More convenient features based on PyTorch (originally Torchure)
TorchFun project was initiated long ago and was published in 2018-6-13.
TorchFun is motivated by the one author loves, who sometimes encounter inconvenience using PyTorch at her work. The author published the project as a python package
TorchFun on PyPi, so that his beloved could access it whenever and wheresoever needed.
The purposed of TorchFun is to provide functions which are important and convenient, but absent in PyTorch, like some layers and visualization utils.
This project has been undergoing in secret so that the author could credit TorchFun to his beloved as a little supprise of help, one day when this project is well-enriched or gets famous.
Interestingly, The original project name given by the author, is
Torchure. That is because he was always multi-tasking trivial affairs in school and got scorched, and when he was learning this new framework in hope to help his beloved, he found plenty of issues/missing-functionalities. He felt that this was totally a torture. So, this project was named "Torchure" to satirize the lame PyTorch (you can still found Torchure in PyPi). And, the author hoped, by developing Torchure, his beloved could feel ease even when encountering the crappy-parts of PyTorch.
This history-of-project was appended recently, because his adorable little beloved wants a supprise immediately, or she will keep rolling on the floor.
To Eggy c71b05bf46d8772e4488335085a2e7fd.
Please visit: https://sorenchiron.github.io/torchfun
pip install torchfun
pip install -U torchfun
used to reshape outputs
flat = Flatten() out = flat(x)
out = flatten(x)
Unfold channel/depth dimensions to enlarge the feature map
Output size is deducted. The size of the unfold square is automatically determined
images: 100x16x16x9. 9=3x3 square subpixel-out: 100x48x48x1
out_channels, channel number of output feature map
Subpixel Layer [Module]
Same functionality as subpixel(x), but with Module interface.
s = Subpixel(out_channels=1) out = s(x)
only deal with torch channel-first image batch,
- x: input data cube, torch tensor or numpy array.
- title: add title to plot. (Default None)
- title can be string, or any string-able object.
- auto_close: (default True)
- Close the pyplot session afterwards.
- Clean the environment just like you had never used matplotlib here.
- if set to False, the plot will remain in the memory for further drawings.
imshow(batch) imshow(batch,title=[a,b,c]) imshow(batch,title='title') imshow(batch,auto_close=False)
TorchFun:imshow:Warning, you are using WebAgg backend for Matplotlib. Please consider windowed display SDKs such as TkAgg backend and GTK* backends.
This means your matplotlib is using web-browser for figure display. We strongly recommend you to use window-based native display because browser-based backends are fragile and tend to crash. You can change the display mamanger for matplotlib each time you execute your script by:
import matplotlib matplotlib.use('TkAgg') # or GTK GTKAgg
or permanantly by editing:
site-packages/matplotlib/mpl-data/matplotlibrc and change backend to
A full list of available backends can be found at:
import matplotlib matplotlib.rcsetup.all_backends
and, the TCL/TK GUI library for
tkinter can be downloaded here.
- arbitrary arguemnts named :
b, or load model
aThe order of the arguments doesn't matter. Example:
>f = open('weight.pts') >load(f,model)
- arbitrary arguemnts named :
a into target
b, or save model
b into target
The order of the arguments doesn't matter.
>f = open('weight.pts') >save(f,model)
Return value: None
Count parameter numer of a module/state_dict/layer/tensor. This function can also print the occupied memory of parameters in MBs
- model_or_dict_or_param: model or state dictionary or parameters()
Return: parameter amount in python-int Returns 0 if datatype not understood
count_parameters(model) count_parameters(state_dict) #all params count_parameters(model.parameters()) #only trainable params count_parameters(weight_tensor) count_parameters(numpy_array)
This is a very useful thing you definitly have been dreaming of.
You can now use packsearch to query names inside any package!
Given an module object as input:
> p = Packsearch(torch) or > p = Packsarch(numpy) whatever
p.search() method. So that you can
search everything inside this package
Packsearch: 35 results found: -------------results start------------- 0 torch.nn.AdaptiveMaxPool1d 1 torch.nn.AdaptiveMaxPool2d 2 torch.nn.AdaptiveMaxPool3d 3 torch.nn.FractionalMaxPool2d 4 torch.nn.MaxPool1d 5 torch.nn.MaxPool2d ...
Given an module object, and search pattern string as input:
> packsearch('maxpoo',torch) output: Packsearch: 35 results found: -------------results start------------- 0 torch.nn.AdaptiveMaxPool1d 1 torch.nn.AdaptiveMaxPool2d 2 torch.nn.AdaptiveMaxPool3d 3 torch.nn.FractionalMaxPool2d 4 torch.nn.MaxPool1d 5 torch.nn.MaxPool2d ...
you can search for everything inside any package
return the summary of all variables.
This is used to detect chaotic changes of weights. You can check the sum_parameters before and after some operations, to know if there is any change made to the params.
I use this function to verify gradient behaviours.
By default, This only hash the trainable parameters!
- module_or_statdict_or_param: torch.nn.module, or model.state_dict(), or model.parameters().
- use_sum: return the sum instead of mean value of all params.
model = MyNet() print(hash_parameters(model)) # see params train_one_step(model) print(hash_parameters(model)) # see if params are updated print(hash_parameters(model.state_dict())) # see if trainable+un-trainable params are updated
force a series of hierachical directories to exist.
force_exist can automatically create directory with any depth.
- dirname: path of the desired directory
- verbose: print every directory creation. default True.
This is a very interesting function. It is used to support arbitrary-arguments-ordering in TorchFun.
Input: The function takes a list of types, and a list of arguments.
Returns: a list of arguments, with the same order as the types-list.
sort_args supports arbitrary-arguments-ordering by itself.
find the source file location of a module arguments:
- module_or_string: target module object, or it's string path like
- open_gui: open the folder with default window-manager.
- module file name, or None