HLS4ML: Supported ML model

[Yi-Hsuan Cheng]

Dear HLS4ML representatives,

 

 

My name is Ian Cheng and I am currently a master student in electrical engineering.

I really appreciate your nice work on the HLS4ML package, and I have a simple question to ask.

 

I have 2 trained models in ONNX and TensorFlow SavedModel format(for object detection, which is Conv2D). Does HLS4ML support such format to be translated?

 

Thanks for your support and wish you have a great weekend.

 

 

Best,

Ian Cheng

 

 

[Yi-Hsuan Cheng]

Dear HLS4ML representative,

 

 

I would like to revise my words. Model can either be ONNX(.onnx), TensorRT(.engine), TensorFlow GraphDef(.pb), TensorFlow Lite(.tflite) and TensorFlow SavedModel format(entire folder).

 

Please kindly advise me if any of the aforementioned format is supported for the conversion.

 

Thanks again and I will look forward to hearing from you.

 

 

 

Best,

Ian Cheng

[Yi-Hsuan Cheng]

Dear Vladimir,

I saw on the Status and Features page, ONNX is only supported for MLP architectures.

And I am working on Conv2D neural network models(CNN). What kind of models are supported for Conv2D? I saw Keras/TensorFlow/QKeras are supported. For TensorFlow, what particular file format is supported?

I am getting the model from YOLOv5, is there any model that is supported from the list here?(https://github.com/ultralytics/yolov5/issues/251)

Thanks agian for your help.

Wish you have a really nice day.

Best,

Ian Cheng 

---

Von: Vladimir Loncar <vladimi...@cern.ch>
Gesendet: Montag, 31. Januar 2022 11:11
An: Cheng, Yi-Hsuan
Cc: hls4m...@gmail.com
Betreff: Re: [hls4ml.help] RE: HLS4ML: Supported ML model

 

Dear Ian,

Thank you for your interest in hls4ml. You should try your model with
the ONNX converter. See the docs page here for the exact function:
https://fastmachinelearning.org/hls4ml/autodoc/hls4ml.converters.html#hls4ml.converters.convert_from_onnx_model

We had some support for frozen GraphDef format (.pb) but that converter
is not maintained and shouldn't be used in its current state. The other
formats you mention are not supported.

Regards,
Vladimir

On 30/01/2022 3:42 pm, Yi-Hsuan Cheng wrote:
> Dear HLS4ML representative,
>
> I would like to revise my words. Model can either be ONNX(.onnx),
> TensorRT(.engine), TensorFlow GraphDef(.pb), TensorFlow Lite(.tflite)
> and TensorFlow SavedModel format(entire folder).
>
> Please kindly advise me if any of the aforementioned format is supported
> for the conversion.
>
> Thanks again and I will look forward to hearing from you.
>
> Best,
>
> Ian Cheng
>

> *From:* Cheng, Yi-Hsuan
> *Sent:* Saturday, January 29, 2022 5:35 PM
> *To:* 'hls4m...@gmail.com' <hls4m...@gmail.com>
> *Subject:* HLS4ML: Supported ML model

>
> Dear HLS4ML representatives,
>
> My name is Ian Cheng and I am currently a master student in electrical
> engineering.
>
> I really appreciate your nice work on the HLS4ML package, and I have a
> simple question to ask.
>
> I have 2 trained models in ONNX and TensorFlow SavedModel format(for
> object detection, which is Conv2D). Does HLS4ML support such format to
> be translated?
>
> Thanks for your support and wish you have a great weekend.
>
> Best,
>
> Ian Cheng
>

> --
> You received this message because you are subscribed to the Google
> Groups "hls4ml help" group.
> To unsubscribe from this group and stop receiving emails from it, send
> an email to hls4mlhelp+...@googlegroups.com
> <mailto:hls4mlhelp+...@googlegroups.com>.
> To view this discussion on the web visit
> https://groups.google.com/d/msgid/hls4mlhelp/927582b9baa84e14a0e78473be1fdb3d%40mailbox.tu-dresden.de
> <https://groups.google.com/d/msgid/hls4mlhelp/927582b9baa84e14a0e78473be1fdb3d%40mailbox.tu-dresden.de?utm_medium=email&utm_source=footer>.