What does util.generic_utils.deserialize_keras_object() function in keras do - machine-learning

In Keras library, activattion.get() calls activattion.deserialize() function which in turn calls util.generic_utils.deserialize_keras_object() function. Can someone explain what happens in deserialize_keras_object() function

As one may read here deserialize_keras_object is responsible for creation of a keras object from:
configuration dictionary - if one is available,
name identifier if a provided one was available.
To understand second point image the following definition:
model.add(Activation("sigmoid"))
What you provide to Activation constructor is a string, not a keras object. In order to make this work - deserialize_keras_object looks up defined names and check if an object called sigmoid is defined and instantiates it.

Related

How to change floating base type of model loaded from urdf file using pydrake?

I am trying to load model from urdf file.
I can achieve this via RigidBodyTree class:
rbtree = RigidBodyTree(file_name, floating_base_type)
Unfortunately, as it said at pydrake.attic reference page, it will be deprecated soon.
I tried to add model using pydrake.multibody.parsing.Parser and pydrake.multibody.plant, but it seems that model is attached only with floating quaternion joint.
Is there a legal way of setting floating base type not with attic API?
For a fixed base, the method you're looking for is MultibodyPlant.WeldFrames(). If plant is the MultibodyPlant object to which you've added your model, and the frame in your model named "my_base_frame" is supposed to be fixed to the world, the appropriate call would be:
plant.WeldFrames(plant.world_frame(), plant.GetFrameByName("my_base_frame"))
Note that this call should be made prior to calling plant.Finalize().
I believe that MultibodyPlant currently does not support roll-pitch-yaw floating bases.

How to call ModelCheckpoint (or any callback) explicitly in Keras?

My ModelCheckpoint is intended to save model each epoch. Unfortunately, I don't use built-in epochs of Model.
How to call ModelCheckpoint not via callback, but explicitly, when my code ends epoch, so that it does all it's job?
I found method on_epoch_end, but can't figure out, where to pass model itself?
Each callback is derived from Callback abstract base class. It has set_model method which is used to pass the corresponding model. Check it in callbacks.py

Add custom loss function in Torch

What are the necessary steps to implement a custom loss function in Torch?
It seems like you have to write an implementation for updateOutput and updateGradInput.
Is that all? So then you basically create a new class:
local CustomCriterion, parent = torch.class('CustomCriterion','nn.Criterion')
and implement the following two functions:
function CustomCriterion:updateOutput(input, target)
function CustomCriterion:updateGradInput(input, target)
Is that correct, or is there more to be done?
Also, for the provided criterions these functions are implemented in C, but I suppose a Lua implementation will also work, albeit possibly a little slower?
I've implemented functions of the form (in pseudo-code)
--assuming input is partitioned in input_a,input_b
-- target is accordingly partitionend in target_a, target_b
f(input)=MSE(input_a,target_a)+ custom_sutff(input_b,target_b)
quite a few times just the way you describe it. So, as far as I know, I think the answer to both your questions is yes.
Basically nn/MSECriterion.lua and this seem to back this up.

Do i need to do g_object_unref() on glib signal parameters?

when i connect a signal to a callback function the callback functions gets passed parameters. Is the reference counter increased before the objects get passed to my callback function or do i have to increase it by myself.
I guess there must be some sort of convention for that because nothing like that is mentioned in the documentation of gtk or libgobject.
Generally, you do not assume a reference on an object when it is passed to your callback. You only assume a reference when the object is the return value of a method which is annotated with "transfer full". You can see these annotations in the documentation.
(I say "generally" because there may always be badly constructed libraries whose API violates these guidelines. You can't do a whole lot about that, though.)

What does the function state.tokenize in 'CodeMirror' mean?

I want to know the function performed by tokenize in codemirror.
codemirror highlights text by calling a tokenizer function, passing it a context ("state"), and a pointer to the current location in the file that needs to be highlighted ("stream"). The job of this function is to advance the stream past the next token, and to return the type of the token. This is described fairly well in the codemirror api documentation here: http://codemirror.net/doc/manual.html#modeapi
In the case of xml.js (which you referenced in a comment), it has multiple tokenizer functions. Depending on the context, it will set the "tokenize" attribute of the state to refer to one of the tokenizer functions. Then it will use whichever function is pointed by by state.tokenize to find the next token in the stream.

Resources