blueoil

blueoil copied to clipboard

Remove resize from bisenet augmentor

2

## What this patch does to fix the issue. Resize is not an augmentor and it is a duplicate of Resize in preprocessor. https://github.com/blue-oil/blueoil/blob/da0e68b1f319a4cc23044b594aa399c2a2f245d0/blueoil/configs/core/segmentation/lm_bisenet_quantize_camvid.py#L72-L75 ## Link to any relevant issues...

yd8534976

Remove config yaml

2

The `config.yaml` file contains the same information as the `config.py` file. So, it is possible to remove the code related to `config.yaml`. This is stated in https://github.com/blue-oil/blueoil/issues/1224. A second point...

joelN123

The usual, expected behaviour is that when the training has finished, all the resource (GPU) is freed up and the docker container doesn't run anymore. And this does happen in...

joelN123

Stop to save config.yaml

2

Currently, we are saving `config.yaml` the contents is same as Python config. IMO, nobody are using `config.yaml`. https://github.com/blue-oil/blueoil/blob/master/blueoil/utils/config.py#L214-L236

iizukak

Create prototype of unit-test

5

Related to #743 Using googletest ([https://github.com/google/googletest]) and DownloadProject ([https://github.com/Crascit/DownloadProject]) All .cpp files in "tests/converter/unit-test/src" are unit test file of converter C++ runtime. And all header files in "tests/converter/unit-test/include" are dummy...

lm-kajihara

Network::run() should take void* (or byte*) instead of float*

1

We want to take uint8_t, uint16_t or something else typed buffer for performance, memory usage optimization. Furthermore, we may want to take multiple input/output, it can be mixed-typed buffers. But...

primenumber

We do not use `output_event.py`. So we can remove. https://github.com/blue-oil/blueoil/blob/master/blueoil/cmd/output_event.py Related to this issue, We also can remove `pytablewriter` dependency, Because `pytablewriter` is only used from `output_event.py`.

iizukak

Since `pass_propagate_format` doesn't consider `Reshape` operator, convert will fail. Insert `Cast` operator before `Reshape` operator when input is quantized.

primenumber

In the latest blueoil, the following error occurs when `blueoil train`: ``` Traceback (most recent call last): File "/usr/local/lib/python3.6/dist-packages/pkg_resources/__init__.py", line 567, in _build_master ws.require(__requires__) File "/usr/local/lib/python3.6/dist-packages/pkg_resources/__init__.py", line 884, in require...

Suikaba

Since we already have TensorView class, we can rewrite all parts which depends on MatrixView class with TensorView class.

tkng