Created
June 15, 2018 13:01
-
-
Save mdouze/a204a3460393a630a8708a2a4f01f5fb to your computer and use it in GitHub Desktop.
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
(fair_env) matthijs@devfair0144:~/src/low-shot-with-diffusion$ python diffusion.py --mode test --nlabeled 2 --seed 1 --nbg 1000000 --niter 3 | |
========================== run on Test | |
load train + test set | |
nb of eval classes 311 | |
/public/apps/anaconda2/5.0.1/envs/fair_env/lib/python2.7/site-packages/h5py/__init__.py:36: FutureWarning: Conversion of the second argument of issubdtype from `float` to `np.floating` is deprecated. In future, it will be treated as `np.float64 == np.dtype(float).type`. | |
from ._conv import register_converters as _register_converters | |
open /checkpoint/matthijs/low-shot/features/val.hdf5 | |
read /checkpoint/matthijs/low-shot/features/val_features.npy | |
features loaded in 3.415 s | |
nb of eval classes 311 | |
open /checkpoint/matthijs/low-shot/features/train.hdf5 | |
read /checkpoint/matthijs/low-shot/features/train_features.npy | |
features loaded in 87.034 s | |
selecting images, seed=1 | |
load /checkpoint/matthijs/low-shot/features/PCAR256.vt | |
dataset sizes: Xtr (250860, 256) (250860 labeled), Xte (25350, 256), 507 classes (eval on 25350 examples, 507 classes, 311 novel classes) | |
memmapping /checkpoint/matthijs/low-shot/features//f100m/concatenated_PCAR256.raw | |
[0.000 s, 0.36 GiB] make knn graph for 250860+1000000 k=30 d=256 | |
[0.000 s, 0.36 GiB] make distractor graph for ndis=1000000 k=30 | |
[0.000 s, 0.36 GiB] fname_base= /checkpoint/matthijs/low-shot/knngraph/ndis1000000 | |
[0.001 s, 0.36 GiB] load /checkpoint/matthijs/low-shot/knngraph/ndis1000000.index | |
moving to 2 GPUs | |
done in 6.657 s | |
[7.566 s, 3.07 GiB] mmap /checkpoint/matthijs/low-shot/knngraph/ndis1000000_k32_D_11.float32 /checkpoint/matthijs/low-shot/knngraph/ndis1000000_k32_I_11.int32 | |
[0.006 s, 3.07 GiB] distractor graph ready | |
[7.573 s, 3.07 GiB] spherifying all descriptors | |
[0.337 s, 3.08 GiB] search labelled (250860, 256) | |
search 245760:250860 / 250860 | |
[7.780 s, 3.08 GiB] search test (25350, 256) | |
[1.132 s, 2.58 GiB] make Flat index for labelled | |
moving to 2 GPUs | |
done in 0.328 s | |
[0.495 s, 3.08 GiB] alloc output | |
[0.000 s, 3.08 GiB] search labelled | |
[2.215 s, 3.08 GiB] search distractors | |
search 999424:1000000 / 1000000 | |
[9.513 s, 4.08 GiB] search test | |
[0.361 s, 3.58 GiB] merge results | |
[0.080 s, 3.55 GiB] graph done | |
[3.55 GiB] begin diffusion | |
knngraph_to_CSRMatrix in: 3.55 GiB | |
A: 3.55 GiB | |
B: 4.32 GiB | |
C: 4.79 GiB | |
D: 4.79 GiB, matrix nnz=64074274 | |
[1.233 s, 4.79 GiB] preproc done | |
[0.000 s, 4.79 GiB] Classes 0:507 | |
[20.490 s, 7.20 GiB] start iter | |
[25.131 s, 7.25 GiB] iter 0 val nnz 0.042 | |
[28.888 s, 7.29 GiB] iter 1 val nnz 0.213 | |
[33.233 s, 7.39 GiB] iter 2 val nnz 0.499 | |
intial state nnz 2.71 top-5 accuracy: 0.412 (0.093 on novel) | |
iter 0/3 nnz 21.15 top-5 accuracy: 0.587 (0.370 on novel) | |
iter 1/3 nnz 108.02 top-5 accuracy: 0.672 (0.531 on novel) | |
iter 2/3 nnz 252.89 top-5 accuracy: 0.678 (0.544 on novel) |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment