{ "cells": [ { "cell_type": "code", "execution_count": 1, "metadata": {}, "outputs": [], "source": [ "from fastai import *\n", "from fastai.vision import *\n", "from fastai.vision.models import efficientnet\n", "from fastai.utils.ipython import *\n", "from fastai.callbacks.tracker import SaveModelCallback\n", "from sklearn.model_selection import StratifiedKFold\n", "import matplotlib.pyplot as plt\n", "from functools import partial" ] }, { "cell_type": "code", "execution_count": 2, "metadata": {}, "outputs": [], "source": [ "%reload_ext autoreload\n", "%autoreload 2\n", "%matplotlib inline" ] }, { "cell_type": "code", "execution_count": 3, "metadata": {}, "outputs": [ { "data": { "text/plain": [ "[PosixPath('../../../Dataset/Herlev Dataset/best-rn34-herlev-multiclass-5fold.pkl'),\n", " PosixPath('../../../Dataset/Herlev Dataset/best-vgg19-herlev-multiclass-5fold.pkl'),\n", " PosixPath('../../../Dataset/Herlev Dataset/best-effb3-herlev-multiclass.pkl'),\n", " PosixPath('../../../Dataset/Herlev Dataset/abnormal_moderate-dysplastic'),\n", " PosixPath('../../../Dataset/Herlev Dataset/normal_superficiel'),\n", " PosixPath('../../../Dataset/Herlev Dataset/best-rn101-herlev-multiclass-5fold.pkl'),\n", " PosixPath('../../../Dataset/Herlev Dataset/abnormal_light-dysplastic'),\n", " PosixPath('../../../Dataset/Herlev Dataset/abnormal_severe-dysplastic'),\n", " PosixPath('../../../Dataset/Herlev Dataset/normal_columnar'),\n", " PosixPath('../../../Dataset/Herlev Dataset/normal_intermediate'),\n", " PosixPath('../../../Dataset/Herlev Dataset/abnormal_carcinoma-in-situ'),\n", " PosixPath('../../../Dataset/Herlev Dataset/models'),\n", " PosixPath('../../../Dataset/Herlev Dataset/best-effb4-herlev-multiclass.pkl')]" ] }, "execution_count": 3, "metadata": {}, "output_type": "execute_result" } ], "source": [ "path = Path(\".\")\n", "data_path = path / \"..\" / \"..\" / \"..\" / \"Dataset\" / \"Herlev Dataset\"\n", "data_path.ls()" ] }, { "cell_type": "code", "execution_count": 4, "metadata": {}, "outputs": [ { "data": { "text/plain": [ "LabelLists;\n", "\n", "Train: LabelList (917 items)\n", "x: ImageList\n", "Image (3, 83, 146),Image (3, 106, 116),Image (3, 129, 119),Image (3, 108, 110),Image (3, 209, 173)\n", "y: CategoryList\n", "abnormal_moderate-dysplastic,abnormal_moderate-dysplastic,abnormal_moderate-dysplastic,abnormal_moderate-dysplastic,abnormal_moderate-dysplastic\n", "Path: ../../../Dataset/Herlev Dataset;\n", "\n", "Valid: LabelList (0 items)\n", "x: ImageList\n", "\n", "y: CategoryList\n", "\n", "Path: ../../../Dataset/Herlev Dataset;\n", "\n", "Test: None" ] }, "execution_count": 4, "metadata": {}, "output_type": "execute_result" } ], "source": [ "data_init = (ImageList.from_folder(data_path)\n", " .split_none()\n", " .label_from_folder())\n", "data_init" ] }, { "cell_type": "code", "execution_count": 5, "metadata": {}, "outputs": [ { "data": { "text/plain": [ "StratifiedKFold(n_splits=5, random_state=0, shuffle=True)" ] }, "execution_count": 5, "metadata": {}, "output_type": "execute_result" } ], "source": [ "skf = StratifiedKFold(n_splits=5, shuffle=True, random_state=0)\n", "skf" ] }, { "cell_type": "code", "execution_count": 6, "metadata": {}, "outputs": [], "source": [ "tfms = get_transforms(flip_vert=True, max_warp=0.0, max_rotate=30.0)" ] }, { "cell_type": "code", "execution_count": 7, "metadata": {}, "outputs": [ { "data": { "text/plain": [ "[ .new_type at 0x7fedf854e378>>,\n", " Precision(average='macro', pos_label=1, eps=1e-09),\n", " Recall(average='macro', pos_label=1, eps=1e-09),\n", " FBeta(average='macro', pos_label=1, eps=1e-09, beta=2),\n", " KappaScore(weights='quadratic')]" ] }, "execution_count": 7, "metadata": {}, "output_type": "execute_result" } ], "source": [ "our_metrics = [accuracy, Precision(average=\"macro\"), Recall(average=\"macro\"), FBeta(average=\"macro\"), KappaScore(weights=\"quadratic\")]\n", "our_metrics" ] }, { "cell_type": "code", "execution_count": 8, "metadata": {}, "outputs": [], "source": [ "idxs = [[train_idx, val_idx] for train_idx, val_idx in skf.split(data_init.x.items, data_init.y.items)]" ] }, { "cell_type": "code", "execution_count": 9, "metadata": {}, "outputs": [], "source": [ "def model_callback(model, model_name):\n", " return [SaveModelCallback(model, every=\"improvement\", monitor=\"accuracy\", name=model_name)]" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "# Fold-1" ] }, { "cell_type": "code", "execution_count": 10, "metadata": {}, "outputs": [ { "data": { "text/plain": [ "ImageDataBunch;\n", "\n", "Train: LabelList (733 items)\n", "x: ImageList\n", "Image (3, 64, 64),Image (3, 64, 64),Image (3, 64, 64),Image (3, 64, 64),Image (3, 64, 64)\n", "y: CategoryList\n", "abnormal_moderate-dysplastic,abnormal_moderate-dysplastic,abnormal_moderate-dysplastic,abnormal_moderate-dysplastic,abnormal_moderate-dysplastic\n", "Path: ../../../Dataset/Herlev Dataset;\n", "\n", "Valid: LabelList (184 items)\n", "x: ImageList\n", "Image (3, 64, 64),Image (3, 64, 64),Image (3, 64, 64),Image (3, 64, 64),Image (3, 64, 64)\n", "y: CategoryList\n", "abnormal_moderate-dysplastic,abnormal_moderate-dysplastic,abnormal_moderate-dysplastic,abnormal_moderate-dysplastic,abnormal_moderate-dysplastic\n", "Path: ../../../Dataset/Herlev Dataset;\n", "\n", "Test: None" ] }, "execution_count": 10, "metadata": {}, "output_type": "execute_result" } ], "source": [ "fold_idxs = idxs[0]\n", "fold_data = (ImageList.from_folder(data_path)\n", " .split_by_idxs(fold_idxs[0], fold_idxs[1])\n", " .label_from_folder()\n", " .transform(tfms, size=64)\n", " .databunch(bs=64)\n", " .normalize(imagenet_stats))\n", "fold_data" ] }, { "cell_type": "code", "execution_count": 11, "metadata": {}, "outputs": [ { "name": "stderr", "output_type": "stream", "text": [ "Downloading: \"https://github.com/lukemelas/EfficientNet-PyTorch/releases/download/1.0/efficientnet-b5-b6417697.pth\" to /home/fadege/.cache/torch/checkpoints/efficientnet-b5-b6417697.pth\n" ] }, { "data": { "application/vnd.jupyter.widget-view+json": { "model_id": "c2b977adab10418189fae5d58f63e23d", "version_major": 2, "version_minor": 0 }, "text/plain": [ "HBox(children=(FloatProgress(value=0.0, max=122410125.0), HTML(value='')))" ] }, "metadata": {}, "output_type": "display_data" }, { "name": "stdout", "output_type": "stream", "text": [ "\n", "Loaded pretrained weights for efficientnet-b5\n" ] } ], "source": [ "learner = Learner(fold_data, efficientnet.EfficientNetB5(fold_data), metrics=our_metrics).to_fp16()\n", "learner = learner.split([learner.model._conv_stem, learner.model._blocks, learner.model._conv_head])" ] }, { "cell_type": "code", "execution_count": 12, "metadata": {}, "outputs": [ { "data": { "text/html": [ "\n", "
\n", " \n", " \n", " 70.00% [7/10 00:25<00:10]\n", "
\n", " \n", "\n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", "
epochtrain_lossvalid_lossaccuracyprecisionrecallf_betakappa_scoretime
02.006588#na#00:03
12.009895#na#00:03
22.003915#na#00:03
31.977159#na#00:03
41.898656#na#00:03
51.831242#na#00:04
62.445978#na#00:03

\n", "\n", "

\n", " \n", " \n", " 81.82% [9/11 00:03<00:00 7.0030]\n", "
\n", " " ], "text/plain": [ "" ] }, "metadata": {}, "output_type": "display_data" }, { "name": "stdout", "output_type": "stream", "text": [ "LR Finder is complete, type {learner_name}.recorder.plot() to see the graph.\n" ] }, { "data": { "image/png": "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\n", "text/plain": [ "
" ] }, "metadata": { "needs_background": "light" }, "output_type": "display_data" } ], "source": [ "learner.lr_find()\n", "learner.recorder.plot()" ] }, { "cell_type": "code", "execution_count": 13, "metadata": {}, "outputs": [ { "data": { "text/html": [ "\n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", "
epochtrain_lossvalid_lossaccuracyprecisionrecallf_betakappa_scoretime
01.9903601.9557840.1250000.1774690.1118670.077404-0.06673000:04
11.9583511.9562110.1576090.1665490.2013870.1519440.07753700:04
21.9127861.9571300.163043nan0.2388020.1677710.11743600:04
31.8590081.9595830.173913nan0.2650710.1835630.09973700:04
41.7950491.9417800.211957nan0.2896650.2186640.14927100:04
51.7102431.8503570.3097830.4242120.3824320.3201280.25126300:04
61.6155061.7297590.3913040.4418950.4615940.4021270.36590200:05
71.5219301.6014140.4293480.4502410.4856840.4456790.45769900:04
81.4313931.4601890.4673910.4895860.5042080.4872220.47878400:05
91.3506631.3926840.4782610.5117740.5026860.4952230.56877000:04
101.2665181.3929080.4782610.5222010.5140040.5085870.58474600:04
111.2045511.4320150.4456520.4822650.4657120.4609160.55899800:04
121.1532221.4088520.4836960.5534310.5172710.5174630.60672800:04
131.1068471.4595910.5108700.5570140.5341200.5319860.61669400:04
141.0488551.4798030.5000000.5449900.5311380.5299350.63413400:05
151.0126621.3793260.5489130.6068860.5685970.5690380.69281500:05
160.9694141.4837050.5163040.5988340.5345760.5351110.64352500:04
170.9246981.4313690.5054350.5748750.5439430.5470390.63497600:05
180.8775961.4640910.5434780.6222410.5755490.5786860.66397500:04
190.8361841.4877320.5434780.6139530.5683910.5701230.70369400:04
200.7968611.4789100.5326090.6076340.5641490.5683680.65862100:04
210.7618741.3855020.5706520.6378700.5944060.5971000.69387800:04
220.7352161.3570680.5815220.6346060.5962040.5975900.69592800:04
230.6978141.2824190.5869570.6682540.6162080.6208020.74159200:05
240.6681681.2656480.5815220.6570630.6156290.6208930.71823700:05
250.6437551.2859930.5815220.6438790.6120260.6160600.72599100:05
260.6179761.3983140.5978260.6646230.6210450.6252820.73640000:04
270.5856271.4013490.5923910.6593240.6209480.6239860.71949100:04
280.5636901.3950570.6032610.6611000.6319370.6348210.72230400:04
290.5345361.4751080.5978260.6516850.6240910.6250610.71447600:04
300.5134451.4286720.5869570.6546990.6087060.6115180.72047100:05
310.4860831.3540750.6195650.6953060.6359810.6406110.74928500:04
320.4673671.3513640.5760870.6496840.5997250.6058760.73477300:04
330.4402701.3803610.5706520.6314320.5938790.5988540.72293400:05
340.4243451.3644960.5760870.6328080.5979380.6023480.71335700:04
350.4074311.3390230.5706520.6055010.5822920.5849140.69779800:04
360.3919561.3539800.5869570.6351510.6100410.6123340.69132100:04
370.3710031.4002850.5706520.6259010.5955570.5985900.71033200:04
380.3612531.3990590.5706520.6270440.5964580.5997850.71293200:05
390.3485181.3926290.5652170.6250000.5936960.5972360.71070300:04
400.3376951.3798400.5706520.6358010.5984580.6031120.72125400:04
410.3330431.3628100.5815220.6486570.6068830.6120310.72723800:04
420.3260531.3613780.5760870.6454160.6032200.6080670.73070800:04
430.3272511.3486990.5869570.6540890.6107440.6156590.73593500:04
440.3198061.3388870.5869570.6540890.6107440.6156590.73593500:05
450.3153301.3258390.5869570.6540890.6107440.6156590.73593500:05
460.3054441.3190570.5869570.6540890.6107440.6156590.73593500:04
470.3029001.3142760.5869570.6543980.6107440.6157190.73813700:04
480.2968841.3060980.5815220.6490530.6070810.6118510.72413700:04
490.2920641.3058400.5706520.6313890.5993590.6029510.70616800:05
" ], "text/plain": [ "" ] }, "metadata": {}, "output_type": "display_data" }, { "name": "stdout", "output_type": "stream", "text": [ "Better model found at epoch 0 with accuracy value: 0.125.\n", "Better model found at epoch 1 with accuracy value: 0.15760870277881622.\n", "Better model found at epoch 2 with accuracy value: 0.16304348409175873.\n", "Better model found at epoch 3 with accuracy value: 0.17391304671764374.\n", "Better model found at epoch 4 with accuracy value: 0.21195651590824127.\n", "Better model found at epoch 5 with accuracy value: 0.30978259444236755.\n", "Better model found at epoch 6 with accuracy value: 0.3913043439388275.\n", "Better model found at epoch 7 with accuracy value: 0.42934781312942505.\n", "Better model found at epoch 8 with accuracy value: 0.46739131212234497.\n", "Better model found at epoch 9 with accuracy value: 0.47826087474823.\n", "Better model found at epoch 12 with accuracy value: 0.4836956560611725.\n", "Better model found at epoch 13 with accuracy value: 0.510869562625885.\n", "Better model found at epoch 15 with accuracy value: 0.5489130616188049.\n", "Better model found at epoch 21 with accuracy value: 0.570652186870575.\n", "Better model found at epoch 22 with accuracy value: 0.58152174949646.\n", "Better model found at epoch 23 with accuracy value: 0.5869565010070801.\n", "Better model found at epoch 26 with accuracy value: 0.5978260636329651.\n", "Better model found at epoch 28 with accuracy value: 0.60326087474823.\n", "Better model found at epoch 31 with accuracy value: 0.6195651888847351.\n" ] } ], "source": [ "learner.fit_one_cycle(50, max_lr=slice(2e-03), callbacks=model_callback(learner, \"best-effb5-herlev-multiclass-fold1-stage1\"))\n", "learner.save(\"last-effb5-herlev-multiclass-fold1-stage1\")" ] }, { "cell_type": "code", "execution_count": 14, "metadata": {}, "outputs": [ { "data": { "text/html": [ "\n", "
\n", " \n", " \n", " 60.00% [6/10 00:21<00:14]\n", "
\n", " \n", "\n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", "
epochtrain_lossvalid_lossaccuracyprecisionrecallf_betakappa_scoretime
00.380424#na#00:03
10.405748#na#00:03
20.394624#na#00:03
30.389265#na#00:03
40.397556#na#00:03
50.785674#na#00:03

