Eunwoo Shin

icon indicating an email [email protected]

icon company Intel icon location Seoul

Results 10 comments of Eunwoo Shin

[HPO] fix bug occurred when running HPO without batch size in search space

run ote-test

Overriding issue for optional dataclass argument

are there any updates here?

Register signal handler for mp.Manager in MemCacheHandlerForMP

@vinnamkim I implemented again based on your comments. Could you check again?

Use model_api action classification model wrapper class

> @eunwoosh this PR is not required for 2.1, so please retarget back to develop Ok I changed target branch to develop

Set mem_cache_img_min_size to input_size if it's none

> Could you elaborate why you changed `max` -> `min`? `max` might be semantically right as the cache would downscale if input image is larger than the criteria which is...

Problems in multi-card distributed training

Hi @nowbug , thanks for finding the issue. First of all I want to say that OTX 2.0 currently doesn't validate distributed training, so it can be a little bit...

Upgrade NNCF and OV versions

Here is validation result. Benchmark medium dataset was used. Task | Model | FP32 acc diff | INT8 acc diff -- | -- | -- | -- multi_class_cls | deit_tiny...

Feature Request: Support for 16 bit integer or 32 bit float single channel images

Hi @j99ca, As Harim said, configurable input size won't be included in OTX 2.0 release, but we have a plan to enable it later. When it's enabled isn't decided but...

Feature Request: Support custom and non-square input sizes

Hi @j99ca , #3759 is preparation step for configurable input size. That PR just enables transforms in recipe to use `$(input_size)`. I'm now implementing configurable input size using #3759. Currently,...

Feature Request: Support custom and non-square input sizes

#3788 is merged. OTX supports non-square input size now.