Inclusion of ANN method #130
Conversation
ezyrb/ann.py
Outdated
|
|
||
| Example: | ||
| >>> import ezyrb | ||
| >>> import torch |
There was a problem hiding this comment.
These torch imports are not needed in the example
ezyrb/ann.py
Outdated
|
|
||
| def __init__(self): | ||
| self.trained_epoch = 0 | ||
| # number of already trained iterations |
There was a problem hiding this comment.
the single comments go above the code line, not below
There was a problem hiding this comment.
Moreover, I totally suggest to remove such comments, since the class variables should be properly explained in the class documentation. No need for duplication!
There was a problem hiding this comment.
These comments have been ideated as a guideline. So maybe they could be useful if changes are needed by a new user.
There was a problem hiding this comment.
That is the exact purpose of the documentation: helping new users. Please remove the comments.
ezyrb/ann.py
Outdated
| >>> import numpy as np | ||
| >>> | ||
| >>> x = np.random.uniform(-1, 1, size=(4, 2)) | ||
| >>> from ann import ANN |
There was a problem hiding this comment.
Not tested, but it looks wrong. Have you check the example?
There was a problem hiding this comment.
Yes, I checked in the Prove.pdf file I attached in my "273cad3" commit. I report a demonstration below. (Out[14] is the training loss at termination).
There was a problem hiding this comment.
Hope that the following convinces you: fresh installation of your commit, and tried the import you said be correct.
Take into account that testing the code means for all (or at least for the large majority) the environments, not only yours (or your directory, as in this case). Also please, check carefully the code, especially after someone's telling you there is probably a mistake.
(base) ~ [ python3 -m pip uninstall ezyrb
Found existing installation: ezyrb 1.0
Uninstalling ezyrb-1.0:
Would remove:
/data/local/lib/python3.8/site-packages/ezyrb-1.0-py3.8.egg
Proceed (y/n)? y
Successfully uninstalled ezyrb-1.0
(base) ~ [ python3 -m pip install git+git://github.com/ggurioli/EZyRB.git@012ee67c4a05d8b26153b633f78e96a868258d16
Defaulting to user installation because normal site-packages is not writeable
Collecting git+git://github.com/ggurioli/EZyRB.git@012ee67c4a05d8b26153b633f78e96a868258d16
Cloning git://github.com/ggurioli/EZyRB.git (to revision 012ee67c4a05d8b26153b633f78e96a868258d16) to /tmp/pip-req-build-ag4g3xrr
Running command git clone -q git://github.com/ggurioli/EZyRB.git /tmp/pip-req-build-ag4g3xrr
Requirement already satisfied: future in /data/local/lib/python3.8/site-packages (from ezyrb==1.1) (0.18.2)
Requirement already satisfied: numpy in /usr/lib/python3.8/site-packages (from ezyrb==1.1) (1.18.4)
Requirement already satisfied: scipy in /usr/lib/python3.8/site-packages (from ezyrb==1.1) (1.4.1)
Requirement already satisfied: matplotlib in /usr/lib/python3.8/site-packages (from ezyrb==1.1) (3.2.1)
Requirement already satisfied: GPy in /data/local/lib/python3.8/site-packages (from ezyrb==1.1) (1.9.9)
Requirement already satisfied: cycler>=0.10 in /usr/lib/python3.8/site-packages (from matplotlib->ezyrb==1.1) (0.10.0)
Requirement already satisfied: kiwisolver>=1.0.1 in /usr/lib/python3.8/site-packages (from matplotlib->ezyrb==1.1) (1.1.0)
Requirement already satisfied: pyparsing!=2.0.4,!=2.1.2,!=2.1.6,>=2.0.1 in /usr/lib/python3.8/site-packages (from matplotlib->ezyrb==1.1) (2.4.7)
Requirement already satisfied: python-dateutil>=2.1 in /usr/lib/python3.8/site-packages (from matplotlib->ezyrb==1.1) (2.8.1)
Requirement already satisfied: paramz>=0.9.0 in /data/local/lib/python3.8/site-packages (from GPy->ezyrb==1.1) (0.9.5)
Requirement already satisfied: six in /usr/lib/python3.8/site-packages (from GPy->ezyrb==1.1) (1.14.0)
Requirement already satisfied: setuptools in /data/local/lib/python3.8/site-packages (from kiwisolver>=1.0.1->matplotlib->ezyrb==1.1) (46.1.3)
Requirement already satisfied: decorator>=4.0.10 in /usr/lib/python3.8/site-packages (from paramz>=0.9.0->GPy->ezyrb==1.1) (4.4.2)
Building wheels for collected packages: ezyrb
Building wheel for ezyrb (setup.py) ... done
Created wheel for ezyrb: filename=ezyrb-1.1-py3-none-any.whl size=12613 sha256=8efae9831e3b19c55bbd161a3b43e525307cab3882d2d6dd19bd014abc5e63a4
Stored in directory: /data/.cache/pip/wheels/f3/91/7b/d288de7188c19541e855435b77de759352c0c3575ff48e7b92
Successfully built ezyrb
Installing collected packages: ezyrb
Successfully installed ezyrb-1.1
(base) ~ [ python3
Python 3.8.2 (default, Apr 8 2020, 14:31:25)
[GCC 9.3.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import ezyrb
>>> from ann import ANN
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
ModuleNotFoundError: No module named 'ann'
There was a problem hiding this comment.
Ok, now I understand what you mean. I've modified the init.py, installed my last commit, and tested the example as reported below. Is it fine?
ezyrb/ann.py
Outdated
| # loss values extraction (type: float) | ||
| loss.backward() | ||
| old_loss = loss.item() |
There was a problem hiding this comment.
It seems to be required for the subsequent self.optimizer.step(). I tried to remove "loss.backward()" but the training is not performing as expected (the magnitude of the training loss at termination is meaningful).
There was a problem hiding this comment.
backward is for sure a fundamental step. I meant the old_loss variable.
There was a problem hiding this comment.
That's true, thanks.
ezyrb/ann.py
Outdated
| # update of the number of trained iterations | ||
| self.trained_epoch += niter | ||
| return loss.item() | ||
|
|
||
| def predict(self,new_point): |
|
@ggurioli Can you implement also a minimal unittest module for the new class? (I suggest to trivially "translate" the test for GPR class https://github.com/mathLab/EZyRB/blob/master/tests/test_gpr.py) |
ezyrb/ann.py
Outdated
| Example: | ||
| >>> import numpy as np | ||
| >>> from ann import ANN | ||
| >>> x = np.random.uniform(-1, 1, size = (4, 2)) |
There was a problem hiding this comment.
size=(4, 2) no space here, since it's an argument.
|
@ggurioli Several commits have similar names, which makes difficult navigate in the history. Please stash your commits (https://git-scm.com/book/it/v2/Git-Tools-Stashing-and-Cleaning) in order to clean the Git history. |
Inclusion of the ann.py, coding the class ANN for the definition of a basic ANN method.
Prove.pdf