\n", "\n", "

\n", " \n", " \n", " 36.36% [4/11 00:01<00:03 1.4637]\n", "
\n", " " ], "text/plain": [ "" ] }, "metadata": {}, "output_type": "display_data" }, { "name": "stdout", "output_type": "stream", "text": [ "LR Finder is complete, type {learner_name}.recorder.plot() to see the graph.\n" ] }, { "data": { "image/png": "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\n", "text/plain": [ "
" ] }, "metadata": { "needs_background": "light" }, "output_type": "display_data" } ], "source": [ "learner.load(\"best-effb5-herlev-multiclass-fold1-stage1\")\n", "learner = to_fp16(learner)\n", "learner.unfreeze()\n", "learner.lr_find()\n", "learner.recorder.plot() " ] }, { "cell_type": "code", "execution_count": 15, "metadata": {}, "outputs": [ { "data": { "text/html": [ "\n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", "
epochtrain_lossvalid_lossaccuracyprecisionrecallf_betakappa_scoretime
00.3466711.3376070.6141300.6917150.6321200.6368820.74533500:04
10.3767721.3266030.5978260.6752130.6189330.6237470.73200200:05
20.3627271.3166830.6032610.6801780.6236950.6284210.73694100:04
30.3660561.3127700.6086960.6857780.6308380.6359500.74596600:04
40.3737741.2972710.6141300.6887940.6356000.6406930.75802700:04
50.3775771.2826930.6086960.6820080.6317390.6366900.75088300:05
60.3735651.2765350.6086960.6816170.6317390.6366500.74823400:05
70.3674361.2796130.5978260.6710390.6233140.6281370.74189100:04
80.3670591.2882910.5869570.6637670.6105460.6165290.71515100:04
90.3620631.2627470.6086960.6759030.6386980.6433960.74350100:04
100.3571631.2669930.5978260.6602600.6182900.6234850.72465800:05
110.3508171.2767150.6032610.6664900.6219530.6273700.73492700:05
120.3440251.2922600.5706520.6494660.6037390.6090960.72457200:04
130.3525311.2935990.5923910.6675170.6282140.6327780.74574800:04
140.3486401.2989800.5869570.6584140.6232540.6269130.73618400:04
150.3413981.2917970.5597830.6390860.5972990.6024050.72044000:04
160.3286001.2871070.5706520.6512510.6048230.6111110.73444800:05
170.3197921.3340820.5706520.6400900.6104640.6136990.70712400:04
180.3202921.3120920.5597830.6277770.5963990.6006530.69679800:04
190.3168581.3205320.5652170.6311270.6000620.6042700.70962000:04
200.3070701.3202360.5543480.6244830.5881570.5927850.70208900:05
210.3068621.3258040.5760870.6478650.6066850.6123960.72107000:04
220.3012611.3316100.5706520.6438610.5995420.6055250.71742300:04
230.3012341.3397810.5597830.6372160.5929190.5987000.69994900:04
240.2929191.3383860.5597830.6351180.5927210.5975780.69179500:04
250.2824421.3478280.5597830.6471970.5927210.5984130.70186400:05
260.2886041.3368120.5652170.6422940.5965820.6017310.69141400:04
270.2809471.3377350.5652170.6394960.5954830.6006450.69751000:04
280.2719101.3509790.5652170.6340600.5965820.6010680.69155600:04
290.2640881.3536930.5652170.6390370.5965820.6014650.69008000:04
300.2584291.3513300.5652170.6377940.5954830.6004940.69015200:05
310.2506391.3557690.5597830.6334470.5916220.5965610.68039500:04
320.2446781.3720110.5706520.6361300.5991460.6037040.68261100:05
330.2456431.3741030.5597830.6308380.5925230.5969610.67968600:04
340.2437811.3658630.5652170.6388190.5963840.6011600.67354400:04
350.2383521.3638860.5652170.6401600.5961860.6009160.67853500:04
360.2344731.3624950.5597830.6375620.5923250.5970020.67975600:05
370.2362371.3688740.5489130.6288930.5839000.5888340.67245100:05
380.2407951.3644440.5543480.6302710.5879590.5929290.66633200:04
390.2371811.3703900.5543480.6302710.5879590.5929290.66633200:04
400.2440471.3752330.5597830.6349970.5916220.5967520.67653500:04
410.2387061.3698120.5652170.6381410.5954830.6005880.67796600:04
420.2341971.3693460.5706520.6347920.5995420.6040520.68476200:04
430.2357311.3687400.5760870.6393920.6032050.6077820.69118000:04
440.2346151.3632870.5760870.6407000.6030070.6076680.69361900:04
450.2263091.3587310.5760870.6447210.6032050.6081980.68344500:05
460.2289741.3613800.5760870.6393920.6032050.6077820.69118000:04
470.2263451.3629960.5706520.6359340.5993440.6039770.68714700:04
480.2238341.3624780.5760870.6447210.6032050.6081980.68344500:04
490.2199321.3617050.5760870.6447210.6032050.6081980.68344500:04
" ], "text/plain": [ "" ] }, "metadata": {}, "output_type": "display_data" }, { "name": "stdout", "output_type": "stream", "text": [ "Better model found at epoch 0 with accuracy value: 0.614130437374115.\n" ] } ], "source": [ "learner.fit_one_cycle(50, max_lr=slice(4.2e-04), callbacks=model_callback(learner, \"best-effb5-herlev-multiclass-fold1-stage2\"))\n", "learner.save(\"last-effb5-herlev-multiclass-fold1-stage2\")" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "# Fold-2" ] }, { "cell_type": "code", "execution_count": 16, "metadata": {}, "outputs": [ { "data": { "text/plain": [ "ImageDataBunch;\n", "\n", "Train: LabelList (733 items)\n", "x: ImageList\n", "Image (3, 64, 64),Image (3, 64, 64),Image (3, 64, 64),Image (3, 64, 64),Image (3, 64, 64)\n", "y: CategoryList\n", "abnormal_moderate-dysplastic,abnormal_moderate-dysplastic,abnormal_moderate-dysplastic,abnormal_moderate-dysplastic,abnormal_moderate-dysplastic\n", "Path: ../../../Dataset/Herlev Dataset;\n", "\n", "Valid: LabelList (184 items)\n", "x: ImageList\n", "Image (3, 64, 64),Image (3, 64, 64),Image (3, 64, 64),Image (3, 64, 64),Image (3, 64, 64)\n", "y: CategoryList\n", "abnormal_moderate-dysplastic,abnormal_moderate-dysplastic,abnormal_moderate-dysplastic,abnormal_moderate-dysplastic,abnormal_moderate-dysplastic\n", "Path: ../../../Dataset/Herlev Dataset;\n", "\n", "Test: None" ] }, "execution_count": 16, "metadata": {}, "output_type": "execute_result" } ], "source": [ "fold_idxs = idxs[1]\n", "fold_data = (ImageList.from_folder(data_path)\n", " .split_by_idxs(fold_idxs[0], fold_idxs[1])\n", " .label_from_folder()\n", " .transform(tfms, size=64)\n", " .databunch(bs=64)\n", " .normalize(imagenet_stats))\n", "fold_data" ] }, { "cell_type": "code", "execution_count": 17, "metadata": {}, "outputs": [ { "data": { "text/html": [ "\n", "
\n", " \n", " \n", " 70.00% [7/10 00:23<00:09]\n", "
\n", " \n", "\n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", "
epochtrain_lossvalid_lossaccuracyprecisionrecallf_betakappa_scoretime
00.656594#na#00:02
10.616238#na#00:02
20.635108#na#00:03
30.642984#na#00:03
40.625624#na#00:03
50.648236#na#00:03
61.035056#na#00:03

\n", "\n", "

\n", " \n", " \n", " 45.45% [5/11 00:02<00:02 2.1597]\n", "
\n", " " ], "text/plain": [ "" ] }, "metadata": {}, "output_type": "display_data" }, { "name": "stdout", "output_type": "stream", "text": [ "LR Finder is complete, type {learner_name}.recorder.plot() to see the graph.\n" ] }, { "data": { "image/png": "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\n", "text/plain": [ "
" ] }, "metadata": { "needs_background": "light" }, "output_type": "display_data" } ], "source": [ "learner.load(\"best-effb5-herlev-multiclass-fold1-stage2\")\n", "learner = to_fp16(learner)\n", "learner.data = fold_data\n", "learner.freeze()\n", "learner = to_fp16(learner)\n", "learner.lr_find()\n", "learner.recorder.plot()" ] }, { "cell_type": "code", "execution_count": 18, "metadata": {}, "outputs": [ { "data": { "text/html": [ "\n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", "
epochtrain_lossvalid_lossaccuracyprecisionrecallf_betakappa_scoretime
00.6637910.1936600.9293480.9368830.9448460.9427490.94807900:03
10.6716830.1908850.9293480.9368830.9448460.9427490.94807900:04
20.6805610.1875130.9293480.9368830.9448460.9427490.94807900:04
30.6427490.1839240.9293480.9368830.9448460.9427490.94807900:03
40.6404580.1891630.9293480.9368830.9448460.9427490.94807900:03
50.6259870.2030830.9239130.9273960.9381220.9343940.95401600:04
60.6054690.2105050.9130430.9201150.9296970.9261750.96198300:03
70.5890270.2176720.9076090.9136250.9194930.9162460.96128200:04
80.5701120.2176370.9293480.9335550.9417850.9390110.97359400:04
90.5602520.2151100.9293480.9339410.9417850.9392170.97116400:04
100.5482460.2335240.9130430.9215030.9296970.9266020.96191500:03
110.5431290.2222560.9293480.9377330.9461430.9433840.96862400:04
120.5382720.2466560.8967390.9114460.9174260.9128060.96378500:04
130.5361020.2466260.8967390.9054030.9066870.9042330.94759100:04
140.5430960.2488180.9130430.9201540.9309940.9271280.95571200:04
150.5276270.2290390.9130430.9164700.9244530.9208760.95593600:04
160.5171100.2274210.9239130.9295390.9372210.9347490.95722500:04
170.5018820.2420330.9021740.9124840.9221730.9185010.95378200:04
180.4957430.2329140.9130430.9204630.9298950.9266060.95557300:03
190.4834060.2370830.9076090.9125850.9196910.9162500.95497600:04
200.4790900.2239480.9184780.9264920.9335580.9311280.95632100:03
210.4779080.2233000.9293480.9393050.9485240.9459150.95789400:04
220.4710650.2224880.9293480.9393050.9485240.9459150.95789400:03
230.4676440.2222620.9293480.9393050.9485240.9459150.95789400:04
240.4641160.2206820.9293480.9393050.9485240.9459150.95789400:03
250.4623860.2217390.9293480.9393050.9485240.9459150.95789400:04
260.4606660.2227200.9293480.9393050.9485240.9459150.95789400:03
270.4530610.2215610.9293480.9393050.9485240.9459150.95789400:03
280.4491050.2216540.9293480.9393050.9485240.9459150.95789400:03
290.4588630.2200080.9293480.9393050.9485240.9459150.95789400:03
" ], "text/plain": [ "" ] }, "metadata": {}, "output_type": "display_data" }, { "name": "stdout", "output_type": "stream", "text": [ "Better model found at epoch 0 with accuracy value: 0.929347813129425.\n" ] } ], "source": [ "learner.fit_one_cycle(30, max_lr=slice(1e-03), callbacks=model_callback(learner, \"best-effb5-herlev-multiclass-fold2-stage1\"))\n", "learner.save(\"last-effb5-herlev-multiclass-fold2-stage1\")" ] }, { "cell_type": "code", "execution_count": 19, "metadata": {}, "outputs": [ { "data": { "text/html": [ "\n", "
\n", " \n", " \n", " 70.00% [7/10 00:25<00:10]\n", "
\n", " \n", "\n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", "
epochtrain_lossvalid_lossaccuracyprecisionrecallf_betakappa_scoretime
00.724178#na#00:03
10.697326#na#00:03
20.677409#na#00:03
30.635879#na#00:03
40.629408#na#00:03
50.939369#na#00:03
62.328744#na#00:03

