Comments (1)
For the first part (where the all alive parameters equals to 112422),
the nonzero numbers mean the number of nonzero weight values.
for the second part (after Huffman encoding),
the numbers mean the number of Bytes needed to store weights.
Note that I didn't encode the biases so they use precisely 4 Bytes for each weight.
However, for normal weights (fc1.weight, fc2.weight, fc3.weight),
I used either CSC or CSR matrix representation (https://en.wikipedia.org/wiki/Sparse_matrix).
(The paper also uses these representations with a minor variation)
So they use a bit fewer Bytes per each weight.
For example, for fc1 layer, 858868/235200 ~= 3.65. It uses approximately 3.65 bytes per weight
for fc2 layer, 35108/30000 ~= 1.17. It uses approximately 1.17 bytes per weight.
(Note that this efficiency comes from the sparsity of fc2 weights)
Maybe I'll add more information about the numbers in the future.
from deep-compression-pytorch.
Related Issues (13)
- Model Size Problem HOT 2
- AttributeError: 'ReLU' object has no attribute 'weight'
- problem in dataloader HOT 2
- dimension error
- weight_shared.py run false
- Using Deep-Compression in MobileNet-V2
- running bug: HOT 1
- pytorch version HOT 1
- Using LeNet5 HOT 1
- some problem in quantization HOT 3
- Darknet weights Pruning, Quantisation and Huffman Encoding!
- Custom architecture HOT 1
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 deep-compression-pytorch.