amused-512-ONNX-WebGPU

amused/amused-512 converted to ONNX for running in a web browser with ONNX Runtime Web's WebGPU backend, as used by Quick Inference.

Changes from the original

• Converted to ONNX (opset 17) with PyTorch's exporter: text_encoder, transformer and vq_decoder.
• The text encoder returns its penultimate hidden layer and pooled projection, as aMUSEd uses them.
• The transformer runs the prompt and the negative prompt together and returns their guided mix (uncond + guidance · (cond − uncond)), one row of logits per image token; float16 weights and output.
• The VQ decoder takes token ids and returns the image in [−1, 1]; float32.
• Sampling and the masking schedule run outside the graphs (as diffusers' AmusedScheduler).

These are modified files: their outputs are close to, but not identical to, the original model's.

Licence and use restrictions

Licensed under the CreativeML Open RAIL++-M License, as the original (LICENSE.md). Its use-based restrictions (paragraph 5 and Attachment A) apply to these files and to anything made from them: you may not use them for the uses listed in Attachment A, and if you distribute them or a derivative you must include those restrictions as an enforceable provision and give recipients a copy of the licence.