merge from Qiushi's fork to yewsg/yews: add polarity and focal_mechanism models #18

zjzzqs · 2020-07-28T01:43:15Z

add polarity and focal_mechanism models

Protect trademarks and logos

* Fix bug in travis.yml * add codecov badge * add python 3.7 to travis.ci * Add python 3.7 travis image * add download badge

Add appveyor.yml for windows CI

* add more test to transforms * replace torch and numpy in module by direct import * add tests for transform correctness

* init docs by sphinx. * Update documentation theme to blue * add doc to README

commit ce4b445 Author: Lijun Zhu <gatechzhu@gmail.com> Date: Wed Apr 17 13:12:43 2019 -0400 yews.transform under cover with 100% coverage. commit 2cf6108 Author: Lijun Zhu <lijunzh@users.noreply.github.com> Date: Tue Apr 16 09:01:48 2019 -0400 use www subdomain for docs commit 4c0b060 Author: Lijun Zhu <gatechzhu@gmail.com> Date: Wed Apr 17 13:15:07 2019 -0400 add is_dataset to check dataset-like objects commit 5a7b0e8 Author: Lijun Zhu <gatechzhu@gmail.com> Date: Tue Apr 16 21:46:35 2019 -0400 Refactorize yews.datasets

commit cbb2b6b Author: Lijun Zhu <gatechzhu@gmail.com> Date: Wed Apr 17 15:04:07 2019 -0400 rename module to avoid python built-ins commit e9ebf46 Author: Lijun Zhu <gatechzhu@gmail.com> Date: Wed Apr 17 15:27:41 2019 -0400 yews.files under cover. commit 744daae Author: Lijun Zhu <gatechzhu@gmail.com> Date: Wed Apr 17 15:04:24 2019 -0400 yews.datasets.dirs under cover commit 1f8be24 Author: Lijun Zhu <gatechzhu@gmail.com> Date: Wed Apr 17 15:04:07 2019 -0400 rename module to avoid python built-ins commit df8c897 Author: Lijun Zhu <gatechzhu@gmail.com> Date: Wed Apr 17 13:44:15 2019 -0400 add test to yews.datasets commit ce4b445 Author: Lijun Zhu <gatechzhu@gmail.com> Date: Wed Apr 17 13:12:43 2019 -0400 yews.transform under cover with 100% coverage. commit 2cf6108 Author: Lijun Zhu <lijunzh@users.noreply.github.com> Date: Tue Apr 16 09:01:48 2019 -0400 use www subdomain for docs commit 4c0b060 Author: Lijun Zhu <gatechzhu@gmail.com> Date: Wed Apr 17 13:15:07 2019 -0400 add is_dataset to check dataset-like objects commit 5a7b0e8 Author: Lijun Zhu <gatechzhu@gmail.com> Date: Tue Apr 16 21:46:35 2019 -0400 Refactorize yews.datasets

Use both lijunzhu and pytorch channels.

commit efe8105c558319d8145b0033f9c108466ca9ad97 Author: Lijun Zhu <gatechzhu@gmail.com> Date: Wed Apr 17 21:57:49 2019 -0400 automate building process commit b9b5ba8f86129ce422d0d521e26064c12bbb88e8 Author: Lijun Zhu <gatechzhu@gmail.com> Date: Wed Apr 17 21:57:07 2019 -0400 hide usage of scipy until necessary commit 9a7c981781ef4f32b346988e493feaf6be02b9dc Author: Lijun Zhu <gatechzhu@gmail.com> Date: Wed Apr 17 20:36:35 2019 -0400 comply with PyPI rst requirement.

…d-cam with cpic_v3, 2000->1000->7->fc

…ews/transforms/functional.py as a backup

…rms/transforms.py

lijunzh

Please take a look at my reviews and change accordingly. Let me know if you have questions.

lijunzh · 2020-07-28T12:01:25Z

.travis.yml

@@ -5,7 +5,10 @@ matrix:
  include:
    - python: "3.6"
    - python: "3.7"
+<<<<<<< HEAD


This is a text generated by git when solving conflicts. Needs to be removed.

Resolve. Please check it again. Thanks.

lijunzh · 2020-07-28T12:03:05Z

README.rst

@@ -16,15 +16,10 @@ applying deep learning techniques on seismic waveform data.



-.. image:: https://travis-ci.org/yewsg/yews.svg?branch=master
-    :target: https://travis-ci.org/yewsg/yews


Why do we want to remove the badges here?

