Sample Integrated#

This notebook will do a simple pipeline based on fastai with transfer learning concepts using resnet architecture. Typical ideas I want to implement here

  1. Get a working pipeline with the current dataset ( using Fastai)

  2. Do a submission on Kaggle

  3. Do some iterations using simple data augmentations

We will split the train dataset into 80-20 split. The idea is to have an actual measurement of model performance looking at the training data. However, trick for kaggle here is to train the model on entire training dataset before submitting results and reviewing score on the test.

Note

  • Big Images are not easy to fit on GPU.

  • Resizing big image to smaller sizes tends to loose certain features

  • Next ideas

    • Apply Digit Cleaner concept

    • Create Clean Dataset / Visualize and review accuracy of digit cleaner

    • Figure out way to do a bounding box

    • After that we can apply a few ideas

      • Apply resnet after digit cleaner ( current data had 0.93 error_rate after 1 epoch)

      • Do bounding box and break down dataset into individual digits. Then do clustering to 10 categories. Then use labels to calculate sum_digit

      • Split into individual digit images -> Resize -> Merge (permutations) into single image then train model on new dataset

Imports#

%load_ext autoreload
%autoreload 2

The autoreload extension is already loaded. To reload it, use:
  %reload_ext autoreload

import torch
torch.cuda.empty_cache()
!nvidia-smi

Tue Mar 15 09:43:54 2022       
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 470.103.01   Driver Version: 470.103.01   CUDA Version: 11.4     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|                               |                      |               MIG M. |
|===============================+======================+======================|
|   0  Tesla K80           Off  | 00000000:00:1E.0 Off |                    0 |
| N/A   67C    P0    61W / 149W |   1316MiB / 11441MiB |      0%      Default |
|                               |                      |                  N/A |
+-------------------------------+----------------------+----------------------+
                                                                               
+-----------------------------------------------------------------------------+
| Processes:                                                                  |
|  GPU   GI   CI        PID   Type   Process name                  GPU Memory |
|        ID   ID                                                   Usage      |
|=============================================================================|
|    0   N/A  N/A    173195      C   ...da/envs/aiking/bin/python     1313MiB |
+-----------------------------------------------------------------------------+

import fastai
from fastai.vision.all import *

from fastai.vision.all import *
from fastai.vision.widgets import *
from aiking.data.external import * #We need to import this after fastai modules
import warnings
from matplotlib import cm
import shutil
import pathlib

warnings.filterwarnings("ignore")

path = untar_data("kaggle_competitions::ultra-mnist"); path

pd.Series((path).ls())

0                /Landmark2/pdo/aiking/data/ultra-mnist/sample_black_cropped_resized
1                                  /Landmark2/pdo/aiking/data/ultra-mnist/test_black
2     /Landmark2/pdo/aiking/data/ultra-mnist/sample_black_cropped_resized_integrated
3                            /Landmark2/pdo/aiking/data/ultra-mnist/train_sample.csv
4                                  /Landmark2/pdo/aiking/data/ultra-mnist/sample.csv
5                                   /Landmark2/pdo/aiking/data/ultra-mnist/train.csv
6                                      /Landmark2/pdo/aiking/data/ultra-mnist/models
7                                       /Landmark2/pdo/aiking/data/ultra-mnist/train
8                             /Landmark2/pdo/aiking/data/ultra-mnist/train_train.csv
9                                /Landmark2/pdo/aiking/data/ultra-mnist/sample_black
10                                       /Landmark2/pdo/aiking/data/ultra-mnist/test
11                               /Landmark2/pdo/aiking/data/ultra-mnist/train_black2
12                       /Landmark2/pdo/aiking/data/ultra-mnist/sample_black_cropped
13                            /Landmark2/pdo/aiking/data/ultra-mnist/valid_train.csv
14                            /Landmark2/pdo/aiking/data/ultra-mnist/ultra-mnist.zip
15                      /Landmark2/pdo/aiking/data/ultra-mnist/sample_submission.csv
16                                /Landmark2/pdo/aiking/data/ultra-mnist/train_black
17                                /Landmark2/pdo/aiking/data/ultra-mnist/test_black2
18                                     /Landmark2/pdo/aiking/data/ultra-mnist/sample
19                            /Landmark2/pdo/aiking/data/ultra-mnist/sample_intg.csv
20                           /Landmark2/pdo/aiking/data/ultra-mnist/valid_sample.csv
dtype: object

# !mkdir {path}/'sample'

path/'sample_black_cropped_resized_integrated'
# path/'train'

Path('/Landmark2/pdo/aiking/data/ultra-mnist/sample_black_cropped_resized_integrated')

(path/"sample_black_cropped_resized_integrated").ls()

(#560) [Path('/Landmark2/pdo/aiking/data/ultra-mnist/sample_black_cropped_resized_integrated/mptgfndbad'),Path('/Landmark2/pdo/aiking/data/ultra-mnist/sample_black_cropped_resized_integrated/uznkagrqln'),Path('/Landmark2/pdo/aiking/data/ultra-mnist/sample_black_cropped_resized_integrated/gkuhpigoxr'),Path('/Landmark2/pdo/aiking/data/ultra-mnist/sample_black_cropped_resized_integrated/tvkpbnawop'),Path('/Landmark2/pdo/aiking/data/ultra-mnist/sample_black_cropped_resized_integrated/jfsumbgtsg'),Path('/Landmark2/pdo/aiking/data/ultra-mnist/sample_black_cropped_resized_integrated/efstasnumi'),Path('/Landmark2/pdo/aiking/data/ultra-mnist/sample_black_cropped_resized_integrated/dhxwgmgutm'),Path('/Landmark2/pdo/aiking/data/ultra-mnist/sample_black_cropped_resized_integrated/xbgfseogty'),Path('/Landmark2/pdo/aiking/data/ultra-mnist/sample_black_cropped_resized_integrated/ytlwuwflrt'),Path('/Landmark2/pdo/aiking/data/ultra-mnist/sample_black_cropped_resized_integrated/nphunajnco')...]

Form new train and valid df#

# df_train = pd.read_csv(path/'train_sample.csv')
# df_train['id'] = df_train['id'] + ".jpeg"
# df_train

df_sample_intg = pd.read_csv(path/"sample_intg.csv"); df_sample_intg
df_sample_train = pd.read_csv(path/"train_sample.csv"); df_sample_train
df_sample_valid = pd.read_csv(path/"valid_sample.csv"); df_sample_valid
df_sample = pd.concat([df_sample_train, df_sample_valid])
# df_sample_intg['digit_sum'] = None
df_sample_intg['dstype'] = None
# df_sample_intg.shape, df_sample_intg.join(df_sample_train.set_index('id'), on='id').shape
train_filter = df_sample_intg['id'].isin(df_sample_train.id)
valid_filter = df_sample_intg['id'].isin(df_sample_valid.id)
df_sample_intg.loc[train_filter, 'dstype'] = 'train'
df_sample_intg.loc[valid_filter, 'dstype'] = 'valid'
df_sample_intg['digit_sum'] = df_sample_intg.join(df_sample.set_index('id'), on='id')['digit_sum']
df_sample_intg
# df_sample_intg.loc[valid_filter, 'digit_sum'] = df_sample_intg[valid_filter].join(df_sample_valid.set_index('id'), on='id')['digit_sum']
# df_sample_intg.dtypes, df_sample_train.dtypes
path fname id dstype digit_sum
0 /Landmark2/pdo/aiking/data/ultra-mnist/sample_black_cropped_resized_integrated/mptgfndbad/mptgfndbad_doocaoebo.jpeg mptgfndbad_doocaoebo.jpeg mptgfndbad train 9
1 /Landmark2/pdo/aiking/data/ultra-mnist/sample_black_cropped_resized_integrated/mptgfndbad/mptgfndbad_ocboeoaod.jpeg mptgfndbad_ocboeoaod.jpeg mptgfndbad train 9
2 /Landmark2/pdo/aiking/data/ultra-mnist/sample_black_cropped_resized_integrated/mptgfndbad/mptgfndbad_cdaoooebo.jpeg mptgfndbad_cdaoooebo.jpeg mptgfndbad train 9
3 /Landmark2/pdo/aiking/data/ultra-mnist/sample_black_cropped_resized_integrated/mptgfndbad/mptgfndbad_bacoooeod.jpeg mptgfndbad_bacoooeod.jpeg mptgfndbad train 9
4 /Landmark2/pdo/aiking/data/ultra-mnist/sample_black_cropped_resized_integrated/mptgfndbad/mptgfndbad_oocabodoe.jpeg mptgfndbad_oocabodoe.jpeg mptgfndbad train 9
... ... ... ... ... ...
5029 /Landmark2/pdo/aiking/data/ultra-mnist/sample_black_cropped_resized_integrated/tvqvvpnnpf/tvqvvpnnpf_dcboaoooo.jpeg tvqvvpnnpf_dcboaoooo.jpeg tvqvvpnnpf train 21
5030 /Landmark2/pdo/aiking/data/ultra-mnist/sample_black_cropped_resized_integrated/tvqvvpnnpf/tvqvvpnnpf_adcoooobo.jpeg tvqvvpnnpf_adcoooobo.jpeg tvqvvpnnpf train 21
5031 /Landmark2/pdo/aiking/data/ultra-mnist/sample_black_cropped_resized_integrated/tvqvvpnnpf/tvqvvpnnpf_ooocodboa.jpeg tvqvvpnnpf_ooocodboa.jpeg tvqvvpnnpf train 21
5032 /Landmark2/pdo/aiking/data/ultra-mnist/sample_black_cropped_resized_integrated/tvqvvpnnpf/tvqvvpnnpf_oocaoodob.jpeg tvqvvpnnpf_oocaoodob.jpeg tvqvvpnnpf train 21
5033 /Landmark2/pdo/aiking/data/ultra-mnist/sample_black_cropped_resized_integrated/tvqvvpnnpf/tvqvvpnnpf_oaodoocbo.jpeg tvqvvpnnpf_oaodoocbo.jpeg tvqvvpnnpf train 21

