9DPpg0CXptUJqEXZHGG5ebtRqXKtIfR3vKT6kTePejYtNr92yNrVb1r6ifW4etv0pKxeFycuEyctU7HUPXw+b6rsSWjpWoylEVkoWKcdSCLw28JL04zHHiZ8Tz7Avhl2SHhwRTECzgCL1+If4F3FG1RHtYDQa4Nyhc8wdMZdzB87RakQrbv7+5kuuN+nThCZ9mhQp51nLE89adiuVVHn0XqQyICEhge+/L4vgYxpHk52aza/3/1ok3TfYlxHTRzDp3KQizkVjP9rBlAHawVQ8RITlTy8vMkVr8jHRuHfj/HVXBbh5uBHcKRhXd9fyNLPKU20dzKfjRvF+5DBOHdzP+5HD+HTcqEuuT548mSlTphScv/TSS7z77rtMnDiRdu3aERoayrx58wryRkdHEx4ezocffkheXh4TJ06kc+fOhIWF8eWXX5brs1U3ilvroZTCr74feVl5RdLDbg/DCKGrcTTV1sFkZ2Zc9TwyMpIffvih4PyHH36gbt26xMXFsXXrVpYvX87EiRM5ceIEb731Fr169SIuLo7HH3+cr7/+mpo1a7Jp0yY2bdrE1KlTOXToULk8V3UiYXUCv/7nV94Pfp8j0UeKXO/yUBdM3sXPkmjKBz3IewU6dOjA6dOnOX78OElJSfj7+xMXF8eYMWNwdXUlKCiIPn36sGnTJvz8/C4p++eff7Jt2zZ++uknAC5cuMC+ffsICQlxxqNUWZJ3JVO7ZW3u/utuApoXncnROJ9q5WA+HTeqoKWiXFwR8z/NZ+XiyvuRxhSku6cXD8/6kVGjRvHTTz9x8uRJIiMjrW6FiAiffvopgwYNKjmzpkR2/7Kb5D3J9JzU85L0ThOs01/XOI9q1UUq3A0q7FwuP8/PFxkZydy5c/npp58YNWoUvXr1Yt68eeTl5ZGUlMSaNWvo0qULvr6+pKT8s/py0KBBfP755+TkGEvB9+7dS1pamiMfrUoT3DmYtre2dbYZGjuoVi0YW2nbti0pKSnUr1+fevXqMXLkSNatW0f79u1RSvHOO+9wzTXXULt2bVxdXWnfvj3jx4/n0UcfJSEhgY4dOyIi1KlTh19++cXZj1Phyc3KJeqlKAa8PgDl8s8grF99v6uU0lRktIMpge3btxccK6V49913effddy/JYzKZWLny0o1hb7zxBm+88Ua52FhVcHV3pWbDmuRl5+Hmqb+aVYFq1UVy9/QqOFYul653KHxeOJ+mbBERElYnsPWprRxde/SSa0opOv9fZ+1cqhDV6jf58KwfC47zB3TzEXMeT85bXN4mVTui34hm68yt1B1Zl+BOwc42R+NgqpWD0TgGMQvb52xn/YfruXj0In4N/ej2eDdCx4ReMpYC0OXBLvSc1JM1f63Rq2arAZVCtsRy7WlLuT1KqVLP/17eDdLdIvsQszDvpnksvn8xJ2JPkHY6jROxJ/j1P78yrdu0IitsPWt5Fht/RFM1KbcWjCUi3RTgXxgBvzcppRZZYsDk8xzwg4h8rpRqgxE7ponleDTQFggGliulWorIpXPNNpDfXYqKiiJSd43sZvuc7RxcfpCctEujs+Wm53Jyy0niZsTR8d7iY59oqj6VRbZkBDDXEpv3ELDfUp/Gyaz/cH0R55KPOddMzBcx5WyRpiKhLt9V6rAbKXULMFhE7rWcjwW6ishDhfLUA/4E/LHIlohIrFLqM2C9iHxryfc18LuI/HTZPQpkS9zc3CKWLVtWol2pqanUqFGjLB6x3KkItp/aeorME5mk7UzDv0/RAEsuJheCwoKKpFcE2+2lMtsO9tvfr1+/dBHxsaVMRRvkHQPMFJH3lVLdgW+UUu2sLSwiX2GoQ+Lj4yN9+/YtsUxUVBTW5KuIVATbv3rqK07EngAg6dekItfrRdQjMiaySHpFsN1eKrPtUL72VwrZEivLapxAt8e7YfIpfseyycdE9ye6l7NFmopEpZAtseQbrZTyUEqFYMiWbCw3yzUFZJ7P5Oi6fxbIhY4JpenApkWcjMnHRLN/NaPdaKsboOXGR00+crYJ1YZKIVsC7FBK/YCh9JgLPFiaGSSN7eTl5BHzRQzRr0UTflc4DbsbDUrlooicH0n83HjWfbCOi4kX8WvgR/cnutNudDu2f7+dDZ9sIC87j/pd6zP0f0M5uOwgsY/EsttrN96B3ty54k4yzmaw8O6FnDt4DpO3iRu+uoGgsCCiXoriwpELnDt4jgtHLtDtsW50fcRQHJ5741wuHr1IbmYuXR/tSsR/IjDnmVl0zyKOxxxHKUX43eF0f1y3opyGrVKQleWjpWPLluz0bFlw5wI5ue2k1WVO7zwt3w/7XnKzc0VEZPEDi2XLzC3yQYMPZMn3S0REJP1MuoiI/PbQb7LqpVUiInJwxUH5vP3nIiKy6sVVMq37NMnJzJG0pDR5O+Dtgvryy2anZ8uUtlMkLTlNjsUck9kDZxfYkHEuo4hdHzb+0Manv5TK/J0RKV/p2Io2yKupoJi8TNw460abyhxacYjjsceZ2nkqALkZuRzbcIzGvRvjVc9Y2OgVYPw8+tdRbv35VgBC+oeQcSaDrItZALQY2gI3DzfcPNzwqetD2qk0/Br4seGTDexesBuAi0cvcnbfWWq3qs25g+dY8vASWg5tSbN/NwNgzetr2PmjseQq5XgKX4R/AUDDHg0ZOmVoaV6N5ipoB6MpwvnD5zkRe4LWN7UuVT0iQvtx7Rn45sCCtD2/7mHH3B021VNYVEy5Ksy5ZhKiEji4/CD3rLsHk7eJmX1nkpuZi5e/FxO2TmD/H/uJ+SKGHT/sYMT0EfR+tje9n+0NGGMwE+ImlOrZNNah12xrimDONXMx8WKp62k6oCm7ftpF2mkj2FbG2QyCwoI4vOYwGScyCtIAGvVqxLbvtgGQEJWAd6A3Hn5XVhfMvJCJl78XJm8TybuTSVyfCEB6cjpiFtrc3Ib+r/XnxOYTpX4Ojf3oFoymCAHNAgoGUktDnTZ16PdaP7759zeIWXA1uXL9lOsZ9tUwFj2yiMNvHcanrg9jl42l70t9WXj3Qj4P+xyTd8ndseaDmxP7RSxTWk+hdqvaNOjWAICLxy6y8K6FBTrNA94cUOrn0JQCWwdtKstHD/JaR3Zatpzecbr0xthIdX/vzqQ8B3l1F6mak7ghkc1fb3a2GZoqiu4iVXNC+oUQ0k/LqWgcg27BaDQah6EdTDXidPxplj6+FKM7rdE4Hu1gHEBSUhJdu3alQ4cOREdH06RJE5KTk51qU/Qb0czqN4s6beo41Q5N9UKPwTiAFStWEBoayrRp05xtSgG+9X25f8v9+DXQGkMa68nONbNy9yl+irUveIFuwVxGWloaQ4cOpX379rRr14558+Zx443/rMlYtmwZI0eOBKBGjRo8++yztG/fnm7dunHq1Cni4