Add inference demos

[coyotte508]

Add demos on https://huggingface.co/huggingfacejs (feel free to contribute demos, or to ask joining the organization)

Natural Language processing

• Fill mask
• Summarization
• Question answering
• Table question answering
• Text classification
• Text generation - demo
• Text2Text generation
• Token classification
• Named entity recognition
• Translation
• Zero-shot classification
• Conversational
• Feature extraction
• Sentence Similarity

Audio

• Automatic speech recognition - demo
• Audio classification - demo

Computer Vision

• Image classification - demo
• Object detection - demo
• Image segmentation
• Text to image
• Image to image
• Image to text - demo

Multimodal

• Document question answering - demo
• Visual question answering - demo

Tabular

• Tabular regression

[fikstoday]

Recently used facebook/nllb-200-distilled-600M as model for translation task. Documentation was not clear
After looking at other examples and many errors later, this worked.

let result = await hf.translation({
        model: "facebook/nllb-200-distilled-600M",
        inputs: "how is the weather like in Gaborone",
        parameters : {
        src_lang: "eng_Latn",
        tgt_lang: "sot_Latn" }
      })

[julien-c]

@fikstoday want to open a PR to add it to the docs?

[fikstoday]

Yes Yes
Where would this snippet be best placed?

[ArsalaBangash]

I'm working on an object detection demo on this space

It's currently using Vue because I have faster prototyping speed with it, but it can be taken out once the demos are ready.

I'm actively creating more inference demos for my work with Scrimba, so I'm ready to be more involved with this issue.

[coyotte508]

I'm working on an object detection demo on this space

It's currently using Vue because I have faster prototyping speed with it, but it can be taken out once the demos are ready.

I'm actively creating more inference demos for my work with Scrimba, so I'm ready to be more involved with this issue.

Nice! When it's ready we can duplicate it to https://huggingface.co/huggingfacejs

Vue is no problem, it's nice to have different frameworks

[ArsalaBangash]

My Object Detection Demo is ready for review.

I welcome your feedback!

I'll be tackling image classification next.

[coyotte508]

When clicking on "detect objects" directly, without doing anything else, I get this error:

[ArsalaBangash]

@coyotte508 I addressed the issue and pushed my changes. It should work now. The models don't always detect objects so you may have to shuffle a couple times.

[coyotte508]

Thank you, it works really nice!

Duplicated here: https://huggingface.co/spaces/huggingfacejs/object-detection-vue/tree/main

[ArsalaBangash]

The Image Classification Demo is ready to review!

[julien-c]

really cool Space @ArsalaBangash!

[ArsalaBangash]

Demos for Speech Recognition and Audio Classification are ready to review!

[ArsalaBangash]

I experimented with adding params for response length and temperature in this Text Summarization Demo, but I'd like some ML guidance on the values for the temperature param.

Beyond a value of 1, the results started looking like cgfrS$#%G~VygHB YT5>d%~l;mK~nj$#$#SD^X~C GFgjbNnV Gg~ ?

I have a few questions:

1. Is it correct to label temperature as a param that controls randomness?
2. What are reasonable options for the temperature param, and does it depend on the model?
3. What other detailed params can I showcase that are easily explained without using deeper ML terminology?

[coyotte508]

I experimented with adding params for response length and temperature in this Text Summarization Demo, but I'd like some ML guidance on the values for the temperature param.

Beyond a value of 1, the results started looking like cgfrS$#%G~VygHB YT5>d%~l;mK~nj$#$#SD^X~C GFgjbNnV Gg~ ?

I have a few questions:

1. Is it correct to label temperature as a param that controls randomness?

2. What are reasonable options for the temperature param, and does it depend on the model?

3. What other detailed params can I showcase that are easily explained without using deeper ML terminology?

cc @radames maybe?