Troubleshooting

Padding is undefined

If specifying a patch size but not padding, you will likely encounter artifacting in the upscaled image.

Most of the time, this artifacting is undesired. To resolve the artifacting, add an explicit padding:

upscaler.upscale('/path/to/img', {
  patchSize: 64,
  padding: 4,
})

If you would like to keep artifacting but hide the warning message, pass an explicit padding value of 0:

upscaler.upscale('/path/to/img', {
  patchSize: 64,
  padding: 0,
})

Progress Specified Without Patch Size

If you've specified a progress callback but are not specifying patchSize in the call to upscale, the progress callback will never be called. progress callbacks only occur when patchSize is provided.

In order to have your progress callback be called, provide explicit patch sizes:

upscaler.upscale('/path/to/img', {
  patchSize: 64,
  progress: ...
})

Unexpected Token

If you're in a Node.js environment, you may encounter something like this:

/node_modules/upscaler/dist/browser/esm/index.js:1
export { default, } from './upscaler';
^^^^^^

SyntaxError: Unexpected token 'export'

This likely means one of two things:

• You are using upscaler, instead of upscaler/node; check out the guide on Node.js here.
• You are using import syntax instead of require syntax; if so, try switching to require('upscaler'). For more information on this, see this Github issue.

Missing Model Path

If you see an error like:

Error: You must provide a "path" when providing a model definition

You've passed a null or undefined path argument in the model argument to UpscalerJS:

const upscaler = new Upscaler({
  model: {
    path: null,
  },
})

Ensure you pass a valid path argument in the model payload. See the guide on models for more information.

Missing Model Scale

If you see an error like:

Error: You must provide a "scale" when providing a model definition

You've passed a null or undefined scale argument in the model argument to UpscalerJS:

const upscaler = new Upscaler({
  model: {
    scale: null,
  },
})

Every model must have an explicit scale defined.

Ensure you pass a valid scale argument in the model payload. See the guide on models for more information.

Invalid Warmup Value

If you see an error like:

Error: Invalid value passed to warmup in warmupSizes: foo

It means you've called .warmup with an invalid value:

upscaler.warmup('foo')

Ensure you're passing one of the following:

• { patchSize: number; padding: number }
• [width, height]
• { patchSize: number; padding: number }[]
• [width, height][]

For more information, see the guide on warm ups, or review the API documentation on the warmup method.

Environment disallows String Input

If you see an error like:

Error: Environment does not support a string URL as an input format.

This means that the environment UpscalerJS is running in does not have access to Image. This means that it cannot construct an image and load its src.

Examples of such environments include web workers and the like (e.g., Cloudflare workers).

To get around this error, pass your data into UpscalerJS as a tensor. For an example, check out the guide on web workers.

Invalid Predicted Tensor

UpscalerJS only supports models that return rank 4 tensors representing image-like data.

You may see this message if you're using a custom model that returns a rank 3 image, or non-image-like data (like text).

If you believe this is in error, or you have a particular use case you think would be appropriate for UpscalerJS, please open a Github issue.

Environment disallows Base64

If you see an error like:

Error: Environment does not support base64 as an output format.

This means that the environment UpscalerJS is running in does not have access to Image or document. This means that it cannot automatically transform a tensor into a base64 representation.

Examples of such environments include web workers and the like (e.g., Cloudflare workers).

To get around this error, specify tensor as the output format of your data, like:

upscaler.upscale(tensor, {
  output: 'tensor',
})

Note: If you've neglected to provide an output argument and you're running the browser version of UpscalerJS, you will need to explicitly provide the output argument. This is because the default argument is base64.

For an example, check out the guide on web workers.

Invalid Model Type

Acceptable model types include:

• layers (default)
• graph

These correspond to the model types supported by Tensorflow.js.

If you don't know what kind of model you have, you may be able to inspect your model.json. If the JSON file has an entry like below, it's a layers model:

{"format": "layers-model", 

If it has an entry like below, it's a graph model:

{"format": "graph-model", 

Error with Model Input Shape

This error implies that the given model does not have an input layer compatible with a rank 4 tensor. It may be expected a rank 3 tensor, or something else entirely.

UpscalerJS only supports models whose input layers are set up to accept rank 4 tensors.

If you believe this is in error, or you have a particular use case you think would be appropriate for UpscalerJS, please open a Github issue.