uL473//y8KFCwkPDycjI+OS+j/44APatWtHu3bt+OgjY1fvu+++yyeffALA448/Tv/+/QFYuXIlt99+e5k8V/i4cO1cNFYhImxLPM+LC+Pp+sZyJny7ma2J5+2qSzuYy1i6dCnBwcFs3bqV+Ph4Bg8ezO7du0lKMoIpzZgxg7vvvhswnFG3bt3YunUrvXv3ZurUqYSHh/PKK68QGRlJXFwcXl7/BBOPjY1lxowZbNiwgfXr1zN16lS2bNlCr169iI6OBiAmJobU1FRycnKIjo6md+/e5f8SNNWS0xcz+XL1AQZ9tIbhn/3NnE1Hua55IDPGd2bd5P521akdzGWEhoaybNkyJk2aRHR0NDVr1mTs2LF8++23nD9/nnXr1jFkyBAA3N3dGTbM0FeKiIggISHhqnX/9ddfjBw5Eh8fH2rUqMFNN91EdHQ0ERERxMbGcvHiRTw8POjevTsxMTFER0fTq1cvm59h35J9RL0cZXM5TfUjMyePX7ceZ/yMjXR7cwVv/r4bHw83Xh/Zjk3PDGTKbR3pd21d3FztcxXlOgajlBoMfIwRD2aaiLx12fUPgX6WU2+grojUslzLA/J1XI+IyHBH2NiyZUs2b97MkiVLeO655xgwYAD33nsvN9xwA56enowaNQo3N+O1mUwmlDJ0f1xdXcnNzbXrniaTiZCQEGbOnMl1111HWFgYq1atYv/+/bRubfuGw4Y9GuLftGh8XI0GjC5Q7OFz/BSbyOJtx0nJzKVeTU8e6NuMmzo2oFmdsos3XKFkS0Tk8UL5HwY6FKoiQ0TCHW3n8ePHCQgI4I477qBWrVpMmzaN4OBggoODee2111i+fLnddffq1Yvx48czefJkRIQFCxbwzTffFFx77733mD59OqGhoTzxxBNEREQUODBb8KzpiWdNT7vt1FRNjp3PYMHmRL75K4NTf6zF0+TCkHb1uLljA7o3q42ri+3ftZIozxZMgWwJgFIqX7Zk5xXyjwFeLCfbCti+fTsTJ07ExcUFk8nE559/DsDtt99OUlKSXS2KfDp27Mj48ePp0sVQXLn33nvp0MHwob169eL111+ne/fu+Pj44OnpaVf3SKMpTFpWLn/sOMnPmxNZe+AMItDK34Unh4QyJPQafD2Lj6dcVlQo2ZJCeRsD64EGYgmNqZTKBeIwQma+JSK/FFPOYbIlH3/8Mc2bN2fo0IoTnCjf9vTEdCRH8AmxSVHCqVRm6Q9H224W4WymkJkLmblCZp7lOK9w2hV+Wo6zcoWMPMi1CGvW8VJcF+xGj/pueJvTq61sST6jgZ/k0ri7jUXkmFKqKbBSKbVdRA4ULiQ2ypbkpaSwfN48ekVG4urre8V8ERER+Pj4MHfuXDw8rhyjpLzJl5/YNX8XWSlZhPd1eA+yzKjM0h+OsP3Y+Qz+2pfEmn3J/L0/mfPpxYvZ5ePqovBxd6WGhxs+Hm54e7kR5OGKj7tx7uPhio+HGzXc3egSEkDnJgG4WLpA5fnuy9PB2CI9Mhp4sHCCiByz/DyolIrCGJ85ULSo9aRGRSFZWaRGrabmDcOumC82NrY0t3E4pY08pyl/0rJyWX/wDNH7konel8SBJCMoV5CfBwNbB9GxkT81vUz4ePzjRAznYTgODzcXu8bnypvydDAFsiUYjmU0cNvlmZRS12IoO64rlOYPpItIllIqEOgBvGOPESJS8Is5v+AXiOjIhV9+KXAwha9rNGWF2SzEH79A9L5k1uxNYvORc+TkCZ4mF7qG1GZMl0b0blmHFnVrVKnvX0WTLQHD8cyVSweHWgNfKqXMGGt33io8+2Qt0R8tJz3xFGE1E1AKMmJiIKIj6Zs2cfK11xGBbRea4N0giF6PDSy5Qiew7bttHF59mBu+usHZpmhK4Pj5DP7al8yafUn8vT+Zc5ZuT5t6ftzdM4TeLeoQ0dgfT5Orky11HOU6BiMiS4All6W9cNn5S8WUWwuElvLeZGeb2Z9aj8zdu2ix/2fy/09IdjZnv/2Wfc1vJrFBPa7Nzq2QLZmDyw+y7KlljF021tmmaIohPdvo9qzZm8xf+5PZfzoVgLq+HvS/NojeLQO5rlkgdXwrzjieo6mog7xljlKK/hP/hevUGHbQH+XqRvM98wBD4W1/q0gS6/WmbUdf+tzXqcI5F4DjMce5ec7N1G1X19mmaCyICJuPnGfuxiMs3naCjJw8PNxc6Nq0NqM7N6RXizq0DKpa3R5bqDYOBgwn0+e+Trh8t4Ptf/VG8nLxAKPlUq83oT2D6HV7mwr7Zeg5uaezTdBYuJCew4IticzZeJQ9p1LwcXflxg7BXB9aj85NAqp0t8cWqpWDAcPJhNc7zfmT0Rxt0J+6vmmcbtCahiejCQ/uhVJtnW2ipoIiImxKOMdX27KIXb6crFwzYQ1q8uZNodzQPpgaHtXuz6lEquUbubhoEc33/MnRa/5ZKdt8zzwuLjqPX3/7do1qqi7n0rL5eXMiczYe4UBSGl5uMKpTI0Z3bkS7+jWdbV6Fplo6mKyEBBJHvgxn/0lLHPkyzXZ+7zyjiiFxQyKJ6xLp9lg3Z5tS7RAR1h08w9yNR1kaf5LsPDMdGtXinVvCqHl+P4MGlmrOodpQ7RyMiHD89nfYtzKRsP4NyKt7nGs8g9m2Erxue5umFWj2yK+BH9eEX+NsM6oVyalZ/BSbyLxNRzmUnIafpxu3dW3E6C4NufYaI2BXVFSp1ndWK6qVgxER/vpxH9sszqXnqBasXn2cnqNaALBtpSE/2nNUC3r06MHatWvL5L7jx49n2LBh3HLLLTaV86vvh199HYXO0ZjNwt8Hkpmz8QjLdp4iJ0/o3MSfh/s35/rQenrAthRUGwdTnHPJb6kopYo4mb///ttptmocT2ZOHpsSzhK9L5nf409w9GwGtbxN3Nm9CWO6NKR53SvvTdNYT7WJaKeUwsPLrYhzKXy956gWhPVvgIeXG76+vqSmpjJgwAA6duxIaGgoCxcuLMg/e/ZswsLCaN++PWPHGgvfEhIS6N+/P2FhYQwYMIAjR44U5F++fDmdOnWiZcuWLF68+Ip25uXl8dRTT9GuXTvCwsL49NNPASOQeIcOHQgNDeXuu+8mKyuLjRs3MmrUqIKyUVFRBRH2NJciIuw+eZGpaw4y9usNtH/5T8Z+vZEZfx+icYAPH48OZ/3TA3h+WBvtXMqQatOCAehyQ9OrrtDNdzL51z09PVmwYAF+fn4kJyfTrVs3hg8fzs6dO3nttddYu3YtgYGBnD1rjBY//PDDjBs3jnHjxjF9+nQeeeQRfvnFiCqRkJDAxo0bOXDgAP369WP//v14ehYNCvX06KfZdWYXcXFxuLm5cfbsWTIzMxk/fjwrVqygZcuW3HnnnXz++edERETw2WefkZaWho+PD/PmzWP06NEOenuVj6SULP7an0T03mSi9yeTlJIFQPO6NbitayN6t6hD16YBeLtXqz+DcqXavdmSBnALXxcRnnnmGdasWYOLiwvHjh3j1KlTrFy5klGjRhEYGAhAQEAAAOvWrWP+/PkAjB07lv/+978Fdd166624uLjQokULmjZtyu7duwkPLxpeYdupbUx4YEJBWM6AgAC2bt1KSEgILVu2BGDcuHFMmTKF8PBwBg8ezK+//sott9zCb7/9xjvv2LUHtEqQmZNHTMI5oi1hD3aduAiAv7eJni3q0KtFIL1aBFKvplcJNWnKimrnYGzhu+++IykpidjYWEwmE02aNCEzM9OuuorrkhWHTx0ffAKtj+kzevRoPvvsMwICAujUqRO+V4lrU9UQEfacSrFsKExmw8EzZOWaMbkqIhr7M3FQK3q3qEPbYL+CWCia8qXajMHYw4ULF6hbty4mk4lVq1Zx+PBhAPr378+PP/7ImTNnAAq6SNdddx1z584FDOdUOOTljz/+iNls5sCBAxw8eJBWrVoVe89//etffPnllwUBxM+ePUurVq1ISEhg//79AHzzzTf06dMHgD59+rB582amTp1abbpHqVm5Fs2eFQz+KJrXftvF8fMZ3Na1ETPGdybuhX8z9z/debBfc0Ib1NTOxYnoFswVUEpx++23c8MNNxAaGkqnTp249tprAWjbti3PPvssffr0wdXVlQ4dOjBz5kw+/fRT7rrrLt59913q1KnDjBkzCupr1KgRXbp04eLFi3zxxRfFjr+AEad37969hIWFYTKZuO+++3jooYeYMWMGo0aNIjc3l86dOzNhwgTWrVuHq6srw4YNY+bMmcyaNatc3o0ziT92gYe+38yRs+kMCa1HnxZ16NkikOBauttTIRGRcvsAg4E9wH5gcjHXP8SIuxsH7AXOF7o2Dthn+Ywr6V7e3t5iDatWrSqSlpycLI0aNbKqfFmwccpGuZB4weZyxdleWbDVdrPZLLPXHpIWzyyRrq8vlw0HzzjGMCuozO9dxH77gTSx8W++UsiWKKUCMBQGOmFEV4i1lD1X1nYeP36cvn378tRTT5V11cVyatspVr+ymr+s8JQAACAASURBVNDb9NLzK3EhI4en529jyfaT9G1Vhw9uDSfAx93ZZmmsoLLIlgwClonIWUvZZRitoTllbWRwcDB79+4t62qL8McffzBp0iTOHzqPu6873/b9lpCQEBYsWODwe1cmth49z0NzNnPifCZPD7mW+3o11WMqlYhKIVuilHoK8BSR1yzXn8cQYnvvsnIOky1xFKn7U/Fu5I2Lu+3j7c62vTSUZLuI8OfhXH7Yk00tD8UD7T1o7l8xluxX5vcO9ttf1WVLSkRslC2BCiCfUYpbO932UnA128+nZ/PUj9tYvvsUA1sH8d6oMGp5V5wuUWV+76BlS6CobMkxLv1TbABElaFtmgpC7OFzPDJnC6dTMnlhWBvu6tGkwuxu19hOea6DKZAtUUq5YziRRZdnKk62BEOJ4N9KKX+LhMm/LWmaKoLZLHyx+gC3frkOFxf4acJ13N0zRDuXSk6lkC0RkbNKqVcxnBTAK/kDvpWRA8sO4FPHR8d6sXAmNYsnf9xK1J4krg+9hrduDsPPwZrJmvKhUsiWWNKnA9MdZlw5knUxC3c9zQrAhoNneGTuFs6l5/Dqje24o2sj3WqpQtjkYJRSLoCLiOQ6yJ5qQZub2zjbBKdjFuHTFfv4cPleGtf2Yfr4zrQN1vFtqxpWOxhL9+ZFIFspdQb4RESmOcwyTZUlKSWL92My2XFmLyPCg3l9ZKiOyF9FseW3+iQQKiInlVL1gDeUUg2u1KXRaIpj7f5kHp0Xx/k0M2/fHMqtnRrqLlEVxpZZpFTgNICInADuAW5yhFFVETELGecynG2GU5m36Qi3f72Bml4mXuzuRWRnPd5S1bHFwXwO/KiUam45bwSkl71JVZPTO04zb+Q8Z5vhNNbsTeKZBfH0alGHRQ/1oIGvjhRSHbD6tywi/wO+A6Yppc5h7Ijeo5QapZRq4SgDqwpBoUHcueJOZ5vhFPacTOHB7zbTom4N/nd7Rx2ishph078REZkvIn2BOkBHYCVwHfBl2ZtW9XBxrX7/tU+nZHL3zE14ubsyfXxnPZhbzbDrt22Zpt5m+VT9KEcau8jIzuO+WTGcTcvmh/u766BQ1RD970TjEMxm4fF5cWw7doGvxnYitIFe41IdqX5t9nLm7IGz7Pz5SiFvqi5vL93N0h0neW5oG/7VJsjZ5michHYwDib2q1iObbzSpvGqyfcbjvDlmoPc2b0xd/do4mxzNE7ElpW8HsDNQJPC5UTklbI3q2pgzjWz7Ztt1Wr2aM3eJJ5fGE/fVnV4YVgbvc6lmmPLGMxC4AIQC2Q5xpyqRV5OHgPeHECd1nWcbUq5sOdkCv9nmY7+7LaOuFXDWTPNpdjiYBqIyGCHWVIFMXmZCB9XVL2xKpI/He3t7sqMu/R0tMbAln8xa5VSpQp9r5QarJTao5Tar5SafIU8tyqldiqldiilvi+UnqeUirN8igSq0jiPwtPR08d31tKsmgJs+TfTExivlDqE0UVSgIhImDWFrZEtsawIfhroISLnlFJ1C1WRISLVozlQiTCbhcfmbSmYjm5XX09Ha/7BFgczpJT3ska25D5gSr7ekYicLuU9nUZedh6u7hUjCr4jeWvpbv7YcYoXhunpaE1RbJItscTDbQEU6J6KyBory5YoW6KU+gVD0bEHRljNl0RkqeVaLobiYy7wloj8Usw9KoRsSXpiOrte3kXE1IgyrfdynC2fsepIDrN2ZjOgkRt3tHa3acbI2baXhspsO5SvbIktsq/3AtuBc8AqIANYaUP5W4Bphc7HAp9dlmcxsAAwASHAUaCW5Vp9y8+mQALQ7Gr3K410bFmQk5HjkHoL40wJ06g9p6Xp07/J+OkbJCc3z+bylVl+tTLbLlK+0rG2DPI+CnQGDotIPwxZ1/M2lLdGtiQRWCQiOSJyCKM10wJARI5Zfh7EkCzpYMO9yx03z6o7i5K/O7plkC+f6ulozVWw5ZuRKSKZYCy6E5HdQCsbylsjW/ILFv0jpVQg0BI4aJEr8SiU3oMrS85qHMjpi8Z0tI+HK9PHd9LT0ZqrYsu3I1EpVQvDCSyzxIQ5bG1hsU62JF//aCeQB0wUkTNKqeuAL5VSZgyn+JYUmn3SlA/p2bncO9uYjv5xQnc9Ha0pEasdjIiMtBy+pJRaBfgBS225mZQgW2Lp5z1h+RTOsxYo1Rqc8iL9TDqu7q54+Ho425QyJc+yO3r7sQtM1dPRGiuxZS9SJ+BZoLGlnAJeB6xaB1Nd2DFvB1kXs+g5uaezTSlT3vp9V8F09EA9Ha2xElu6SN8BEzFmksyOMafy0/n/OjvbhDJn9roEpkYfYlz3xtyld0drbMAWB5Mk/8i7aqoBJy5k8PKinSzdcZL+19bleb07WmMjtjiYF5VS04AVFNpNLSLzy9yqCs6iRYvYuXMnkycXu52q0pObZ2bWusN88Ocecs3CxEGtuK9XUz0drbEZWxzMXcC1GIvg8rtIAlQ7BzN8+HCGDx/ubDMcwtaj53n2l+3EH7tI31Z1eGV4OxrV9na2WZpKii0OprOI2LLupVKRlpbGrbfeSmJiInl5eTz//PNMmjSJcePG8euvv5KTk8OPP/7Itddey8yZM4mJieGzzz5j/PjxeHp6EhMTw9nTZ3n/vfcZeevIkm9YwbiYmcP7f+xh9vrD1KnhwZTbOnJ96DW6S6QpFbaGa6iyqu1Lly4lODiYrVu3Eh8fz+DBRuibwMBANm/ezAMPPMB7771XbNmEhAQ2btzImLwxPPjIg2RmZpan6aVCRPht2wkGvr+a2esPc2e3xix/sg9Dw+qVvXPJTYeEOZBbvRUuqxO2OJhuQJwlnss2pdR2pdQ2RxlW3oSGhrJs2TImTZpEdHQ0NWsa6zxuuslQx42IiCAhIaHYsrfeeitpp9KomVmT5i2bs3v37vIyu1QcOZPOXTM38eD3m6nr58HCB3vw8oh2+Hmayv5m5+Pht7aw4R74rQ3kVR4nrLEfW7pIVTqaXcuWLdm8eTNLlizhueeeY8CAAQB4eBgL5lxdXcnNzS22rFKKzHOZREyIYHH04grfrcjONTM1+iCfrNiHm4vihWFtuLN7Y8cM4orA3s8gbpLFqQikHQbzLtizH1o+CJb3dW7DfM5EzUTycvBuEk7w6NdI3RXNyUXvgNmMaw1/mj46h9y08xz7ZiLZyUdQ7l7Uv+1NvBq05tTiD8k5d4zs5KNknz1GYP97COx3FwCHv7iP7HPHkdwsAvvdTUDP2xBzHonf/peMw9tAKQK630rggHvL/h1UY6xyMMr4izGLyFEH2+M0jh8/TkBAAHfccQe1atVi2rRpVpf98ccfGTduHE3ubsLBWQdp1ariDlVtSjjLM/O3s+90KoPbXsOLw9s4bsl/XiasHgFJf0Ne4W6RgJghbjIc+xX6LCLz9BEuxC6m2VM/o1xNHJvzLOc3LuDUovdo+sQPuAc2IjfN2Ft7evEHeDZsS+MJU0nd8zeJs5+gxTO/A5B18gAhj83FnJXG3pf7Ubv3HShXE/XHvoubTy3M2Znsf/sG/MKHkHM2kdzzJ2n5vBHWIy/9gmPeQzXGKgcjIqKUWkIlWa5vD9u3b2fixIm4uLhgMpn4/PPPueWWW6wq26hRI7p06cLFixf54osv8PT0LLlQOXMuLZu3ft/NvJij1K/lxdfjOjGgtYNX5GYch1MrQYpv+ZGXZlzPOEbqnr/JOLqd/W8bs3Pm7EzSE+LwbtEF98BGALj51AIg7cAmGv/HUCuu0aoHeannyMtIAcC3XX9cTB64mDxwq1Gb3IvJmPzrcWbVDC5u/QOAnHMnyE46hEdQM7KTj3B83gv4tutPjda9Hfk2qiW2dJE2K6U6i8gmh1njRAYNGsSgQYMuSSs85tKpUyeioqIAGD9+POPHjy+4NnDgQL744otysNJ2RISfNx/jjSW7uJiRw/19mvLogBblI0Bfoyl4XQPpiVfO41XPyCer8e96C9fcOKng0sVtyzkfa9vaTuXm/s+JiytiziV17zpS9/xFs4kLcHH34uCHkZhzsnD1rknzZ5aSums1Z6O/48LmxTQYW/xAvsY+bOl0dwXWKaUOVMVB3qrI/tOpjJm6nqd+3EpIoA+LH+nJ00Nal49zyafJHaCuMGisTMZ1oMa1PbiwZQm5KckA5Kadx7P+taTv20h28pGCNACf5l04t3EBAKl71+Fawx9XL98rmmDOSMHVqyYu7l5kntxP+qEtRn2pZ0HM1OxwPUHDnyLj6I4yeWTNP9jyTRtUcpbqx8yZMzm06hCJGxJp0LWBs80pYPnOUzzwXSxeJlfevCmUyE4NcXFxwuBz49Gw51PIyyl6zcVkXAc867Uk6IanOPTpWDCbwdWN+qNfpf7tb3L4q/tBBDff2oQ88h11hz7OsW8msu+1QSh3Lxrc+cFVTajRpg9nor9l78v98QhqhneIEass5/xJEr95yhiIBq4Z8d+yfXaNTeEaDhcXkxcbYsIopQYDH2PEg5kmIm8Vk+dW4CWMVcJbReQ2S/o44DlLttdEZJa19y3C2YOw9jPY9gOEPAVv3AFht8J1D0FAU5ury0nPQTnjj/cKXEjP4ekF22lR15fZ93QhsIYTQ0fUCgNTDWO85XJMflDrn2G9Wp1uoFanG4pk823b75JzN59aNJ4wtUi+oGGPX3KeP3gLEPLQ7GLNa/H0kmLTNWWDLeEa7sUIm9kAI/h2N2Ad0N/K8nbLliilAoAXgU4YjifWUvactfYDtH1xKZ1zYvmf6WPcyMVd5RkXslPI3jST3E3f8n85j7LJFMGOl62flW85tKUtZjicN3/fxdm0bGaM7+xc5wLGFHTnz+HwnEvTz/tD+/8VTFFrqia2dJHyY/KuF5F+SqlrgTdsKF8a2ZJBwDIROWspuwxjXc5l39qrUzv7OP9z/xhvVVT51l3l4U4e/zN9zODsIg2rSsO6A2eYu+ko9/dpWnGCQjUcaXwKExUFDfs6wxpNOWKLg8kUkUylVEFMXqWULQs+6mOoBOSTiDFwXJiWAEqpv7lUtqS4svUvv8FlsiUFsz75PBYqbOCZS/5ppnoEE9Xq5YJzEcOTXl62IpKamnqJndl5wvN/Z1DXW9HR/SRRUaecZ1wJXG57ZaIy2w7la3+5xeS1wZ4WGIG/GwBrbJGrFZGvgK8AfHx8pG/fvpdcT3lxNL7q0n0wUa1epu+eFy/NJ1743nbSduvLmaioKAo/4ztLd3Mq/QDf3duVHs0DnWeYFVxue2WiMtsO5Wu/1dPUIjJSRM6LyEvA88DXwI023Ks0siXWlC0RH6zbZOeD9ftkVj6/kpTjKbaaUubsPH6RL9ccZFREgwrvXDTVB6sdjDK4Qyn1goisxhjotUUr2m7ZEv5RG/C3zGT925JmE2lYtyQ+DetX4gaFBWHydsDmQBvIMwuT52/D39vEs0NbO9UWjaYwtiy0+x/QHRhjOU/BmBWyChHJBfJlS3YBP+TLliil8qM3/QGcsciWrMIiW2IZ3H0Vw0ltAl7JH/C1hV/yepAtV9eLzhZX5udZH7C77ai2eNZy7taAGX8fYlviBV4a3pZa3u4lF9BoyglbxmC6ikhHpdQWAMs0sk3fZntlSyzXpgPTbbnf5UzNG8rNrtG4k3fFPLm48XXe9YwrzY3KkaNn03n/z70MbF2XoaH1nG2ORnMJtrRgcixrWQRAKVWHSqYucESC+L+cR0kXjyItmWxxJV08+L