5034 rows × 5 columns


ImageDataLoaders.from_df?

Signature:
ImageDataLoaders.from_df(
    df,
    path='.',
    valid_pct=0.2,
    seed=None,
    fn_col=0,
    folder=None,
    suff='',
    label_col=1,
    label_delim=None,
    y_block=None,
    valid_col=None,
    item_tfms=None,
    batch_tfms=None,
    bs=64,
    val_bs=None,
    shuffle=True,
    device=None,
)
Docstring: Create from `df` using `fn_col` and `label_col`
File:      /opt/anaconda/envs/aiking/lib/python3.9/site-packages/fastai/vision/data.py
Type:      method

df_sample_intg['path']= df_sample_intg['path'].apply(lambda p : pathlib.Path(p).relative_to(path))
df_train = df_sample_intg[df_sample_intg['dstype'] =='train']
df_valid = df_sample_intg[df_sample_intg['dstype'] =='valid']
df_valid
path fname id dstype digit_sum
36 sample_black_cropped_resized_integrated/jfsumbgtsg/jfsumbgtsg_aoodooecb.jpeg jfsumbgtsg_aoodooecb.jpeg jfsumbgtsg valid 9
37 sample_black_cropped_resized_integrated/jfsumbgtsg/jfsumbgtsg_oaooobecd.jpeg jfsumbgtsg_oaooobecd.jpeg jfsumbgtsg valid 9
38 sample_black_cropped_resized_integrated/jfsumbgtsg/jfsumbgtsg_oobcaeood.jpeg jfsumbgtsg_oobcaeood.jpeg jfsumbgtsg valid 9
39 sample_black_cropped_resized_integrated/jfsumbgtsg/jfsumbgtsg_doeocboao.jpeg jfsumbgtsg_doeocboao.jpeg jfsumbgtsg valid 9
40 sample_black_cropped_resized_integrated/jfsumbgtsg/jfsumbgtsg_oaoobcode.jpeg jfsumbgtsg_oaoobcode.jpeg jfsumbgtsg valid 9
... ... ... ... ... ...
4957 sample_black_cropped_resized_integrated/mdggejfmfw/mdggejfmfw_obafcooed.jpeg mdggejfmfw_obafcooed.jpeg mdggejfmfw valid 16
4958 sample_black_cropped_resized_integrated/mdggejfmfw/mdggejfmfw_abdoeofco.jpeg mdggejfmfw_abdoeofco.jpeg mdggejfmfw valid 16
4959 sample_black_cropped_resized_integrated/mdggejfmfw/mdggejfmfw_bcefaoodo.jpeg mdggejfmfw_bcefaoodo.jpeg mdggejfmfw valid 16
4960 sample_black_cropped_resized_integrated/mdggejfmfw/mdggejfmfw_eafoobdco.jpeg mdggejfmfw_eafoobdco.jpeg mdggejfmfw valid 16
4961 sample_black_cropped_resized_integrated/mdggejfmfw/mdggejfmfw_bocaoodef.jpeg mdggejfmfw_bocaoodef.jpeg mdggejfmfw valid 16

