Some optional parameters have a default of None, which is more akin to nil or the Pythonic None than 0. These parameters need to be made into pointers so that the unset value is nil rather than 0.

As an example, see the ConversationalParameters MaxTime struct. https://github.com/TannerKvarfordt/hfapigo/blob/790bb2db9734e57d6d82e160195f20d1ccf20eb2/conversational.go#L57-L59

The Issue

hfapigo calls to fill_mask endpoints do not support multi-input, multi-mask arguments. Supported arguments are multi-input or multi-mask, but not both at the same time.

To be clear, some HuggingFace models do support these calls, but this library does not.

For example, the following calls are supported by this library.

// Supported
hfapigo.SendFillMaskRequest(hfapigo.RecommendedFillMaskModel, &hfapigo.FillMaskRequest{
	Inputs:  [
		"The answer to life, the universe, and everything is [MASK].",
        ],
})

// Supported
hfapigo.SendFillMaskRequest(hfapigo.RecommendedFillMaskModel, &hfapigo.FillMaskRequest{
	Inputs:  [
		"The answer to life, the universe, and everything is [MASK].",
		"So long, and thanks for all the [MASK].",
        ],
})

// Supported
hfapigo.SendFillMaskRequest(hfapigo.RecommendedFillMaskModel, &hfapigo.FillMaskRequest{
	Inputs:  [
		"The answer to life, the [MASK], and everything is [MASK].",
        ],
})

And the following calls are not supported by this library.

// NOT Supported
hfapigo.SendFillMaskRequest(hfapigo.RecommendedFillMaskModel, &hfapigo.FillMaskRequest{
	Inputs:  [
		"The answer to life, the [MASK], and everything is [MASK].",
                "So [MASK] and thanks for all the [MASK]",
        ],
})

// NOT Supported
hfapigo.SendFillMaskRequest(hfapigo.RecommendedFillMaskModel, &hfapigo.FillMaskRequest{
	Inputs:  [
		"The answer to life, the [MASK], and everything is [MASK].",
                "The answer to life, the [MASK], and everything is 42.",
        ],
})

The reason these types of calls are not supported is that it quickly becomes non-trivial to deduce the structure of the response. I'm sure it is possible, but I don't foresee myself having time to work on this in the near future.

Workaround

If you need to make a call of this type to the API, you can build your FillMaskRequest and manually make the call to MakeHFAPIRequest, which will give you the appropriate JSON response. From there it is up to you to deduce the structure of the returned JSON and unmarshal it.

As of the initial writing of this library, the feature extraction endpoint doesn't seem to be well defined. At the moment it seems like the request structure varies from model to model. That may or may not be the case, and its likely that I just don't understand what's going on, but unfortunately I don't have the free time to expend on figuring out what's going on. I'll wait until the docs become more explicit to implement this endpoint.