This repository has been archived by the owner on Nov 17, 2023. It is now read-only.
Define loss function in Python #1868
Merged
Conversation
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
|
Looks nice, will this works naturally with the data parallel case? |
|
Also cc @jermainewang Maybe minpy could be used here as well ? |
|
@tqchen yes, the computation graph module is independent of the loss module. The first part can do data parallel as usual, the latter part can also do whatever they like. I guess for example people could implement the loss function with pycuda etc. |
|
I guess it might be possible to define a minpy module that wraps Amy minpy function and utilize the auto grad feature of minpy. |
|
Yes, that is what I mean. Could be interesting for a separate PR |
Squashed commits: [960414a] simplify the user interface for defining python loss module [dbb797e] test python loss on gpu [f5d94a8] fix lint errors (+1 squashed commit) Squashed commits: [5d7c7ea] add an example to define loss function in Python
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
We already have the ability to define symbolic operators in Python. But due to multi-threading issue, this feature cause a lot of pain to implement in other languages like Julia (e.g. dmlc/MXNet.jl#16). With the new module system, we can achieve similar thing (at least for the case of simple linear chaining) by defining a loss module directly in Python.
This PR is a simple demo of defining a multiclass-hinge loss and train on MNIST.
Later on (???) I will try to port the module system to Julia.