Completed Model Development #3
Conversation
| ## Training Steps | ||
|
|
||
| 1. Clone the repository | ||
| 2. Create a data, logs and models directory in the root |
There was a problem hiding this comment.
@sshivaditya2019 Instead of explaining this in the documentation, can you create the directories data, logs and models, and remove this step from the README? Just to make it as easy as possible to reproduce the steps.
There was a problem hiding this comment.
Ok will add the same
There was a problem hiding this comment.
@sshivaditya2019 Or generate these folders using Python code. That would probably be the most robust solution?
|
|
||
| 1. Clone the repository | ||
| 2. Create a data, logs and models directory in the root | ||
| 3. Create a sub directory named nummodel inside the models and logs directory |
There was a problem hiding this comment.
@sshivaditya2019 Same comment as above.
There was a problem hiding this comment.
@sshivaditya2019 I'm getting this error message when running train.py:
Data Prepping
Data Shuffled
Data Generation completed
Total training Samples:40000
Toal validation samples: 5000
Total test samples: 5000
Traceback (most recent call last):
File "src/nummodel/train.py", line 134, in <module>
main()
File "src/nummodel/train.py", line 35, in main
train_img_paths,train_labels,validation_img_paths,validation_labels,test_img_paths,test_labels = get_lab()
File "src/nummodel/data.py", line 100, in get_lab
train_img_paths, train_labels = get_image_paths_and_labels(train_samples)
File "src/nummodel/data.py", line 94, in get_image_paths_and_labels
img.save(pa, "PNG")
File "venv/lib/python3.8/site-packages/PIL/Image.py", line 2297, in save
fp = builtins.open(filename, "w+b")
FileNotFoundError: [Errno 2] No such file or directory: 'data/data/nums/10060_6595.png'Here is how to reproduce:
pip install tensorflow
pip install matplotlib
pip install tf2onnx
mkdir data
python3 src/nummodel/train.py| ## Training Steps | ||
|
|
||
| 1. Clone the repository | ||
| 2. Create a data, logs and models directory in the root |
There was a problem hiding this comment.
@sshivaditya2019 Or generate these folders using Python code. That would probably be the most robust solution?
The tflite still does not support the Greedy CTC Decoder. Working on an alternative. The onnx version works on android using onnx-runtime-android(aar)