Python InferInput: set_data_from_numpy is not necessary in some cases.

[RRoundTable]

Objective

In python, InferInput only support set_data_from_numpy. It means that users have to convert their data to numpy array even though it it not necessary.

So, I added an interface that can set raw_content without converting to numpy array.

Detail

When model in triton server can process data that has bytes datatype, User don't have to converting it to numpy array.

For example, if we use dali as preprocessor, dali can process bytes data.

[matthewkotila]

Hi @RRoundTable, thanks for making these changes. Can you fill out and email the CLA as described here?:

https://github.com/triton-inference-server/server/blob/main/CONTRIBUTING.md#contributor-license-agreement-cla

[RRoundTable]

Hi @RRoundTable, thanks for making these changes. Can you fill out and email the CLA as described here?:

https://github.com/triton-inference-server/server/blob/main/CONTRIBUTING.md#contributor-license-agreement-cla

Sorry. I checked late. I will fill out and email the CLA and let you know after that.

[matthewkotila]

@matthewkotila: Hi @RRoundTable, thanks for making these changes. Can you fill out and email the CLA as described here?: https://github.com/triton-inference-server/server/blob/main/CONTRIBUTING.md#contributor-license-agreement-cla

@RRoundTable: Sorry. I checked late. I will fill out and email the CLA and let you know after that.

No worries 🙏

[RRoundTable]

Hi. I send an email with CCLA. Thanks.

[matthewkotila]

@RRoundTable I see the PR was closed--do you still want to contribute these changes (or anything to the triton-inference-server organization)?

[RRoundTable]

@matthewkotila No I don't Thanks!