\n", "\n", "

\n", " \n", " \n", " 0.00% [0/11 00:00<00:00]\n", "
\n", " " ], "text/plain": [ "" ] }, "metadata": {}, "output_type": "display_data" }, { "name": "stdout", "output_type": "stream", "text": [ "LR Finder is complete, type {learner_name}.recorder.plot() to see the graph.\n" ] }, { "data": { "image/png": "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\n", "text/plain": [ "
" ] }, "metadata": { "needs_background": "light" }, "output_type": "display_data" } ], "source": [ "learner.load(\"best-effb5-herlev-multiclass-fold2-stage1\")\n", "learner = to_fp16(learner)\n", "learner.unfreeze()\n", "learner.lr_find()\n", "learner.recorder.plot()" ] }, { "cell_type": "code", "execution_count": 20, "metadata": {}, "outputs": [ { "data": { "text/html": [ "\n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", "
epochtrain_lossvalid_lossaccuracyprecisionrecallf_betakappa_scoretime
00.6318080.1899400.9293480.9368830.9448460.9427490.94807900:04
10.6377310.1871240.9347830.9430800.9485090.9471310.94881600:04
20.6505880.1861080.9347830.9430800.9485090.9471310.94881600:04
30.6449680.1845370.9347830.9430800.9485090.9471310.94881600:04
40.6529030.1855640.9402170.9480000.9556520.9537000.96191000:04
50.6248410.1861960.9402170.9480000.9556520.9537000.96191000:04
60.6040120.1797790.9402170.9480000.9556520.9537000.96191000:04
70.6049380.1765940.9456520.9542970.9595130.9582960.95631500:04
80.5953850.1904370.9402170.9496560.9523700.9516360.95293300:04
90.5734800.2021010.9184780.9326020.9318430.9317010.93018200:04
100.5486540.1974600.9239130.9367080.9368020.9364220.93177300:04
110.5301180.2035730.9184780.9314040.9320740.9314450.92509700:04
120.5006320.2127030.9239130.9301820.9337750.9321100.93531600:04
130.4780550.2205950.9021740.9110490.9123460.9115960.91020100:04
140.4626100.2298240.9130430.9218480.9196720.9197940.91784000:04
150.4462800.2475140.9021740.9123580.9136290.9126820.91445800:04
160.4353080.2487360.9076090.9198940.9160090.9163180.91058600:04
170.4162190.2558200.9076090.9193360.9182070.9175290.92209400:04
180.4043650.2297820.9184780.9317490.9322720.9314690.91926100:04
190.3870660.2413280.9130430.9220550.9209690.9206680.91426400:04
200.3710790.2580740.9130430.9220550.9209690.9206680.91426400:04
210.3592930.2790780.8804350.8904050.8887910.8879940.90223600:04
220.3491100.2677120.9021740.9091900.9069040.9064430.91279900:04
230.3394190.2645740.9021740.9121240.9136100.9129520.90521100:05
240.3313910.2442330.8967390.9051710.9021420.9018900.90502000:04
250.3249700.2569540.8858700.8904200.8899090.8894930.91525400:04
260.3137100.2780740.8913040.9041980.8994330.8993640.91634600:05
270.3018860.3054990.8858700.8965460.8968690.8955350.91645500:05
280.2908660.3032160.8913040.9011480.9003340.8994810.93275800:05
290.2769870.3011790.9021740.9131700.9113380.9105330.93073800:05
300.2624410.3124200.8858700.8903340.8919280.8908760.92160000:04
310.2539560.3224790.8804350.8927430.8906460.8899050.90303300:04
320.2374120.3253130.8750000.8888730.8858840.8852060.90190500:04
330.2263200.3268230.8804350.8992660.8884290.8893920.91480900:04
340.2252440.3131060.8858700.9038930.8931910.8940910.90979400:04
350.2225220.3016030.8858700.8949590.8873300.8883320.91958000:04
360.2151860.2968770.8913040.8947710.8957360.8951640.92054800:04
370.2152790.2909190.8804350.8833810.8884100.8866900.91904300:04
380.2123630.2898740.8858700.8876570.8933360.8913890.91973500:04
390.2106600.2905020.8750000.8792060.8825310.8807990.90002800:04
400.2098260.2951020.8804350.8834030.8896730.8873970.91277400:04
410.2111400.2960570.8858700.8903750.8933360.8919230.91344200:04
420.2107440.2947290.8858700.8903750.8933360.8919230.91344200:04
430.2045840.2950630.8750000.8819570.8834840.8824780.91208300:04
440.2025940.2940230.8750000.8819570.8834840.8824780.91208300:04
450.1957460.2926550.8750000.8819570.8834840.8824780.91208300:04
460.2063660.2901710.8695650.8780640.8763410.8758360.89935800:04
470.1940740.2890330.8750000.8820610.8800040.8797830.90622600:04
480.1885430.2889500.8750000.8820610.8800040.8797830.90622600:04
490.1861140.2887720.8804350.8861810.8871470.8864370.91903100:04
" ], "text/plain": [ "" ] }, "metadata": {}, "output_type": "display_data" }, { "name": "stdout", "output_type": "stream", "text": [ "Better model found at epoch 0 with accuracy value: 0.929347813129425.\n", "Better model found at epoch 1 with accuracy value: 0.9347826242446899.\n", "Better model found at epoch 4 with accuracy value: 0.9402173757553101.\n", "Better model found at epoch 7 with accuracy value: 0.945652186870575.\n" ] } ], "source": [ "learner.fit_one_cycle(50, max_lr=slice(1e-04, 2e-04), callbacks=model_callback(learner, \"best-effb5-herlev-multiclass-fold2-stage2\"))\n", "learner.save(\"last-effb5-herlev-multiclass-fold2-stage2\")" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "# Fold-3" ] }, { "cell_type": "code", "execution_count": 21, "metadata": {}, "outputs": [ { "data": { "text/plain": [ "ImageDataBunch;\n", "\n", "Train: LabelList (734 items)\n", "x: ImageList\n", "Image (3, 64, 64),Image (3, 64, 64),Image (3, 64, 64),Image (3, 64, 64),Image (3, 64, 64)\n", "y: CategoryList\n", "abnormal_moderate-dysplastic,abnormal_moderate-dysplastic,abnormal_moderate-dysplastic,abnormal_moderate-dysplastic,abnormal_moderate-dysplastic\n", "Path: ../../../Dataset/Herlev Dataset;\n", "\n", "Valid: LabelList (183 items)\n", "x: ImageList\n", "Image (3, 64, 64),Image (3, 64, 64),Image (3, 64, 64),Image (3, 64, 64),Image (3, 64, 64)\n", "y: CategoryList\n", "abnormal_moderate-dysplastic,abnormal_moderate-dysplastic,abnormal_moderate-dysplastic,abnormal_moderate-dysplastic,abnormal_moderate-dysplastic\n", "Path: ../../../Dataset/Herlev Dataset;\n", "\n", "Test: None" ] }, "execution_count": 21, "metadata": {}, "output_type": "execute_result" } ], "source": [ "fold_idxs = idxs[2]\n", "fold_data = (ImageList.from_folder(data_path)\n", " .split_by_idxs(fold_idxs[0], fold_idxs[1])\n", " .label_from_folder()\n", " .transform(tfms, size=64)\n", " .databunch(bs=64)\n", " .normalize(imagenet_stats))\n", "fold_data" ] }, { "cell_type": "code", "execution_count": 22, "metadata": {}, "outputs": [ { "data": { "text/html": [ "\n", "
\n", " \n", " \n", " 70.00% [7/10 00:19<00:08]\n", "
\n", " \n", "\n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", "
epochtrain_lossvalid_lossaccuracyprecisionrecallf_betakappa_scoretime
00.552787#na#00:02
10.531320#na#00:02
20.551429#na#00:02
30.553952#na#00:02
40.548774#na#00:03
50.559560#na#00:02
60.989472#na#00:02

\n", "\n", "

\n", " \n", " \n", " 27.27% [3/11 00:01<00:04 1.7004]\n", "
\n", " " ], "text/plain": [ "" ] }, "metadata": {}, "output_type": "display_data" }, { "name": "stdout", "output_type": "stream", "text": [ "LR Finder is complete, type {learner_name}.recorder.plot() to see the graph.\n" ] }, { "data": { "image/png": "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\n", "text/plain": [ "
" ] }, "metadata": { "needs_background": "light" }, "output_type": "display_data" } ], "source": [ "learner.load(\"best-effb5-herlev-multiclass-fold2-stage2\")\n", "learner = to_fp16(learner)\n", "learner.data = fold_data\n", "learner.freeze()\n", "learner = to_fp16(learner)\n", "learner.lr_find()\n", "learner.recorder.plot()" ] }, { "cell_type": "code", "execution_count": 23, "metadata": {}, "outputs": [ { "data": { "text/html": [ "\n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", "
epochtrain_lossvalid_lossaccuracyprecisionrecallf_betakappa_scoretime
00.5424840.0920490.9781420.9821930.9806240.9805940.98069600:04
10.5510230.0929430.9781420.9821930.9806240.9805940.98069600:03
20.5367150.0964470.9781420.9821930.9806240.9805940.98069600:04
30.5328460.1043840.9726780.9785360.9756980.9757330.97988900:03
40.5416460.1256560.9617490.9682290.9632530.9634580.97570100:03
50.5499380.1687000.9508200.9560490.9588670.9578120.96486500:03
60.5433420.2066110.9180330.9364500.9289620.9293580.93889000:03
70.5492710.1933210.9180330.9308820.9312740.9307140.93897600:03
80.5343040.1840570.9180330.9358110.9306450.9307750.92723300:04
90.5312600.1767570.9344260.9467720.9442050.9444490.95412100:04
100.5279500.1813630.9344260.9493300.9430150.9437670.94711600:03
110.5377310.1690440.9562840.9593090.9587990.9583450.97027900:03
120.5408640.1545710.9672130.9718250.9734810.9730220.97243400:03
130.5411420.2282880.9344260.9396940.9378560.9372560.95124200:03
140.5342200.2013360.9344260.9443520.9416600.9411810.94721900:04
150.5202020.1742400.9344260.9427330.9430150.9424000.94070800:03
160.5057140.2281310.9071040.9089340.9107080.9085840.92095600:03
170.5002250.2142290.9453550.9468920.9455600.9448640.94255300:04
180.4899800.2137410.9344260.9397090.9370630.9364490.94686300:04
190.4847480.2209470.9289620.9309680.9257400.9252420.94743000:04
200.4771730.1976040.9289620.9275250.9242660.9243780.94358500:04
210.4704900.1994240.9289620.9298790.9278160.9276050.94360000:03
220.4637680.1918580.9344260.9328550.9317850.9314390.94704900:04
230.4471410.1897420.9289620.9275470.9282130.9277120.93977600:04
240.4353550.1869650.9344260.9382610.9384170.9376180.94030900:03
250.4396970.1856280.9344260.9382610.9384170.9376180.94030900:04
260.4392510.1882370.9344260.9382610.9384170.9376180.94030900:04
270.4389700.1886410.9344260.9382610.9384170.9376180.94030900:03
280.4218820.1894010.9344260.9382610.9384170.9376180.94030900:04
290.4104170.1920970.9344260.9382610.9384170.9376180.94030900:04
" ], "text/plain": [ "" ] }, "metadata": {}, "output_type": "display_data" }, { "name": "stdout", "output_type": "stream", "text": [ "Better model found at epoch 0 with accuracy value: 0.9781420826911926.\n" ] } ], "source": [ "learner.fit_one_cycle(30, max_lr=slice(2.5e-03), callbacks=model_callback(learner, \"best-effb5-herlev-multiclass-fold3-stage1\"))\n", "learner.save(\"last-effb5-herlev-multiclass-fold3-stage1\")" ] }, { "cell_type": "code", "execution_count": 24, "metadata": {}, "outputs": [ { "data": { "text/html": [ "\n", "
\n", " \n", " \n", " 60.00% [6/10 00:22<00:14]\n", "
\n", " \n", "\n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", "
epochtrain_lossvalid_lossaccuracyprecisionrecallf_betakappa_scoretime
00.543508#na#00:04
10.537442#na#00:03
20.540773#na#00:03
30.514560#na#00:03
40.502647#na#00:03
50.767722#na#00:03

