lutzroeder / netron Goto Github PK
View Code? Open in Web Editor NEWVisualizer for neural network, deep learning and machine learning models
Home Page: https://netron.app
License: MIT License
Visualizer for neural network, deep learning and machine learning models
Home Page: https://netron.app
License: MIT License
Hi there, this is a really nice tool.
I noticed an issue with the way attribute values are displayed.
Take a look at this file (exported from CNTK): CNTK-ConvNet-trained.pb.zip
Notice the node Plus30
and its attribures broadcast
, axis
. They both show up as 0 in Netron, even though they should both be set to 1.
Similarly many Constant
ops in the graph have different value
attributes, but they all show up as 0.
Using version 0.6.0
When dealing with a large graph, it would be great to hit Ctrl-F/Cmd-F and open a search dialog.
All operations matching search could be highlighted in the graph, if only one is found zoom and pan could bring it into focus.
It would be great to search by name, but attribute values, operation types, etc. could also generate matches.
I'm wonder wit thise code... is amazing. I have a version with the posibility to add a background grid and the color of the grid can be changed.
Don't see how i cand send you. But if you are interested i can e-mail to yo.
Thanks!
DP: Sorry my english isn't very good!
import onnx
import netron
#netron.serve_file("output/alexnet.onnx")
model = onnx.load("output/alexnet.onnx")
#onnx.checker.check_model(model)
#print(onnx.helper.printable_graph(model.graph))
ImportError Traceback (most recent call last)
in ()
1 import onnx
----> 2 import netron
3
4 #netron.serve_file("output/alexnet.onnx")
5 # Load the ONNX model
/usr/local/lib/python3.5/dist-packages/netron/init.py in ()
1 #!/usr/bin/python
2
----> 3 from netron import serve_data
4 from netron import serve_file
ImportError: cannot import name 'serve_data'
Example: Inception_v2/model.db
This is the problem... and the solution in StackOverflow
This example is very beatifull... but i can't have the skills to understand this code, to do it my self.
Can you add a 90ΒΊ between entities? because if you add so much entities the connector lines can be confused between entities.
DP: sorry my english isn't very good.
Investigate adding canvas limits and using browser scrollbars.
now it's run in localhost:8080,i want run it in the * ,and have i can stop the Processing
Hi,
I use virtualenv and when I install netron, the python path is wrong in the netron executable script. I have to change that first line in the netron script (which indicates where python is located) to the python executable in the virtualenv. Just to let you know this could be a nuisance for others using virtualenv...
Best wishes,
Tian Jin.
Examples:
Exermote.onnx.zip
DocumentClassification.onnx.zip
Kmeans.onnx.zip
squeezenet1.1_shape.onnx.zip
sparse_const.zip
bertsquad-10.zip
bvlc_alexnet_opset_3.onnx.zip
resnet50_opset_9.onnx.zip
ssd.onnx.zip
maskrcnn.onnx.zip
yolov3.onnx.zip
keras2onnx.onnx.zip
vgg19.onnx.zip
bidaf-9.onnx.zip
inception_v1_opset_9.onnx.zip
inception_v2_opset_6.onnx.zip
inception_v2_opset_9.onnx.zip
gpt2.onnx.zip
gpt2-10.onnx.zip
bert-base-uncased.onnx.zip
transformer.onnx.zip
maskDet_opt.onnx.zip
issue_845.onnx.zip
input_0.pb.zip
fp8.onnx.zip
bfloat16.onnx.zip
arcface-resnet100.onnx.zip [source]
gather.json.zip
nms_base_component.json.zip
candy.json.zip
optional_type.json.zip
zipmap_int64float.json.zip
int4.int32_data.pb.zip
test_lstm_with_peepholes.onnx.zip
null_graph.onnx.zip
phi3-mini-128k-instruct-cuda-fp16.onnx.zip
test_LSTM_tanh_bidirectional.onnx.zip
test_RNN_bidirectional_one_layer_relu.onnx.zip
Tools:
onnx_strip_weights.zip
root@instance-1:/opt/jupyter/study_pytorch/lpth/important_rnn/onnx/Netron-1.1.0# python3 setup.py build
running build
running build_py
error: can't copy 'node_modules/protobufjs/dist/protobuf.js': doesn't exist or not a regular file
Could you declare a license for this project? This project cannot be used without an open source license, so if you could update the project with your intended license, that would be great.
It would be great to use some colors to add more information to the visual representation of graph. Constants could be one color, activation functions another color, etc.
When you do doble click over an element and you have the shift key pressed... create an error:
This is the info that I can get from the browser:
"Line: 1203
Error: Unable to get value of the property 'hitTest': object is null or undefined"
DP: Sorry my English isn't very good.
This field is populated by front-ends with human-readable information about each node. For example, Pytorch populates this with stack trace information. Can we make this field accessible in the visualization?
We try to use it as a preferred extension because .pb is quite ambiguous. Should be just a matter of changing a filter in the open dialog.
python netron.py model.db
Operator is always visible and show open node details page. Drilling into documentation should happen from node details page.
on macOS:
netron my_model.onnx
A declarative, efficient, and flexible JavaScript library for building user interfaces.
π Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
The Web framework for perfectionists with deadlines.
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. πππ
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
We are working to build community through open source technology. NB: members must have two-factor auth.
Open source projects and samples from Microsoft.
Google β€οΈ Open Source for everyone.
Alibaba Open Source for everyone
Data-Driven Documents codes.
China tencent open source team.