SaeHie Park

[one-cmds] Reduce test time

2

Like current `luci-value-test` takes long time to test and #12071 is to provide alternative method to reduce time, it would be wonderful if we could reduce test duration somehow.

CFE module to provide kernel(compute) codes

7

This issue is added to discuss to provide(separate) kernel(compute) codes to reduce module responsibility and to provide common module for it's purpose. From https://github.com/Samsung/ONE/pull/10082#discussion_r1028644584 and https://github.com/Samsung/ONE/issues/10066#issuecomment-1322766629

[mio-circle] Reduce duplicates

1

from https://github.com/Samsung/ONE/pull/12195#pullrequestreview-1764529666 by @mhs4670go --- I come up with two options to reduce duplicate codes (of course we don't have to reduce them) 1. If we pull out duplicate APIs...

[circle2circle] loco::must_cast() failed to cast with lack of information

2

We sometimes get > circle2circle: ERROR: loco::must_cast() failed to cast: PN4luci11CircleConstE with lack of node information for finding which node may caused the problem. It would be better to provide...

[luci-interpreter] Unknown Op of "Unsupported type." message

3

Current `luci-interpreter` shows just "Unsupported type." without which Op has the problem. Lets' show some more information.

Fix dagre.js for abnormal wide edges

6

This will update dagre.js for abnormal wide edges

[CFe] Support Relu0To1 Op

2

Let's support `Relu0To1` Op which is $`f(x) = max(0, min(1, x))`$ - [x] res/tfliterecipe - [x] tflchef - [x] tfldump - [x] tflite2circle - [x] circledump - [x] luci -...

[one-cmds/one-import] Fix "--verbose" option

3

~`one-import` and `one-import-*` doesn't provide `--verbose` option.~ - `one-import-onnx` does have `--verbose` but changing `TF_CPP_MIN_LOG_LEVEL` doesn't seem to work these have `sys.tracebacklimit = 0` to silence unnecessary messages, but for...

[compiler] Upgrade ONNXRuntime

6

ONNX 1.16.0 was released which provides INT16 QuantizeLinear/DequantizeLinear. Let's upgrade!

DRAFT upgrade onnx

2

on-going draft to upgrade onnx and onnxruntime.