\n", "\n", "

\n", " \n", " \n", " 36.36% [4/11 00:02<00:03 1.1947]\n", "
\n", " " ], "text/plain": [ "" ] }, "metadata": {}, "output_type": "display_data" }, { "name": "stdout", "output_type": "stream", "text": [ "LR Finder is complete, type {learner_name}.recorder.plot() to see the graph.\n" ] }, { "data": { "image/png": "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\n", "text/plain": [ "
" ] }, "metadata": { "needs_background": "light" }, "output_type": "display_data" } ], "source": [ "learner.load(\"best-effb5-herlev-multiclass-fold3-stage1\")\n", "learner = to_fp16(learner)\n", "learner.unfreeze()\n", "learner.lr_find()\n", "learner.recorder.plot()" ] }, { "cell_type": "code", "execution_count": 25, "metadata": {}, "outputs": [ { "data": { "text/html": [ "\n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", "
epochtrain_lossvalid_lossaccuracyprecisionrecallf_betakappa_scoretime
00.5474750.0921970.9781420.9821930.9806240.9805940.98069600:04
10.5575840.0937830.9781420.9821930.9806240.9805940.98069600:04
20.5463330.0940850.9781420.9821930.9806240.9805940.98069600:04
30.5310150.0947160.9781420.9821930.9806240.9805940.98069600:04
40.5387220.0943910.9781420.9821930.9806240.9805940.98069600:04
50.5316890.0942750.9781420.9821930.9806240.9805940.98069600:04
60.5347780.0958460.9726780.9785360.9756980.9757330.97988900:04
70.5186680.0978640.9726780.9785360.9756980.9757330.97988900:04
80.5155040.0977790.9726780.9785360.9756980.9757330.97988900:04
90.5008740.0995820.9726780.9771010.9770530.9768370.97327100:05
100.4954580.0981950.9726780.9771010.9770530.9768370.97327100:04
110.4978250.0979500.9781420.9861720.9807880.9814690.97634200:04
120.4968660.1042160.9672130.9794940.9683430.9698960.97464900:04
130.4820800.1034130.9672130.9794940.9683430.9698960.97464900:04
140.4755410.1063810.9726780.9786530.9731050.9738670.97982100:04
150.4756640.1096000.9672130.9756330.9683430.9692370.97212000:05
160.4656180.1069600.9781420.9861720.9780310.9793300.98315400:04
170.4588670.1038240.9781420.9861720.9780310.9793300.98315400:04
180.4528450.1066930.9726780.9786530.9731050.9738670.97982100:04
190.4541160.1084600.9672130.9756330.9683430.9692370.97212000:04
200.4536420.1112670.9617490.9727470.9635810.9645600.96434200:04
210.4462870.1134100.9617490.9727470.9635810.9645600.96434200:05
220.4331030.1133780.9672130.9756330.9683430.9692370.97212000:04
230.4332630.1129200.9726780.9786530.9731050.9738670.97982100:04
240.4253730.1143820.9672130.9756330.9683430.9692370.97212000:04
250.4204650.1167080.9672130.9735510.9683430.9689900.97634200:04
260.4139590.1149340.9617490.9682120.9647720.9652650.96888900:04
270.4121710.1143070.9617490.9682120.9647720.9652650.96888900:04
280.4100670.1146180.9617490.9682120.9647720.9652650.96888900:04
290.4094560.1149300.9672130.9718230.9722910.9719810.96978300:04
300.4057340.1150230.9672130.9718230.9722910.9719810.96978300:04
310.3987770.1158910.9617490.9682120.9647720.9652650.96888900:04
320.4005080.1167810.9562840.9632730.9600100.9603680.96535600:04
330.3927020.1160300.9617490.9682120.9647720.9652650.96888900:04
340.3926930.1155440.9617490.9682120.9647720.9652650.96888900:05
350.3965580.1170970.9617490.9682120.9647720.9652650.96888900:04
360.3931570.1177270.9617490.9682120.9647720.9652650.96888900:05
370.3922420.1172800.9617490.9682120.9647720.9652650.96888900:05
380.3917630.1165270.9617490.9682120.9647720.9652650.96888900:04
390.4004870.1157890.9617490.9682120.9647720.9652650.96888900:04
" ], "text/plain": [ "" ] }, "metadata": {}, "output_type": "display_data" }, { "name": "stdout", "output_type": "stream", "text": [ "Better model found at epoch 0 with accuracy value: 0.9781420826911926.\n" ] } ], "source": [ "learner.fit_one_cycle(40, max_lr=slice(2.5e-04), callbacks=model_callback(learner, \"best-effb5-herlev-multiclass-fold3-stage2\"))\n", "learner.save(\"last-effb5-herlev-multiclass-fold3-stage2\")" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "# Fold-4" ] }, { "cell_type": "code", "execution_count": 26, "metadata": {}, "outputs": [ { "data": { "text/plain": [ "ImageDataBunch;\n", "\n", "Train: LabelList (734 items)\n", "x: ImageList\n", "Image (3, 64, 64),Image (3, 64, 64),Image (3, 64, 64),Image (3, 64, 64),Image (3, 64, 64)\n", "y: CategoryList\n", "abnormal_moderate-dysplastic,abnormal_moderate-dysplastic,abnormal_moderate-dysplastic,abnormal_moderate-dysplastic,abnormal_moderate-dysplastic\n", "Path: ../../../Dataset/Herlev Dataset;\n", "\n", "Valid: LabelList (183 items)\n", "x: ImageList\n", "Image (3, 64, 64),Image (3, 64, 64),Image (3, 64, 64),Image (3, 64, 64),Image (3, 64, 64)\n", "y: CategoryList\n", "abnormal_moderate-dysplastic,abnormal_moderate-dysplastic,abnormal_moderate-dysplastic,abnormal_moderate-dysplastic,abnormal_moderate-dysplastic\n", "Path: ../../../Dataset/Herlev Dataset;\n", "\n", "Test: None" ] }, "execution_count": 26, "metadata": {}, "output_type": "execute_result" } ], "source": [ "fold_idxs = idxs[3]\n", "fold_data = (ImageList.from_folder(data_path)\n", " .split_by_idxs(fold_idxs[0], fold_idxs[1])\n", " .label_from_folder()\n", " .transform(tfms, size=64)\n", " .databunch(bs=64)\n", " .normalize(imagenet_stats))\n", "fold_data" ] }, { "cell_type": "code", "execution_count": 27, "metadata": {}, "outputs": [ { "data": { "text/html": [ "\n", "
\n", " \n", " \n", " 70.00% [7/10 00:20<00:08]\n", "
\n", " \n", "\n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", "
epochtrain_lossvalid_lossaccuracyprecisionrecallf_betakappa_scoretime
00.592700#na#00:03
10.565768#na#00:02
20.558862#na#00:02
30.552209#na#00:03
40.557785#na#00:02
50.578329#na#00:02
60.970533#na#00:02

\n", "\n", "

\n", " \n", " \n", " 27.27% [3/11 00:01<00:03 1.6721]\n", "
\n", " " ], "text/plain": [ "" ] }, "metadata": {}, "output_type": "display_data" }, { "name": "stdout", "output_type": "stream", "text": [ "LR Finder is complete, type {learner_name}.recorder.plot() to see the graph.\n" ] }, { "data": { "image/png": "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\n", "text/plain": [ "
" ] }, "metadata": { "needs_background": "light" }, "output_type": "display_data" } ], "source": [ "learner.load(\"best-effb5-herlev-multiclass-fold3-stage2\")\n", "learner = to_fp16(learner)\n", "learner.data = fold_data\n", "learner.freeze()\n", "learner = to_fp16(learner)\n", "learner.lr_find()\n", "learner.recorder.plot()" ] }, { "cell_type": "code", "execution_count": 28, "metadata": {}, "outputs": [ { "data": { "text/html": [ "\n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", "
epochtrain_lossvalid_lossaccuracyprecisionrecallf_betakappa_scoretime
00.5610030.1018980.9726780.9788990.9760470.9764190.98915600:03
10.5645710.1018870.9726780.9788990.9760470.9764190.98915600:04
20.5388950.1009990.9726780.9788990.9760470.9764190.98915600:04
30.5424560.1011950.9726780.9782150.9760470.9763500.98250900:04
40.5516150.1005500.9781420.9829260.9800150.9805090.98336900:04
50.5618770.1009780.9781420.9829260.9800150.9805090.98336900:03
60.5587790.1012480.9781420.9829260.9800150.9805090.98336900:04
70.5521960.1019230.9781420.9829260.9800150.9805090.98336900:04
80.5554510.1006990.9781420.9829260.9800150.9805090.98336900:04
90.5451940.1013020.9781420.9829260.9800150.9805090.98336900:04
100.5362510.1008030.9781420.9829260.9800150.9805090.98336900:04
110.5444390.1015400.9781420.9829260.9800150.9805090.98336900:04
120.5444880.1030500.9726780.9782370.9764440.9766750.98253600:04
130.5419820.1035880.9726780.9782370.9764440.9766750.98253600:04
140.5406300.1043430.9726780.9782370.9764440.9766750.98253600:03
150.5427600.1042070.9726780.9782370.9764440.9766750.98253600:04
160.5462790.1040790.9726780.9782370.9764440.9766750.98253600:03
170.5361070.1040250.9726780.9782370.9764440.9766750.98253600:04
180.5289550.1045380.9726780.9782370.9764440.9766750.98253600:03
190.5380880.1046070.9726780.9782370.9764440.9766750.98253600:04
200.5390760.1048720.9726780.9782370.9764440.9766750.98253600:03
210.5385640.1050740.9726780.9782370.9764440.9766750.98253600:03
220.5465890.1051200.9726780.9782370.9764440.9766750.98253600:04
230.5430170.1044200.9726780.9782370.9764440.9766750.98253600:04
240.5423760.1051020.9672130.9738220.9724760.9725290.98167500:03
250.5468340.1050040.9726780.9782370.9764440.9766750.98253600:03
260.5321920.1057690.9672130.9738220.9724760.9725290.98167500:04
270.5320560.1058030.9672130.9738220.9724760.9725290.98167500:04
280.5390690.1052510.9726780.9782370.9764440.9766750.98253600:04
290.5449180.1041150.9726780.9782370.9764440.9766750.98253600:03
" ], "text/plain": [ "" ] }, "metadata": {}, "output_type": "display_data" }, { "name": "stdout", "output_type": "stream", "text": [ "Better model found at epoch 0 with accuracy value: 0.9726775884628296.\n", "Better model found at epoch 4 with accuracy value: 0.9781420826911926.\n" ] } ], "source": [ "learner.fit_one_cycle(30, max_lr=slice(8.5e-06), callbacks=model_callback(learner, \"best-effb5-herlev-multiclass-fold4-stage1\"))\n", "learner.save(\"last-effb5-herlev-multiclass-fold4-stage1\")" ] }, { "cell_type": "code", "execution_count": 29, "metadata": {}, "outputs": [ { "data": { "text/html": [ "\n", "
\n", " \n", " \n", " 60.00% [6/10 00:22<00:14]\n", "
\n", " \n", "\n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", "
epochtrain_lossvalid_lossaccuracyprecisionrecallf_betakappa_scoretime
00.584816#na#00:03
10.562936#na#00:03
20.550384#na#00:04
30.531703#na#00:03
40.540769#na#00:03
50.760071#na#00:03

\n", "\n", "