1007 rows × 5 columns

dls = ImageDataLoaders.from_df(df_train,path, folder=None, valid_pct=0.2, fn_col=0, label_col=4, bs=16); dls

<fastai.data.core.DataLoaders at 0x152ef20e5760>

dls.show_batch()
../../_images/06_sample_integrated_17_0.png 
!nvidia-smi

Tue Mar 15 09:44:09 2022       
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 470.103.01   Driver Version: 470.103.01   CUDA Version: 11.4     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|                               |                      |               MIG M. |
|===============================+======================+======================|
|   0  Tesla K80           Off  | 00000000:00:1E.0 Off |                    0 |
| N/A   68C    P0    62W / 149W |   1316MiB / 11441MiB |      0%      Default |
|                               |                      |                  N/A |
+-------------------------------+----------------------+----------------------+
                                                                               
+-----------------------------------------------------------------------------+
| Processes:                                                                  |
|  GPU   GI   CI        PID   Type   Process name                  GPU Memory |
|        ID   ID                                                   Usage      |
|=============================================================================|
|    0   N/A  N/A    173195      C   ...da/envs/aiking/bin/python     1313MiB |
+-----------------------------------------------------------------------------+

learn = cnn_learner(dls, resnet34, metrics=[error_rate, accuracy]); learn

<fastai.learner.Learner at 0x152edfc161c0>

learn.fine_tune(120, freeze_epochs=50)
epoch train_loss valid_loss error_rate accuracy time
0 4.799863 3.852269 0.934161 0.065839 00:17
1 4.440562 3.630288 0.913043 0.086957 00:18
2 4.137702 3.485804 0.910559 0.089441 00:18
3 3.930142 3.326438 0.881988 0.118012 00:18
4 3.772889 3.232335 0.875776 0.124224 00:18
5 3.537516 3.166939 0.872050 0.127950 00:18
6 3.261934 3.110428 0.859627 0.140373 00:18
7 3.226398 3.069852 0.857143 0.142857 00:17
8 3.125710 3.024059 0.842236 0.157764 00:17
9 2.930925 3.010297 0.827329 0.172671 00:17
10 2.793818 2.934250 0.831056 0.168944 00:17
11 2.674159 2.938103 0.821118 0.178882 00:17
12 2.607552 2.868927 0.804969 0.195031 00:17
13 2.445592 2.868701 0.811180 0.188820 00:17
14 2.324407 2.838203 0.796273 0.203727 00:17
15 2.280806 2.842952 0.806211 0.193789 00:17
16 2.259488 2.829938 0.780124 0.219876 00:17
17 2.162574 2.797639 0.763975 0.236025 00:17
18 2.081153 2.750004 0.751553 0.248447 00:17
19 1.971656 2.738195 0.757764 0.242236 00:17
20 1.952935 2.759722 0.786335 0.213665 00:17
21 1.873172 2.714657 0.762733 0.237267 00:17
22 1.861983 2.709193 0.734161 0.265839 00:17
23 1.857842 2.599679 0.727950 0.272050 00:17
24 1.770387 2.656592 0.734161 0.265839 00:17
25 1.717701 2.600894 0.732919 0.267081 00:17
26 1.651577 2.582834 0.696894 0.303106 00:17
27 1.646941 2.565307 0.719255 0.280745 00:17
28 1.605862 2.606378 0.709317 0.290683 00:17
29 1.578137 2.604069 0.690683 0.309317 00:17
30 1.487164 2.589584 0.688199 0.311801 00:17
31 1.499904 2.622078 0.685714 0.314286 00:17
32 1.416947 2.607370 0.695652 0.304348 00:17
33 1.483090 2.651856 0.689441 0.310559 00:17