Resolve. Please check it again. Thanks.

lijunzh · 2020-07-28T12:05:06Z

src/yews/models/__init__.py

@@ -1,2 +1,4 @@
 from .cpic import *
+from .polarity import *
+from .focal_mechanism import *



This is OK now. But we will later want to selectively import model constructors into init.py so that we can the limit usage of model classes into restricted ways that has been tested.

lijunzh · 2020-07-28T12:06:05Z

src/yews/models/cpic.py


 model_urls = {
    'cpic_v1': 'https://www.dropbox.com/s/ckb4glf35agi9xa/cpic_v1_wenchuan-bdd92da2.pth?dl=1',
-    'cpic_v2': 'https://www.dropbox.com/s/kyiuprnn8014fs5/cpic_v2_wenchuan-ee92060a.pth?dl=1'
+    'cpic_v2': 'https://www.dropbox.com/s/kyiuprnn8014fs5/cpic_v2_wenchuan-ee92060a.pth?dl=1',
+    'cpic_v3': 'xxxx'


Are you planning to upload a pretrained model for cpic_v3? You can use '' instead of 'xxxx' for place holder.

Resolve. Please check it again. Thanks.

lijunzh · 2020-07-28T12:12:50Z

src/yews/train/functional.py

@@ -12,6 +12,7 @@ def get_torch_device():

 def model_on_device(model, device):
    return torch.nn.DataParallel(model.to(device))
+    #return torch.nn.DataParallel(model,device_ids=[0]).to("cuda")


I understand that you want to select which GPU to use. However, that is precisely the reason why I passed in device as an argument. You may consider using something below:

device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu") model_on_device(model, device)

In general, you want to control the selection of GPU in your main script instead of inside the trainer class. Thus, it is better to be passed in as an argument.

Resolve. Please check it again. Thanks.

lijunzh · 2020-07-28T12:13:23Z

src/yews/transforms/transforms.py

+<<<<<<< HEAD
+from scipy import signal
+=======
+>>>>>>> upstream/master



Same here. Fix the git artifects.

Resolve. Please check it again. Thanks.

lijunzh · 2020-07-28T12:14:49Z

src/yews/transforms/transforms.py

+        b, a = signal.butter(order, [low, high], btype='bandpass')
+        wav = signal.filtfilt(b, a, wav, axis=-1, padtype=None, padlen=None, irlen=None)
+
+        return wav


Please consider adding unit tests to the transforms added here. You only needs some basic cases to cover the straightforward usages.

Resolve. Please check it again. Thanks.

lijunzh · 2020-07-28T12:15:36Z

src/yews/models/focal_mechanism.py

@@ -0,0 +1,331 @@
+import torch.nn as nn


Did not check the correctness here. Since you are working on the model, you konw better than me.

lijunzh · 2020-07-28T12:17:20Z

examples/wenchuan_cpic.training.py

+    trainer = Trainer(model, CrossEntropyLoss(), lr=0.1)
+    trainer.validate(val_loader, print_freq=100)
+
+    print("Now: end : " + str(datetime.datetime.now()))


I am in the process of revamping the underlying training tools (multi-process up to 8 GPUs and a logging system). We may need to change this later. In the meantime, feel free to change it as long as it does not error out.

…device(cuda:0 if torch.cuda.is_available() else cpu) model_on_device(model, device)

lijunzh · 2020-08-01T20:46:29Z

@zjzzqs I've merged your changes. There are still some places that we can discuss and improve. But, we should do it in another PR next time.

* master: merge from Qiushi's fork to yewsg/yews: add polarity and focal_mechanism models (#18)

lijunzh and others added 30 commits April 12, 2019 12:34

Replace MIT license by Apache 2.0

03819dd

Protect trademarks and logos

include license and readme for pip package

50694af

add tests for yews.transform.functional

8c83e13

try travis from torchvision

f142a87

Add travis.ci badge

076dec9

Fix bug in travis.yml (#1)

01f8d19

* Fix bug in travis.yml * add codecov badge * add python 3.7 to travis.ci * Add python 3.7 travis image * add download badge

update logo color

cc5dc0b

try torchvision’s sphinx setup

b43159b

Update logo

ff88395

Add appveyor.yml for Windows CI (#3)