+cRzkilUOWQ0R4ZsF2XBS8MqJdhQ8Toal+2OJgPsEQpg9SSr0O/IVt62AqBFHmcAZnv8WcvP6kiBeIMWs0J68/g7PfIspsy7CSc1l7PJfofclMGnItwbW0yqKm4mHLVoHvlFKxwABL0o0isssxZjkGHw9X0rLyOCJBvJh7Fy/m3sWTksv4rK+L5LOGvYv3knUxi9DbrjyT3qRJE2JiYggMLNuZnTOpWczZnU3HRrW4o2vjMq1boykrbNkq4AlcD/TC6Bq5K6UOiUiliX1Y3PL/qKgoEm7va1M9ubm5uLm5ceTvI7jXcM6g6quLd5KencfbN4c5ZxOjRmMFtgzyzsaYOfrEcn4b8A0wqqyNcjavvvoq3377LXXq1KFhw4ZERESwePFiwsPD+euvvxgzZgwtW7bkyalP4uLjQoNVDfjuu+8ICgrizJkzjBkzhmPHjtG9e3fEslO3OBISEhgyZAg9e/Zk7dq11K9fn4ULF+Ll5UVcXBwTJkwgPT2dZs2aMX36dPz9/enbty+1G7dk8Z+r6NKzH/dFfkCHDh2Ijo4mLS2N2bNn8+abb7J9+3YiIyN57bXXyvHNaTSXYssYTDsRuUdEVlk+9wFtHWWYs9i0aRM///wzW7du5ffffycmJqbgWnZ2NjExMTz55JP07NmTZQuWsSF6A6NHj+add94B4OWXX6Znz57s2LGDkSNHcuTIkaveb9++fTz44IPs2LGDWrVq8fPPPwNw55138vbbb7Nt2zZCQ0N5+WVjO0OeWfh7zyl6TZzGCw8YKwbc3d2JiYlhwoQJjBgxgilTphAfH8/MmTM5c+aMI16TRmMVtka06yYi6wGUUl2BmBLKVDr+/vtvRowYgaenJ56entxwwz/hAyIjIwuOExMTefLVJzlx4gTZ2dmEhIQAsGbNGubPN7Tohg4dir+//1XvFxISQni4MbCcr1xw4cIFzp8/T58+fQAYN24co0YZDcWj59JRYUN4++ZQUg4Z8b7yReBCQ0Np27Yt9eoZ09VNmzbl6NGj1K5du9TvRaOxB1taMBEY2kgJSqkEjFANnatTZDsfH5+C44cffpiHHnqI7du38+WXX9qthZSvWgBXVy4A2HLkHCcvZDK0YwgRjQOK1OHi4nJJfS4uLletT6NxNLY4mMFACNDH8gmxpA0DikYJqqT06NGDX3/9lczMTFJTU1m8eHGx+S5cuED9+saG7lmz/ol91bt3b77//nsAfv/9d86dsylkDQA1a9bE39+f6OhoAL755ht69urN0/O34+7mwthuetZIUzkosYuklOoMHBWRw5bzO4GbMXZSv2TPkv2KTOfOnRk+fDhhYWEEBQURGhpaIMJWmJdeeolRo0bh7+9P//79OXToEAAvvvgiY8aMoW3btlx33XU0atTILjtmzZpVMMjbtGlTrrvrORatO0VIoA8+HuUYU1ejKQ0ictUPsBkIsBz3Bo5jOJhXgZ9KKu+sj7e3t1jDqlWriqSlpKSIiEhaWppERERIbGxskTzHY49L1MtRVt2jtOw/nSItnlki//fdpXYUZ3tlQdvuPOy1H0gTG/8OrflX6Cr/tFIiga9E5GfgZ6VUXFk7vIrAf/7zH3bu3ElmZibjxo2jY8eORfL41vclZECIw20xm4Wn52/Hy92Vl26ocpN2miqOVQ5GKeUmxm7oAVgixtlQvtKRP4ZyNWoE1aBGUA2r6jtz5kyBFG1hVqxYUeIMz9xNR9l46Czv3BxGHV8nx9fVaGzEGgcxB1itlEoGMoBoAKVUc0BrbVpB7dq1iYuzvbF36mImby7ZxXXNajOqU8WRRNForKXEWSQReR14EpgJ9LT0xfLLPmzLzZRSg5VSe5RS+5VSk4u5Pl4plaSUirN87i10La9Qum1yf5WUFxbGk51n5o2RoXqntKZSYq029fpi0vbaciNrZEsszBORh4qpIkNEKs9W51KyNP4Ef+w4xeQh19Ik0KfkAhpNBcSWdTClpUC2RESygXzZkkrJ/Nvnk34m3SF1X8jI4YWFO2gb7Me9PR0/kKzROAr1T4/HwTdS6hZgsIjcazkfixEl76FCecYDbwJJGAG/HxeRo5ZruRhxgHOBt0Tkl2LuUVi2JGLZsmWXZylCamoqNWpYN1hbmHOx56jZviYubmXvo2fGZ7E6MZcXu3vSpOaVQ0fYa3tFQNvuPOy1v1+/fukiYltz2tZ5bXs/wC0YcrH552OBzy7LUxvwsBzfD6wsdK2+5WdTIAFodrX7lWYdjDNZdyBZGk9aLK//trPEvBXNdlvQtjuP8lwHU55dpBKlR8QI8J0vuzgNY/9T/rVjlp8HgSiggyONdQZms/D8L/E0CvDm8YEVS45Wo7GH8nQwJcqWKKUKR60ejqE+gEWuxMNyHAj04FLJ2SrBmn1J7DudypP/bomXu3VR9TSaiky5LZQTkVylVL5siSswXSyyJUCMiCwCHrFImOQCZ4HxluKtgS+VUmYMp/iWFJ19qvTMXneYOr4eDGmn1QE0VYNyXYkrJcuWPA08XUy5tYDVErKOJubLGNxruBN2e1iZ1Xn4TBqr9pzmkf4tcHfAwLFG4wz0N9kOjm86Tk5aTpnWOXvdYVyV4rau9u2+1mgqItrB2EHKsRR86/uWWX3p2bn8EHOUIaH1CPJzrkqkRlOWVMnNio5m4DsDqdmwaIwYe1mw5RgpmbmM664DSWmqFtrB2EFQaNkpP4oIs9cepm2wHxGNrx6/V6OpbFTrLtLBCwfZfXY3hy4ccpoN6w+eZc+pFMZd10RvaNRUOaqtgxERXvj7Bcxi5oW/X7iqftHllGUg7VlrE/D3NjG8fXCZ1anRVBSqrYNZdngZe88ZG8L3nNvD8iPLC669+uqrtGrVip49ezJmzBjee+89+vbty2OPPUanTp34+OOP+fXXX+natSsdOnRg4MCBnDp1CoDVq1cTHh5OeHg4HTp0ICUlhaioKHr37s3QoUNp1aoVEyZMwGw2c+x8Bn/uPElk50Z4mvTCOk3Vo1qOwaTnpPPq+lfJyM0AICM3g1fWvUKP4B7siNtRILyWk5NDx44diYgwdixkZ2cz59U5HF17lA49O7B+/XqUUkybNo133nmH999/n/fee48pU6bQo0cPUlNT8fQ0ZoU2btzIzp07ady4MYMHD2b+/PkcrNEOgDu66alpTdWkWjqYKXFTCpxLPhm5GUyJm4Lpb9NVhdcahDagdsvaJCYmEhkZWUR4rUePHjzxxBPcfvvt3HTTTTRoYESi69KlC02bNgVgzJgxRK2JZm2dmgxsHUQDf+9yenKNpnypdl2kgxcOMm/PPLLysi5Jz8rLYt6eeZzJuLLUqo+PD14BXgQ0C7ii8NrkyZOZNm0aGRkZ9OjRg927dwMUGcBNSE7jbFo2469rUrYPqNFUIKqVg8kf2M0xF78KN8ecQ4xXTKmE1w4cOEBoaCiTJk2ic+fOBQ5m48aNHDp0CLPZzLx58zjp1YQWdWvQvZmWddVUXaqVgzmacpT45HjMYi72ulnMnAg4QZ9BfQgLC2PIkCElCq9FREQQGBhYkP7RRx/Rrl07wsLCMJlMDBkyBDAE3R566CFat26Nb51gkgLbc6eemtZUcarVGExD34a0C2zH9uTtxToZF+VCaGAorzzzCh+++SHp6en07t2biIgI7rvvvkvyjhgxghEjikb8/PTTT4u9t5+fX0Fr6NG5W9iz6zQ3dahfBk+l0VRcqlULRinFKz1eweRiKva6ycXEKz1e4f777yc8PJyOHTty8803FwiviQhfRXxFbpb962BOp2SyZPsJbunUQEvAaqo85foNV0oNBj7GiAczTUTeuuz6eOBd/ol095mITLNcGwc8Z0l/TURmYQdNazYlslVkkYFeD1cPIltFElIz5KrCazfOuhE3Gx1D37596du3LwBzNhwlJ0+4s3sTe8zXaCoV5daCKSRbMgRoA4xRSrUpJus8EQm3fPKdSwDwItAVQ53gRaWU3Rt3Hgx/EC83r0vSvNy8eDD8wZKegbrt6tp7W7JzzXy34TB9WtYhREuRaKoBlUW2ZBCwTETOisg5YBkw2F5DvE3ePN/t+QIn4+XmxQvdX8Db5Nj1KH/sOMnplCw9Na2pNpRnF6k+cLTQeSJGi+RyblZK9eZS2ZLiyhYZIb1MtoSoqKgrGuOOOw/4PkANcw0e8H0A0yETUYeunL8s+GR9BnW9FXJiB1EnSx/xMzU19arPWJHRtjuP8rS/oo0y/grMEZEspdT9wCygv7WFReQr4CsAHx8fyR/3uBKNLjTizxV/MqjfIEJqOlbgLP7YBfYt/Yvnhramf6+mZVJnVFQUJT1jRUXb7jzK0/7KIltSYll7aFqzKdcGXGu1c1nxzAri58Xbda/Z6xLwMrkyqlPDEvNqNFWFSiFbgqFE8G+LfIk/8G9LWrlyOv60zTNIAOfSslkYd5yRHetT06v4KXKNpipSKWRLROSsUupVDCcF8IqInC0v2/NJOZ6Cb7DtsXjnbjpKVq6ZcXpqWlPNqBSyJZZr04HpDjWwBEYvHI13oG0zTXlm4dv1h+netDatrim7QOEaTWWgWq3kLS1+9f1s7iIt33WKY+czGHedDuitqX5oB+NgZq9LILimJwNbl12gcI2msqAdjAPZdyqFv/ef4fZujXFz1a9aU/3Q33oHMmtdAu5uLozpokNiaqon2sFYyeavN7PyuZVW57+YmcP8zccY3j6YAB93B1qm0VRcKtpK3gpL6JhQm8I0/BSTSHp2np6a1lRrtIOxEpO3CZO3dYvkzGZh9roEOjaqRWiDspOY1WgqG7qL5ADW7Esi4Uw64/SuaU01RzsYBzBrbQJ1fD0Y0q5eyZk1miqMdjBlTEJyGlF7k7itSyPc3fTr1VRv9F+Alfxw8w8cjj5cYr5v1h/GVSlu66qnpjUaPchrJTdMu6HEQd60rFx+iDnKkNB6BPl5lpNlGk3FRTsYK/Hy9yoxzy9xx0jJzGVcd73vSKMB3UUqM0SEWWsTaBvsR0Rju+ORazRVinJ1MEqpwUqpPUqp/UqpyVfJd7NSSpRSnSznTZRSGUqpOMvni/Kz2jrWHzzL3lOpjNNqjRpNAeXWRSokW/IvjKDdm5RSi0Rk52X5fIFHgQ2XVXFARMLLxVg7mLU2AX9vE8PbBzvbFI2mwlARZUteBd4GMsvRtquSnZbNR00+uuL1Y+cz+HPnSSI7N8LT5FqOlmk0FZsKJVuilOoINBSR35RSEbadoAAADQpJREFUEy8rH6KU2gJcBJ4TkejLb2CLbEk+1kg4ZJ3JIiMl44r5Fh3IxizQTI4TFXWyxHuWFZVZPkPb7jzK1X4RKZcPcAuGXGz++VgMadj8cxcgCmhiOY8COlmOPYDaluMIDEfld7X7eXt7izWsWrWqxDxJu5PkkxafXPH6kI/WyMgpf1l1v7LEGtsrKtp252Gv/UCa2Ph3X5FkS3yBdkCUUioB6AYsUkp1EpEsETkDICKxwAGgZblYDQQ0D+CedfcUe+3ImXR2nriotwVoNMVQYWRLROSCiASKSBMRaQKsB4aLSIxSqo5lkBilVFOgBXCwvAx3cXXBu3bxwb5/jz8BwOB215SXORpNpaGiyZZcid7AK0qpHMAMTBAnyJYUx+/xJwmtX5OGAY7VtdZoKiMVSrbksvS+hY5/Bn52qHF2cPx8BnFHzzNxUCtnm6LRVEj0St5SsDTemDEaortHGk2xaAdjBes/Ws/qV1YXSV8af5JWQb40rVPDCVZpNBUfvdnRCjpN6IQ513xJ2umUTDYdPssj/Vs4ySqNpuKjHYwVuHkWfU1/7jiFCFwfqqenNZorobtIdvJ7/AmaBvrQMkh3jzSaK6EdjB2cS8tm/cGzDG53jd45rdFcBe1g7GDZzlPkmUWv3tVoSkA7GCuYPXA2xzb9s6vh9/gTNPD3ol19PydapdFUfPQgrxVELogsGOi9mJnDX/uTGa8DS2k0JaIdjBV4+HoUHK/YdYqcPGGw7h5pNCWiu0g28vv2kwT5edChYS1nm6LRVHi0g7GBtKxcVu9NYnDba3Bx0d0jjaYktIOxgag9SWTlmhmiF9dpNFahHUwJpJ5K5ZPmnwCwJP4EgTXc6dwkwMlWaTSVg0ohW2JJe9pSbo9SalD5WAxZF7NQSpGZk8eq3af5V5trcNXdI43GKiqFbIlSqg1GBLy2QDCwXCnVUkTyHG131sUs3H3dWbM3ifTsPB2aQaOxgcoiWzICmGuJzXsI2G+pz+HU61CP8avHszT+JDW9THRvVrs8bqvRVAkqi2xJfYwYvYXL1r/8BoVlSwBRSmVYYZcbkGtFPgDcX7I2Z7lgk+0VDG2787DX/pIF2ou5UYVAKeUCfACMt7cOEfkK+MrG+8aISKeSc1Y8tO3OoTLbDuVrf3k6GFtkSwCuwZAt+f/2zj/GrqKK458v/UEbfrXYpiGgKSAVK2JFJATaUpAgLRaoGGqCtIioBERACSHBPwrRSGmoAgZIULSStvKzESogRdsCLVgp3e4uBVrBQiD9gVIrICKW4x/nPHd2eW/3ve2+vW/z5pO8vHnnzp17Zu7c82bmzpk5vYpzM5lMAzIgti2JeF+VtKekg/FtS9b0o+6ZTKYXDIhtSyLe3cAGvO94cR++QaqpS9VgZN2LYSDrDv2ov3xHyEwmk+l78kzeTCZTN7KByWQydaNpDUy1bgtFImmzpDZJLZKeCdn+kpZJ2hTfI0MuSTdFflpjTlF/63uHpO2S2hNZzfpKmh3xN0maXaDucyS9HuXfImlacqys60oR9UrSRyUtl7RB0nOSLg158WVvZk33wQeZXwIOAYYC64HxRetVRs/NwKgusuuBqyJ8FTA3wtOAhwEBxwJ/KkDfycBRQHtv9QX2B16O75ERHlmQ7nOAK8rEHR91Zk/g4KhLg4qqV8ABwFER3gfYGDoWXvbN2oKp1m2hETkDWBDhBcCZifzX5jwNjJDUr+tKmNnjwJtdxLXq+0VgmZm9aWY7gGXAqQXpXolKriuF1Csz22Jmz0b4LeB5fKZ74WXfrAamnNvCh1wPGgADHpW0NtwgAMaY2ZYIbwXGRLhR81Srvo2Wj+9EN+KOUheDBtZd0ljgs7izcOFl36wGZqAw0cyOAqYCF0uanB40b9cOmHkGA01f4FbgUGACsAW4oVh1ukfS3sB9wGVm9s/0WFFl36wGZkC4HpjZ6/G9HViCN8G3lbo+8b09ojdqnmrVt2HyYWbbzGyXmX0A3E6HB3/D6S5pCG5cFprZ/SEuvOyb1cB067bQCEjaK9bGQdJewClAO65naXR/NvDbCD8AzIo3BMcCO5PmcZHUqu/vgVMkjYwuySkh63e6jGHNwMsfKruuFFKvJAn4BfC8mc1PDhVf9vUe4W7UDz6SvhEf9b+6aH3K6HcI/hZiPfBcSUfgI8AfgE3AY8D+IRe+oNdLQBtwdAE6L8a7Eu/j/fdv9EZf4Hx84PQvwNcL1P3O0K01HsoDkvhXh+4vAlOLrFfARLz70wq0xGdaI5R9dhXIZDJ1o1m7SJlMph/IBiaTydSNbGAymUzdyAYmk8nUjWxgMplM3WgqAyNpV+IZ29Jbb1dJK5RsClfjuVMkHZf8vlDSrN6kVSbtcZIeCk/YZyXdLWlMz2d2SmOEpIu6OT5c0kr5Pld1Q+5JPqqe14jrzAsP5Hm9PP/n8n27uovz/3tca92RNCH14u4m3tvVptlNGqMlPbK76aQ0zK4C/cS7ZjahYB2mAG8DqwHM7La+SFTSMOB3wPfM7MGQTQFGA9tqSGoEcBFwS4Xj5wP3Wz9setdbJA02s2q35fgWPj+kV/kxswuqiLM793gCcDTw0G6kURVm9oakLZKON7NVfZVo03yAt8vITgXuSX5PAZZG+FbgGXyi2zVJnBXE5KQ0TeArwK8iPB13OFuHT3IaA4zFnc5exydDTSJZEgCvTE/jE6aWEK7ycb25+GzRjcCkMvk4H/eQLZfvYcAv8UlV64ATQ/6pSLMlrnkY7gH8bsjmlUlrNTA2KasVwL3AC8BCOpZh3UwsNYE/ICsiPAf37H0CeAX4Mr6sQBvwCDAkOb8kXwN8POSj8Snxf47P8Um6dwKrgMVddBYwD5+J2wbMDPkDwK7I68wu51SrZ6e6APwInxz5NO5sWErriiT+jXHNduCYkB8DPBX3ZzXwCXzJh1eBN0o6Ansn97IVOKuHa1cqrxPomJS3Dtgn5GcAt/TZM1f0Q9+fn6QytSQ3bHDcxL0So/K1CJdmPg6KinFkDQZmJB0P2wXADV0rW5nK1wqcEOFrgZ8m1yudPw14rEze5gOXVsj39/FF1gEOj/wOA24Gzgn5UHxjrbEka6J0SWcosDX5PQXYifus7BEPyMQ4tpnKBuZJYAjwGeBfxExY3KiemZxfmr08iw6jvyi5xsfw6fGldNcCw8vofRa+9MAg3NC/SszKpcyfTo16pnXBgOkRvh74QZl7vAK4PcKTS2UN7AsMjvDJwH0RPg/4WaLX3FK9KNWzHq5dqbwepMPY7J1c+0Cgra+eudxFAqLfOV3SvcBpwJVx6OxYJmEwvqjPeNwIVMNBwF3hzzIU+Gt3kSXtB4wws5UhWgDck0QpObCtxY1ALUzEjQlm9oKkV4BxuEG4WtJBeLdnk7u1VGQU8I8usjVm9lrkoSV0e7IHfR42s/clteEPfanf30bnvC1Ovn8S4ZOB8Yme+4YXMcADZlZuN8+JeKtmF+4AuBL4PD37CVWrZ4n/AEsjvBbfh70ci8HXoJG0r6QR+EJRCyQdhhuLIRXOPRn3cSLS2NHDtSuV1ypgvqSF+L1/LY5vx/d/7xOaapC3G34DnA2chG+h8lY4sV0BfMHMjsTHN4aVOTf1tUiP34z/83wa+HaFc2vhvfjeRfmxs+eAz9WSoJktAk7Hu0QPSTqph1Pe5cP5eC8Jp7r9l476VfYccy/l9y3+OoEP6Jw3KxPeAzjWzCbE50AzKw1wvtOD/rVSrZ4l0jiV7hN8eNkEw/dkX25mR+Dd61rrS6Vrly0vM7sOb1kPB1ZJOjziD8Pvc5+QDYyzEl8u8Zu4sQFvsr4D7Iw3MVMrnLtN0iflW9/OSOT70eHqPjuRv4X/W3XCzHYCOyRNCtG5oVe1LAKOk3RaSSBpsqQj8HGEc0I2Dm8qvyjpEOBlM7sJ97Q9spJ+oeMOYFAMKPfEZjoM3lk15CNlZvL9VIQfBS4pRZBUzaD9E8BMSYMkjca7JkVu3DcTQNJE3JN5J53ry3lJ3K73YxlwcelHsghWJcqWl6RDzazNzObiYzMlAzOODq/x3abZDMzwLq+prwOIpvNS3IgsDdl6fPDrBfzhrTSqflWcsxr3xi0xB7hH0lrgb4n8QWBGXH8SnZkNzJPUig/4XlttxqJr8CXgknhNvQF/G/QG/kZoj2jq3wWcZ2bv4a229ujaHIEPEv8d/0drr/Dq9lG8y9ET1wA3yhcr7+0bp5FRFpcCl4fsu8DR8lXmNgAXVpHOErxrux74I3ClmW3tpU59wb8lrQNuw722wcdNfhzytOWzHO/itEiaCfwQL5d2SeuBE3u4VqXyuizSaMU9yB8O+Yl4a71PyN7UmZqQr0B/uZmdW7Qumb5H0uPAGcnYzm7RbC2YzG5ivrj08npPtMv0P9F9nN9XxgVyCyaTydSR3ILJZDJ1IxuYTCZTN7KByWQydSMbmEwmUzeygclkMnXjf/tF7rNQ25dwAAAAAElFTkSuQmCC\n",
|
||
"text/plain": [
|
||
"<Figure size 288x288 with 2 Axes>"
|
||
]
|
||
},
|
||
"metadata": {
|
||
"needs_background": "light"
|
||
},
|
||
"output_type": "display_data"
|
||
},
|
||
{
|
||
"name": "stdout",
|
||
"output_type": "stream",
|
||
"text": [
|
||
"cifar100\n"
|
||
]
|
||
},
|
||
{
|
||
"data": {
|
||
"image/png": "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\n",
|
||
"text/plain": [
|
||
"<Figure size 288x288 with 2 Axes>"
|
||
]
|
||
},
|
||
"metadata": {
|
||
"needs_background": "light"
|
||
},
|
||
"output_type": "display_data"
|
||
},
|
||
{
|
||
"name": "stdout",
|
||
"output_type": "stream",
|
||
"text": [
|
||
"ImageNet16-120\n"
|
||
]
|
||
},
|
||
{
|
||
"data": {
|
||
"image/png": "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\n",
|
||
"text/plain": [
|
||
"<Figure size 288x288 with 2 Axes>"
|
||
]
|
||
},
|
||
"metadata": {
|
||
"needs_background": "light"
|
||
},
|
||
"output_type": "display_data"
|
||
}
|
||
],
|
||
"source": [
|
||
"import itertools\n",
|
||
"sh={\n",
|
||
" 'cifar10':{\n",
|
||
" 'jacob_cov':(40,-.02),\n",
|
||
" 'synflow':(40,.015),\n",
|
||
" 'snip':(10,.02),\n",
|
||
" 'grad_norm':(40,-.025),\n",
|
||
" 'grasp':(40,.0),\n",
|
||
" 'vote':(60,.0)\n",
|
||
" },\n",
|
||
" 'cifar100':{\n",
|
||
" 'jacob_cov':(40,.0),\n",
|
||
" 'synflow':(40,.0),\n",
|
||
" 'snip':(35,-.02),\n",
|
||
" 'grad_norm':(35,.02),\n",
|
||
" 'grasp':(35,.0),\n",
|
||
" 'vote':(60,.0)\n",
|
||
" },\n",
|
||
" 'ImageNet16-120':{\n",
|
||
" 'jacob_cov':(40,.0),\n",
|
||
" 'synflow':(40,.0),\n",
|
||
" 'snip':(40,.02),\n",
|
||
" 'grad_norm':(40,-.01),\n",
|
||
" 'grasp':(40,-.03),\n",
|
||
" 'vote':(60,.0)\n",
|
||
" },\n",
|
||
"}\n",
|
||
"markers = {'synflow':'*','jacob_cov':'x','snip':'o','grad_norm':'s','fisher':'+','grasp':'d', 'vote':'P'}\n",
|
||
"for ds,slow in dslow.items():\n",
|
||
" plt.figure(figsize=(4,4))\n",
|
||
" x=range(0,196*41,196)\n",
|
||
" plt.plot(x,slow)\n",
|
||
" for k,v in allc[ds.upper() if 'cifar' in ds else ds].items():\n",
|
||
" if v < 0.4:\n",
|
||
" continue\n",
|
||
" plt.scatter(1,v, marker=markers[k], s=100)\n",
|
||
" plt.text(1+sh[ds][k][0],v+sh[ds][k][1],f'{k}',horizontalalignment='left')\n",
|
||
" \n",
|
||
" \n",
|
||
" k='vote'\n",
|
||
" v = votes[ds]\n",
|
||
" plt.scatter(3,v, marker=markers['vote'], s=100)\n",
|
||
" plt.text(1+sh[ds][k][0],v+sh[ds][k][1],f'{k}',horizontalalignment='left')\n",
|
||
" \n",
|
||
" if ds == 'cifar10':\n",
|
||
" x2 = [c/3.3 for c in x]\n",
|
||
" plt.plot(x2,econas[3][0:epx], label='econas $r_{16}c_8$', linestyle=ls[4], color='purple', linewidth=1)\n",
|
||
" p=15\n",
|
||
" plt.scatter(p*196/d, econas[3][p], marker='o', color='purple', s=60)\n",
|
||
" plt.annotate('econas+',(p*196/d+30, econas[3][p]-0.02), horizontalalignment='left', color='purple')\n",
|
||
" x3 = [c/4 for c in x]\n",
|
||
" p=20\n",
|
||
" plt.scatter(p*196/4, econas[0][p], marker='p', color='orange', s=80)\n",
|
||
" plt.annotate(f'econas',(p*196/4+50, econas[0][p]-0.005), horizontalalignment='left', color='chocolate')\n",
|
||
"\n",
|
||
" \n",
|
||
" #plt.legend()\n",
|
||
" plt.grid()\n",
|
||
"\n",
|
||
" ax1 = plt.gca()\n",
|
||
" ax1.set_xlim(-100,196*11)\n",
|
||
" ax1.set_ylim(0.4,0.85)\n",
|
||
" ax2 = ax1.twiny()\n",
|
||
" ax2.set_xticks(range(0,12))\n",
|
||
" ax1.set_xlabel(\"Evaluation Cost (number of minibatches)\")\n",
|
||
" ax2.set_xlabel(\"Epochs\")\n",
|
||
"\n",
|
||
" ax1.set_ylabel('Spearman $\\\\rho$')\n",
|
||
" #plt.xscale('log')\n",
|
||
" print(ds)\n",
|
||
" plt.tight_layout()\n",
|
||
" plt.savefig(f'nb2'+(ds if 'cifar' in ds else 'im16')+'.pdf')\n",
|
||
" plt.show()"
|
||
]
|
||
},
|
||
{
|
||
"cell_type": "code",
|
||
"execution_count": null,
|
||
"metadata": {},
|
||
"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": "3.7.6"
|
||
}
|
||
},
|
||
"nbformat": 4,
|
||
"nbformat_minor": 4
|
||
}
|