34 1.375965 2.557366 0.667081 0.332919 00:17
35 1.371571 2.497138 0.633540 0.366460 00:17
36 1.211827 2.438924 0.645963 0.354037 00:17
37 1.239312 2.473795 0.618634 0.381366 00:17
38 1.246859 2.457524 0.619876 0.380124 00:17
39 1.235076 2.392699 0.612422 0.387578 00:17
40 1.121472 2.471771 0.634783 0.365217 00:17
41 1.137847 2.335734 0.627329 0.372671 00:17
42 1.139235 2.393010 0.619876 0.380124 00:17
43 1.119598 2.397207 0.611180 0.388820 00:17
44 1.060512 2.402379 0.614907 0.385093 00:17
45 1.056118 2.368381 0.598758 0.401242 00:17
46 1.030649 2.285016 0.583851 0.416149 00:17
47 0.960971 2.176330 0.563975 0.436025 00:17
48 1.022968 2.215484 0.582609 0.417391 00:17
49 0.937664 2.248081 0.550311 0.449689 00:17
epoch train_loss valid_loss error_rate accuracy time
0 0.867752 2.070304 0.532919 0.467081 00:24
1 0.657931 1.962824 0.499379 0.500621 00:24
2 0.491939 1.833532 0.460870 0.539130 00:24
3 0.429447 1.721688 0.448447 0.551553 00:24
4 0.359826 1.722428 0.438509 0.561491 00:24
5 0.301921 1.702332 0.439752 0.560248 00:24
6 0.297945 1.634223 0.429814 0.570186 00:24
7 0.272402 1.613680 0.423602 0.576398 00:24
8 0.217226 1.498233 0.408696 0.591304 00:24
9 0.192630 1.502558 0.407453 0.592547 00:24
10 0.168445 1.563167 0.424845 0.575155 00:24
11 0.213405 1.507696 0.375155 0.624845 00:24
12 0.232360 1.438052 0.355280 0.644720 00:24
13 0.195828 1.398571 0.345342 0.654658 00:24
14 0.187390 1.284458 0.339130 0.660870 00:24
15 0.185759 1.265104 0.341615 0.658385 00:24
16 0.201981 1.307698 0.325466 0.674534 00:24
17 0.164799 1.397388 0.350311 0.649689 00:24
18 0.233455 1.324180 0.322981 0.677019 00:24
19 0.197242 1.124726 0.299379 0.700621 00:24
20 0.194921 1.244918 0.318012 0.681988 00:24
21 0.232981 1.216448 0.298137 0.701863 00:24
22 0.208103 1.070318 0.267081 0.732919 00:25
23 0.190553 1.139177 0.304348 0.695652 00:25
24 0.195688 1.171082 0.274534 0.725466 00:25
25 0.171583 1.080241 0.259627 0.740373 00:24
26 0.232525 0.954305 0.243478 0.756522 00:24
27 0.155711 0.969420 0.263354 0.736646 00:24
28 0.133733 0.837546 0.212422 0.787578 00:25
29 0.195431 0.987912 0.221118 0.778882 00:25
30 0.214681 0.939399 0.249689 0.750311 00:26
31 0.124636 0.754483 0.200000 0.800000 00:25
32 0.168588 0.819754 0.209938 0.790062 00:26
33 0.161661 0.791658 0.200000 0.800000 00:24
34 0.109011 0.782671 0.197516 0.802484 00:24
35 0.192774 0.789440 0.183851 0.816149 00:24
36 0.116204 0.625333 0.185093 0.814907 00:24
37 0.115642 0.688804 0.178882 0.821118 00:24
38 0.163575 0.828804 0.183851 0.816149 00:24
39 0.099515 0.641896 0.160248 0.839752 00:24
40 0.097081 0.602731 0.165217 0.834783 00:24
41 0.121348 0.636313 0.156522 0.843478 00:25
42 0.073731 0.656654 0.142857 0.857143 00:24
43 0.092405 0.526649 0.132919 0.867081 00:25
44 0.068598 0.442036 0.118012 0.881988 00:24
45 0.086852 0.523134 0.139130 0.860870 00:24
46 0.078918 0.480999 0.134161 0.865839 00:24
47 0.065713 0.465827 0.124224 0.875776 00:25
48 0.075608 0.439734 0.108075 0.891925 00:24
49 0.076840 0.388335 0.113043 0.886957 00:24
50 0.054057 0.341099 0.095652 0.904348 00:24
51 0.034011 0.340049 0.101863 0.898137 00:24
52 0.057932 0.397950 0.100621 0.899379 00:24
53 0.059346 0.382785 0.111801 0.888199 00:24
54 0.051964 0.452730 0.113043 0.886957 00:25
55 0.038493 0.320593 0.093168 0.906832 00:25
56 0.062239 0.304702 0.090683 0.909317 00:24