11e2cee

Add appveyor.yml for windows CI

add appveyor badge

2e4703c

add bages for anaconda cloud and pypi

7538b25

move badge below the title

b37c503

remove line between logo and title

f74ca1d

Add more test to transforms (#2)

8b300c8

* add more test to transforms * replace torch and numpy in module by direct import * add tests for transform correctness

Add instructin to install pytorch first

210fa17

Update conda command

c95ad7f

Uploading PyTorch builds to lijunzhu channel

50e0868

not import yews in docs

ae94d69

Create initial docs (yewsg#4)

7516965

* init docs by sphinx. * Update documentation theme to blue * add doc to README

use www subdomain for docs

6c86d34

Add transform to convert label to int

b2051a9

Improve conda installation

35a559d

Use both lijunzhu and pytorch channels.

automate release

bf621c6

improve docs.

4f5bbd3

move docs to a separate repo.

c6edbc5

bump version to 0.0.3

720b2e8

zjzzqs added 17 commits April 27, 2020 01:32

add vgg style model cpicv3, stop at 4 for grad-cam

1e464a8

vgg style model FmV2 stop at 8*8

0b28b77

vgg style model FmV2 stop at 8*8

a8e95ee

update cpic.py and wenchuan_cpic.training.py based on the test of gra…

6f4014b

…d-cam with cpic_v3, 2000->1000->7->fc

forget why, so just add a comment #wav = wav.astype(float) into src/y…

d2b0be3

…ews/transforms/functional.py as a backup

add RemoveMean RemoveTrend Taper BandpassFilter into src/yews/transfo…

3935031

…rms/transforms.py

add polarity_cnn_lstm from Zijian Li

773bae9

input 600->300

2c717e7

rm data in example

9120bd9

fix super

4b2ae51

update example

1047605

update cnn_lstm

b2b40f1

add a line for LSTM which can only use one gpu

898d3f5

need to be updated, how to read the pretrained model

e21445f

Resolve merge conflict

35d670d

Resolve merge conflict

20a4124

delete train.py.bak

2f27820

lijunzh self-requested a review July 28, 2020 12:00

lijunzh reviewed Jul 28, 2020

View reviewed changes

zjzzqs added 6 commits August 1, 2020 02:38

fix bug <<<<<<< ======= >>>>>>>

6d84a52

recover some image and target

2297e5f

nothing important

8805b78

xxxx to null link

e019e7c

delete the commit of using 1 gpu, in the future, use: device = torch.…

9ed1303

…device(cuda:0 if torch.cuda.is_available() else cpu) model_on_device(model, device)

remove RemoveMean, change Taper and BandpassFilter

2dadda1

zjzzqs requested a review from lijunzh August 1, 2020 06:57

lijunzh approved these changes Aug 1, 2020

View reviewed changes

lijunzh merged commit 730a51e into yewsg:master Aug 1, 2020

lijunzh added a commit that referenced this pull request Aug 3, 2020

Merge branch 'master' into chunk

2fe4743

* master: merge from Qiushi's fork to yewsg/yews: add polarity and focal_mechanism models (#18)

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

merge from Qiushi's fork to yewsg/yews: add polarity and focal_mechanism models #18

merge from Qiushi's fork to yewsg/yews: add polarity and focal_mechanism models #18

zjzzqs commented Jul 28, 2020

lijunzh left a comment

lijunzh Jul 28, 2020

zjzzqs Aug 1, 2020

lijunzh Jul 28, 2020

zjzzqs Aug 1, 2020

lijunzh Jul 28, 2020

lijunzh Jul 28, 2020

zjzzqs Aug 1, 2020

lijunzh Jul 28, 2020

zjzzqs Aug 1, 2020

lijunzh Jul 28, 2020

zjzzqs Aug 1, 2020

lijunzh Jul 28, 2020

zjzzqs Aug 1, 2020

lijunzh Jul 28, 2020

lijunzh Jul 28, 2020

lijunzh commented Aug 1, 2020

		@@ -16,15 +16,10 @@ applying deep learning techniques on seismic waveform data.



		.. image:: https://travis-ci.org/yewsg/yews.svg?branch=master
		:target: https://travis-ci.org/yewsg/yews

merge from Qiushi's fork to yewsg/yews: add polarity and focal_mechanism models #18

merge from Qiushi's fork to yewsg/yews: add polarity and focal_mechanism models #18

Conversation

zjzzqs commented Jul 28, 2020

lijunzh left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

lijunzh commented Aug 1, 2020