Modify inference engine

[chunyang-wen]

It seems that engine module under inference package is not well developed. This PR tries to

• Replace assert with raise and additional information. Actually assert can be turned off sometimes.
• Remove useless import
• Fix a bug about return_tuple
• modify the function name of _check_quantize_setting to _init_quantization_setting

More documentation about this module is needed.

[chunyang-wen]

ping @RezaYazdaniAminabadi

[RezaYazdaniAminabadi]

Hi @chunyang-wen

Thanks for going through the inference engine-API and make these nice modifications. I will put more information on the parameters description.
Thanks,
Reza

[chunyang-wen]

So should we merge this PR first and then you add further improvement or I just close it?

[RezaYazdaniAminabadi]

So should we merge this PR first and then you add further improvement or I just close it?

No, let me please add them and we can merge it soon. Thanks

[chunyang-wen]

So should we merge this PR first and then you add further improvement or I just close it?

No, let me please add them and we can merge it soon. Thanks

Sure, thanks.

[chunyang-wen]

Have no idea why the check fails. From other build break logs, it seems that certain problems happen to the distributed runtime, such as port conflict or connection reset by peer.

[chunyang-wen]

@jeffra please merge or open auto merge. Thanks.

[jeffra]

We had a CI runner failure earlier due to some new runners being misconfigured. Should be fixed now.