\n", " \n", " \n", " 45.45% [5/11 00:02<00:02 1.3520]\n", "
\n", " " ], "text/plain": [ "" ] }, "metadata": {}, "output_type": "display_data" }, { "name": "stdout", "output_type": "stream", "text": [ "LR Finder is complete, type {learner_name}.recorder.plot() to see the graph.\n" ] }, { "data": { "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYgAAAEGCAYAAAB/+QKOAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjEsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+j8jraAAAgAElEQVR4nO3deXxU9b3/8dcnCWENa8K+BBBERBYFRHFDq1K1Yuu1xVtbvbZS22p/rbe2drnWar31trfVerWttkXrUvcNFatWW3eFAGELixgIJAGSkEAWss/n98dMdIhDFpLJTJL38/GYh3O2OZ85hrzzPed7vsfcHRERkcYSYl2AiIjEJwWEiIhEpIAQEZGIFBAiIhKRAkJERCJKinUB7SU1NdXT09NjXYaISKeyatWqIndPi7SsywREeno6GRkZsS5DRKRTMbOcwy3TKSYREYlIASEiIhEpIEREJCIFhIiIRKSAEBGRiBQQIiISkQJCREQiimpAmNlCM9tiZtvM7IYIy8ea2T/NbI2ZrTOz88KW/Si03RYzOzeadYqIdFZPrcrl0RU7o/LZUQsIM0sE7gY+C0wFLjWzqY1W+ynwuLvPAhYDvw9tOzU0fSywEPh96PNERCTMg+/n8GxmXlQ+O5otiLnANnfPdvca4FFgUaN1HOgfej8AyA+9XwQ86u7V7r4d2Bb6PBERCakPOFv2lDF1xICofH40A2IUsCtsOjc0L9xNwGVmlgssB65txbaY2RIzyzCzjMLCwvaqW0SkU9ixr4LK2nqOGZESlc+P9UXqS4H73X00cB7woJm1uCZ3v9fdZ7v77LS0iGNNiYh0WVn5pQBMHdm/mTWPTDQH68sDxoRNjw7NC/c1gtcYcPf3zKwXkNrCbUVEurVNu0vpkWhMGtr5WhArgUlmNt7MkgledF7WaJ2dwFkAZnYM0AsoDK232Mx6mtl4YBKwIoq1ioh0Olm7S5mY1o/kpOj8Ko9aC8Ld68zsGuBlIBFY6u4bzexmIMPdlwH/CfzJzL5H8IL1Fe7uwEYzexzIAuqAb7t7fbRqFRHpjLLySzllUmrUPj+qz4Nw9+UELz6Hz7sx7H0WMP8w294K3BrN+kREOqui8moKyqqZOiI61x8g9hepRUTkCGzaHbpArYAQEZFwDT2YjlFAiIhIuKzdpYwc0ItBfZOjtg8FhIhIJ7Rpd2lUWw+ggBAR6XSqauv5qLAiajfINVBAiIh0Mlv3llEf8KheoAYFhIhIp9PQg0mnmERE5BBZ+aX0TU5k7OA+Ud2PAkJEpJPJCl2gTkiwqO5HASEi0okEAs6m3WVRP70ECggRkU4lt6SS8uq6qPdgAgWEiEinkrX7ABDdITYaKCBERDqRrN1lJBgcPTw6z4AIp4AQEelEsvJLmZDWj149EqO+LwWEiEgnsml3aYecXgIFhIhIp7H/YA15+ys7pAcTKCBERDqNTbvLADqkBxMoIEREOo2sDnhIUDgFhIhIJ5GVX0pqv56kpfTskP0pIEREOolNu0s77PQSKCBERDqFmroAHxaUddjpJVBAiIh0Clv3llFb72pBiIjIoVbuKAZg9rhBHbZPBYSISCewckcxowb2ZuTA3h22TwWEiEicc3dW7ihhTnrHtR4gygFhZgvNbIuZbTOzGyIsv93MMkOvrWa2P2xZfdiyZdGsU0QknuXsO0hhWTVzxg/u0P0mReuDzSwRuBs4G8gFVprZMnfPaljH3b8Xtv61wKywj6h095nRqk9EpLNYEbr+MCe9YwMimi2IucA2d8929xrgUWBRE+tfCjwSxXpERDqljB3FDOzTg6PS+nXofqMZEKOAXWHTuaF5n2Jm44DxwOths3uZWYaZvW9mFx1muyWhdTIKCwvbq24RkbiyckcJs8cNivozqBuLl4vUi4En3b0+bN44d58N/Dtwh5lNbLyRu9/r7rPdfXZaWlpH1Soi0mEKy6rZXlTR4aeXILoBkQeMCZseHZoXyWIanV5y97zQf7OBf3Ho9QkRkW4ho+H6QwdfoIboBsRKYJKZjTezZIIh8KneSGY2BRgEvBc2b5CZ9Qy9TwXmA1mNtxUR6epW7CimV48Epo0c0OH7jlovJnevM7NrgJeBRGCpu280s5uBDHdvCIvFwKPu7mGbHwPcY2YBgiF2W3jvJxGR7iJjRwkzxwwkOanjrwhELSAA3H05sLzRvBsbTd8UYbt3geOiWZuISLwrr65jY/4Bvr3gqJjsP14uUouISCNrdpYQ8I6//6GBAkJEJE6t3F5MgsHxHThAXzgFhIhInFq5o4SpI/vTr2dUrwYclgJCRCQO1dQFWLOrJGanl0ABISISlzbkH6CqNqCAEBGRQ2XEaIC+cAoIEZE4tGJ7CeNT+5KW0jNmNSggRETiTCDgrMop7tDHi0aigBARiTMfFZZTcrA2JuMvhVNAiIjEmZU7SoDYXn8ABYSISNx596Mi0lJ6kj6kT0zrUECIiMSR2voAb2wtZMHRaZh17AOCGlNAiIjEkYwdJZRV1XHmlGGxLkUBISIST17fvJfkxAROnZQa61IUECIi8eS1zQXMmziEvjEafymcAkJEJE5sL6ogu7CCs6YMjXUpgAJCRCRuvLZpLwBnKiBERCTc65sLmDysH2MGx7Z7awMFhIhIHCitqmXF9uK46L3UQAEhIhIH3txaSF3A+cwx8XF6CRQQIiJx4fVNBQzs04NZY2M7QF84BYSISIzVB5x/bilgwdFDSUyI7d3T4RQQIiIxlrmrhJKDtXHTe6mBAkJEJMZe21RAYoJx2uS0WJdyCAWEiEiMvbapgDnpgxjQu0esSzmEAkJEJIZ2FR9ky94yPnNM/HRvbRDVgDCzhWa2xcy2mdkNEZbfbmaZoddWM9sftuxyM/sw9Lo8mnWKiMTKP7cUAPFz93S4qI0GZWaJwN3A2UAusNLMlrl7VsM67v69sPWvBWaF3g8GfgbMBhxYFdq2JFr1iojEwmubChif2pcJaf1iXcqnRLMFMRfY5u7Z7l4DPAosamL9S4FHQu/PBV519+JQKLwKLIxirSIiHa6qtp73svex4Oj4az1AdANiFLArbDo3NO9TzGwcMB54vTXbmtkSM8sws4zCwsJ2KVpEpKOs3bWfmroAJ08cEutSIoqXi9SLgSfdvb41G7n7ve4+291np6XFV/cwEZHmZOQEz5qfMC5+7p4OF82AyAPGhE2PDs2LZDGfnF5q7bYiIp1Sxo5ijhraj0F9k2NdSkTRDIiVwCQzG29myQRDYFnjlcxsCjAIeC9s9svAOWY2yMwGAeeE5omIdAmBgLMqp4TZcdp6gCj2YnL3OjO7huAv9kRgqbtvNLObgQx3bwiLxcCj7u5h2xab2S0EQwbgZncvjlatIiId7cOCckqr6uL29BJEMSAA3H05sLzRvBsbTd90mG2XAkujVpyISAxl5AT/5p2TPjjGlRxevFykFhHpVlbtKCG1XzLjhsTH0+MiUUCIiMTAypxiThg3CLP4Gd67MQWEiEgHKyitYldxZVyfXgIFhIhIh4v3+x8aKCBERDrYyh3F9ExK4NiRA2JdSpMUECIiHWxVTgkzxwwkOSm+fwXHd3UiIl3MwZo6NuaXMjs9vk8vgQJCRKRDZe7cT33AmT0uvi9QgwJCRKRDZeSUYAbHj1ULQkREwmTklDB5aAoD+sTX86cjUUCIiHSQ+oCzOqeEEzrB9QdQQIiIdJgte8oor65jjgJCRETCrQoN0NcZLlCDAkJEpMOs3FHC0JSejB7UO9altEiLAsLM+ppZQuj9ZDO70Mzi/wqLiEgcWZVTwpz0wXE9QF+4lrYg3gR6mdko4BXgK8D90SpKRKSryd9fSd7+yrgffylcSwPC3P0g8AXg9+5+CXBs9MoSEelaGgbo6wx3UDdocUCY2UnAl4EXQ/MSo1OSiEjX8+SqXFL79eSYEf1jXUqLtTQgvgv8CHgm9FzpCcA/o1eWiEjXkZVfyptbC/mP+en0SOw8fYNa9Exqd38DeAMgdLG6yN2/E83CRES6ij+9lU2f5EQuO3FcrEtplZb2YvqbmfU3s77ABiDLzK6PbmkiIp1f3v5Klq3N59K5YzvF8BrhWtrWmerupcBFwEvAeII9mUREpAlL394OwJWnjI9xJa3X0oDoEbrv4SJgmbvXAh69skREOr8DB2t5ZMVOLpwxklEDO8fNceFaGhD3ADuAvsCbZjYOKI1WUSIiXcFDH+RwsKaeJadNiHUpR6SlF6nvBO4Mm5VjZguiU5KISOdXVVvPfe/s4LTJaZ2qa2u4ll6kHmBmvzWzjNDrNwRbE81tt9DMtpjZNjO74TDrfNHMssxso5n9LWx+vZllhl7LWvyNRETiwDNr8igqr+bqTtp6gBa2IIClBHsvfTE0/RXgPoJ3VkdkZonA3cDZQC6w0syWuXtW2DqTCN5fMd/dS8xsaNhHVLr7zBZ/ExGROBEIOH96M5tpo/pz0sQhsS7niLU0ICa6+8Vh0z83s8xmtpkLbHP3bAAzexRYBGSFrXMVcLe7lwC4e0EL6xERiVuvbtpLdlEF/3fprE4zMF8kLb1IXWlmpzRMmNl8oLKZbUYBu8Kmc0Pzwk0GJpvZO2b2vpktDFvWK3Q6630zu6iFdYqIxNxf3trOmMG9+ey04bEupU1a2oK4GnjAzAaEpkuAy9tp/5OAM4DRBHtIHefu+4Fx7p4XGtbjdTNb7+4fhW9sZkuAJQBjx45th3JERNqmrj7Aml0lfO2UCSR1omE1ImlR9e6+1t1nANOB6e4+Czizmc3ygDFh06ND88LlErqvwt23A1sJBgbunhf6bzbwL2BWhLrudffZ7j47LS2tJV9FRCSqdpVUUlvvTExrth9P3GtVvLl7aeiOaoDrmll9JTDJzMabWTKwGGjcG+lZgq0HzCyV4CmnbDMbZGY9w+bP59BrFyIicSm7sByACV0gIFp6iimSJq+8uHudmV0DvExwaPCloZFgbwYy3H1ZaNk5ZpYF1APXu/s+MzsZuMfMAgRD7Lbw3k8iIvEqu7ACgAmp/WJcSdu1JSCaHWrD3ZcDyxvNuzHsvRNsiVzXaJ13gePaUJuISExkF5UzqE8PBvVNjnUpbdZkQJhZGZGDwIDON7CIiEiUfVRYwYS0zt96gGYCwt1TOqoQEZGuILuwggVHd41OM527D5aISBwpraqlqLy6y7QgFBAiIu3k4wvUXaAHEyggRETaTUMX165wDwQoIERE2k12YQWJCcbYwQoIEREJk11UzphBvUlO6hq/WrvGtxARiQPZXaiLKyggRETaRSDgbC+qYEJq1zi9BAoIEZF2kbe/kuq6gFoQIiJyqOyirtXFFRQQIiLtoiuN4tpAASEi0g62F1WQ0jOJtH49Y11Ku1FAiIi0g2APpr6d+hnUjSkgRETaQXZheZe6QA0KCBGRNjtYU0f+gaou1cUVFBAiIm22/eMeTGpBiIhImK42imsDBYSISBtlF1ZgBuN1iklERMJlF5UzckBvevVIjHUp7UoBISLSRg1dXLsaBYSISBu4e7CLaxc7vQQKCBGRNikoq6aipr7L9WACBYSISJt81AXHYGqggBARaYNPuriqBSEiImGyCyvo1SOBEf17xbqUdhfVgDCzhWa2xcy2mdkNh1nni2aWZWYbzexvYfMvN7MPQ6/Lo1mniMiRyi4qZ3xqPxISus4gfQ2SovXBZpYI3A2cDeQCK81smbtnha0zCfgRMN/dS8xsaGj+YOBnwGzAgVWhbUuiVa+IyJHILqzguNEDYl1GVESzBTEX2Obu2e5eAzwKLGq0zlXA3Q2/+N29IDT/XOBVdy8OLXsVWBjFWkVEWq26rp7ckoNM7IJdXCG6ATEK2BU2nRuaF24yMNnM3jGz981sYSu2xcyWmFmGmWUUFha2Y+kiIs3L2XeQgHfNC9QQ+4vUScAk4AzgUuBPZjawpRu7+73uPtvdZ6elpUWpRBGRyDbtLgVgogKi1fKAMWHTo0PzwuUCy9y91t23A1sJBkZLthURiamXN+4htV8yx4xIiXUpURHNgFgJTDKz8WaWDCwGljVa51mCrQfMLJXgKads4GXgHDMbZGaDgHNC80RE4kJ5dR2vby7gvONGkJQY65Mx0RG1XkzuXmdm1xD8xZ4ILHX3jWZ2M5Dh7sv4JAiygHrgenffB2BmtxAMGYCb3b04WrWKiLTWa5v2UlUb4HMzRsa6lKiJWkAAuPtyYHmjeTeGvXfgutCr8bZLgaXRrE9E5Eg9vzaf4f17ccLYQbEuJWq6ZrtIRCSKDhys5Y2thVwwfUSXvEGugQJCRKSVXs7aQ229c0EXPr0ECggRkVZ7Yd1uxgzuzYwuegd1AwUEUFMXoKYuEOsyRKQT2FdezTvbivjc9JGYdd3TS6CAYOe+g5z43/9g+frdsS5FRDqBlzbsoT7gXDC9a59eAgUEowf1pm/PJJ5eo/vwRKR5L6zLZ2Ja3y57c1y4bh8QCQnG52eN4u0PC9lbWhXrckQkju0treKD7cVc0A1OL4ECAoDPzxpFwOG5TLUiROTwlq/fjTt8bsaIWJfSIRQQBEdinDlmIE+vVkCIyOE9vzafY0b056ihXf/0EiggPnbx8aPYvKeMrPzSdvm8v2/Yw/97dA3FFTXt8nkiElu5JQdZvXM/F0zvHq0HUEB87ILpI+mRaDyzJrdNn7P7QCVXPZDB1Q+t4rnMfH767HqCI4qISGf24rpgT8fPdYPeSw0UECGD+iaz4OihPJuZT1196++JqA84f313B2f/9k3e+rCQGz47hevOnszy9Xt4LjM/ChWLSEfYXlTBDU+t4zevbOX4sQMZO6RPrEvqMFEdrK+z+cLxo3glay9vbyvijKOHtni7rXvL+OFT61izcz+nTkrl1ouOY+yQPtQHnDe2FvJfz21g7vjBjBzYu8WfWVVbz/cey+QrJ43j5ImpR/J1RKQNNuQd4A//+ojlG3bTIzGBL84ZzTULJsW6rA6lgAizYMpQBvTuwTNr8locEO99tI+rHsggOSmBO740k0UzP+n+lphg/PaLM/js797i+ifX8uCVJ7Z4YK+l72znpQ172FNaxTPfUkCIdJTiihquezyTf20pJKVnElefPpEr548nLaVnrEvrcDrFFKZnUiIXTB/Byxv3UF5d1+z6r23ay+X3rWDEgF4s/86pXDRr1Kf6Ro8b0pefnj+Vd7bt46/v7WhRHQVlVdz9+jYG9unBmp37Wb2z5Ai+jYi0VnVdPd94MIN3P9rHDxYezTs/OpMfLpzSLcMBFBCf8oXjR1NVG+ClZobeeC4zj288uIopw1N47BsnMXxAr8Oue+ncMSw4Oo3bXtrMtoLyZmv4zctbqakP8NDXTiSlZxL3vbOjtV9DRFrJ3fnhk+tYuaOE31wyg2+dcRT9e/WIdVkxpYBo5PixA0kf0qfJeyIeej+H7z6WyQnjBvHw109kcN/kJj/TzPifi6fTJzmR6x7PpLaJi+Ab8g7w+KpdXHFyOtNGDeBLc8awfP1udh+oPOLvJCLNu/O1bTybmc/3z5ncpZ8S1xoKiEbMjM/PGs172fvI23/oL+XKmnru/uc2fvrsBs48eih/vXIuKS38C2No/17c+vnjWJd7gJuWbaQ+8Omur+7OLS9kMahPMtecGbwYdvnJ6bg7D7yX0/YvJyIRPZeZx+3/2MrFx4/m2wuOinU5cUMXqSP4/KxR3P6PrfzpzWymDE9hbe5+1u46wJa9ZdQHnAtnjOQ3X5xBj1Y+qPy840bwjdMmcM+b2ew5UMXvLp1Fv56f/C94eeNePthezC8umsaA3sHgGTO4D+dMHc7fPtjJd86cRO/kxHb9riLdXcaOYq5/Yh0njh/ML79wXLcYY6ml1IKIYOyQPsxNH8z97+7ghqfX8+K63Qzpl8w3T5/I0itmc8eXZrY6HBr86LxjuGXRsfxrayGX/PE98kOtlOq6ev57+SYmD+vH4jljDtnmylPGc6CylqfbeBMfwJY9ZXzzoVUf71ekO8vZV8GSB1cxalBv/njZCSQn6VdiOLUgDuPXl0xnbe4Bjhs1gPQhfdr1r4qvnJTO2CF9uebh1Vx09zv8+fLZvPfRPnYWH+TBr80lqVH4zEkfxLRR/Vn69nYunTP2iJ+Be7Cmjm8+vIrswgrKq+t44Mq5+mtJuq3CsmouX7qCgDtLr5jDoGauJXZHisvDGDekLxfOGMn41L5R+SV6+uQ0nvrWycEbcO55jztf+5Czpgzl1Elpn1rXzLhy/ng+KqzgrW1FR7zPnz23ke1FFfzbCaN568MiHlmxqy1fQaTTOlBZy1eXrmBvaTV/uXwO41P7xrqkuKSAiKHJw1J49tvzOWZEf2rrnR+ff8xh1z1/+gjSUnqy9O3tR7SvZ9fk8cSqXK5dcBS/ung6848awq0vZrGr+OCRli/SKVXW1PP1v65kW0EZf/zKCZwwblCsS4pbCogYS0vpyePfOIk3f7CAiWn9Drtez6REvjJvHG9sLWRbQVmr9rG9qIKfPLOeuemD+c5Zk0hICHa7BfjhU+sIROhRJdIV1dYH+PbfVpORU8JvvziT0yd/usUun9A1iDjQIzGhyRvtGnz5xLHc9c9t/OaVrXzj9ImkD+nDwD5Nnzetrqvn2kdW0yMpgTsWz/z4+sboQX34yflT+fEz63l4xU6+Mm9cm75D3v5KXlq/mwOVtew/WMuByuCrsraeL84ew8XHf/ouc5GOFAg4339iLa9vLuDWz0/TvQ4toIDoRIb068llJ477eJwmgP69kkhP7cu4IX2ZPmoAx48LXtDumRTsDvs/L21hQ14pf/rq7E8NFnjp3DG8tGE3v1y+idMnpR3xKJXbCsq49E8fUFhWTYJB/949GNi7BwN69+BgTT3ff2Itz2Xm8d+fP44xgyPvo6yqlnW5B5g3YQiJR3gRXuRw3J2fP7+R5zLzuf7co/nyiW37g6i7sGg+q8DMFgK/AxKBP7v7bY2WXwH8Gmi4bfkud/9zaFk9sD40f6e7X9jUvmbPnu0ZGRntWH18cne2FZSzY99BcvZVkLPvIDv2VZBdWPHxjX3JiQlMG9WfiWn9eGJVLlecnM5NFx4b8fPy91dy7u1vMnVkfx65al6re0ht2VPGl//8PmA8cOVcpgxPOeQzAgHnoQ9y+J+XNuPA9ecezVdPSicxwQgEnPe37+PJjFxe2rCHytp6rj59Ijd8dsqRHh6RT6kPOD99dj2PrNjF108Zz0/OP0at2TBmtsrdZ0dcFq2AMLNEYCtwNpALrAQudfessHWuAGa7+zURti9398OflG+kuwREUwrKqlids581O0tYlVPCurwDTBmewhNXn/RxiyKSx1fu4gdPreP/nTWJ735mUov/8WzMP8Blf/6A5KQE/nbVvCavoeTtr+Qnz6znX1sKmTV2IKcelcozmXnsKq4kpWcSn5s5kvKqOpatzeeuf5/FBe30UBZ3p6o2oBsMu6mq2nq++2gmf9+4h2sWHMV/njNZ4dBIUwERzVNMc4Ft7p4dKuJRYBGQ1eRWcsSGpvRi4bThLJw2HICaugAJxqfuq2jsktmjeXtbEb977UO27CnjV5dMb3aQsvW5B7jsLx/QNzmRv101j/RmugmOGtib+66Yw7OZedz8fBaZu/Yzf2Iq3z/naM49dji9eiRSUxcgb38l1z+xjgmp/Zg6sn/rDkAj7s4PnlzHc2vzuezEcXzzjInddlTO7qisqpYlD6zivex93HjBVK48ZXysS+p0otmC+Ddgobt/PTT9FeDE8NZCqAXxS6CQYGvje+6+K7SsDsgE6oDb3P3ZCPtYAiwBGDt27Ak5ORqv6Ei5O39+azu3/X0zowb25vdfPp5powZEXHfNzhK+unQF/Xv14NEl8w57XeFwyqqCF6+Hpnz6wnxBaRWfu+ttkpMSWPbtU9p089Kdr33Ib1/dyuxxg1i9s4SeSYl89eRxXH3aRN0U1cUVlVdzxX0r2Ly7jP+9ZAYXzRoV65LiVqxOMbUkIIYA5e5ebWbfAL7k7meGlo1y9zwzmwC8Dpzl7h8dbn86xdQ+VuUUc83f1rCvvIYbPzeVL584lvqAk7lrP29uLeSNrYWsyzvAmEF9eGTJPEa14il5LbV6ZwmL73mfEycM5r4r5jTbAork2TV5fPexTL5w/Ch+c8kMthdV8LvXPmTZ2nz69EjkylPG843TJx4yFpZ0DTn7Krh86Qr2lFbxh8tOYEErng7ZHcUqIE4CbnL3c0PTPwJw918eZv1EoNjdP/Vnq5ndD7zg7k8ebn8KiPZTXFHD9x7L5I2thUwfPYDtRRWUVdWRYDBzzEBOm5zGv584NmILoL08tnInP3xqPUtOm8CPz/vkBsLqunrySiqpqg1wzIiUiOeTV2wv5rI/f8Dx4wbywJUnHjK+zta9Zdzxj60sX7+HUQN78z8XT+eUSXpiX1fx/Np8fvz0ehISjKVXzOaEcYNjXVLci1VAJBE8bXQWwV5KK4F/d/eNYeuMcPfdofefB37o7vPMbBBwMNSySAXeAxaFX+BuTAHRvgIB549vfsRza/KZNTYYCvMnpjKgT8c9QOW/nt3Ag+/ncPbUYRRX1JBbcpC9pdUfL586oj9LTpvA+dNHfDx44vaiCj7/+3cY3DeZp7958mHvE1mVU8L1T64lu7CCS+eO4cfnHdPiodsl/lTW1PPz5zfy6MpdzBo7kDsXz2r1qc/uKiYBEdrxecAdBLu5LnX3W83sZiDD3ZeZ2S+BCwleZygGvunum83sZOAeIEDwbu873P0vTe1LAdH11NQF+O5ja1iXGzylNXpQb0aH/ltZW8/97+5gW0E5Iwb04sr541k4bThf+csHlFbV8cy3TmbckKYvnFfV1n88rPvw/r345cXTdWdtJ7R5TynX/G0NHxWWc/XpE7nu7MlHPNpydxSzgOhICojuJxBw3thayL1vZvNe9j6AYJfbr5/I7PSWn1pYs7OE659cx7aCcj5zzFBOGDeYY0f259iR/RnSr+N6PVXV1vPL5ZsYN6Qvl8werRZNCzy2cic3PreRlF49uP1LMyIOdilNU0BIl7c+9wAPvZ/DZ6YO4+ypw1q9fVVtPXe9vo1nM/PILfnkWRkjBvRi2qgBnDVlKAunDT/sKaIuZwMAAA0hSURBVKtdxQdZtjafLXvKWHLahMP2ADscd+d7j2XybGY+AP16JnHJ7NFccXJ6sy2h7mpVTgn/9sd3mT8xldu/NFNdmI+QAkKkFfYfrCErv5SN+aVsyD/A6p0l7CquJCnBOGVSKucfN4Jzjh1OfcB5cV0+z2Xmk5FTAkDf5ESq6gJ8e8FRXLPgqBY/gKahS+715x7NqZNSue+dHbywLp+6gHPWlKFcffrEVrWKurrqunrOv/NtDlbX8cp1p6s3WhsoIETawN3ZkFfKC+vzeWHtbvL2V5KcmEDAnbqAM3lYPxbNHMWFM0aS0iuJnz+fxTNr8jhmRH/+95LpHDuy6dbE82vzufaRNR93yW3omVVQWsVD7+fw8Ac7KT5Yw7fPOIrvfmbSEXX77Wp+++pW7nztQ+77jznqxtpGCgiRduIevCfkpQ17SDDjwhkjI3a3fWXjHn78zAb2H6zhmjOP4ltnRG5NrNlZwuJ732f66AE89PUTIw6JUllTz8+WbeDxjFzmTRjMnYtnMbR/9LoYx7ste8q44P/e4vzjRnDH4lmxLqfTU0CIxEBJRQ03hUYQHdI3mYXThnP+9BGcOD44Ym3e/koW3fUOfZITefbb8xnczN3dT2Ts4r+e20C/nj2489KZnDwxOvdvuHvcjldUH3Au/sO77Cw+yD+uO73ZYybNU0CIxNCbWwt5PGMXr20qoLK2nrSUnpw3bTgfbC8mb38lz3zrZI4amtKiz9qyp4xvPryKHUUVfOesSZw2OY0EMxLNMIPEBGP0oN5H1ANqb2kVt764iTe2FnLd2ZO5bN64uBt6/S9vb+eWF7L43eKZLJqp4TPagwJCJA4crKnj9c0FvLhuN69vLqAu4Nx3xRxOa+W9FxXVdfz4mfU8F+rx1NiA3j344cIpLJ4zpkXDt9fWB/jruzu4/dWt1AacY4ansDb3ALPGDuSXXziOKcPbNmhie9lVfJBzbn+TeRMGs/SKOXHbyulsFBAicaa8uo7i8pojfkiTu7Nm134OVNYSCDgBh4A7tfUBHnwvhw+2F3P82IH84qLjmhwV94Psfdz43Ea27C1jwdFp3HThsYwd3IfnMvO5+YUsSitrufr0iVxz5lH06hG7IdPdna8uXcHqnBJeue70qIwB1l0pIES6EXfn6dV5/PfyTeyvrOWKk9P53tmTSUowPios58O95WzdW8aG/FLe3FrIqIG9+dnnpnL21GGH/FVeXFHDL17M4unVeYxP7csF00cwNKUnaSm9GNa/J0P792JoSs+o37VcH3Bue2kTf3prOzcvOpavnpQe1f11NwoIkW5o/8EafvXyFh5ZsZPePRKpqq0nEPrnnpRgpKf25bPThvOtM45q8oFKb39YxM0vbGRbQfnH2zcYmtKTOy+dxbwJQ6L2Ha59ZA1vfVjEZfPGcvOF01r91ENpmgJCpBtbvbOEx1bsYtiAXkwe1o/Jw1JIH9K3xTfxNairD1BcUcPe0moKyqrYW1rNn9/OJmffQW5YOIWvnzq+Xa8LbN5TypIHVrH7QCW3LJrG4rlj2+2z5RMKCBGJirKqWq5/Yh1/37iH844bzq/+bUa73NW8fP1uvv/EWvr1TOIPl53ACeMGtUO1EklTAaFbMkXkiKX06sEfLjueH583hb9v2MOiu95mW0HZEX9eTV2A217azLceXs2U4Sk8f+0pCocYUkCISJuYGUtOm8jDX5/HgcpaFt31Dr/6+2a2F1W06nO2FZRz8R/e5Y9vfMSlc8fyyJJ5DOvGd4zHA51iEpF2s+dAFTc+t4F/bNpLwGHu+MF8cfYYzjtuOH2SI596cnce/mAnv3gxi149ErntC8excNqIDq68+9I1CBHpUHtLq3hqdS5PZOSyvaiCfj2TOH1yGpOHpXDU0H5MGtaP9CF9OVBZyw+fWsfrmws4dVIq/3vJDLUaOpgCQkRiwt1ZuaOExzN28cH2feSWVNLwKycxwUhOTKDenR99dgqXn5SuLqwx0FRAaBB1EYkaM2Pu+MHMHR98lkVlTT3ZReVsKwi+isqr+Y/545k8rGVjUUnHUkCISIfpnZzIsSMHNPuMDIkP6sUkIiIRKSBERCQiBYSIiESkgBARkYgUECIiEpECQkREIlJAiIhIRAoIERGJqMsMtWFmhUAOMAA4cJjVIi1rybzw6fD3qUDREZZ8OE3VfyTr63i0fLmOR/PLmpvX1LGJ9fFoybptPR6tmW7v43GkPxvj3D0t4hru3qVewL2tWdaSeeHTjd5ndGT9R7K+joeOR0cej2aOTUyPR0vWbevxaM10ex+P9v7ZcPcueYrp+VYua8m855tY1t5a+/nNra/j0fLlOh7NL2tuXlPHJhpa8/ktWbetx6O10+2pvX82us4pplgwsww/zCiI3ZGOx6F0PA6l43GoznA8umILoiPdG+sC4oyOx6F0PA6l43GouD8eakGIiEhEakGIiEhECggREYlIARFiZkvNrMDMNhzBtieY2Xoz22Zmd5qZhS271sw2m9lGM/tV+1YdPdE4HmZ2k5nlmVlm6HVe+1ceHdH6+Qgt/08zczNLbb+KoytKPx+3mNm60M/GK2Y2sv0rj44oHY9fh353rDOzZ8xsYPtX3jQFxCfuBxYe4bZ/AK4CJoVeCwHMbAGwCJjh7scC/9v2MjvM/bTz8Qi53d1nhl7L21Zih7qfKBwPMxsDnAPsbGN9He1+2v94/Nrdp7v7TOAF4Ma2FtmB7qf9j8erwDR3nw5sBX7UxhpbTQER4u5vAsXh88xsopn93cxWmdlbZjal8XZmNgLo7+7ve/CK/wPARaHF3wRuc/fq0D4Kovst2k+UjkenFcXjcTvwA6BT9RaJxvFw99KwVfvSiY5JlI7HK+5eF1r1fWB0dL/FpykgmnYvcK27nwB8H/h9hHVGAblh07mheQCTgVPN7AMze8PM5kS12uhr6/EAuCbUZF5qZoOiV2qHaNPxMLNFQJ67r412oR2kzT8fZnarme0CvkznakFE0h7/XhpcCbzU7hU2I6mjd9hZmFk/4GTgibBTxj1b+TFJwGBgHjAHeNzMJngn7FvcTsfjD8AtBP8yvAX4DcEf/E6nrcfDzPoAPyZ4eqnTa6efD9z9J8BPzOxHwDXAz9qtyA7UXscj9Fk/AeqAh9unupZTQBxeArA/dD70Y2aWCKwKTS4j+EsvvOk3GsgLvc8Fng4FwgozCxAcoKswmoVHSZuPh7vvDdvuTwTPM3dWbT0eE4HxwNrQL5DRwGozm+vue6JcezS0x7+XcA8Dy+mkAUE7HQ8zuwK4ADgrJn9YtudgUZ39BaQDG8Km3wUuCb03ghebI223gmArwQg2A88Lzb8auDn0fjKwi9DNiZ3hFYXjMSJsne8Bj8b6O8byeDRaZweQGuvvGOOfj0lh61wLPBnr7xjj47EQyALSYvadYn1Q4+UFPALsBmoJ/uX/NYJ/4f0dWBv6H3XjYbadDWwAPgLuaggBIBl4KLRsNXBmrL9njI/Hg8B6YB3Bv55GdNT3icfj0WidThUQUfr5eCo0fx3BgeRGxfp7xvh4bCP4R2Vm6PXHjv5eGmpDREQiUi8mERGJSAEhIiIRKSBERCQiBYSIiESkgBARkYgUENKlmVl5B+/v3Xb6nDPM7EBoZNPNZtbsQI9mdpGZTW2P/YuAAkKkVcysydEH3P3kdtzdWx68E3cWcIGZzW9m/YsABYS0GwWEdDuHG2XTzD4XGlhxjZn9w8yGhebfZGYPmtk7wIOh6aVm9i8zyzaz74R9dnnov2eElj8ZagE8HDbO/3mheatC4/83OeSIu1cSvFGqYZC/q8xspZmtNbOnzKyPmZ0MXAj8OtTqmNiS0URFmqKAkO7ocKNsvg3Mc/dZwKMEh+FuMBX4jLtfGpqeApwLzAV+ZmY9IuxnFvDd0LYTgPlm1gu4B/hsaP9pzRUbGvV2EvBmaNbT7j7H3WcAm4Cvufu7BO9Ov96Dz9r4qInvKdIiGqxPupVmRtkcDTwWGqM/Gdgetumy0F/yDV704HM+qs2sABjGocM2A6xw99zQfjMJjtVTDmS7e8NnPwIsOUy5p5rZWoLhcId/MojfNDP7BTAQ6Ae83MrvKdIiCgjpbiKOshnyf8Bv3X2ZmZ0B3BS2rKLRutVh7+uJ/G+pJes05S13v8DMxgPvm9nj7p5J8OllF7n72tBon2dE2Lap7ynSIjrFJN2KB59att3MLgGwoBmhxQP4ZKjly6NUwhZggpmlh6a/1NwGodbGbcAPQ7NSgN2h01pfDlu1LLSsue8p0iIKCOnq+phZbtjrOoK/VL8WOn2zkeBzwyHYYnjCzFYBRdEoJnSa6lvA30P7KQMOtGDTPwKnhYLlv4APgHeAzWHrPApcH7rIPpHDf0+RFtForiIdzMz6uXt5qFfT3cCH7n57rOsSaUwtCJGOd1XoovVGgqe17olxPSIRqQUhIiIRqQUhIiIRKSBERCQiBYSIiESkgBARkYgUECIiEtH/B4X8Ja5sLsQ0AAAAAElFTkSuQmCC\n", "text/plain": [ "
" ] }, "metadata": { "needs_background": "light" }, "output_type": "display_data" } ], "source": [ "learner.load(\"best-effb5-herlev-multiclass-fold4-stage1\")\n", "learner = to_fp16(learner)\n", "learner.unfreeze()\n", "learner.lr_find()\n", "learner.recorder.plot()" ] }, { "cell_type": "code", "execution_count": 30, "metadata": {}, "outputs": [ { "data": { "text/html": [ "\n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", "
epochtrain_lossvalid_lossaccuracyprecisionrecallf_betakappa_scoretime
00.5308840.1006640.9726780.9782150.9760470.9763500.98250900:04
10.5307460.0998450.9781420.9829260.9800150.9805090.98336900:04
20.5162670.1004070.9781420.9829260.9800150.9805090.98336900:04
30.5297570.1005150.9726780.9782150.9760470.9763500.98250900:05
40.5077250.1015160.9726780.9782150.9760470.9763500.98250900:05
50.4995580.1053160.9672130.9737940.9720790.9721670.98164700:04
60.4983380.1084590.9672130.9737940.9720790.9721670.98164700:04
70.4962780.1127110.9726780.9782150.9760470.9763500.98250900:04
80.4926950.1149860.9672130.9737600.9712850.9716020.97907600:05
90.4848690.1242780.9672130.9738220.9724760.9725290.98167500:05
100.4759710.1314900.9562840.9603580.9580260.9581520.97990900:05
110.4631460.1331810.9562840.9566260.9581900.9575390.96724900:04
120.4583500.1280870.9617490.9610190.9617610.9613600.96809100:04
130.4572440.1330940.9617490.9610190.9617610.9613600.96809100:04
140.4479990.1387390.9617490.9610190.9617610.9613600.96809100:04
150.4341610.1441770.9562840.9570680.9575100.9571290.96111900:04
160.4317120.1474760.9617490.9648940.9622720.9623670.98093900:04
170.4238110.1531930.9508200.9433180.9479860.9468240.96413600:04
180.4148930.1479790.9508200.9467540.9470280.9469430.96668700:04
190.4192140.1524530.9562840.9515960.9506000.9507560.96752200:04
200.4005540.1493610.9508200.9480100.9456730.9460040.96669100:04
210.3930650.1513780.9508200.9480100.9456730.9460040.96669100:04
220.3755660.1444160.9508200.9481820.9482660.9479190.97340100:04
230.3725130.1453520.9562840.9486850.9545470.9530680.97183200:05
240.3685030.1449750.9562840.9504890.9541500.9532630.97173800:04
250.3559160.1408570.9617490.9552390.9581180.9574150.97261300:04
260.3537190.1375620.9617490.9585130.9579540.9578660.98513100:04
270.3492540.1377870.9617490.9594360.9579540.9581060.98755000:04
280.3478710.1370380.9562840.9549120.9530280.9530570.98430600:04
290.3375070.1371390.9508200.9474810.9482660.9477720.97100200:04
300.3390450.1369890.9562840.9504770.9531920.9523780.96948800:04
310.3356890.1377970.9562840.9504770.9531920.9523780.96948800:04
320.3377630.1379060.9562840.9504770.9531920.9523780.96948800:05
330.3322440.1371660.9562840.9504770.9531920.9523780.96948800:04
340.3348660.1365450.9562840.9504770.9531920.9523780.96948800:04
350.3294670.1377270.9562840.9504770.9531920.9523780.96948800:04
360.3212190.1382220.9562840.9504770.9531920.9523780.96948800:04
370.3152500.1371210.9562840.9510820.9531920.9525810.97182800:04
380.3144610.1381910.9562840.9504770.9531920.9523780.96948800:04
390.3167120.1379750.9562840.9504770.9531920.9523780.96948800:04
" ], "text/plain": [ "" ] }, "metadata": {}, "output_type": "display_data" }, { "name": "stdout", "output_type": "stream", "text": [ "Better model found at epoch 0 with accuracy value: 0.9726775884628296.\n", "Better model found at epoch 1 with accuracy value: 0.9781420826911926.\n" ] } ], "source": [ "learner.fit_one_cycle(40, max_lr=slice(5e-04), callbacks=model_callback(learner, \"best-effb5-herlev-multiclass-fold4-stage2\"))\n", "learner.save(\"last-effb5-herlev-multiclass-fold4-stage2\")" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "# Fold-5" ] }, { "cell_type": "code", "execution_count": 31, "metadata": {}, "outputs": [ { "data": { "text/plain": [ "ImageDataBunch;\n", "\n", "Train: LabelList (734 items)\n", "x: ImageList\n", "Image (3, 64, 64),Image (3, 64, 64),Image (3, 64, 64),Image (3, 64, 64),Image (3, 64, 64)\n", "y: CategoryList\n", "abnormal_moderate-dysplastic,abnormal_moderate-dysplastic,abnormal_moderate-dysplastic,abnormal_moderate-dysplastic,abnormal_moderate-dysplastic\n", "Path: ../../../Dataset/Herlev Dataset;\n", "\n", "Valid: LabelList (183 items)\n", "x: ImageList\n", "Image (3, 64, 64),Image (3, 64, 64),Image (3, 64, 64),Image (3, 64, 64),Image (3, 64, 64)\n", "y: CategoryList\n", "abnormal_moderate-dysplastic,abnormal_moderate-dysplastic,abnormal_moderate-dysplastic,abnormal_moderate-dysplastic,abnormal_moderate-dysplastic\n", "Path: ../../../Dataset/Herlev Dataset;\n", "\n", "Test: None" ] }, "execution_count": 31, "metadata": {}, "output_type": "execute_result" } ], "source": [ "fold_idxs = idxs[4]\n", "fold_data = (ImageList.from_folder(data_path)\n", " .split_by_idxs(fold_idxs[0], fold_idxs[1])\n", " .label_from_folder()\n", " .transform(tfms, size=64)\n", " .databunch(bs=64)\n", " .normalize(imagenet_stats))\n", "fold_data" ] }, { "cell_type": "code", "execution_count": 32, "metadata": {}, "outputs": [ { "data": { "text/html": [ "\n", "
\n", " \n", " \n", " 70.00% [7/10 00:19<00:08]\n", "
\n", " \n", "\n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", "
epochtrain_lossvalid_lossaccuracyprecisionrecallf_betakappa_scoretime
00.572817#na#00:02
10.545696#na#00:02
20.536563#na#00:02
30.539627#na#00:02
40.521721#na#00:02
50.553229#na#00:02
60.998182#na#00:02

