Comments (4)
This is likely related to #25
from gradientaccumulator.
Update: This does not happen after loading. It happens after model wrapping. It is likely due to the usage of tf.Variable
for creating the gradients, which should be accumulated (see here).
Also note that by calling the reinit_grad_accum
method, externally, further increases the nontrainable weights count (see here).
Both behaviours are wrong and needs proper solutions.
from gradientaccumulator.
We may have to do something like this.
from gradientaccumulator.
Unfortunately, double parameters during gradient accumulation is expected behavior due to the accumulation of gradients for each parameter separately using the gradient_accumulation variables.
from gradientaccumulator.
Related Issues (20)
- Use tf.function on train_step HOT 11
- 0.5.1, tf 2.11 error for accuoptimizer HOT 8
- Replacing AccumBatchNormalization not working as intended HOT 2
- ConvNeXt not compatible with Model wrapper HOT 1
- No mixed precision support with GradientAccumulateOptimizer? HOT 7
- Replacing BN layer with AccumBN layer results in poorer convergence
- confusion over how to use this module HOT 2
- Dummy issue to test auto-assign
- Dummy issue to test auto-assign
- Test HOT 1
- Review and potentially simplify the implementation HOT 25
- raise ValueError('Optimizer must have a "lr" attribute.') HOT 5
- AccumBN is not compatible with 3D ops e.g. Conv3D HOT 1
- Mixed precision not working as intended with AccumBatchNormalization HOT 8
- Add linting to improve code style HOT 1
- Unit test for optimizer invariance in distributed trainings HOT 2
- Optimizer wrapper not working as intended HOT 2
- Optimizer wrapper not compatible with tf==2.6 HOT 2
- AttributeError using Optimizer wrapper with tf==2.4 HOT 1
- Optimizer wrapper performance is dependent on tensorflow version HOT 3
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from gradientaccumulator.