57 0.033538 0.266728 0.079503 0.920497 00:24
58 0.035955 0.304774 0.098137 0.901863 00:25
59 0.036616 0.333432 0.093168 0.906832 00:26
60 0.044442 0.376899 0.106832 0.893168 00:25
61 0.038577 0.285362 0.090683 0.909317 00:25
62 0.022169 0.328253 0.085714 0.914286 00:25
63 0.022791 0.391093 0.103106 0.896894 00:25
64 0.042478 0.299289 0.081988 0.918012 00:25
65 0.035514 0.238172 0.073292 0.926708 00:25
66 0.022597 0.263564 0.086957 0.913043 00:24
67 0.020295 0.218213 0.065839 0.934161 00:24
68 0.035894 0.282178 0.074534 0.925466 00:25
69 0.017982 0.260687 0.072050 0.927950 00:24
70 0.014254 0.238768 0.074534 0.925466 00:25
71 0.018478 0.199174 0.064596 0.935404 00:24
72 0.020054 0.219791 0.059627 0.940373 00:24
73 0.012720 0.224474 0.065839 0.934161 00:24
74 0.010848 0.215075 0.060870 0.939130 00:24
75 0.010696 0.227063 0.069565 0.930435 00:24
76 0.005756 0.189376 0.050932 0.949068 00:25
77 0.010287 0.185252 0.057143 0.942857 00:24
78 0.015689 0.136085 0.039752 0.960248 00:24
79 0.005426 0.115787 0.039752 0.960248 00:24
80 0.009698 0.151767 0.048447 0.951553 00:25
81 0.005204 0.110152 0.034783 0.965217 00:25
82 0.004047 0.114986 0.034783 0.965217 00:24
83 0.004158 0.163543 0.053416 0.946584 00:24
84 0.009039 0.138988 0.043478 0.956522 00:24
85 0.006754 0.191037 0.052174 0.947826 00:25
86 0.003982 0.195469 0.049689 0.950311 00:24
87 0.003866 0.177618 0.049689 0.950311 00:25
88 0.004492 0.174466 0.043478 0.956522 00:25
89 0.004034 0.152141 0.043478 0.956522 00:24
90 0.001440 0.186529 0.050932 0.949068 00:25
91 0.003572 0.189599 0.050932 0.949068 00:25
92 0.001508 0.168098 0.047205 0.952795 00:25
93 0.001858 0.148741 0.034783 0.965217 00:25
94 0.000744 0.138702 0.040994 0.959006 00:25
95 0.002585 0.141475 0.039752 0.960248 00:25
96 0.004510 0.135052 0.031056 0.968944 00:25
97 0.003152 0.136238 0.044720 0.955280 00:25
98 0.000858 0.136676 0.034783 0.965217 00:25
99 0.000519 0.143548 0.050932 0.949068 00:25
100 0.001485 0.119592 0.038509 0.961491 00:25
101 0.001062 0.121021 0.040994 0.959006 00:25
102 0.000551 0.099376 0.036025 0.963975 00:25
103 0.000613 0.107080 0.042236 0.957764 00:25
104 0.000308 0.108137 0.036025 0.963975 00:25
105 0.000468 0.101996 0.039752 0.960248 00:25
106 0.001191 0.096008 0.031056 0.968944 00:25
107 0.000355 0.108981 0.037267 0.962733 00:25
108 0.000911 0.101906 0.036025 0.963975 00:25
109 0.000985 0.106482 0.037267 0.962733 00:24
110 0.000230 0.105506 0.032298 0.967702 00:25
111 0.000306 0.099976 0.033540 0.966460 00:25
112 0.005453 0.099953 0.032298 0.967702 00:25
113 0.003114 0.102232 0.033540 0.966460 00:25
114 0.001258 0.102980 0.032298 0.967702 00:25
115 0.000452 0.102922 0.026087 0.973913 00:25
116 0.000287 0.099809 0.033540 0.966460 00:25
117 0.001599 0.097145 0.034783 0.965217 00:25
118 0.001247 0.103972 0.032298 0.967702 00:25
119 0.000188 0.105384 0.037267 0.962733 00:25 
!pwd
# learn.save("ultramnist_a0.pkl")
learn.export()

/Landmark2/pdo/Code/ALDAOps/aiking/book/Competitions/ultramnist

learn?

Signature:      learn(event_name)
Type:           Learner
String form:    <fastai.learner.Learner object at 0x152edfc161c0>
File:           /opt/anaconda/envs/aiking/lib/python3.9/site-packages/fastai/learner.py
Docstring:      Group together a `model`, some `dls` and a `loss_func` to handle training
Call docstring: Call `event_name` for all `Callback`s in `self.cbs`