\n", "\n", "

\n", " \n", " \n", " 36.36% [4/11 00:01<00:02 1.8143]\n", "
\n", " " ], "text/plain": [ "" ] }, "metadata": {}, "output_type": "display_data" }, { "name": "stdout", "output_type": "stream", "text": [ "LR Finder is complete, type {learner_name}.recorder.plot() to see the graph.\n" ] }, { "data": { "image/png": "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\n", "text/plain": [ "
" ] }, "metadata": { "needs_background": "light" }, "output_type": "display_data" } ], "source": [ "learner.load(\"best-effb5-herlev-multiclass-fold4-stage2\")\n", "learner = to_fp16(learner)\n", "learner.data = fold_data\n", "learner.freeze()\n", "learner = to_fp16(learner)\n", "learner.lr_find()\n", "learner.recorder.plot()" ] }, { "cell_type": "code", "execution_count": 33, "metadata": {}, "outputs": [ { "data": { "text/html": [ "\n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", "
epochtrain_lossvalid_lossaccuracyprecisionrecallf_betakappa_scoretime
00.4968630.1178810.9836070.9879310.9868130.9869980.97744800:03
10.4963740.1185570.9890710.9930310.9904760.9908280.98489300:03
20.5002260.1183900.9836070.9891700.9855500.9860700.98407900:03
30.5043590.1251580.9781420.9841370.9815820.9819330.98321400:03
40.5032610.1403550.9562840.9661340.9650560.9650980.97989000:03
50.5094620.1517040.9453550.9570000.9574250.9570760.97160300:03
60.5188520.1571980.9453550.9581170.9583830.9575030.97151100:03
70.5222920.1540190.9562840.9626790.9578620.9584110.97345100:03
80.5130400.1760040.9289620.9372690.9411580.9392870.96266300:04
90.5137000.1567650.9453550.9551020.9567720.9563280.96509700:04
100.5192980.2224620.9180330.9251060.9179180.9171420.96776000:03
110.5120310.1983080.9398910.9565690.9506060.9507900.96096100:03
120.5064430.2381450.9234970.9383700.9373960.9363730.94640000:03
130.5011380.2265500.9180330.9263480.9231110.9220810.95782800:03
140.5071010.2325620.9289620.9402530.9292170.9292290.96709500:03
150.5034310.2166550.9398910.9522170.9524990.9519350.97063300:04
160.5045690.2187160.9289620.9458280.9413880.9415140.95141400:03
170.4946850.1954410.9398910.9528730.9483660.9491560.95843400:03
180.4827830.1875470.9344260.9483170.9453560.9457200.95739700:03
190.4695290.1954930.9234970.9437570.9374190.9375750.95292600:03
200.4640900.1986570.9180330.9396380.9324930.9328540.94563100:03
210.4552880.1840040.9234970.9431020.9345460.9356540.95497500:04
220.4511770.1881410.9617490.9717690.9665020.9671050.97385200:04
230.4417570.1911070.9508200.9628090.9588710.9590580.97216100:03
240.4428650.1943550.9508200.9628090.9588710.9590580.97216100:03
250.4344990.1965480.9398910.9514180.9502820.9501230.97052700:03
260.4220210.1981010.9453550.9554830.9552080.9547340.97136800:03
270.4130120.2004640.9508200.9595110.9591760.9589140.97225500:04
280.4075120.1998540.9508200.9595110.9591760.9589140.97225500:03
290.3986530.2010260.9508200.9595110.9591760.9589140.97225500:04
" ], "text/plain": [ "" ] }, "metadata": {}, "output_type": "display_data" }, { "name": "stdout", "output_type": "stream", "text": [ "Better model found at epoch 0 with accuracy value: 0.9836065769195557.\n", "Better model found at epoch 1 with accuracy value: 0.9890710115432739.\n" ] } ], "source": [ "learner.fit_one_cycle(30, max_lr=slice(2e-03), callbacks=model_callback(learner, \"best-effb5-herlev-multiclass-fold5-stage1\"))\n", "learner.save(\"last-effb5-herlev-multiclass-fold5-stage1\")" ] }, { "cell_type": "code", "execution_count": 34, "metadata": {}, "outputs": [ { "data": { "text/html": [ "\n", "
\n", " \n", " \n", " 60.00% [6/10 00:22<00:14]\n", "
\n", " \n", "\n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", "
epochtrain_lossvalid_lossaccuracyprecisionrecallf_betakappa_scoretime
00.514498#na#00:03
10.545023#na#00:03
20.528558#na#00:03
30.528379#na#00:04
40.531895#na#00:04
50.806348#na#00:03

\n", "\n", "

\n", " \n", " \n", " 45.45% [5/11 00:02<00:02 1.4245]\n", "
\n", " " ], "text/plain": [ "" ] }, "metadata": {}, "output_type": "display_data" }, { "name": "stdout", "output_type": "stream", "text": [ "LR Finder is complete, type {learner_name}.recorder.plot() to see the graph.\n" ] }, { "data": { "image/png": "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\n", "text/plain": [ "
" ] }, "metadata": { "needs_background": "light" }, "output_type": "display_data" } ], "source": [ "learner.load(\"best-effb5-herlev-multiclass-fold5-stage1\")\n", "learner = to_fp16(learner)\n", "learner.unfreeze()\n", "learner.lr_find()\n", "learner.recorder.plot()" ] }, { "cell_type": "code", "execution_count": 35, "metadata": {}, "outputs": [ { "data": { "text/html": [ "\n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", "
epochtrain_lossvalid_lossaccuracyprecisionrecallf_betakappa_scoretime
00.4800320.1198660.9781420.9831690.9828450.9828270.97657500:04
10.4998260.1198700.9726780.9790370.9779190.9781040.97577800:04
20.5288380.1220670.9836070.9891700.9855500.9860700.98407900:04
30.5177210.1235560.9781420.9841370.9815820.9819330.98321400:04
40.5123470.1260010.9726780.9789610.9782240.9783300.97581500:04
50.5016150.1299630.9726780.9773340.9768200.9765790.96969100:04
60.4908280.1354520.9726780.9773340.9768200.9765790.96969100:04
70.4716290.1456230.9726780.9752080.9713780.9717720.98246400:04
80.4563990.1455270.9562840.9657450.9660140.9655550.97985700:04
90.4503490.1491080.9617490.9670130.9637460.9639870.97825400:04
100.4305860.1382010.9617490.9699040.9699820.9697620.97402400:04
110.4236480.1359720.9672130.9718600.9674090.9678430.97909100:04
120.4190640.1280130.9672130.9752430.9726870.9730390.98153500:04
130.4010830.1388230.9617490.9682660.9690240.9687020.98073000:04
140.3995820.1445510.9562840.9673250.9642630.9645620.97301000:04
150.4022340.1415600.9562840.9662610.9650560.9650700.97982300:04
160.3941920.1476070.9617490.9706350.9690240.9692050.98069500:04
170.3817950.1433680.9562840.9680270.9642630.9646200.97296700:04
180.3758060.1471830.9562840.9669120.9650560.9651070.97979100:04
190.3648300.1480360.9672130.9749510.9739510.9739190.98150400:04
200.3538310.1483980.9617490.9707350.9702880.9700470.98066200:04
210.3445220.1484950.9562840.9668270.9663190.9658900.97978900:05
220.3373450.1458800.9562840.9668270.9663190.9658900.97978900:04
230.3309010.1439370.9617490.9707350.9702880.9700470.98066200:04
240.3326890.1439600.9617490.9707350.9702880.9700470.98066200:04
250.3213370.1449620.9617490.9707350.9702880.9700470.98066200:04
260.3165410.1458030.9562840.9668270.9663190.9658900.97978900:04
270.3090860.1456720.9562840.9663040.9653610.9653540.97985400:05
280.3187530.1458170.9562840.9668270.9663190.9658900.97978900:05
290.3088650.1457070.9508200.9621890.9613930.9612310.97898100:05
" ], "text/plain": [ "" ] }, "metadata": {}, "output_type": "display_data" }, { "name": "stdout", "output_type": "stream", "text": [ "Better model found at epoch 0 with accuracy value: 0.9781420826911926.\n", "Better model found at epoch 2 with accuracy value: 0.9836065769195557.\n" ] } ], "source": [ "learner.fit_one_cycle(30, max_lr=slice(6.5e-04), callbacks=model_callback(learner, \"best-effb5-herlev-multiclass-fold5-stage2\"))\n", "learner.save(\"last-effb5-herlev-multiclass-fold5-stage2\")" ] }, { "cell_type": "code", "execution_count": 36, "metadata": {}, "outputs": [], "source": [ "learner.load(\"best-effb5-herlev-multiclass-fold5-stage2\")\n", "learner.freeze()\n", "learner.export(\"best-effb5-herlev-multiclass.pkl\")" ] }, { "cell_type": "code", "execution_count": null, "metadata": {}, "outputs": [], "source": [] }, { "cell_type": "code", "execution_count": null, "metadata": {}, "outputs": [], "source": [] }, { "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": "3.6.9" } }, "nbformat": 4, "nbformat_minor": 4 }